{ "paper_id": "2020", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T07:12:14.010235Z" }, "title": "Annotating Croatian Semantic Type Coercions in CROATPAS", "authors": [ { "first": "Costanza", "middle": [], "last": "Marini", "suffix": "", "affiliation": { "laboratory": "", "institution": "University of Pavia", "location": { "addrLine": "65", "postCode": "27100", "settlement": "Pavia", "region": "PV" } }, "email": "costanza.marini93@gmail.com" }, { "first": "Elisabetta", "middle": [], "last": "Je\u017eek", "suffix": "", "affiliation": { "laboratory": "", "institution": "University of Pavia", "location": { "addrLine": "65", "postCode": "27100", "settlement": "Pavia", "region": "PV" } }, "email": "jezek@unipv.it" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "This short research paper presents the results of a corpus-based metonymy annotation exercise on a sample of 101 Croatian verb entries-corresponding to 457 patters and over 20,000 corpus lines-taken from CROATPAS (Marini & Je\u017eek, 2019), a digital repository of verb argument structures manually annotated with Semantic Type labels on their argument slots following a methodology inspired by", "pdf_parse": { "paper_id": "2020", "_pdf_hash": "", "abstract": [ { "text": "This short research paper presents the results of a corpus-based metonymy annotation exercise on a sample of 101 Croatian verb entries-corresponding to 457 patters and over 20,000 corpus lines-taken from CROATPAS (Marini & Je\u017eek, 2019), a digital repository of verb argument structures manually annotated with Semantic Type labels on their argument slots following a methodology inspired by", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Abstract", "sec_num": null } ], "body_text": [ { "text": "If we look at the lexicon in its whole, it is possible to identify systematic alternations of meaning that apply not only to single lexical instances but entire classes of words, i.e. patterns of so-called regular polysemy (Apresjan, 1973) . Some common alternations are author/work; product/producer; event/food or container/content.", "cite_spans": [ { "start": 223, "end": 239, "text": "(Apresjan, 1973)", "ref_id": "BIBREF0" } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "When dealing with these alternations, however, it is necessary to distinguish between metonymic and inherent polysemy. In metonymic shifts, meaning is extended by conceptual contiguity and a change of referent is required, since one entity is used to denote another which is conceptually associated with it (Je\u017eek, 2016: 59) . This is the case, for instance, of the alternation container/content, exemplified by sentences such as \"I would have eaten the whole fridge\", where fridge actually stands for the food it contains.", "cite_spans": [ { "start": 307, "end": 324, "text": "(Je\u017eek, 2016: 59)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "In the case of inherent polysemy, on the other hand, there is no sense extension nor change of referent, but only one ontologically complex entity. This is the case, for instance, of alternations such as information source/ artifact as in \"The book I am reading weighs one kilo\" (Pustejovsky & Je\u017eek, 2008: 185) , where the lexical item book can be understood at the same time as the information it contains and a heavy object. The possibility for more than one of the senses of a complex entity to be activated simultaneously is called co-predication and is a prerogative of inherently polysemous words.", "cite_spans": [ { "start": 279, "end": 311, "text": "(Pustejovsky & Je\u017eek, 2008: 185)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "In this paper, we are going to present the first results of a metonymy annotation exercise on a sample of Croatian verbs taken from the Croatian Typed Predicate Argument Structures resource (CROATPAS, Marini & Je\u017eek, 2019 ) (see section 2.1). Since the resource rests on Generative Lexicon Theory (Pustejovsky, 1995 (Pustejovsky, & 1998 Pustejovsky & Je\u017eek, 2008) , metonymies are annotated and analysed as Semantic Type Coercions (see section 2.2). The set of semantic labels used for the annotation and the sample choice are covered in section 2.3 and 2.4, respectively.", "cite_spans": [ { "start": 201, "end": 221, "text": "Marini & Je\u017eek, 2019", "ref_id": "BIBREF7" }, { "start": 297, "end": 315, "text": "(Pustejovsky, 1995", "ref_id": "BIBREF8" }, { "start": 316, "end": 336, "text": "(Pustejovsky, & 1998", "ref_id": "BIBREF9" }, { "start": 337, "end": 363, "text": "Pustejovsky & Je\u017eek, 2008)", "ref_id": "BIBREF10" } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "1 Its first release will contain approximately 200 Croatian verb entries and will be accessible by 2020 on the website of University of Pavia: https://cla.unipv.it/?page_id=53723.