Printer Friendly

Anglicisms and German near-synonyms. What lexical co-occurrence reveals about their meanings.

1. Introduction

It is a well-known fact that borrowing is driven not only by naming need or the need to fill lexical gaps in the receptor language. Quite frequently, a loan word displays a meaning component or a connotative nuance that is not inherent to a semantically similar native word and thus allows for lexical differentiation. Anglicisms in German are a case in point. For example, while lexical items like Quad, beamen or smart were required to denote concepts for which German lacked adequate expressions, loan words such as Drink, Song or shoppen semantically correspond with Getrank, Lied, and einkaufen. Nevertheless, these pairs are not completely synonymous because there are very subtle contrasts. (1)

Although there is a vast amount of literature on anglicisms in German, contrastive analyses are restricted to the studies of Yang (1990), Kettemann (2006) and Onysko & Winter-Froemel (2011) so far, and apart from Yang's attempt to describe semantic contrasts in terms of word-field theory, a theoretical framework is missing (cf. Baeskow (2018) for an overview of these studies). Moreover, computer-based corpus analyses are still at a very early stage in this field. While English loan words and the contexts in which they occur had to be manually collected and saved on filing cards by the time Yang wrote his thesis on anglicisms in the German newsmagazine Der Spiegel, Kettemann's analyses benefit from COSMAS II, a linguistic corpus provided by the IDS Mannheim, whose concordances allow him to describe the distribution of cool, shoppen, Event and their native near-equivalents kuhl, einkaufen and Ereignis in German newspapers. The analyses performed by Onysko & Winter-Froemel (2011), which aim at a pragmatic classification of anglicisms in German, are based on Spiegel 2000 -- an electronic resource comprising Spiegel editions from 1994 up to 2000. (2) While these studies--like the present article--focus on anglicisms and German near-synonyms, Sosnizka (2014) compares collocational patterns of the source and target language in electronically accessible German and American business and news magazines by using WordSmith Tools. Her analyses follow the tradition of British contextualism (e.g. Firth 1957, Jones & Sinclair 1974).

The aim of this study is to present results of innovative contrastive analyses performed on the basis of lexical co-occurrence in large quantities of text. The database for this study, which will be described in detail in section 2, comprises three journalistic corpora (the Spiegel-, Focus- and Zeit-Archiv) and two web-based corpora (Twitter and the German version of WaCky). Proceeding from co-occurrence matrices generated from these corpora for selected anglicisms and potential German equivalents, it will be argued that semantically relevant lexical items which are among the 30 most frequent co-occurrences (3) of a key word in at least two corpora used for this study reflect some generic knowledge we associate with the concept the key word denotes and thus help to identify lexical contrasts as well as semantic overlap. Furthermore, it will be shown that despite a certain degree of semantic overlap, the anglicisms convey specific meaning components or shades of meaning and thus allow for semantic and/or connotative differentiation. The search for semantically relevant information in the co-occurrence matrices was guided by the notion of qualia structure (henceforth abbreviated as QS), which is an essential component of Pustejovsky's (1996) Generative Lexicon.

The article is structured as follows: Section 2 provides an overview of the Generative Lexicon and the connotative framework. In section 3, the corpora as well as the methods developed for the contrastive analyses will be introduced, and it will be shown that qualia structures are a suitable device for information retrieval. In section 4, representative contrastive analyses of Dealer vs. Handler, Drink vs. Getrank, Song vs. Lied, Chanson, Arie, Skyline vs. Stadtsilhouette and shoppen vs. einkaufen will be presented. The article ends with a conclusion in section 6.

2. The theoretical framework

Referring to the Aristotelian philosophy, Moravcsik (1975) emphasizes that the ideas we have about objects, processes etc. are based on understanding, which goes beyond mathematical propositions and generic knowledge. Understanding a concept involves understanding that it is part of the extra-linguistic reality, where it interacts with other concepts. Moravcsik's (1981) account of understanding is intensional and suggests that the capacity to identify extensions is not part of the speakers' linguistic competence. For example, although speakers generally know the concept DISEASE, they are not necessarily able to identify instances of it without consulting a physician. While the extension remains constant, the intensional representation of concepts is incomplete, or, more precisely, it allows for variation. In particular, it is enriched from childhood to adolescence, and further change may be due to new scientific insights. Just like an adult's intensional representations of concepts differ from those of a child, the knowledge of an expert is more specific than that of the layperson. Variation as to intensional knowledge is possible even among different generations of scientists. Nevertheless, conceptual representations that differ in specificity are not mutually exclusive, but complementary, and incompleteness does not necessarily constitute an obstacle to communication.

2.1 From Aristotelian aitiae to qualia structures

In the Aristotelian philosophy, understanding a concept means knowing its aitiae, i.e. the four generative factors which define its distinctive properties, constituency, function, and coming into existence. These factors provide a common scheme for the representation of intensional knowledge--independently of its specificity. The essence of the aitiational framework is that semantic competence is measured not by the ability to distinguish an entity x from all other entities in the world, but by the degree to which an individual is able to distinguish the entity x from some family of elements y, z, w--i.e. from elements of the same semantic field. Thus, semantic competence involves for example the ability to distinguish bees from wasps, hornets etc. or to differentiate between verbs of motion such as run, walk, or gallop (Moravcsik 1981: 22-23). This assumption makes the aitiational approach a solid basis for contrastive analyses.

In Pustejovsky's Generative Lexicon (henceforth abbreviated as GL), the ensemble of generative factors, which are referred to as 'modes of explanation' (1996:76), makes up the qualia structure of a concept. (4)

* FORMAL: that which distinguishes it in a larger domain;

* CONSTITUTIVE: the relation between an object and its constituent parts;

* TELIC: its purpose and function;

* AGENTIVE: factors involved in its origin or "bringing it about".

Qualia structures, which provide a multi-dimensional representation of concepts in the form of well-defined types and relational structures (Pustejovsky 1996: 78), interact with three other levels of representation: an argument structure, an (extended) event structure, and a lexical inheritance structure. As far as argument structures are concerned, Pustejovsky (1996: 63-64) distinguishes four types of arguments, namely TRUE ARGUMENTS, which are syntactically realized, DEFAULT ARGUMENTS, i.e. "parameters which participate in the logical expressions in the qualia, but which are not necessarily expressed syntactically", e.g. John built a house out of bricks (1996: 63), SHADOW ARGUMENTS, which are semantically incorporated into the lexical item (e.g. to butter), and TRUE ADJUNCTS, which allow for temporal or spatial modification and which are not part of a lexical item's semantic representation.

As pointed out already by Moravcsik, qualia structures, which define the prototypical meaning components of concepts, constitute a suitable device for representing fine-grained contrasts between semantically similar lexical items. For example, according to Pustejovsky (1996: 77-79), the English nouns novel and dictionary are semantically related (i.e. both objects are books in a general sense). Nevertheless, they differ not only with respect to their content, but also with respect to their purpose (we read a novel, but we consult a dictionary) and their coming into being (i.e. a novel is written, whereas a dictionary is compiled). These differences, which are reflected by collocational patterns, are represented at the CONSTITUTIVE, TELIC and AGENTIVE quale respectively, as illustrated below.

(1) [mathematical expression not reproducible]

(2) [mathematical expression not reproducible]

Both novel and dictionary have an argument (y) in their argument structure that is of the ontological type 'physicalobject' and corresponds to the referential argument <R> introduced by Williams (1981) for nouns. The argument structures additionally display two default arguments D-ARG1 and D-ARG2 of the type 'human', whose variables (x) and (z), like (y), are bound in the qualia structures. In (1), (x) is related to a reading event which defines the function of the concept NOVEL, whereas (z) actively participates in the writing event that causes the coming into existence of (y). Thus, (x) and (z) correspond to the reader and the author respectively. The structure of the text associated with novel is specified in the CONSTITUTIVE quale. The qualia value 'narrative' contrasts with the value 'information', which defines the internal structure of a dictionary. Further contrasts emerge from the TELIC and AGENTIVE quales, where (x) is the agent of a consulting event and (z) actively participates in an event of compiling (y).

Qualia structures have been designed to account for context-specific interpretations of lexical items. For example, a sentence like Mary began the novel is perfectly interpretable in GL theory because the contextual sense 'began reading' is inferable from the reading event in the TELIC quale of novel--the complement of the polysemous verb to begin (cf. (1)). This mechanism is referred to as true complement coercion by Pustejovsky (1996: 115-117).

As pointed out above, the dynamics of intensional knowledge does not generally affect the exchange of information. According to Moravcsik (1981: 24), "there is enough overlap among the aitiational schemes of speakers to make communication, in most cases, possible." If speakers share information as to aitiational schemes (or qualia structures in Pustejovsky's terminology), we should expect that indicators of this common intensional knowledge are retrievable from corpora.

Evidence for this assumption comes from computational linguistics and empirical ontology. An early attempt to extract qualia-related information from an electronic corpus is made by Anick & Pustejovsky (1990). Their search is based on "collocational patterns" which linguistically reflect the content of qualia structures. For example, the collocations read a book, read a tape, use the mouse or with the mouse are suggestive of the TELIC role associated with the concepts BOOK, TAPE, or MOUSE. Anick & Pustejovsky's results were designed for information retrieval applications.

More recent works have shown that an automatic extraction of qualia relations from large corpora is possible as well. In particular, Cimiano & Wenderoth (2005), whose intention is to reveal the impact of qualia structures on natural language processes and to support the lexicographer's work, developed a system which automatically produces the qualia structures (it acquired by learning) from the World Wide Web. Poesio & Almuhareb (2008) attempt to evaluate the relevance of attributes extracted from the Web for the description of concepts, with attributes being conceived of as essential properties of concepts as defined in AI, linguistics and philosophy--including qualia roles.

