2020. 2. 19. 17:43ㆍ카테고리 없음
1The case of language borrowing was investigated in depth by Gusmani (1983), who argues that a linguistic loan is an interference phenomenon, connected with contact and mutual influence of different languages. According to his study, the motivations behind the origin of a loan lie in the individual act of a speaker or of a group of speakers. The need to resort to a foreign alternative derives from the prestige held by the latter against an equivalent word in the mother tongue of the speaker (or from the absence all together of an alternative, as in our work: “ Se mi vede, Miki mi banna ( spawnare /spo’nare/). 5 From Hanks, Pustejovsky 2005, a pattern is intended here as an argument structure with specificatio 8The next part of our research involved the analysis of the semantic patterns for each lemma, thus compiling one or more data-driven senses for every verb. The senses obtained were classified according to Verb Net’s semantic class hierarchy. The assumptions underlying this investigation are grounded on Corpus Pattern Analysis (CPA) and Computational Lexicography (Hanks 2008; Hanks 2012; Jezek 2011).
9Verb patterns have – in general – the following structure, where: (1) Spammare 2b AgentPERSON V spammare (ThemeARTEFACT ABSTRACT). 10We have chosen all uppercase for the semantic type, and first letter uppercase for the thematic role, extended to every argument of the verb. Round brackets contain the possible optional arguments of the verb. 6 The resource is not yet available for public consultation. 11After extracting the semantic patterns for each lemma from the corpus, we stored the information in an electronic lexicon, built using the software Personal Lexicon 2.7.1, a language learning resource developed by Alexander Smith between 2007 and 2015.
The software comes both in free and registered versions, the current lexicon has been compiled – and it will be consultable – using the free version. 12The lexicon is designed to give a precise account of every semantic feature and every meaning variation of the verb loans. As the reader will see observing Figure 1, each entry is characterized by the following elements (some of them pre-named in the software). 17Clicking on each one produces the list of entries belonging to that particular class; this list appears in the third section of the lexicon, the Lexical Items column storing all the entries ordered alphabetically. 18In this section we report the results of both the semantic analysis of the loans and the annotation task. 19The lexicon contains 157 senses for a total of 90 verbs. As shown in Table 1, the 157 senses have been classified into 3 groups according to three main criteria about the degree of semantic conservativeness of the loan:.
The meaning remains the same as the original verb. The meaning remains linked to the original one, but it diversifies to some degree.
The meaning changes to the point that it becomes a new meaning altogether Group # Type Numbers Group 1 Same sense 88 New verb form 11 Group 2 Diversified sense 25 Group 3 New sense 26 New v. And new sense 7 Senses 157. 12 We choose Geertzen, J. (2012) online resource for agreement evaluation.
Dizionario Ceco Italiano Pdf To Jpg Free
23The semantic annotation task was conducted following the methodology of Pustejovksy and Stubbs (2012); only a sample of 440 random occurrences was annotated by 9 groups of anannotators, each constituted by 3 people. They were given guidelines explaining the method and the tagsets, and they were asked to separately annotate the semantic type and the thematic role of each verbal argument. We used Fleiss’ k algorithm to calculate the agreement (Artstein and Poesio, 2008), the values being interpreted according to Landis and Koch (1977). We already said that the results have been only partially positive, in particular – as for the thematic role – only one group reached the 0.6 threshold considered acceptable with semantic annotation, the others showing moderate agreement and fair agreement (one group only). For the semantic type annotation, three groups reached the 0.6 value, four groups showed moderate agreement and two groups showed fair agreement. Nonetheless, we could make interesting linguistic considerations. “ Prima di lanciare il tutto ho overcloccato la scheda video”.
25We feel that the only partially satisfying results may depend on the tricky lexical meaning of each loan. It is clearly easier to annotate the argument structure of a well-known verb like potenziare, rather than the one of the loan over-cloccare ( potenziare and overcloccare being almost synonyms). The thematic role level is the most problematic, obtaining substantial agreement only in one case; the semantic type level on the other end is perceived as a less abstract, more transparent concept and the annotation is slightly better, with three groups over 0.6. What is really interesting is that the group which performed best with thematic roles is also the one which did worst with semantic types. Moreover, the groups which performed best with semantic types showed only fair to moderate agreement in thematic roles.
We observe a – general and group wise – performance improvement with the semantic type level.