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "CROATPAS (Marini & Je\u017eek, 2019 ) -short for Croatian Typed Predicate Argument Structure resource -is a digital dictionary of Croatian verbs focusing on verbal polysemy, which is currently being developed at the University of Pavia 1 next to its Italian sister project TPAS (Je\u017eek et al., 2014) . CROATPAS consists in a repository of verb valency structures whose argument slots have been manually annotated with a set of semantic labels called Semantic Types (henceforth SemTypes), following a corpus-based lexicographic methodology inspired by Corpus Pattern Analysis (CPA, Hanks, 2004 Hanks, & 2013 Hanks & Pustejovsky, 2005) .", "cite_spans": [ { "start": 9, "end": 30, "text": "(Marini & Je\u017eek, 2019", "ref_id": "BIBREF7" }, { "start": 273, "end": 293, "text": "(Je\u017eek et al., 2014)", "ref_id": "BIBREF4" }, { "start": 575, "end": 586, "text": "Hanks, 2004", "ref_id": null }, { "start": 587, "end": 600, "text": "Hanks, & 2013", "ref_id": null }, { "start": 601, "end": 627, "text": "Hanks & Pustejovsky, 2005)", "ref_id": "BIBREF2" } ], "ref_spans": [], "eq_spans": [], "section": "The CROATPAS resource", "sec_num": "2.1" }, { "text": "From a theoretical point of view, CPA rests on the Theory of Norms and Exploitations (TNE, Hanks 2004 & 2013 , which differentiates between two types of word uses: conventional ones -the norms -and deviations from such norms -the exploitations. When applying CPA, lexicographers traditionally focus on identifying normal word usage by mapping standard meanings onto their syntagmatic patterns of use.", "cite_spans": [ { "start": 85, "end": 101, "text": "(TNE, Hanks 2004", "ref_id": null }, { "start": 102, "end": 108, "text": "& 2013", "ref_id": "BIBREF1" } ], "ref_spans": [], "eq_spans": [], "section": "The CROATPAS resource", "sec_num": "2.1" }, { "text": "In CROATPAS, our CPA-inspired methodology consists in the following four steps: 1) sampling 250 random concordances from a representative corpus of Standard Croatian for each verb entry, namely the Croatian Web as Corpus (Ljube\u0161i\u0107 & Klubi\u010dka, 2014) ; 2) manually disambiguating its different senses and 3) associating the right SemTypes to the argument slots found in each sensebound valency structure. The fourth and last step is only possible thanks to our editing environment SKEMA, which is connected to the Croatian Web as Corpus through the Sketch Engine corpus management platform (Kilgarriff et al., 2014) and enables annotators to create patterns for each retrieved verb sense, such as the ones in (English, to drink) As you can see from the patterns above, the first sense of the Croatian verb piti (English, to drink) is the most obvious one, namely that of an [Animate] drinking a [Beverage] . However, if a [Human] is told to be drinking a [Drug] -such as a pill or antibiotics (Croatian, tabletu and antibiotike) -then he or she is simply ingesting or swallowing them. Finally, if we talk of a [Human] drinking (without specifying any direct object), he or she is by default ingesting an alcoholic drink.", "cite_spans": [ { "start": 221, "end": 248, "text": "(Ljube\u0161i\u0107 & Klubi\u010dka, 2014)", "ref_id": "BIBREF5" }, { "start": 588, "end": 613, "text": "(Kilgarriff et al., 2014)", "ref_id": "BIBREF6" }, { "start": 893, "end": 903, "text": "[Beverage]", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "The CROATPAS resource", "sec_num": "2.1" }, { "text": "In addition to verbal polysemy, CROATPAS also allows lexicographers to annotate metonymic arguments by adding specific sub patterns to existing verb senses (see Figure 2 ). from the Croatian verb po\u010deti (English, to begin) Despite involving the same verb sense as pattern 1, the metonymic sub pattern 1.1.m is linked only to those concordance lines where there is a mismatch in the SemType of the direct object: namely [Document] instead of [Activity] . This mismatch signals that a metonymic shift is taking place, which in Generative Lexicon Theory takes the name of Semantic Type coercion (Pustejovsky & Je\u017eek, 2008; Je\u017eek & Quochi, 2010) . In order to explain this concept, let us look at a couple of sentences provided by Pustejovsky (1995: 115-6 ) starring a good translational equivalent of the Croatian verb po\u010deti, namely:", "cite_spans": [ { "start": 441, "end": 451, "text": "[Activity]", "ref_id": null }, { "start": 592, "end": 619, "text": "(Pustejovsky & Je\u017eek, 2008;", "ref_id": "BIBREF10" }, { "start": 620, "end": 641, "text": "Je\u017eek & Quochi, 2010)", "ref_id": "BIBREF4" }, { "start": 727, "end": 751, "text": "Pustejovsky (1995: 115-6", "ref_id": null } ], "ref_spans": [ { "start": 161, "end": 169, "text": "Figure 2", "ref_id": "FIGREF2" } ], "eq_spans": [], "section": "Annotating Semantic Type Coercions", "sec_num": "2.2" }, { "text": "(1) John began reading a book.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Annotating Semantic Type Coercions", "sec_num": "2.2" }, { "text": "(2) John began a book.