Although the aims of these works are quite different from the aims of the present article, it becomes clear that aitiational schemes anticipated by Moravscik (1981: 17) "as a psychological and semantic claim" are an integral part of language use rather than theoretical constructs. The case studies in section 4 it will show that qualia-related information is also traceable in co-occurrence matrices and helps to formalize semantic contrasts between anglicisms and German near-synonyms.

2.2 Non-denotative information

Lexical items do not only have a referential function, but also convey rather heterogeneous facets of meaning which are traditionally referred to as connotations. Following Erdmann (1900), Ludwig (1986, 2002), Yang (1990: 45-46) and Fries (2007) it is assumed here that connotations comprise (a) all kinds of associations which make up the 'Nebensinn' ('by-sense') of a word, (b) its 'Gefuhlswert' ('emotional value'), (c) stylistic information, and (d) communicative-pragmatic information.

If associations generally arise in the context of a lexical item, i.e. if they are not formed in the minds of individual speakers, they are part of its connotative information content and thus constitute its 'Nebensinn' in the sense of Erdmann (1900: 82). Erdmann illustrates this by-sense for the pair Krieger "warrior" and Soldat 'soldier'. While the former is associated with fight and battle, the latter rather evokes the image of barracks and parade ground. Additionally, the noun Krieger is archaic and only of historical relevance.

Emotions are defined by Fries (2007: 298) as clusters of subjective-psychological experience and motor behaviour which are encoded in linguistic or indexical signs. As far as the emotional predisposition (abbreviated as EM) is concerned, Fries (2007: 309) introduces three independent dimensions, namely (a) emotional polarity EMpol+, EMpol-, EMpol0, (b) emotional expectation EMexp-, EMexp+, and (c) emotional intensity EMint+, EMint-. Although this classification was developed for German linguistic signs such as leider 'unfortunately', hoffentlich 'hopefully', pfui 'ugh', Bewunderung 'admiration' etc., it is also suited to describe the emotional potential of anglicisms. Emotional values are either word-inherent (e.g. [Wellness.sub.EMpol+], [stalken.sub.EMpol-] "to stalk", [Wow!.sub.EMexp-]) or unfold in the context. For example, the anglicism Crash has a pejorative connotation in colloquial speech, where it refers to a collision, or in the jargon of Stock Exchange, but it is an emotionally neutral technical term in linguistics. Lexical items are predisposed to occur in particular domains of communication, and their use may be determined by the social norm. Lexicographically, Ludwig (1986: 187-193) distinguishes between the communicative predisposition, which determines stylistic properties at the word-level (e.g., Gesicht 'face' is stylistically neutral, whereas Antlitz 'countenance' and Visage 'mug' range above and below 'neutral' respectively), and communicative-pragmatic markers, which assign lexical items to specific domains of communication (e.g. 'technical', 'group-specific'), signal a speaker's or author's emotional attitude, or predict temporal or regional restrictions for the use of a lexeme (e.g. 'archaic', 'Afro-American English').

In German, anglicisms are often used for stylistic or communicative-pragmatic reasons. Stylistic analyses of anglicisms in German have been strongly influenced by Galinsky (1963). An important notion is that of colouring ('Kolorit'), which is comparable to timbre in music. If the author of a text uses an anglicism in order to evoke a particular setting, to characterize a certain group of speakers or depict a new trend, local, social, or technical colouring (5) is being created (e.g. Sheriff, cool, Wellness-Shop). These stylistic devices enable him or her to convey a more vivid impression of his/her subject (cf. Pfitzner 1978, chapter III, Yang 1990, chapter 4, Donalies 1992: 104-106, Leutloff 2003, Gotzeler 2008: 282-289). In German youth language, anglicisms are primarily associated with social functions, which are discussed in detail by Androutsopoulos (1998, chapter 7).

3. Computerlinguistic methods and their application

As far as research on anglicisms is concerned, the studies described in this article are innovative in two ways. First, the contrastive analyses are performed within a well-established theoretical framework, and secondly, the relevant data were extracted from very large heterogeneous corpora. The texts were accessed via a searchable database (KANG) comprising ten subcorpora, which was built by Jurgen Rolshoven and his colleagues from the University of Cologne. Within the scope of a research project on anglicisms in German, co-occurrence matrices for anglicisms and semantically similar German words were automatically generated from five subcorpora--the Spiegel-, Focus- and Zeit-Archiv, Twitter, and WaCky. The Spiegel-Archiv (documented period: 1946-2015, data volume: 1.7 GB) is a collection of texts from the German newsmagazine Der Spiegel, which has always been a source of lexical innovation (cf. Carstensen 1965: 22, Onysko 2007: 98). A more recent newsmagazine is Focus, which was first published in 1993. The Focus-Archiv (data volume: 375 MB) comprises articles from the first edition up to 2015. The Zeit-Archiv (documented period: 1946-2016, data volume: 1.3 GB) is based on articles from the supra-regional German newspaper Die Zeit. The data for these three subcorpora were extracted using the open online access to the archive of each of the papers. Apart from the articles' raw text, the publishing year and month as well as the name of the source were saved in the corpus as meta data. The Twitter corpus (2011-2015), whose tweets were collected via the Twitter-API (6), was kindly provided by Prof. Dr. Chris Biemann (University of Darmstadt, now University of Hamburg) and Dr. Martin Riedl (University of Darmstadt). It is regularly updated and used for further processing at the University of Cologne. After filtering it for relevant German texts it now consists of approximately ten million tweets with a data seize of 3.5 GB including some minor meta data such as the publishing date. KANG also provides access to the German version of the Web-As-Corpus Kool Yinitiative (WaCky), which was designed by Marco Baroni and colleagues (7) to crawl the web. The German WaCky corpus represents the largest part of the data (9.5 GB) used to build KANG. (8) All the data were standardized and merged into the searchable database.

The co-occurrence matrices were generated separately for each corpus by the computer linguists from the University of Cologne. By generating co-occurrence matrices from journalistic texts and web-based corpora it was ensured that the co-occurrences are not text-type specific. Each matrix displays a hierarchy of thirty lexical items that occur most frequently in the context of (a) an anglicism and (b) a potential German equivalent in a given corpus. The following extract of the co-occurrence matrix for the key word Drink from the Spiegel-Archiv is intended to provide a first impression.

The co-occurrence values were determined by counting the occurrences of all the words in a certain context window around the term and sorting them by frequency. Semantically irrelevant words (stop words, i.e. functional items, as opposed to lexical items) and compounds were removed from the texts before applying the context window. The window breadth was fixed to [+ or -]5, i.e. the span to be considered consists of five words to the left and five words to the right of a key word. The given values are the ratios of found contexts containing the respective co-occurrence.

The ranking of the co-occurrences in terms of co-occurrence values is less relevant for the present study because it is assumed here that looking for common co-occurrences in the five corpora more efficiently contributes to the identification of concept-defining information. The basic criterion is that a word appears among the "Top 30" co-occurrences as described above, independently of its position in the matrix. As observed by Baroni & Lenci (2008), token frequency may be accidental or result from fixed expressions. Their analyses have shown for example that the fixed expression year of the tiger occurs much more frequently in their corpora than the pattern tail and tiger, which signals a semantic relation and allows for different realizations (e.g. tail of the tiger, tigers have tails, tigers with tails etc.). Thus, the analyses performed in this study are qualitative rather than quantitative.

Co-occurrence matrices deliver the raw material for knowledge representations. They do not immediately reveal recurrent patterns such as qualia relations or other pieces of lexical information, but each matrix contains clues as to the interpretation of its key word. If these clues recur in at least two co-occurrence matrices from the corpora used for this study, we may assume that they are lexically relevant. The search for useful lexical information in the co-occurrence matrices was primarily guided by the qualia-based approach. (9) Co-occurrences, which also include word-forms (e.g. singt, sang, gesungen in the case of Lied 'Song'), may be suggestive of qualia relations, qualia values, ontological types, hyponyms, arguments, fixed collocations, idioms, or connotative information.

Lexical items that contribute to the definition of a key word were manually extracted from the co-occurrence matrices and make up the co-occurrence profile (10) of the key word. While a co-occurrence matrix also contains irrelevant information, a co-occurrence profile is a subset of all the co-occurrences automatically identified for a key word which only contains interpretable 'slices' of knowledge.

Co-occurrences may be irrelevant for different reasons. For example, unlike Bar, Soft and Energy are irrelevant for the semantics of Drink (cf. Table 1) because they are part of the highly frequent compounds Soft Drink and Energy Drink. Although compounds were automatically excluded from the co-occurrence analyses, these sequences could not be prevented from entering the matrices because they do not follow the German orthographic convention to write compounds in one word (i.e. Softdrink, Energydrink). Despite their relative frequency, co-occurrences such as ersten 'first', Uhr 'clock' or Mann 'man' are irrelevant as well because they cannot be systematically related to the concept DRINK.

Co-occurrence profiles provide an appropriate starting point for the construction of knowledge representations and for the performance of contrastive analyses because they reveal contrasts as well as semantic overlap between loan words and semantically similar native equivalents. Significantly, competing near-synonyms do not imply lexical redundancy, but signal a need for semantic or connotative differentiation, which may be very subtle.

4. Contrastive analyses

Loan words frequently undergo meaning specification when they enter the receptor language. The reason is that borrowing takes place at the level of parole. In a concrete communicative situation, a particular meaning component of a loan word is required rather than its entire meaning spectrum. Thus, speakers of the receptor language tend to borrow a lexical unit, i.e. a form and one meaning component related to this form (cf. Yang 1990: 46, 167, Gevaudan 2002: 25, Onysko 2007: 16-17, Winter-Froemel 2011: 213). However, as observed by Carstensen (1965: 256), English loan words--especially older ones--do not preserve a static meaning in German, but rather undergo semantic change. In particular, they are subject to meaning extension, which manifests itself most obviously in compounding. In Baeskow (2018), the well-integrated loan words Drink, Dealer and Job were analysed as to their behaviour in the head position of hybrid N+N compounds, which were automatically extracted from the Spiegel- and Zeit-Archiv and manually selected from Cosmas II, the linguistic corpus provided by the IDS Mannheim (11). It was shown that the specific meaning the head-forming nouns were associated with when they were borrowed from English may be contextually overridden by German modifiers and that meaning extension by modification coincides with a semantic approximation of the head-forming anglicisms and German near-synonyms, e.g. Handler 'trader', Getrank 'drink' and Beruf 'profession' in the case of Dealer, Drink and Job. A natural next step was to analyse the semantic behaviour of selected anglicisms beyond compounding. The basic question was whether there is a general tendency for anglicisms to extend their meaning in the course of time, so that the contrasts which distinguished them from their German equivalents are gradually blurred. Apart from two of the previously analysed nouns, namely Dealer and Drink, the analyses include the verbal anglicism shoppen examined by Kettemann (2006) in rather limited contexts, and two nominal anglicisms which have not yet been subject to contrastive analyses, namely Song and Skyline.

4.1 Handler vs. Dealer

According to the Anglizismen-Worterbuch compiled by Carstensen & Busse (1993-1996) (12), the agent noun Dealer, which is polysemous in English, was borrowed with the very specific meaning component "someone who sells illegal drugs".

In Baeskow (2018) it was shown that this meaning component, which is also prevalent in compounds of the type Drogendealer, Rauschgiftdealer (both meaning 'drug dealer') or Heroindealer, may be overridden by native modifiers. Examples are Autodealer "car dealer", Knoblauchdealer 'garlic dealer', Klingeltondealer 'ring-tone dealer' or Plattendealer 'record dealer', which are not (or only metaphorically) related to the drug scene. Although compounds like these are stylistically marked so far, they suggest that the meaning of Dealer at least partially overlaps with the meaning of a native agent noun that has a more general meaning, namely Handler 'trader', but does this local, i.e. context-specific approximation also hold beyond compounding? Is it justified to state that Dealer has undergone semantic extension in the receptor language? Let us begin by looking at the co-occurrence profile generated for Handler, which provides significant clues as to the generic knowledge associated with this concept.

In particular, this profile reveals an activity prototypically associated with HANDLER. This activity is realized by the verb verkaufen 'sell' and the past participle verkauft, which are recurrent in the corpora. Since the activity of selling defines a professional occupation, it determines the content of the TELIC quale of HANDLER. Moreover, since this activity requires a human agent, the referent of this person-denoting noun must be of the type 'Person (x)', which is specified in the FORMAL quale.

As far as the object of transaction is concerned, four corpora provide the collective noun Ware 'merchandise', which fits the position of the internal argument (y) opened by verkaufen. Compounds like Obsthandler 'fruiterer', Autohandler 'car dealer', Buchhandler 'bookseller', Antiquitatenhandler 'antique dealer', Haushaltswarenhandler 'hardware dealer' and many others show that the very general qualia value 'Ware' allows for specification. However, since the syntactic realization of (y) is not obligatory in the context of Handler (in Twitter it is not among the "Top 30" co-occurrences), it is considered here to be a default argument as defined in section 2. Further co-occurrences which are logically related to the concept HANDLER are Kunde(n) 'customer(s)', Markt 'market', Preis(e) 'price(s)', Geld 'money', and Geschaft(e). The noun Geschaft either refers to a shop or to a transaction business. Markt is polysemous, too, because it either denotes a concrete or an abstract location. In the first reading, it functions as an adjunct which spatially locates the event of selling goods associated with HANDLER. In its abstract reading, it refers to the interplay of supply and demand in which the trader is involved. The following sentences from WaCky illustrate certain constellations which the co-occurrences can enter with their key word (italics and translation by HB).

(3) a. Auf dem Markt in der Fu[beta]gangerzone verkaufen die Handler Krimskrams fur wenig Geld.

'In the market place in the pedestrian zone, the traders are selling odds and ends for little money.'

b. Zunachst muss ein Handler seine Kunden identifizieren.

'First, a trader needs to identify his/her customers.'

c. Zu beiden Seiten boten Handler ihre Ware lautstark an.

'Traders loudly offered their goods on both sides.'

d. Habt ihr gehandelt oder hat der Handler sofort den Preis genannt?

'Did you bargain or did the trader quote the price immediately?'

(4) a. Wenn an den Wochenenden auf dem Markt rund 30 Handler ihre Waren anbieten, [...]

'When approximately 30 traders offer their goods in the market place at weekends, [...]'

b. Zusatzlich entsteht ein neuer Markt fur Handler von Emissionsberechtigungen.

'Additionally, a new market for traders of emission rights is developing.'

(5) a. Viele Handler hatten ihre Geschafte geoffnet.

'Many traders had opened their shops.'

b. Bei uns tragt der Handler die Beweislast, mit wem er ein Geschaft eingegangen ist.

'At our place, the trader bears the burden of proof with whom he has entered into a business.'

These findings suggest that automatically generated co-occurrence matrices do not display arbitrary information. The qualia-based framework facilitates the evaluation of co-occurrences and helps to establish semantic relations between these items and a given key word. In particular, it is a useful guide to the identification of contrasts between semantically similar key words. The following profile indicates that six co-occurrences, namely Geld 'money', Geschaft 'business', Kunden 'customers' Ware 'merchandise', verkaufen 'sell' and verkauft 'sold' are shared by the concepts HANDLER and DEALER. Thus, the referents of both nouns are readily identifiable as participants in a transaction event.

Even more importantly, Table 3 provides a reliable answer to the question whether Dealer underwent meaning extension beyond compounding. Apart from instances of semantic overlap, the co-occurrences suggest that Dealer has preserved the specific meaning component it was associated with at the time it was borrowed. Although Ware "merchandise" occurs in both profiles, the object of transaction is definitely restricted to drugs in the case of Dealer. While Ware is displayed only in the Spiegel-Archiv and WaCky, Drogen is robustly represented in each of the five corpora. In addition, there are various instantiations of this qualia value, namely Heroin, Stoff 'dope', Kokain, and Rauschgift. Even the customers to whom the drugs are sold are further specified by the nouns Junkies and Suchtige 'addicts'.

Moreover, following Pustejovsky's (1996: 229-230) distinction between 'role defining' nominals (e.g. physicist, linguist, violinist) and 'situationally-defined' nominals (e.g. pedestrian, passenger, student), it is argued here that Handler is interpreted generically and that Dealer has a specific interpretation. While Handler constitutes an occupational title, a person referred to as Dealer is identified as such only if he is engaged in selling illegal drugs. The terms 'role-defining' nominals and 'situationally-defined' nominals reflect the more established distinction between individual-level nominals (ILNs) and stage-level nominals (SLNs) (13). While nouns of the former type define the role of an individual independently of the activity performed at the time of reference, the interpretation of nouns of the latter type require the actual performance of characteristic activities. In GL theory, this difference is accounted for by assigning the activities typically associated with ILNs and SLNs to the TELIC and the AGENTIVE quale respectively, as exemplified below for the concepts HANDLER and DEALER.

(6) [mathematical expression not reproducible]

(7) [mathematical expression not reproducible]

Note that the co-occurrence profile displayed in Table 3 also shapes the connotative information content of Dealer. The co-occurrence of this key word with nouns referring to drugs, with Polizei 'police', Fahnder 'investigator', Szene 'scene', and with the nouns Junkies and Suchtige (both of which are inherently specified for <EMpol-> in the sense of Fries 2007) suggests that the activity specified in the AGENTIVE quale is illegal. Here we are dealing with a connotation or Nebensinn "by-sense" in Erdmann's (1900: 82) terminology, which is not inherent to Handler and which evokes negative emotions. (14)

The results presented in this section have shown that there is a partial overlap between the concepts associated with HANDLER and DEALER. Both nouns refer to transactions involving merchandise, customers and money, but the meaning of the latter is more specific because the prototypical object of transaction is specified in its qualia structure and conveys the impression of illegality (along with other co-occurrences which are not part of the profile of HANDLER). The contrastive analyses of Dealer and Handler in compounding (Baeskow 2018) led to similar results, but also revealed that Dealer is quite frequently substituted for Handler (and related native nouns) in the head position of N+N compounds for stylistic purposes in journalistic texts. If the qualia value 'Drogen' is overridden by a modifier that does not refer to drugs, but to objects of value, harmful substances or limited resources, the connotation 'illegal' is preserved by the compound (e.g. Grundstucksdealer 'estate dealer', Sprengstoffdealer 'dealer in explosives', Elfenbeindealer 'ivory dealer'). If the modifier refers to everyday items or food, the compound is either metaphorically/humorously related to the drug scene, or the aspect of illegality is contextually suppressed (e.g. Strumpfdealer 'hosier', Plattendealer 'record dealer', Sonnenbrillendealer 'dealer in sunglasses'). Beyond compounding, however, Dealer and Handler are not (yet) exchangeable, because Dealer is too strongly associated with the distribution of illegal drugs. The frequent occurrence of Dealer in the context of Drogen in all five corpora suggests that this anglicism largely preserved the specific meaning it displayed when it was borrowed from English.

4.2 Getrank vs. Drink

Like Dealer, the noun Drink was subject to meaning specification when it entered the German language in the 19th century. According to Stiven (1936: 71), it was exclusively used to refer to alcoholic beverages. In the AWB, it is primarily defined as "alkoholisches (Misch-)Getrank" ('(mixed) alcoholic beverage'). Furthermore, it is pointed out that this anglicism extended its meaning to denote not only alcoholic beverages, but also "(Misch-)Getranke jeglicher Art" ('all kinds of (mixed) beverages'). The contexts provided in this reference work suggest that instances of meaning extension in compounding (e.g. Cola-Drinks, Suppen-Drink 'soup drink', both 1978) and beyond are not attested before the late 1970s. In Baeskow (2018), new trends as to the use of Drink are identified. First, Drink has become a vogue word in the context of modifiers related to wellness and health, where it unfolds a pseudo-medicinal flavour that is not conveyed by its German near-synonyms (e.g. Darmgesundheitsdrink 'intestinal-health drink', Fett-weg-Drink 'fat-away drink', Herzdrink 'heart drink'). Secondly, Drink is preferred over Getrank (and similar near-synonyms) if the beverage is prepared by mixing the (alcoholic or non-alcoholic) ingredients denoted by the modifier. The mixing event, which is represented as an option in the AWB, is becoming more salient in view of the numerous innovative compounds whose modifiers refer to natural ingredients (e.g. Erdbeer-Bananen-Drink "strawberry-banana drink", Kiwi-Apfel-Drink 'kiwi-apple drink', Karotten-Brokkoli-Drink 'carrot-broccoli drink'). In these compounds and in contexts like those in (8), which are attested even before the 1970s, semantic approximation of Drink and the native noun Getrank is most evident.