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Annotating Semantic Type Coercions", "sec_num": "2.2" }, { "text": "In sentence (1), the verb's second argument -i.e. reading a book -denotes an [Activity], whereas in sentences (2) it denotes a [Document] -a book. We call Semantic Type Coercion the compositional mechanism which enables us to reconstruct the semantics of the second direct object by forcing -i.e. coercing -[Document] into an [Activity] denotation. As pointed out by Je\u017eek & Quochi (2010 : 1465 , coercion always involves an attested Source Type (e.g. [Document]) which is coerced into a Target Type to fit the verb's selectional requirements (e.g. [Activity] ). The shift can involve any argument slot and is graphically represented as follows: [Document] ! [Activity].", "cite_spans": [ { "start": 326, "end": 336, "text": "[Activity]", "ref_id": null }, { "start": 367, "end": 387, "text": "Je\u017eek & Quochi (2010", "ref_id": "BIBREF4" }, { "start": 388, "end": 394, "text": ": 1465", "ref_id": null }, { "start": 549, "end": 559, "text": "[Activity]", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Annotating Semantic Type Coercions", "sec_num": "2.2" }, { "text": "Qualia Exploitation This being said, if we look at Semantic Type Coercions from a sub lexical point of view, they can be considered exploitations 2 of one of the available qualia roles associated with the Source Type not satisfying the verb's selectional requirements (Pustejovsky & Je\u017eek, 2008: 195) .", "cite_spans": [ { "start": 268, "end": 300, "text": "(Pustejovsky & Je\u017eek, 2008: 195)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "2.2.1", "sec_num": null }, { "text": "Qualia structure is one of the four levels of representation involved in the computational apparatus of Generative Lexicon (Pustejovsky, 1995 (Pustejovsky, & 1998 and it consists of the four most important semantic properties of any lexical item: its Formal, Constitutive, Telic and Agentive qualia. The term qualia comes from Latin and is the plural of the word quale, which means \"what kind?\". The qualia structure of the noun sandwich (Pustejovsky & Je\u017eek, 2008 : 185) As we can see in Figure 3 , the Constitutive quale consists of all the parts that make up the entity we are dealing with -in this case, the sandwich's ingredients. The Formal quale answers to the question \"What sort of thing is this?\" -in this case, a [Physical Entity]. The Telic quale -from the Greek word t\u00e9los, i.e. end -expresses the function of the entity denoted by our lexical item -which, for a sandwich, is being eaten. Last but not least, the Agentive quale specifies the entity's origin.", "cite_spans": [ { "start": 123, "end": 141, "text": "(Pustejovsky, 1995", "ref_id": "BIBREF8" }, { "start": 142, "end": 162, "text": "(Pustejovsky, & 1998", "ref_id": "BIBREF9" }, { "start": 438, "end": 464, "text": "(Pustejovsky & Je\u017eek, 2008", "ref_id": "BIBREF10" } ], "ref_spans": [ { "start": 489, "end": 497, "text": "Figure 3", "ref_id": null } ], "eq_spans": [], "section": "2.2.1", "sec_num": null }, { "text": "If we look at the metonymic sub pattern 1.1.m from Figure 2 under this new light, the Semantic Type Coercion [Document] ! [Activity] can be interpreted as an exploitation of either the Telic quale \"reading\" or the Agentive quale \"writing\", both associated with the qualia structure of any document, since we write so that others can read. It will be the broader context to assign the correct interpretation.", "cite_spans": [], "ref_spans": [ { "start": 51, "end": 59, "text": "Figure 2", "ref_id": "FIGREF2" } ], "eq_spans": [], "section": "2.2.1", "sec_num": null }, { "text": "The list of SemTypes used in CROATPAS is taken from the Italian TPAS resource (Je\u017eek et al., 2014) and belongs to the TPAS ontology , a hierarchically organised set of labels originating from the Brandeis Shallow Ontology (Pustejovsky et al., 2004) currently containing 180 bracketed labels, such as [Human], [Document], and so forth.", "cite_spans": [ { "start": 78, "end": 98, "text": "(Je\u017eek et al., 2014)", "ref_id": "BIBREF4" }, { "start": 222, "end": 248, "text": "(Pustejovsky et al., 2004)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "The System of Semantic Type labels", "sec_num": "2.3" }, { "text": "Despite looking like ontological categories, SemTypes are semantic classes obtained by \"manual clustering and generalization over sets of lexical items found in the argument positions\" in valency structures taken from large corpora (Je\u017eek et al, 2014: 891) . They are thus able to mirror the way humans talk about entities, states and events through language.", "cite_spans": [ { "start": 232, "end": 256, "text": "(Je\u017eek et al, 2014: 891)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "The System of Semantic Type labels", "sec_num": "2.3" }, { "text": "According to Generative Lexicon, SemTypes can be divided into three groups depending on their internal structure:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The System of Semantic Type labels", "sec_num": "2.3" }, { "text": "generally use it in the expression \"qualia exploitation\", which pertains to Generative Lexicon terminology. ", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The System of Semantic Type labels", "sec_num": "2.3" }, { "text": "The verb sample 3 we concentrated on for this metonymy annotation study consists of 44 Croatian aspectual verb pairs 4 and 13 biaspectual verbs taken from the CROATPAS resource (Marini & Je\u017eek, 