(8) a. Sie verabreichte siebzehn Kraftfahrern Drinks. Acht erhielten reinen Fruchtsaft, die anderen ein alkoholisches Getrank. (Zeit-Archiv, 6/1952)

'She gave drinks to seventeen drivers. Eight of them got pure fruit juice, the others an alcoholic drink.'

b. So nahm er in der Milchbar des Bundeshauses einen Drink mit einem Parlamentsportier [...] (Spiegel-Archiv, 4/1956)

'Thus, he had a drink with the parliamentary porter in the milk bar of the federal parliament building [...]'

In (9), Mineralwasser 'mineral water' is obviously referred to as Drink for stylistic purposes. If a perfectly appropriate native noun (in this case Getrank) is contextually replaced by an anglicism in order to render an everyday concept more attractive or prestigious, we are dealing with an instance of semantic-stylistic upgrading in the sense of Pfitzner (1978: 194-195). In journalese, this stylistic device may be accompanied by a touch of irony.

(9) a. Mineralwasser ist der absolute In-Drink der Saison. - Dem Durstigen, der im Cafe des "Steigenberger Parkhotels" zu Hamburg ein Mineralwasser bestellt, reicht der Ober eine Karte: die Wasserkarte. (Focus-Archiv, 8/1994)

'Mineral water is the absolute in-drink of the season. - The waiter passes a card to the thirsty guest who orders mineral water at the cafe of the 'Steigenberger Parkhotel' in Hamburg: the water menu.'

b. Schon aus Solidaritat mit Daniel Schreiber kann ich zu beiden Buchern nur einen Drink empfehlen: Mineralwasser. Sage niemand, das sei fantasietotend: Es gibt Mineralwasser zu au[beta]erst fantasievollen Preisen. (Fokus-Archiv, 2/2015) 'If only in solidarity with Daniel Schreiber I can recommend just one drink for both books: Mineral water. Don't tell anyone this is killing fantasy. There is mineral water available for most fanciful prices.'

Despite these instances of semantic approximation, the co-occurrence matrices of both nouns, from which compounds were excluded, indicate that the basic contrast is still prevalent. The co-occurrence matrix generated for Getrank gave rise to the following profile:

To begin with, Getrank is of the ontological type 'Flussigkeit' ('liquid'). This very general piece of information is not directly retrievable from the co-occurrence profile, but it is reflected by the co-hyponyms Cola, Bier 'beer', Wasser 'water', Kaffee 'coffee', Tee 'tea', Wein 'wine', Alkohol, alkoholisches 'alcoholic', Whisky, and Champagner 'champagne'. Further support for the typing of Getrank as 'Flussigkeit' comes from the co-occurrence partners Flasche 'bottle' and Glas 'glass', which denote containers for keeping liquids in. The function typically associated with Getrank, namely trinken 'to drink'

, is directly retrievable from the co-occurrence profile, which displays the infinitive as well as the word forms [trank.sub.[preterite]] and [getrunken.sub.[past participle]]. Significantly, the co-occurrences indicate that Getrank refers to all kinds of beverages, including those containing alcohol. Thus, the constituency remains unspecified in its QS.

(10) [mathematical expression not reproducible]

As expected, the co-occurrence profile of Drink is very similar to that of Getrank. In particular, the referents of both nouns are of the same ontological type and share a common telos. Moreover, there are instances of Whisky, Cola and Coca in both profiles. In spite of these semantic similarities, subtle contrasts between Getrank and Drink become apparent. Consider the following co-occurrence profile:

This profile suggests that Drink typically refers to alcoholic beverages (beyond compounding). First of all, co-hyponyms like Kaffee, Wasser or Tee identified for Getrank, which denote common thirst quenchers, do not surface among the 30 most frequent co-occurrence partners of Drink. Secondly, this anglicism occurs in the context of the noun Bar in each of the corpora used for this study. Since it is part of the generic knowledge that a bar is a location where alcoholic beverages are served (unless the reference is restricted by a modifier, as in the compound Milchbar 'milk bar' (cf (8b)), this relatively stable co-occurrence indicates that {Alkohol} continues to be a defining property which basically distinguishes Drink from its native competitor Getrank. Thirdly, the function associated with Drink (TELIC) is lexically realized by word forms of nehmen (rather than trinken). In German, the phrase einen Drink nehmen (literally 'to take a drink') is a fixed collocation in which Drink refers to an alcoholic beverage. Phrases like ?einen Softdrink/Energydrink/Frucht-drink nehmen are definitely marked.

(11) [mathematical expression not reproducible]

Further collocations in which Drink exclusively refers to alcohol are auf einen Drink 'for a drink' and bei einem Drink 'with a drink'. Since stop words (i.e. function words) were excluded from the automatic co-occurrence analyses, these constructions are not retrievable from Table 5, but they are listed in the AWB (entry for Drink), where they are associated with the image of a social gathering with friends or business partners at which alcoholic beverages are enjoyed. Similarly, a relaxing, informal atmosphere is evoked by the fixed collocation mit einem Drink in der Hand 'with a drink in his/her hand', for which there is an obvious indicator in the co-occurrence profile, namely the noun Hand (Spiegel, Focus, and WaCky). Attention should also be paid to the adverbial abends 'in the evening' and the noun Abend 'evening', which are to be found among the co-occurrences of Drink in five corpora. Since these temporal relations are absent from the profile of Getrank, we may conclude that the social gathering typically (though not exclusively) takes place in the evening when work is done. Concrete examples are provided below:

(12) a. Wenn er abends ausgehen will, hat er die Wahl zwischen 25 Klubs, in denen er seinen Drink nehmen kann. (Spiegel-Archiv, 5/1948)

'When he wants to go out in the evening, he may chose among 25 clubs to have his drink.'

b. Er ladt sie zum Lunch und unterhalt sie des Abends bei festlichem Dinner und Drinks vorm Kamin [...] (Spiegel-Archiv, 9/1987)

'He invites them for lunch, and in the evening he entertains them with a festive dinner and drinks in front of the fireplace.'

c. Ein frischer, angenehm aufmunternder Drink fur den spaten Abend. "Zur Bar-Kultur", sagt Daun, "gehort das richtige Getrank zur richtigen Situation." (Focus-Archiv, 7/2015)

'A fresh, pleasantly encouraging drink for the late evening. "The right drink for the right situation makes up the bar culture", says Daun.'

To sum up, the co-occurrence profile generated for Drink from five extensive corpora suggests that this anglicism preserved its reference to alcoholic beverages in German. However, the qualia value {Alkohol}, which distinguishes it from Getrank, does not have absolute character, but constitutes a prototypical semantic property which may be overridden by a modifier (e.g. Frucht-Drink) or the larger context (cf. (8) and (9)). In spite of the partial semantic overlap of Drink and Getrank, the anglicism provides connotative information which is not inherent to its native competitor.

4.3 Lied vs. Song, Chanson, Arie

Lexical co-occurrence is a suitable device not only for comparing word pairs, but also helps to identify contrasts and similarities between members of a lexical field. For example, the noun Musik 'music' and the Verb singen 'to sing' (including various word-forms such as singt or sang) are frequently represented in the context of the anglicism Song. These lexical items also co-occur with at least three other key words, namely with the native noun Lied (which corresponds with English Song) and two further loanwords--Chanson and Arie 'aria'. Each of these nouns is of the ontological type 'Musikstuck (y)' 'piece of music' (FORMAL), and in each case, 'singen (e, x: human, y)' determines the TELIC quale. However, despite the semantic overlap, Lied, Song, Chanson and Arie are not freely exchangeable. Indicators of subtle contrasts are provided by the individual co-occurrence profiles. Let us begin by comparing Song with the more general native noun Lied.

The first noteworthy contrast revealed by these profiles is related to the coming-into-being process, which is defined at the AGENTIVE quale. In addition to the 47 occurrences of the infinitive schreiben 'to write' in the Spiegel-Archiv, the past participle geschrieben 'written' systematically occurs in the context of Song. It is missing only in Twitter. Although Lied collocates with schreiben, too (ein Lied schreiben), this verb is not among the thirty most frequent co-occurrences of the native near-equivalent. This contrast allows for the conclusion that the concept denoted by Song is more readily associated with the composer (or songwriter, which is another anglicism in German) than the concept denoted by Lied. Similarly, the relatively frequent occurrence of Band in the context of Song indicates that the anglicism is more closely related to the performers than Lied (hence ein Beatles-Song, ein Abba-Song, ein Song von den Stones etc.).