2019) , for a total of 101 verb entries linked to 457 different patterns.", "cite_spans": [ { "start": 177, "end": 199, "text": "(Marini & Je\u017eek, 2019)", "ref_id": "BIBREF7" } ], "ref_spans": [], "eq_spans": [], "section": "Verb choice", "sec_num": "2.4" }, { "text": "Half of the sample is made up of the Croatian translational equivalents of a sample of Italian verbs known 3 See Appendix 1 for a complete list of all the CROATPAS verbs in our sample, together with their TPAS counterparts and English equivalents. In the Italian list, the verbs sentire and guidare appear twice because we decided to create entries for more than one of their Croatian translational equivalents, namely \u010duti (to hear) and osje\u0107ati/osjetiti (to feel) for the first, voditi/provoditi (to lead) and voziti (to drive) for the second. On the other hand, one of the verbs from the original list of Je\u017eek & Quochi (2010) has not been taken into account because its Croatian translational to trigger Semantic Type Coercions, the so-called coercive verbs analysed by Je\u017eek & Quochi (2010) ; while the other half are the Croatian translational equivalents of a selection of Italian verbs belonging to the language's fundamental vocabulary (FO), i.e. a group of 2,000 words with the highest frequency counts covering about 90% of all Italian written and spoken text (Chiari & De Mauro, 2014: 113) . All Croatian translational equivalents were selected consulting the Zanichelli Italian/Croatian bilingual dictionary Croato compatto (Aleksandra \u0160piki\u0107, 2017).", "cite_spans": [ { "start": 608, "end": 629, "text": "Je\u017eek & Quochi (2010)", "ref_id": "BIBREF4" }, { "start": 774, "end": 795, "text": "Je\u017eek & Quochi (2010)", "ref_id": "BIBREF4" }, { "start": 1071, "end": 1101, "text": "(Chiari & De Mauro, 2014: 113)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Verb choice", "sec_num": "2.4" }, { "text": "As a result of our metonymy annotation exercise, we were able to enrich the 457 patterns stored in CROATPAS adding 106 metonymic sub patterns. The metonymic corpus lines justifying these sub patterns are 1,052, a number which is already included in the over 22,000 annotated corpus lines currently linked to the resource. 2 for the full inventory). Since different metonymic shifts can occur in the same pattern and even in the same argument slot, we decided to encode them -when possible -within the same sub pattern, as in Figure 5 . the ones in Figure 6 . We are going to focus on the three highlighted ones. ", "cite_spans": [], "ref_spans": [ { "start": 322, "end": 323, "text": "2", "ref_id": "TABREF5" }, { "start": 525, "end": 533, "text": "Figure 5", "ref_id": null }, { "start": 548, "end": 556, "text": "Figure 6", "ref_id": null } ], "eq_spans": [], "section": "Results", "sec_num": "3." }, { "text": "In our annotation exercise, we managed to identify a total of 179 Semantic Type coercions of 62 different kinds (see Appendix 2 for the full list). : Pattern 2 and its metonymic sub pattern 2.1.m from the Croatian verb ugostiti (English, to host) When saying a sentence like \"Hrvatska \u0107e ugostiti Srbiju u \u010detvrtfinalu\" (which translates to \"Croatia will host Serbia for the quarter final\"), the SemType [Area] is coerced into a [Football Team] , since what the speaker actually means is that the Croatian national team will play against the Serbian one, and not the respective geographical areas.", "cite_spans": [ { "start": 429, "end": 444, "text": "[Football Team]", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "The most frequent Semantic Type coercions", "sec_num": "3.1" }, { "text": "The CROATPAS verbs giving rise to the most Semantic Type coercions are the following: tutnjati (English, to rumble) with 11 coercions to be traced back to only 2 observed patters; odjekivati (English, to echo) with 10 coercions and only 3 patterns; okrenuti (English, to turn) with 9 SemType coercions and 16 patterns, followed by both the perfective and imperfective variant of the Croatian equivalent of to listen -namely slu\u0161ati and poslu\u0161ati -both with 3 recorded senses and 9 metonymic sub patterns each.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The most coercive Croatian verbs", "sec_num": "3.2" }, { "text": "Since after these first five verbs the number of SemType coercions drastically diminishes to 5 or less for the rest of the sample, it is not unreasonable to suggest that verbs of hearing are particularly well suited to trigger metonymic shifts within their valency structure. To give an idea of the mechanisms at play in these sound-focused coercions, take a look at Figure 8 . and provides also some particularly well-suited examples between square brackets, such as vlak (English, train), oluja (English, storm) and motor (English, engine) . In all of these instances, a qualia role of the entity in object position is exploited and coerced into a [Sound] , like in the case of sirene (English, sirens), whose Telic quale is \"producing a sound\".", "cite_spans": [ { "start": 524, "end": 541, "text": "(English, engine)", "ref_id": null }, { "start": 650, "end": 657, "text": "[Sound]", "ref_id": null } ], "ref_spans": [ { "start": 367, "end": 375, "text": "Figure 8", "ref_id": "FIGREF6" } ], "eq_spans": [], "section": "The most coercive Croatian verbs", "sec_num": "3.2" }, { "text": "If we look at the clause roles where Semantic Type coercions take place (see Table 3 ), we can see that approximately half of the observed metonymic shifts take place in the subject slot, nearly 40% involves the verb's direct object and 14% indirect complements. Table 3 : Semantic Type Coercions by clause roles Even though subjects, objects and indirect complements are not equally distributed across the verb sample, the percentages in Table 2 still demonstrate that all argument slots can be good candidates for metonymies to take place.", "cite_spans": [], "ref_spans": [ { "start": 77, "end": 84, "text": "Table 3", "ref_id": null }, { "start": 263, "end": 270, "text": "Table 3", "ref_id": null }, { "start": 439, "end": 446, "text": "Table 2", "ref_id": "TABREF5" } ], "eq_spans": [], "section": "Semantic Type coercions and clause roles", "sec_num": "3.3" }, { "text": "As previously mentioned, Semantic Type Coercions can also be analysed in terms of Source Type and Target Type (Je\u017eek & Quochi: 2010 As for Source Types, as it was to be expected from the data in Table 2 ", "cite_spans": [ { "start": 110, "end": 131, "text": "(Je\u017eek & Quochi: 2010", "ref_id": "BIBREF4" } ], "ref_spans": [ { "start": 195, "end": 202, "text": "Table 2", "ref_id": "TABREF5" } ], "eq_spans": [], "section": "Source Types and Target Types", "sec_num": "3.4" }, { "text": "In this paper, we have presented the first results of a metonymy annotation exercise on a sample of 101 Croatian verb entries taken from the semantic resource CROATPAS (Marini & Je\u017eek, 2019) , a digital repository of verb argument structures manually annotated with Semantic Type labels on their argumental structure. At present, the resource contains 457 patterns and 106 metonymic sub patterns. The overall number of annotated corpus lines is 22,052, of which 1,052 are linked to the 106 metonymic sub patterns they provide evidence for. We explained the mechanism underlying how metonymy works in our chosen framework and provided an overview of the set of semantic labels we used, together with a clarification of our verb choice. Our results show that [Area] ! [Human Group] proves to be our most frequent Semantic Type Coercion, appearing 25 out of 179 times. Sound verbs such as tutnjati (English, to rumble), odjekivati (English, to echo) and slu\u0161ati/poslu\u0161ati (English, to listen) position themselves amongst the most coercive verbs in the sample: a result supported also by the fact that the most frequent Target Type, appearing in 39 coercion instances out of 179, is [Sound] . On the other hand, the most frequent Source Type is [Area], a finding which agrees with the data on the most frequent Semantic Type coercions overall. From a tentative analysis of clause role predisposition to Semantic Type Coercion, all argument slots seem to be able to enable the shift. In order to give a stronger claim to our results and evaluate the CROATPAS resource, we plan on involving other annotators and devise a task to measure the degree of Inter Annotator Agreement. Once evaluated, we believe that our inventory of manually annotated metonymic corpus lines could be used as training data to develop an automatic metonymy recognition method. Current ongoing work is focussed on comparing our results with an equivalent annotation performed in the TPAS resource on the set of Italian verbs which corresponds to the first half of our Croatian sample (Romani, 2020 ", "cite_spans": [ { "start": 168, "end": 190, "text": "(Marini & Je\u017eek, 2019)", "ref_id": "BIBREF7" }, { "start": 1179, "end": 1186, "text": "[Sound]", "ref_id": null }, { "start": 2053, "end": 2066, "text": "(Romani, 2020", "ref_id": "BIBREF11" } ], "ref_spans": [], "eq_spans": [], "section": "Conclusions", "sec_num": "4." }, { "text": "Be aware that term exploitation in this paper may refer to two different frameworks: in section 2.1 it falls within Hank's Theory of Norms and Exploitations, while in section 2.2.1 and 2.3 we", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "Verbs marked by an asterisk (*) appear twice.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null } ], "back_matter": [ { "text": "The Croatian verb entries from CROATPAS used for our Semantic Type Coercion exercise 5 ", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Appendix 1:", "sec_num": null } ], "bib_entries": { "BIBREF0": { "ref_id": "b0", "title": "Regular polysemy", "authors": [ { "first": "Y", "middle": [ "D" ], "last": "Apresjan", "suffix": "" } ], "year": 1973, "venue": "Linguistics, 142", "volume": "", "issue": "", "pages": "5--32", "other_ids": {}, "num": null, "urls": [], "raw_text": "Apresjan, Y.D. (1973). Regular polysemy. In: Linguistics, 142, pp. 5-32.", "links": null }, "BIBREF1": { "ref_id": "b1", "title": "The New Basic Vocabulary of Italian as a Linguistic Resource", "authors": [ { "first": "I", "middle": [], "last": "Chiari", "suffix": "" }, { "first": "T", "middle": [], "last": "De Mauro", "suffix": "" } ], "year": 2004, "venue": "Proceedings of the 1st Italian Conference on Computational Linguistics (CLiC-it)", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Chiari, I. & De Mauro, T. (2014). The New Basic Vocabulary of Italian as a Linguistic Resource. In: Proceedings of the 1st Italian Conference on Computational Linguistics (CLiC-it). Pisa, Italy. Hanks, P. (2004). Corpus Pattern Analysis. In: Proceedings of the 11th EURALEX International Congress. Lorient, France. Hanks, P. (2013). Lexical Analysis: Norms and Exploitations. Cambridge: The MIT Press.", "links": null }, "BIBREF2": { "ref_id": "b2", "title": "A Pattern Dictionary for Natural Language Processing", "authors": [ { "first": "P", "middle": [], "last": "Hanks", "suffix": "" }, { "first": "J", "middle": [], "last": "Pustejovsky", "suffix": "" } ], "year": 2005, "venue": "Revue fran\u00e7aise de linguistique appliqu\u00e9e", "volume": "10", "issue": "", "pages": "63--82", "other_ids": {}, "num": null, "urls": [], "raw_text": "Hanks, P. & Pustejovsky, J. (2005). A Pattern Dictionary for Natural Language Processing. In: Revue fran\u00e7aise de linguistique appliqu\u00e9e, 10 (2), pp. 63-82.", "links": null }, "BIBREF3": { "ref_id": "b3", "title": "The lexicon: An introduction", "authors": [ { "first": "E", "middle": [], "last": "Je\u017eek", "suffix": "" } ], "year": 2016, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Je\u017eek, E. (2016). The lexicon: An introduction. Oxford: Oxford University Press.", "links": null }, "BIBREF4": { "ref_id": "b4", "title": "T-PAS: A resource of Typed Predicate Argument Structures for linguistic analysis and semantic processing", "authors": [ { "first": "E", "middle": [], "last": "Je\u017eek", "suffix": "" }, { "first": "Austria", "middle": [], "last": "Graz", "suffix": "" }, { "first": "E", "middle": [], "last": "Je\u017eek", "suffix": "" }, { "first": "V", "middle": [], "last": "Quochi", "suffix": "" }, { "first": "", "middle": [], "last": "Valletta", "suffix": "" }, { "first": "", "middle": [], "last": "Malta", "suffix": "" }, { "first": "E", "middle": [], "last": "Je\u017eek", "suffix": "" }, { "first": "B", "middle": [], "last": "Magnini", "suffix": "" }, { "first": "A", "middle": [], "last": "Feltracco", "suffix": "" }, { "first": "A", "middle": [], "last": "Bianchini", "suffix": "" }, { "first": "O", "middle": [], "last": "Popescu", "suffix": "" } ], "year": 2010, "venue": "Proceedings of the 7th conference on International Language Resources and Evaluation (LREC)", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Je\u017eek, E. (2019). Sweetening Ontologies Cont'd: Aligning bottom-up with top-down ontologies. In: Proceedings of CREOL 2019. Graz, Austria. Je\u017eek, E. & Quochi, V. (2010). Capturing Coercions in Texts: A First Annotation Exercise. In: Proceedings of the 7th conference on International Language Resources and Evaluation (LREC). Valletta, Malta. Je\u017eek, E., Magnini, B., Feltracco, A., Bianchini, A., Popescu, O. (2014). T-PAS: A resource of Typed Predicate Argument Structures for linguistic analysis and semantic processing. In: Proceedings of the 9th conference on International Language Resources and Evaluation (LREC). Reykjavik, Iceland.", "links": null }, "BIBREF5": { "ref_id": "b5", "title": "{bs, hr, sr} WaC -web corpora of Bosnian", "authors": [ { "first": "&", "middle": [ "F" ], "last": "Ljube\u0161i\u0107", "suffix": "" }, { "first": "", "middle": [], "last": "Klubi\u010dka", "suffix": "" } ], "year": 2014, "venue": "Proceedings of the 9th Web as Corpus Workshop", "volume": "", "issue": "", "pages": "29--35", "other_ids": {}, "num": null, "urls": [], "raw_text": "Ljube\u0161i\u0107 & F. Klubi\u010dka (2014). {bs, hr, sr} WaC -web corpora of Bosnian, Croatian and Serbian. In: Proceedings of the 9th Web as Corpus Workshop, pp. 29\u221235.", "links": null }, "BIBREF6": { "ref_id": "b6", "title": "The Sketch Engine: ten years on", "authors": [ { "first": "O", "middle": [], "last": "Kilgarriff", "suffix": "" }, { "first": "V", "middle": [], "last": "Baisa", "suffix": "" }, { "first": "J", "middle": [], "last": "Bu\u0161ta", "suffix": "" }, { "first": "M", "middle": [], "last": "Jakub\u00ed\u010dek", "suffix": "" }, { "first": "V", "middle": [], "last": "Kov\u00e1r", "suffix": "" }, { "first": "J", "middle": [], "last": "Michelfeit", "suffix": "" }, { "first": "P", "middle": [], "last": "Rychl\u00fd", "suffix": "" }, { "first": "V", "middle": [], "last": "Suchomel", "suffix": "" } ], "year": 2014, "venue": "In: Lexicography", "volume": "1", "issue": "1", "pages": "7--36", "other_ids": {}, "num": null, "urls": [], "raw_text": "Kilgarriff, O., Baisa, V., Bu\u0161ta, J., Jakub\u00ed\u010dek, M., Kov\u00e1r, V., Michelfeit, J., Rychl\u00fd, P., Suchomel, V. (2014). The Sketch Engine: ten years on. In: Lexicography 1(1), pp. 7-36.", "links": null }, "BIBREF7": { "ref_id": "b7", "title": "CROATPAS: A Resource of Corpus-derived Typed Predicate Argument Structures for Croatian", "authors": [ { "first": "C", "middle": [], "last": "Marini", "suffix": "" }, { "first": "E", "middle": [], "last": "Je\u017eek", "suffix": "" } ], "year": 2019, "venue": "Proceedings of the 6th Italian Conference on Computational Linguistics", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Marini, C. & Je\u017eek, E. (2019) CROATPAS: A Resource of Corpus-derived Typed Predicate Argument Structures for Croatian. In: Proceedings of the 6th Italian Conference on Computational Linguistics (CLiC-it).", "links": null }, "BIBREF8": { "ref_id": "b8", "title": "The Generative Lexicon", "authors": [ { "first": "Italy", "middle": [], "last": "Bari", "suffix": "" }, { "first": "J", "middle": [], "last": "Pustejovsky", "suffix": "" } ], "year": 1995, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Bari, Italy. Pustejovsky, J. (1995). The Generative Lexicon. Cambridge: The MIT Press.", "links": null }, "BIBREF9": { "ref_id": "b9", "title": "The semantics of lexical underspecification", "authors": [ { "first": "J", "middle": [], "last": "Pustejovsky", "suffix": "" } ], "year": 1998, "venue": "Folia Linguistica 32", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Pustejovsky, J. (1998). The semantics of lexical underspecification. In: Folia Linguistica 32.", "links": null }, "BIBREF10": { "ref_id": "b10", "title": "Semantic Coercion in Language: Beyond Distributional Analysis", "authors": [ { "first": "J", "middle": [], "last": "Pustejovsky", "suffix": "" }, { "first": "E", "middle": [], "last": "Je\u017eek", "suffix": "" } ], "year": 2008, "venue": "Italian Journal of Linguistics", "volume": "20", "issue": "", "pages": "181--214", "other_ids": {}, "num": null, "urls": [], "raw_text": "Pustejovsky, J. & Je\u017eek, E. (2008). Semantic Coercion in Language: Beyond Distributional Analysis. In: Italian Journal of Linguistics, vol. 20, pp. 181-214.", "links": null }, "BIBREF11": { "ref_id": "b11", "title": "Searching for Metonymies in Natural Language Text. A corpus-based study on a resource for Italian for Verbs", "authors": [ { "first": "E", "middle": [], "last": "Romani", "suffix": "" } ], "year": 2020, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Romani, E. (2020). Searching for Metonymies in Natural Language Text. A corpus-based study on a resource for Italian for Verbs. BA Thesis, University of Pavia.", "links": null }, "BIBREF12": { "ref_id": "b12", "title": "Croato compatto: dizionario croato/italiano e italiano/croato", "authors": [ { "first": "A", "middle": [], "last": "\u0160piki\u0107", "suffix": "" } ], "year": 2017, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "\u0160piki\u0107, A. (2017). Croato compatto: dizionario croato/italiano e italiano/croato, Zanichelli: Bologna.", "links": null }, "BIBREF13": { "ref_id": "b13", "title": "59 (suze | smijeh | smije\u0161ak > Emotion) do not have a proper Source Types but only source lexical items due to the fact that they belong to idiomatic patterns. In the first case, srce (English, heart) can be coerced into a sound since hearts usually have a heartbeat. As for the second case, although the words suze (English, tears), smijeh (English, laughter) and smije\u0161ak (English, smile) are all coerced into the emotions they typically represent, they cannot be grouped into a shared SemType since some of them are", "authors": [], "year": null, "venue": "The Coercions ranked 58 (srce > Sound) and", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "The Coercions ranked 58 (srce > Sound) and 59 (suze | smijeh | smije\u0161ak > Emotion) do not have a proper Source Types but only source lexical items due to the fact that they belong to idiomatic patterns. In the first case, srce (English, heart) can be coerced into a sound since hearts usually have a heartbeat. As for the second case, although the words suze (English, tears), smijeh (English, laughter) and smije\u0161ak (English, smile) are all coerced into the emotions they typically represent, they cannot be grouped into a shared SemType since some of them are [Physical Entities] (e.g. suze), while others are [Activities] (e.g. smijeh and smije\u0161ak).", "links": null } }, "ref_entries": { "FIGREF0": { "uris": null, "text": "Figure 1.", "num": null, "type_str": "figure" }, "FIGREF1": { "uris": null, "text": "The first 3 patterns from the Croatian verb piti", "num": null, "type_str": "figure" }, "FIGREF2": { "uris": null, "text": "Pattern 1 and its metonymic sub pattern 1.1.m", "num": null, "type_str": "figure" }, "FIGREF3": { "uris": null, "text": "Figure 3: The qualia structure of the noun sandwich (Pustejovsky & Je\u017eek, 2008: 185) As we can see in Figure 3, the Constitutive quale consists of all the parts that make up the entity we are dealing with -in this case, the sandwich's ingredients. The Formal quale answers to the question \"What sort of thing is this?