Further co-occurrences which allow for a distinction between Song and Lied are Pop and Rock. These items do not contribute to the semantic structure of Song, but rather provide connotative information because they suggest that we are dealing here with a modern piece of music of anglo-american origin or influence, which may be of international popularity. (15) Like Band, Pop and Rock, the noun Song found its way into other languages, too. Thus, each of these anglicisms has a high international recognition value. On the other hand, Lied co-occurs with two word forms of the adjective alt 'old', namely alten and alte, in four corpora. Depending on the linguistic context, the resulting phrase either receives a literal or a metaphorical interpretation. If a piece of music for singing is old in the literal sense, it is referred to as Lied rather than Song in German, as shown below:

(13) a. Und auch sonst stimmt am alten Lied vom braven Landmann kaum noch eine Verszeile. (Spiegel-Archiv, 11/1977)

'And in other respects, too, hardly any verse line of the old song of the brave countryman is still correct.'

b. Und da war Amir so geruhrt, da[beta] er auf der Stelle "Ich schwebe uber den Baumen" sang, ein altes arabisches Lied. (Spiegel-Archiv, 1/1996)

'And then Amir was so strongly moved that he instantaneously sang 'I am floating above the trees', an old Arabian song.'

A metaphorical interpretation is required for the idiom Es ist das alte Lied 'It's the (same) old story', which means that a well-known problem recurs frequently. In this context, Lied is not replaceable by Song (Es ist *der alte Song).

The use of Song may also be motivated by communicative-pragmatic and stylistic considerations. In the Twitter-Korpus, it frequently occurs in the context of modifiers or intensifiers that are typical of German youth language (e.g. krass 'wicked, cool, awesome' in (14a), Hammer (16) in (14b) and cool in (14c)), and quite a few utterances are accompanied by emoticons, e.g.

(14) a. Noch nie erlebt, dass jemand so krass einen Song angeteasert hat [??] aber hat sich gelohnt (Twitter 8/2015)

'I never heard someone tease a song so wickedly [??] but it was worthwhile.'

b. der [sic!] song dein herz tragt felsen fand ich wahnsinnig gut. der text ist toll und der refrain ist hammer! (Twitter 2012)

'I considered the song 'Dein Herz tragt Felsen' ['Your heart is bearing rocks'] to be incredibly good. The text is great and the refrain is awesome.'

c. bitte nehmt Jedward's [HASHTAG293601174] in eure playlist auf! das ist so ein cooler song und verdient echt eine chance! (Twitter 11/2012)

'Please add Jedward's to your playlist! It's such a cool song and really deserves a chance!'

Tweets like these signal that young speakers in particular consider Song to be less conventional or even more prestigious than Lied. This is what Pfitzner (1978) refers to as 'affect', which typically manifests itself in semantic-stylistic upgrading mentioned already in section 4.2. (17) Furthermore, it is noteworthy that Song also surfaces in a collocation in which Lied is less usual, namely einen Song performen 'to perform a song'. While Lied is absent from the thirty most frequent co-occurrences of performen, Song occurs in the context of this verbal anglicism in three corpora:

A Google-based search restricted to German websites yields 2.010 hits for einen Song performen and only 279 hits for ein Lied performen. (18) In this collocation, Song unfolds a (pseudo-)technical aura which is not conveyed by Lied in its basic reading '(short) piece of music to be sung'. On the other hand, Lied constitutes a real technical term used by music experts to refer to the art form developed by Franz Schubert and continued by composers like Hugo Wolf or Richard Strauss. In this particular reading, Lied was borrowed into other languages such as English or Spanish. (19) This contrast illustrates the distinction between Fachsprache ('technical terminology') and Fachkolorit ('technical colouring'). While Lied in the sense of 'art form' denotes a musicological concept, Song rather constitutes a vogue word that is suited to evoke the impression of expertise.

Interestingly, Song seems to have largely abandoned an early meaning component. According to the AWB (entry for Song), this loanword is first attested in 1798, but only occurred sporadically until the beginning of the 20th century, when it was used to refer to socio-critical cabaret songs in the Brechtian style and hence assumed a very specific meaning as compared to its English model. Although the compound Protestsong still exists in German, there are no indicators in the co-occurrence matrices generated from the five modern corpora which point towards a critical connotation. Thus, we may conclude that the narrow reading imposed on this loanword by Brecht was abandoned in favour of a more general meaning, which entails a semantic approximation of Lied and Song. However, despite this meaning extension, Lied is not completely synonymous with Song because of its connotative shift towards a vogue word in the jargon of light music.

As indicated above, semantic overlap is observable not only for Song and Lied. Consider the co-occurrence profile of Chanson, which belongs to the same lexical field.

The profile of Chanson shares with the profiles of Song and Lied the co-occurrences singen (along with related word-forms) and Musik, which determine the TELIC and the FORMAL quale respectively. More importantly, however, a prototypical property of CHANSON that distinguishes this concept from the concepts SONG and LIED is retrievable from the co-occurrence profile as well, namely its French origin. Considering the fact that the overall corpus frequency of Chanson is relatively low (in the Spiegel-Archiv, for example, there are 3162 instances of Song, but only 431 instances of Chanson), the word-forms franzosische and franzosischen of the adjective franzosisch "French" are well represented in the context of this loanword. The Zeit-Archiv and WaCky additionally display the proper noun Frankreich 'France'. Semantically, the French origin of the concept CHANSON is represented at the CONSTITUTIVE quale.

(15) [mathematical expression not reproducible]

Of course, franzosisch is merely a default value, i.e. a prototypical value which may be contextually overridden. For instance, the co-occurrence profile of Chanson also displays the word-form deutschen 'German' even if this adjective is underrepresented. This finding does not come as a surprise because this type of song has also been cultivated in Germany at least since the Weimar Republic. However, although the chanson is not regionally restricted to France, it is definitely inspired by a French model (16a) and thus suitable to create local colouring (16b).

(16) a. Aznavour: Nein! Das franzosische Chanson ist ein typisch franzosisches Produkt, weil es die Liebe zur franzosischen Sprache ist, die diese Autoren [Gainsbourg, Moustaki, Piaf etc.; HB] erst hervorgebracht hat. (Zeit-Archiv 5/2014)

'Aznavour: No! The French Chanson is a typically French product because it is love for the French language which eventually brought forth these authors.'

b. Dann trallert die Empfangsdame ein Chanson - Erlebnisgastronomie im Stil der "Bouffes Anversois - Cafes chantants" des 19. Jahrhunderts. (Zeit-Archiv, 2/2007)

'Then the receptionist warbles a chanson - event gastronomy in the style of the 19th century "Bouffes Anversois - Cafes chantants".'

Another member of the lexical field examined in this section is Arie 'aria', whose distinctive property is also retrievable from the corpora.

Like Lied, Song and Chanson, Arie displays the co-occurrences singen (as well as related word-forms) and Musik, which provide the most basic components for its qualia structure. One co-occurring lexical item not shared by the other field neighbours is Oper 'opera'. This co-occurrence partner signals that Arie is a relational noun, which is typically part of an opera (20) - even if it is performed independently of the complete work. Following Bouillon et al. (2012: 1529), it is assumed here that part-of relations are represented at the CONSTITUTIVE quale, as shown in (17).

(17) [mathematical expression not reproducible]

Further co-occurrences which are represented less frequently but nevertheless fit the profile of Arie are Rezitativ (Zt 15, WaC 15) and Buhne 'stage' (Sp 7, Zt 7). While a recitative--like an aria--is typically part of an opera (except that it is spoken instead of sung), Buhne is interpretable as an adjunct that spatially locates the event of singing.

While Yang's (1990: 49-55, 92-93) pre-digital identification of semes and distinctive connotative features largely depended on speakers' judgements and dictionary entries, the present analyses benefit from automatically performed co-occurrence analyses which provide clues as to the semantic and/or connontative information content of lexical items and thus help to reveal not only lexical contrasts and semantic overlap between members of word pairs, but also to determine the distribution of members of lexical fields. Moreover, since the co-occurrence matrices were generated from large quantities of text, they ensure that the relations and values that are eventually mapped onto qualia structures are empirically motivated.

4.4 Skyline vs. Stadtsilhouette

Although proper nouns are not part of knowledge representations, they may contribute to the connotative information content of a lexical item. As will be shown in this section, the systematic co-occurrence of a keyword with geographical names may be an indicator of local colouring ('Lokalkolorit'). Evidence comes from the profile of the anglicism Skyline, which is represented below:

Significantly, the noun Skyline, which roughly corresponds with the less frequent German-French hybrid compound Stadtsilhouette (literally 'city silhouette'), co-occurs with New, York, Yorker and Manhattan in all the corpora used for this study. This constellation, which is not available for Stadtsilhouette (die Skyline /?Stadtsilhouette von Manhattan) (21), locates the concept SKYLINE in a typically American setting and makes it a suitable device for local colouring--a stylistic nuance which according to Galinsky (1963: 101) conveys "the impression of American 'atmosphere' on the German listener's or reader's mind." The following sentences from the print media illustrate this effect:

(18) a. Am Fenster zieht die Skyline von Manhattan vorbei. Sein Ziel ist downtown, eine Bank hat ihn zum Lunchvortrag eingeladen. (Spiegel 4/2006)

'The Skyline of Manhattan passed the window. His destination is downtown, a bank invited him for a lunchtalk.'

b. Die Skyline definierte nicht nur die Stadt, sondern auch die Menschen, die zu ihren Fu[beta]en leben. Wolkenkratzer wirken plotzlich zerbrechlich. (Focus 3/2002)

'The skyline defined not only the city, but also the humans living at its feet. Skyscrapers suddenly appear fragile.'

In (18a), the American atmosphere is conveyed by an accumulation of three anglicisms: Skyline, downtown, and the hybrid compound Lunchvortrag. In (18b), the noun Skyline has a strongly symbolic function because it is representative of life in Manhattan before and after 9/11. Both sentences would be less vivid and less authentic if Skyline was replaced by Stadtsilhouette.

Table 11 indicates that the connotative relation between Skyline and New York or Manhattan is transferable to megacities displaying a similar architecture. In particular, Skyline co-occurs with Frankfurt, whose financial district with all its skyscrapers is strongly reminiscent of Manhattan. On account of this similarity, Frankfurt is humorously referred to as Mainhattan (i.e. Manhattan on the Main). The names of other megacities only occur sporadically in the context of Skyline, as the following frequencies show: Hongkong (Sp 9, Zt 4), Londoner (Sp 6), London (Sp 6), Dubai (Foc 4), Shanghai (Tw 3), Chicago (WaC 23).

The link for the connotative transfer is provided by two common nouns which regularly co-occur with Skyline in the corpora, namely Wolkenkratzer (22) and Hochhauser 'high-rise buildings'. Although Hochhauser (unlike Wolkenkratzer) also co-occurs with Stadtsilhouette (Foc 2, Zt 3, WaCky 6), modern buildings of a considerable height do not seem to be salient defining components of this concept. Instead, Stadtsilhouette tends to co-occur with various nouns referring to or related to churches, namely Frauenkirche (Sp 1), Kirchenschiff 'nave' (Foc 1), Kirchen (Zt 1, WaC 8), and Dom (WaC, 8). Another co-occurrence partner is Altstadt "historic city" (Tw 1, WaC 6), which refers to a part of a city that usually consists of historic buildings. In the contexts provided in (19), the substitution of Skyline for Stadtsilhouette would result in stylistic markedness:

(19) a. Steile Staffeln, enge Gassen und spitze Giebel pragen die Stadtsilhouette von der Tubinger Altstadt bis hinauf zum Schloss. (Twitter 11/2012).

'Steep steps, narrow lanes and pointed gables determine the skyline of Tubingen's historic city up to the castle.'

b. Aufragend aus der Masse, geben einige wenige Gebaude der Stadtsilhouette ihren unverkennbaren Schnitt: rechts, im Osten, der machtige Kubus des Alcazar, der Stadtfestung, in der Mitte der Turm der gotischen Kathedrale.

'Looming out of the mass, a few buildings shape the unmistakable skyline: to the right, in the east, the mighty cube of the Alcazar, the town fortress, in the middle the tower of the Gothic cathedral.'

The only place names co-occurring with Stadtsilhouette in more than one corpus are Hamburg (Sp 2, WaC6) and Koln "Cologne" (Zt 2, WaC 5)--two German cities which lack skyscrapers of a dimension typically associated with the United States. Since recently, Hamburg's highest building is the Elbphilharmonie (110 metres) (23), which is still relatively small in comparison with the Empire State Building (381 metres) (24). In Cologne, there are some high-rise buildings, too, but since its panorama is inextricably linked to the two pointed towers of the famous cathedral--the Kolner Dom--the use of Stadtsilhouette constitutes an option.

From a semantic point of view, however, Skyline and Stadtsilhouette are in principle exchangeable because both nouns refer to the outline of a number of buildings seen against the sky and thus exclude the shape of natural objects such as rocks, hills or trees. (25) As shown in this section, the contrast is rather connotatively motivated, and the systematic occurrence of New York and Manhattan in the context of Skyline allows for the conclusion that these proper nouns trigger the use of this anglicism.

4.5 Shoppen vs. einkaufen

Co-occurrence matrices also allow for a comparison of lexical items other than nouns. In this section, two near-synonyms analysed by Kettemann (2006: 177-178) will be revisited, namely shoppen and einkaufen 'to do the shopping'. As indicated in the introduction, Kettemann performed his analyses on the basis of concordances from COSMAS II. Although the contexts were extracted non-automatically, they make interesting predictions. As far as shoppen is concerned, co-occurrences like gerne "with pleasure", relaxen 'to relax', Spa[beta] 'fun', gemutlich 'leisurely' or Freudinnen treffen 'meet (female) friends' suggest that this anglicism refers to a pleasant leisure activity which is not necessarily a targeted activity. In terms of the Generative Lexicon, the TELIC quale of shoppen remains unspecified because there is not always a concrete need for the goods to be purchased. By contrast, the co-occurrences identified by Kettemann for einkaufen signal that this activity is part of the household chores, which have to be done quickly and efficiently. Examples are mussen 'have to' kochen 'to cook', putzen 'to clean', gunstig 'cheap', or Stress. Kettemann's findings were checked against the automatically detected contexts which gave rise to the following co-occurrence profiles:

While both verbs enter a fixed collocation with gehen 'to go' (shoppen gehen, einkaufen gehen) (26) and share the co-occurrence Geld 'money', only shoppen displays the positively connotated items gern(e), Spa[beta] and Freundin(nen), Freund(en) in its co-occurrence profile. In the context of einkaufen, there are co-occurrences such as billig(er) 'cheap(er)', kochen 'to cook', Arbeit 'work', or arbeiten 'to work'. These constellations, some of which are exemplified below, confirm Kettemann's observation that shoppen--unlike einkaufen--refers to a pleasant and relaxing activity.

(20) a. Mit meinen Freunden gehe ich gern in die Stadt shoppen, ins Kino oder ein Eis essen. Das macht immer riesen spa[beta] [sic!] und uberhaupt keine Langeweile. (WaCky)

'I like to go shopping, to go to the cinema or to eat an ice-cream with my [friends.sub.[+fem]]. This is always great fun and never boring.'

b. Morgen wahrscheinlich mit meiner besten Freundin shoppen und stobern bei hugendubel. Ich freu mich jetzt schon. (Twitter, 8/2011)

'Tomorrow probably shopping with my best [friend.sub.[+fem]] and rummaging at Hugendubel's. I am already looking forward to it.'

(21) a. Auf der anderen Seite konnte man sagen, da[beta] ich Kunstler bin und als solcher nicht meine ganze Zeit mit Einkaufen und Kochen verbringen will [...] (Focus, 6/1995)

'On the other hand, one could say that I am an artist and thus do not want to spend all my time doing the shopping or cooking.'

b. ,,Wenn ich nach der Arbeit noch einkaufen mu[beta], dann Klauschen aus der Tagesstatte abholen und mit ihm spielen--das schaffe ich nicht"(Zeit 1/1990)

'If I have to do the shopping after work, to fetch Klauschen from the day-care centre afterwards and to play with him--I don't manage it.'

At the lexical level, this contrast is accounted for by emotional markers in the sense of Fries (2007). While shoppen is inherently specified for <EMpol+>, its native competitor should be considered emotionally neutral. Although einkaufen may be more closely related to household chores, it is in no way a pejorative verb. Following Fries, emotional neutrality will be represented by the feature <[EMpol.sub.0]>.

Another revealing, though less frequent co-occurrence of shoppen is Klamotten, which is colloquial for articles of clothing. Although the goods to be purchased are not restricted to clothes, it is a matter of fact that shoppen is generally avoided in the context of food or domestic articles in German unless a particular stylistic effect is to be achieved. For example, the use of this anglicism in the slogan Die ganze Woche Frische shoppen 'buy freshness throughout the week', which covers the weekly advertising leaflet of a German discounter, is clearly marked and intended to attract the customers' attention. In this case, shoppen is contextually assigned the connotative feature <EMexp->, which encodes the emotion of unexpectedness.

Contrasts are also observable as to the locations in which the events denoted by shoppen and einkaufen typically take place. In the corpora, the noun Supermarkt "supermarket", which is associated with a large supply of goods, self-service and special offers, only occurs in the context of einkaufen--the activity mainly driven by need. On the other hand, the co-occurrence profile of shoppen displays Stadt "city" and --more importantly--the noun Internet. Although the virtual location denoted by Internet is also compatible with einkaufen (im Internet einkaufen), the verb shoppen unfolds its positive connotation even in this context and thus implies the advantages associated with online shopping (sitting comfortably in front of the computer in order to chose a particular article independently of opening hours, using the delivery service, paying by online banking etc.).

To sum up, shoppen differs semantically from einkaufen in that it denotes an activity which is not necessarily targeted. While einkaufen constitutes a telic verb, shoppen is inherently atelic although a transitive, telic use is not generally precluded (e.g. Klamotten, Frische shoppen). From a connotative point of view, shoppen is associated with a pleasant activity, whereas einkaufen is emotionally neutral. Computerlinguistic evidence for this contrast, which was anticipated by Kettemann (2006) on the basis of a small set of data, comes from automatically identified co-occurrence partners like gern(e), Freundin(nen), or Spa[beta].

5. Conclusion

In this article, the anglicisms Dealer, Drink, Song, Skyline and shoppen were compared to semantically close German equivalents. The case studies have shown that contextual information from very large and heterogeneous corpora is a reliable basis for information retrieval and for the performance of contrastive analyses. Although we cannot expect co-occurrence matrices to provide full-fledged semantic representations or encyclopaedic details, prototypical properties of the key words under examination are made accessible to some degree. In order to avoid arbitrariness, the selection of semantically relevant co-occurrences was guided by Pustejovsky's (1996) Generative Lexicon whose multi-dimensional qualia structures provide basic parameters for generic knowledge. These parameters account not only for contrasts between anglicisms and native near-synonyms, but also for instances of semantic approximation.

The automatically performed co-occurrence analyses also revealed lexical properties that are much more difficult to capture out of context, namely connotations, which comprise stylistic, emotional and communicative-pragmatic facets of key words as well as images we typically associate with them. Although connotative nuances are not part of semantic representations, they enrich our knowledge about concepts and allow us to work out very subtle contrasts between loanwords and semantically close native equivalents.

On the whole, the contrastive analyses have shown that despite semantic overlap, the distribution of the anglicisms under consideration and their German near-synonyms is non-arbitrary and largely predictable from their co-occurrence profiles. While Dealer, Drink, Skyline and shoppen preserved the specific meaning components with which they were borrowed from English and unfold connotations that are not conveyed by the native near-synonyms Handler, Getrank, Stadtsilhouette and einkaufen respectively, semantic approximation is most obvious for the pair Lied and Song, the latter of which abandoned its political, socio-critical flavour. Nevertheless, these nouns do not constitute complete synonyms either because Song gained the status of a (pseudo-)technical vogue word which distinguishes it from Lied in its basic and technical reading. As predicted by Weisgerber (1962: 167), there are no instances of complete synonymy in the examples selected from KANG.