\" -in this case, a [Physical Entity]. The Telic quale -from the Greek word t\u00e9los, i.e. end -expresses the function of the entity denoted by our lexical item -which, for a sandwich, is being eaten. Last but not least, the Agentive quale specifies the entity's origin. If we look at the metonymic sub pattern 1.1.m from Figure 2 under this new light, the Semantic Type Coercion [Document] ! [Activity] can be interpreted as an exploitation of either the Telic quale \"reading\" or the Agentive quale \"writing\", both associated with the qualia structure of any document, since we write so that others can read. It will be the broader context to assign the correct interpretation.", "num": null, "type_str": "figure" }, "FIGREF4": { "uris": null, "text": "The top-level of the TPAS system", "num": null, "type_str": "figure" }, "FIGREF5": { "uris": null, "text": "Figure 7: Pattern 2 and its metonymic sub pattern 2.1.m from the Croatian verb ugostiti (English, to host) When saying a sentence like \"Hrvatska \u0107e ugostiti Srbiju u \u010detvrtfinalu\" (which translates to \"Croatia will host Serbia for the quarter final\"), the SemType [Area] is coerced into a [Football Team], since what the speaker actually means is that the Croatian national team will play against the Serbian one, and not the respective geographical areas.", "num": null, "type_str": "figure" }, "FIGREF6": { "uris": null, "text": "Pattern 1 and its metonymic sub pattern 1.1.m from the Croatian verb tutnjati (English, to rumble) As we can see, pattern 1.1.m lists all the SemTypes of the entities whose sound can rumble, roar or echo (e.g. [Vehicle], [Weather Event], [Engine], [Sound Maker]\u2026)", "num": null, "type_str": "figure" }, "TABREF0": { "html": null, "type_str": "table", "text": "Natural Types referring to natural concepts characterised only by a Formal and a Constitutive quale, e.g.[Animal] or[Natural Landscape Feature]; 2) Artifactual or Tensor Types denoting man-made entities usually possessing also a Telic and an Agentive quale to express their purpose and origin, e.g.[Beverage]; 3) Complex Types characterised by multiple Semantic Types clustered together and normally used to denote inherently polysemous lexical items, e.g.[Institution]. Complex Types can either allow for co-predication or, when only one of their senses is used, for Dot Exploitation. Since differentiating between Qualia Exploitation and Dot Exploitation is not always clear-cut, the TPAS ontology keeps track of all acknowledged Complex Types by treating them as cases of multiple inheritance, i.e. by anchoring them to multiple positions within the SemType hierarchical system as inFigure 4,where [Institution] inherits from both [Abstract Entity] and [Human Group].", "num": null, "content": "
If Tensor Types are characterised by an asymmetrical
structure linking their head SemType to a component of its
qualia structure, as in [Beverage \u2297 Telic Activity
(drinking)], Complex Types are generally internally
symmetrical, as in the case of [Institution = Human Group
\u2022 Abstract Entity]. Since a dot is used to link together their
components, Complex Types are also called Dot Objects.
Artifactual Types are those usually instantiating
metonymic shifts via Qualia Exploitation whereas,
" }, "TABREF2": { "html": null, "type_str": "table", "text": "", "num": null, "content": "" }, "TABREF3": { "html": null, "type_str": "table", "text": "Each of them counts as an instance of the Semantic Type Coercion they stand for, which might have other instances in other sub patterns. All of the coercion instances above occur on the direct object slot of pattern 1 of the verb slu\u0161ati (English, to listen to) and are justified by corpus examples such as", "num": null, "content": "
" }, "TABREF5": { "html": null, "type_str": "table", "text": "portrays the 15 most frequent coercions in our inventory. Since we did not extract the number of corpus lines each Semantic Type coercion is exemplified by, the figures in the third column report the coercion instances, i.e. the amount of times each coercion appears in a different sub pattern or in a different argument slot within the same sub pattern.", "num": null, "content": "
RankSemantic Type CoercionCoercion instances
1Area > Human Group25
2Area > Institution21
3Area > Human Group: Football Team6
4Artifact > Activity6
5Business Enterprise > Road Vehicle6
6Musical Composition > Sound6
7Concept > Human Group5
8Sound Maker > Sound5
9Activity > Sound4
10Beverage > Activity4
11Building > Activity4
12Event > Location4
13Food > Activity4
14Bomb > Sound4
15Document > Activity3
" }, "TABREF6": { "html": null, "type_str": "table", "text": "Our 15 most frequent Semantic Type coercions As we can see from the data, the most frequently annotated Semantic Type coercion in our sample happens to be [Area] ! [Human Group], which makes up for 25 out of the 179 attested occurrences of our 62 different Semantic Type Coercions. As for the second and the third most frequent coercions, we can say that they not only share the same Source Type as the most frequent one, but their", "num": null, "content": "
Target Types are also somewhat hierarchically related,
since [Human Group] is one of the constituents of the
Complex Type [Institution] and [Football Team] is a
hyponym of [Human Group]. The metonymic sub pattern
2.1.m in Figure 7 encoding the Semantic Type coercion
[Area] ! [Human Group: Football Team] will give us an
idea of how this specific coercion works.
" } } } }