Acknowledgements

The results presented in this study were obtained during the research project "Anglizismen im Deutschen: Kontextbasierte Interpretation, dynamische Restrukturierung und Generalisierung" kindly supported by the Deutsche Forschungsgemeinschaft (DFG) and reflect the co-operation of the authors. While Jurgen Rolshoven's contribution involves data collection, data processing and the generation of co-occurrence matrices as a type of vector semantics, Heike Baeskow is responsible for the linguistic analyses. We would like to thank Borge Kiss, Peter Seipel and Ipek Cengiz for their assistance in data processing.

References

Androutsopoulos, Jannis. 1998. Deutsche Jugendsprache. Untersuchungen zu ihren Strukturen und Funktionen. Frankfurt a. M. [etc.]: Peter Lang.

Anick, Peter & Pustejovsky, James. 1990. An application of lexical semantics to knowledge acquisition from corpora. In Karlgren, Hans (ed.), Proceedings of the 13th International Conference on Computational Linguistics, 7-12, University of Helsinki, Finland.

Baeskow, Heike (2018): "-Drink or -trunk, that is the question: The distribution of selected anglicisms and German near-synonyms in compounding". Studia Linguistica (Online publication ahead of print; https://doi.org/10.1111/stul.12086).

Baroni, Marco & Lenci, Alessandro. 2008. Concepts and properties in word spaces. In Lenci, Alessandro (ed.), From context to meaning. Distributional models of the lexicon in linguistics and cognitive science. Special Issue of the Italian Journal of Linguistics 20(1). 55-88.

Belica, Cyril. 2011. Semantische Nahe als Ahnlichkeit von Kookkurrenzprofilen. In Abel, Andrea & Zanin, Renata (eds.), Korpora in Lehre und Forschung, 155-178. Bozen: Bozen-Bolzano University Press. Freie Universitat Bozen-Bolzano.

Bernadini, Silvia & Baroni, Marco (eds.). 2006. Wacky! Working papers on the Web as corpus. Bologna: GEDIT. Available online at http://wackybook.sslmit.unibo.it/

Bouillon, Pierrette, Jezek, Elisabetta, Melloni, Chiara & Picton, Aurelie. 2012. Annotating qualia relations in Italian and French complex nominals". In Calzolari, Nicoletta, Chukri, Khalid, Declerck, Thierry, Ugur Dogan, Mehmet, Maegaard, Bente, Mariani, Joseph, Moreno, Asuncion, Odijk, Jan & Piperidis, Stelios (eds.), Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC), 1527-1532. Istanbul: European Languages Resources Association (Elra).

Carlson, Gregory Norman. 1977. Reference to Kinds in English. Massachusetts, Amherst: University of Massachusetts. (Doctoral dissertation).

Carstensen, Broder. 1965. Englische Einflusse auf die deutsche Sprache nach 1945. Heidelberg: Carl Winter Universitatsverlag.

Carstensen, Broder & Busse, Ulrich (eds.): Anglizismen-Worterbuch. Der Einflu[beta] des Englischen auf den deutschen Wortschatz nach 1945. Berlin, New York: De Gruyter. Volume 1 (1993), Volume 2 (1994), Volume 3 (1996); reprinted 2001.

Cimiano, Philipp & Wenderoth, Johanna. 2005. Automatically learning qualia structures from the Web. Proceedings of the ACL-SIGLEX Workshop on Deep Lexical Acquisition, 28-37. Ann Arbor: Association for Computational Linguistics.

Davies, Mark. (2002-) Corpus del Espanol: 100 million words, 1200s-1900s. Available online at http://www.corpusdelespanol.org.

Donalies, Elke 1992. Hippes Hopping und toughe Trendies: Uber neumodische, noch nicht kodifizierte Anglizismen in deutschsprachigen Female-Yuppie-Zeitschriften". Deutsche Sprache: Zeitschrift fur Theorie, Praxis, Dokumentation 20 (2). 97-110.

Erdmann, Karl-Otto. 1900. Die Bedeutung des Wortes. Leipzig: Eduard Avenarius.

Firth, John R. 1957. Modes of meaning. In Firth, John R. (ed.), Papers in Linguistics 1934-1951, 180-215. London: Oxford University Press.

Fries, Norbert. 2007. Die Kodierung von Emotionen in Texten. Teil 1: Grundlagen. Journal of Literary Theory 1(2). 293-337.

Galinsky, Hans 1963. Stylistic aspects of borrowing. A stylistic and comparative view of American elements in Modern German and British English. In Fraenkel, Ernst, Galinsky, Hans, Gerhard, Dietrich & Lang, H.J. (eds.), Jahrbuch fur Amerikastudien, 98-135. Band 8. Heidelberg: Universitatsverlag Winter.

Gevaudan, Paul. 2002. Klassifikation lexikalischer Entwicklungen. Semantische, morphologische und stratische Filiation. [http://homepages.uni-tuebingen.de/paul.gevaudan/Filiation.pdf]

Gotzeler, Christiane. 2008. Anglizismen in der Pressesprache. Alte und neue Bundeslander im Vergleich. Bremen: Hempen Verlag.

Jones, Susan & Sinclair, John M. 1974. English lexical collocations. Cahiers de Lexicologie 24. 15-61.

Kettemann, Bernhard. 2006. Morphologische Integration und semantische Differenzierung durch Anglizismen im Deutschen. In Mair, C. (ed.), Corpora and the history of English. Papers dedicated to Manfred Markus on the occasion of his 65th birthday, 169-182. Heidelberg: Winter.

Leutloff, Kristina. 2003. Anglizismen in deutschen Jugendzeitschriften. Zu ihren stilistischen und kommunikativen Funktionen am Beispiel von Bravo und Yam! Lebende Sprachen 48. 5-13.

Lipka, Leonhard. 2002. English lexicology. Tubingen: Gunter Narr Verlag.

Ludwig, Klaus-Dieter. 1986. Nicht-denotative Informationen lexikalischer Einheiten als Worterbucheintrage. Zeitschrift fur Phonetik, Sprachwissenschaft und Kommunikationsforschung 39(2). 182-194.

Ludwig, Klaus-Dieter. 2002. Registerkonzepte: Ein Uberblick. In Cruse, D. Alan, Hundsnurscher, Franz, Job, Michael & Lutzeier, Peter Rolf (eds.) Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wortern und Wortschatzen. 1. Halbband, 784-793. Berlin [etc.]: de Gruyter.

Moravcsik, Julius. 1975. Aitia as generative factor in Aristotle's philosophy. Dialogue 14. 622-636.

Moravcsik, Julius. 1981. How do words get their meanings? Journal of Philosophy 78. 5-24.

Onysko, Alexander. 2007. Anglicisms in German. Borrowing, lexical productivity, and written codeswitching. Berlin [etc.]: Walter de Gruyter.

Onysko, Alexander & Winter-Froemel, Esme. 2011. Necessary loans--luxury loans? Exploring the pragmatic dimension of borrowing. Journal of Pragmatics 43. 1550-1567.

Oxford English Dictionary. http://dictionary.oed.com/

Pfitzner, Jurgen. 1978. Der Anglizismus im Deutschen. Ein Beitrag zur Bestimmung seiner stilistischen Funktion in der heutigen Presse. Stuttgart: J.B. Metzlersche Verlagsbuchhandlung.

Poesio, Massimo & Almuhareb, Abdulrahman. 2008. Extracting concept descriptions from the Web: the importance of attributes and values. In Buitelaar, Paul & Cimiano, Philipp (eds.), Proceedings of the 2008 conference on ontology learning and population: Bridging the gap between text and knowledge, 29-44. Amsterdam: IOS Press.

Pustejovsky, James. 1996. The Generative Lexicon. Cambridge, MA: The MIT Press.

Sosnizka, Julia. 2014. The collocational behavior of anglicisms in German and American business and news magazines. Gottingen: Cuvillier Verlag.

Stiven, Agnes B. 1936. Englands Einflu[beta] auf den deutschen Wortschatz. Doctoral dissertation, Marburg.

Weisgerber, Leo. 1962. Von den Kraften der deutschen Sprache. Band I: Grundzuge der inhaltbezogenen Grammatik. 3. Auflage. Dusseldorf: Padagogischer Verlag Schwann.

Williams, Edwin. 1981. Argument structure and morphology. Linguistic Review 1. 81-114.

Winter-Froemel, Esme. 2011. Entlehnung in der Kommunikation und im Sprachwandel: Theorie und Analysen zum Franzosischen. Berlin/Boston: de Gruyter.

Yang, Wenliang. 1990. Anglizismen im Deutschen: am Beispiel des Nachrichtenmagazins Der Spiegel. Tubingen: Max Niemeyer Verlag.

Heike Baeskow, Bergische Universitat Wuppertal & Jurgen Rolshoven, Universitat zu Koln

Heike Baeskow Bergische Universitat Wuppertal Fakultat fur Geistes- und Kulturwissenschaften Anglistik/Amerikanistik Gau[beta]stra[beta]e 20 42119 Wuppertal Germany baeskow@uni-wuppertal.de

Jurgen Rolshoven Universitat zu Koln Institut fur Linguistik Abteilung Sprachliche Informationsverarbeitung Albertus-Magnus-Platz 50931 Koln Germany rols@spinfo.uni-koeln.de

(1) As pointed out for example by Weisgerber (1962: 167), complete synonymy is generally avoided.

(2) This reference corpus is also used by Onysko (2007) for a comprehensive analysis of anglicisms in German.

(3) This notion will be applied here to any of the 30 lexical items that occur most frequently in the context of a given key word, i.e. of an anglicism or its German near-synonym. Given this rather broad conception of the linguistic context, the much and controversially discussed term 'collocation' (cf. Sosnizka 2014, chap. 2 for a historical overview) will be restricted to fixed collocations, i.e. to unpredictable combinations of words which have to be learned and memorized as units and which affect the levels of both syntax and word-formation, e.g. Er raucht stark "He smokes heavily", ein starker Raucher "a heavy smoker" (Lipka 2002: 182f).

(4) It should be noted that qualia structures do not necessarily specify all of these generative factors. For example, since natural types such as stone or water are not inherently associated with a function, the TELIC quale remains unspecified (unless some purpose or intention is being imposed on these concepts); cf. Pustejovsky (2003: 375-379).

(5) The corresponding German notions are 'Lokalkolorit', 'Sozialkolorit', and 'Fachkolorit'.

(6) https://dev.twitter.com/streaming/reference/get/statuses/sample

(7) http://wacky.sslmit.unibo.it/lib/exe/fetch.php?media=papers:wacky_2008.pdf

(8) A collection of working papers on the Web as Corpus was published by Bernadini & Baroni (2006). It is online available at http://wackybook.sslmit.unibo.it/.

(9) This approach also supplements generic knowledge that is not retrievable from co-occurrence matrices.

(10) This notion was borrowed from Belica (2011), who uses it in a slightly different way. According to Belica, a co-occurrence profile of an object comprises all the quantitative results of a co-occurrence analysis performed for this object.

(11) http://www.ids-mannheim.de/cosmas2/projekt/einsteiger/was.html

(12) This reference work, which presents anglicisms in context and provides them with precise definitions, will henceforth be abbreviated to AWB.

(13) The distinction between stages and individuals goes back to Carlson (1977)

(14) Although the referent of the emotionally neutral noun Handler may also deal in drugs (cf. Drogenhandler, Rauschgifthandler), the range of goods to be distributed is in principle infinite.

(15) Of course, song texts are not restricted to English. In KANG, the anglicism Song frequently occurs in the context of names of German bands or musicians.

(16) The idiom Das ist/war der Hammer either refers to something extremely positive or extremely negative.

(17) According to Pfitzner (1978: 211), the use of anglicisms to create semantic-stylistic downgrading is relatively restricted. In Author1 (2018), this device is exemplified by expressive compounds like Knochenjob 'back-breaking job', Sklavenjob 'slave job', Routinejob 'routine job', Stressjob 'stress job', Hollenjob 'hell's job', Mistjob 'crap job', or Billigjob 'low-paying job' in which Job is usually not replaced by Beruf 'profession'.

(18) Last access June 20th, 2017

(19) For English, the Oxford English Dictionary (OED) also lists the hybrid compounds lieder-singer and lieder-singing. The Corpus del Espanol constructed by Mark Davies even provides a Spanish agent noun liederista.

(20) Beyond the opera, an aria is also part of a cantata or an oratorium; cf. Seeger, H. (1966): Musiklexikon in zwei Banden. Leipzig: VEB Deutscher Verlag fur Musik.

(21) The profile of Stadtsilhouette will not be displayed here because this noun has a very low corpus frequency. Instead, reference will be made to co-occurrences in the individual corpora.

(22) Wolkenkratzer (literally 'cloud scraper') is a loan rendition of the English noun skyscraper.

(23) http://www.hamburg.de/sehenswuerdigkeiten/4401242/hochhaeuser-in-hamburg/

(24) https://de.wikipedia.org/wiki/Empire_State_Building

(25) In English, by contrast, the denotation of skyline also includes "the line at which the earth or a part of the landscape appears to meet the sky" (OED), e.g. I do love horses moving slowly against a skyline of trees (a1933). Once again, the borrowing process led to meaning specification.

(26) A slightly archaic variant listed in the AWB is shopping gehen.
Table 2: Co-occurrence profile of Handler

HANDLER         Sp   Foc    Zt    Tw     WaC

corpus freq.  9381  3208  3671  1009  10.148
Kunden         345   269   139    31     386
Kunde                 50                 151
verkaufen      287    86   100    16     210
verkauft       163    65    75           183
Markt          214    69    88    11     153
Ware           186    65    79           183
Preis          171    80    55           142
Preise         175    67    57           108
Geschaft       203    62    56
Geschafte      113          50
Geld           158    65    60           166
kaufen         114    60    58    11     197
Kaufer                81    57           210

Table 3: Co-occurrence profile of Dealer

Dealer        Spiegel  Focus  Zeit  Twitter  WaCky

corpus freq.    1.608    454   320      316  2.386
Drogen             97     20    13       23    186
Droge                           10        4     33
Polizei            71     42    17       10    115
Heroin             62     19    13              77
Stoff              65     16     8        6     44
Geld               55      9     9              41
Kokain             43     23              4     34
Junkies            40     12                    29
Suchtige           30     12
Szene              37     11
Geschaft           36     10
Rauschgift         35      9                    27
Kunden             34      8                    27
Fahnder            32     10
Gramm              34     11                    28
kaufen                                    6     26
Ware               28                           29
verkauft                                  5     36
verkaufen                                       37

Table 4: Co-occurrence profile of Getrank

GETRANK        Sp  Foc  Zt  Tw  WaC

Cola           47    6  26  10   74
Coca           28       15
Bier           29   13  27  21  140
Wasser         25   10  16  27   52
Kaffee         22   10  23  19   94
Tee             8       11   7  113
Wein           25   16  18      153
Whisky         19       14
Champagner      9       11
Alkohol        16    4   8
alkoholisches        5      14
Flasche        22       13
Glas           14    6   8       70
trinken        12    8  16  23  105
getrunken       8

Table 5: Co-occurrence profile of Drink

Drink       Spiegel  Focus  Zeit  Twitter  WaCky

corp. freq      490    234   350      613  2.837
Bar              34     15    33        8    161
nehmen           12      6    26              81
nimmt             8            7              42
trinken                                14     38
trinke                                  7
getrunken                              10
Glas                     8                    64
Hand             14      7                    43
Alkohol          10      5                    43
Whisky           10            8
Cola             12            9
Coca                           8
abends            8            7
Abend                    5              8     55
Wodka             7            7

Table 6: Co-occurrence profile of Song

Song             Sp      Foc      Zt       Tw      WaC

corpus freq.      3.162    1.444   1.706    7.380   10.382
Lied             94       34      44               200
Album            70       61      61      211      418
Band            115       59      44               381
Musik            74       46      43               285
Pop              60       29
Rock             66               26               149
schreiben        47
geschrieben      54       37      26               158
singen           72       30      27       85
singt            63       29      48
sing                      32               97
Sanger           47               27
Platte           54               35               122
hei[beta]t      110       30      55       86      161
horen                             33      144      182
gehort                                    132      138
hort                              27       79

Table 7: Co-occurrence profile of Lied

Lied           Sp      Foc       Zt       Tw      WaC

corpus freq.    1.668    1.367    2.485    5.992  875
singen        122      569      320      195       62
singt          60       39      116      103       20
sang           57                54                13
sangen         36                46
gesungen       48                75                14
Musik          26       23       46                23
Text           29                35                17
alte           35       20       90
alten                                              14
hei[beta]t              22       70      156       15
Melodie        27                37                20

Table 8: Co-occurrence profile of performen (extract)

Performen     Sp  Foc  Zt   Tw  WaC

corpus freq.   8   16  19  189  189
SONG                    2   13   16
SONGS                        6    9

Table 9: Co-occurrence profile of Chanson

Chanson         Sp  Foc   Zt  Tw  WaC

corpus freq.   431   62  256  22    1.352
singen          13    3    7
singt           13        17
Sanger          29    3
Sangerin        35        11       40
Song                  2            38
sang                       8
gesungen                   7
franzosische    24    4    9       40
franzosischen   18         7       68
Frankreich                 6       37
Lied            10        13       55
Lieder                             34
Musik           13    4   10       84
deutschen       11    3            35

Table 10: Co-occurrence profile of Arie

ARIE           Sp  Foc   Zt  Tw  WaC

corpus freq.  300   83  341  33  190
Oper           16    5   19        6
singen          9    6   12   3    9
singt          13    4   23   1    9
sang                 4        2
Rezitativ                15       15
Musik           5    5    5        6
Buhne           7         7

Table 11: Co-occurrence profile of Skyline

Skyline         Sp  Foc   Zt   Tw  WaC

corp. freq.    299   92  205  154    1.602
New             36   11   14    9  183
York            25    6   11    6  112
Yorker           8    3         4   35
Stadt           36   12   19    4  154
Manhattan       29    9   11    3  105
Manhattans      14    9    5        35
Blick           28   13   22   10   16
Wolkenkratzer   22         7        43
Frankfurt        8    3   12    9  105
Frankfurter     16    3             78
Hochhauser      12    6    7        28
City            10    4    7    5   45

Table 12: Co-occurrence profile of einkaufen

EINKAUFEN      Sp      Foc   Zt       Tw      WaC

corpus freq.    2.210  871    1.534    4.060   14.956
gehen         148       64  146      554        1.566
Geld           87       21   52       48      443
Supermarkt     52       24   39               405
Kunden         57       37   36               339
billig(er)     85       24   74
kochen                  23   46       49      361
Arbeit                       32               284
arbeiten                     30               262

Table 13: Co-occurrence profile of shoppen

SHOPPEN         Sp  Foc   Zt   Tw      WaC

corpus freq.   268  297  155    2.579    2.109
gehen           31   30   20  383      353
kaufen               13   11            49
Internet        13   16    5            51
Stadt                      8   49       87
gern(e)          9   10    4   45      100
Freundin(nen)              4   51
Freunde(n)                              91
Spa[beta]                      67       84
Geld            14   10    7   59       61
Klamotten                  4   37
COPYRIGHT 2018 The Slovak Association for the Study of English
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2018 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Baeskow, Heike; Rolshoven, Jurgen
Publication:SKASE Journal of Theoretical Linguistics
Article Type:Report
Geographic Code:1USA
Date:Jun 1, 2018
Words:12899
Previous Article:Paradigmatic Morphology Splinters, Combining Forms, and Secreted Affixes.
Next Article:Rules or Constraints? A Cross Methodological Comparison of Approaches to 'Syncope' and 'Vowel Shortening' in Cairene Arabic.
Topics:

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |