| { |
| "paper_id": "W14-0137", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T05:49:52.896035Z" |
| }, |
| "title": "News about the Romanian Wordnet", |
| "authors": [ |
| { |
| "first": "Verginica", |
| "middle": [ |
| "Barbu" |
| ], |
| "last": "Mititelu", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| }, |
| { |
| "first": "Daniel", |
| "middle": [], |
| "last": "\u0218tefan", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Dumitrescu", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "sdumitrescu@racai.ro" |
| }, |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Tufi\u0219", |
| "suffix": "", |
| "affiliation": {}, |
| "email": "tufis@racai.ro" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "There are more than 60 wordnets worldwide; the Romanian wordnet is among those that are maintained and further developed. Begun within the BalkaNet project and further enriched in various (application oriented) projects, it was used in word sense disambiguation, machine translation and question answering with promising results. We present here the latest qualitative and quantitative improvements of our lexical resource, special attention being paid to derivational relations, the latest statistics, as well as the development of an Application Programming Interface, meant to facilitate work with the wordnet, both for its further development purposes and for its use in applications. In the context of creating a common European research infrastructure network, our wordnet is licensed through META-SHARE, being freely available for scientific purposes.", |
| "pdf_parse": { |
| "paper_id": "W14-0137", |
| "_pdf_hash": "", |
| "abstract": [ |
| { |
| "text": "There are more than 60 wordnets worldwide; the Romanian wordnet is among those that are maintained and further developed. Begun within the BalkaNet project and further enriched in various (application oriented) projects, it was used in word sense disambiguation, machine translation and question answering with promising results. We present here the latest qualitative and quantitative improvements of our lexical resource, special attention being paid to derivational relations, the latest statistics, as well as the development of an Application Programming Interface, meant to facilitate work with the wordnet, both for its further development purposes and for its use in applications. In the context of creating a common European research infrastructure network, our wordnet is licensed through META-SHARE, being freely available for scientific purposes.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Abstract", |
| "sec_num": null |
| } |
| ], |
| "body_text": [ |
| { |
| "text": "The development of the Romanian wordnet (RoWN henceforth) started within BalkaNet project 1 . Afterwards, it has been developed and maintained within several projects by the Natural Language Processing (NLP) group of the Romanian Academy Research Institute for Artificial Intelligence (RACAI): ROTEL 2 , STAR 3 , SIR-1 http://www.dblab.upatras.gr/balkanet 2 http://www.ai.ici.ro/rotel_eng/index. htm 3 http://www.racai.ro/star RESDEC 4 , ACCURAT 5 , METANET4U 6 , the Romanian Academy research plan.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Within BalkaNet a core of 18000 synsets was created. They were aligned to the Princeton WordNet (PWN) versions available throughout time, respectively version 2.0 at the end of the project. Among those synsets there were more than 400 that lexicalize concepts specific to the Balkan area. These were implemented in all six languages of the project (Bulgarian, Czech, Greek, Romanian, Serbian, Turkish) and were linked to hypernym synsets, already existing in PWN, so they were not left dangling in the network.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "RoWN contains words belonging both to the general vocabulary and to various domains of activity. Throughout time, we aimed at a complete coverage of the basic common sets from EuroWordNet 7 , of the 1984 corpus 8 , of the newspaper articles corpus NAACL2003 9 , of the Acquis Communautaire corpus and the Eurovoc thesaurus 10 , of VerbNet 3.1 11 , and as much as possible from the ROWikipedia lexical stock.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Two basic development principles have always been followed: the Hierarchy Preservation Principle (HPP) (according to which the hierarchical structure of the concepts in a wordnet is the same irrespective of the natural language for which the wordnet is developed) and the Conceptual Density Principle (which ensures that once a concept is selected to be implemented, all its ancestors up to the unique beginners are also selected, thus preventing the existence of dangling nodes) (Tufi\u015f et al., 2004) . The former principle was the assumption behind our development methodology, namely the expand method. The latter ensured the lack of dangling nodes in the nouns and verbs hierarchies. As a consequence of the way we chose to create our language resource, the lexical density has never been our preoccupation, thus there are many words that do not occur in as many synsets as how many meanings they have. Nevertheless, we do not exclude such an objective from our further developments.", |
| "cite_spans": [ |
| { |
| "start": 480, |
| "end": 500, |
| "text": "(Tufi\u015f et al., 2004)", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "At present, RoWN is aligned to PWN version 3.0. Details about the way we performed the alignment from PWN 2.0 to PWN 3.0 and about the way we solved the encountered problems (the n:1 or 1:n matches between synsets in the two versions) are presented in Tufi\u015f et al. (2013) .", |
| "cite_spans": [ |
| { |
| "start": 252, |
| "end": 271, |
| "text": "Tufi\u015f et al. (2013)", |
| "ref_id": "BIBREF16" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "RoWN is licensed through META-SHARE 12 (). It is free for academic research, but restricted for commercial use.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "In this paper we present the latest qualitative and quantitative improvements of our lexical resource, the latest statistics (Section 3), special attention being paid to derivational relations (Section 4), as well as the development of an Application Programming Interface, meant to facilitate work with the wordnet, both for its further development purposes and for its use in applications (Section 5). Our intentions for further development are included in the Conclusions section. Before proceeding, we enumerate the applications in which our team used RoWN and which, throughout time, influenced our decisions about the concepts to be further implemented in the network. Ion and Tufi\u015f (2009) and Ion and \u015etef\u0103nescu (2011) describe word sense disambiguation (WSD) methods that make use of wordnets: the former is set in a multilingual environment and the WSD is done with the help of aligned word-", |
| "cite_spans": [ |
| { |
| "start": 675, |
| "end": 695, |
| "text": "Ion and Tufi\u015f (2009)", |
| "ref_id": "BIBREF4" |
| }, |
| { |
| "start": 700, |
| "end": 725, |
| "text": "Ion and \u015etef\u0103nescu (2011)", |
| "ref_id": "BIBREF3" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "browse/18", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Uses of RoWN", |
| "sec_num": "2" |
| }, |
| { |
| "text": "nets. The latter is set in a monolingual environment and the WSD is done with the help of the lexical chains established between the cooccurring words in the text, chains whose length is calculated in the wordnet. The assumption is that the shorter the lexical chain, the more similar the words. The length of the lexical chain depends on the number of relations marked in the network. The results in the multilingual environment are reported as better than those in the monolingual one.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Uses of RoWN", |
| "sec_num": "2" |
| }, |
| { |
| "text": "For a Question Answering (QA) system, RoWN was used for expanding the query introduced by the user (Ion et al., 2008) with words semantically related (i.e., synonyms, hypo-and hyperonyms) to the ones it contained. Moreover, RoWN was also used in the last phase, that of ranking the found results by calculating the semantic distance, again as a lexical chain, between the words introduced by the user and those occurring in the text. It was noticed that the relations with the greatest contribution at calculating the score are hyponymy and derivational relations.", |
| "cite_spans": [ |
| { |
| "start": 99, |
| "end": 117, |
| "text": "(Ion et al., 2008)", |
| "ref_id": "BIBREF5" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Uses of RoWN", |
| "sec_num": "2" |
| }, |
| { |
| "text": "Aligned wordnets are valuable sources of cross-language equivalents, especially multiword terms, in machine translation.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Uses of RoWN", |
| "sec_num": "2" |
| }, |
| { |
| "text": "Lately our efforts of implementing new synsets aimed at a complete coverage of VerbNet 3.1, with the prospect of creating a syntactic parser for Romanian.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Latest Quantitative Developments", |
| "sec_num": "3" |
| }, |
| { |
| "text": "The up-to-date statistics about RoWN are presented in Table 1 and 2 below. In the former, PoS stands for part of speech, S for synset, L for literal, UL for unique literals and NL for nonlexicalized synsets. Obeying the HPP stated above implies the transfer of the hierarchies from PWN into RoWN. The lack of perfect equivalences among languages is widely known; nevertheless, we chose to disregard it. Moreover, there are lexical gaps in all languages. We call them nonlexicalized concepts and represent them as empty synsets. For example, for the PWN verbal synset {zip_up:1} (gloss: close with a zipper) there is no literal in the corresponding Romanian synset. However, such synsets do not lack relations: the corresponding ones from PWN are transferred into RoWN. Nouns 41063 56532 52009 1839 Verbs 10397 16484 14210 759 Adjective 4822 8203 7407 79 Adverbs 3066 4019 3248 110 TOTAL 59348 85238 75656 2787 It is worth noticing that antonymy, which is a lexical relation in PWN, is represented as a semantic one in RoWN. The conceptual opposition between the synsets is more useful in various applications than the mere antonymy between two literals. With the exception of attribute relation, all the others enumerated in Table 2 link synsets with literals of the same part of speech. A path between two words of a different part of speech, about which any speaker would say they are related, although not impossible to find, would be too long, thus providing wrong information about the similarity between those words.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 54, |
| "end": 61, |
| "text": "Table 1", |
| "ref_id": "TABREF0" |
| }, |
| { |
| "start": 769, |
| "end": 902, |
| "text": "Nouns 41063 56532 52009 1839 Verbs 10397 16484 14210 759 Adjective 4822 8203 7407 79 Adverbs 3066 4019 3248 110 TOTAL", |
| "ref_id": "TABREF0" |
| }, |
| { |
| "start": 1241, |
| "end": 1248, |
| "text": "Table 2", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Latest Quantitative Developments", |
| "sec_num": "3" |
| }, |
| { |
| "text": "Using RoWN in applications, as presented above, showed unnatural lexical chains, such as one of the possible chains between inventator -inventor\u2016 and inventa -to invent\u2016:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Derivational Relations", |
| "sec_num": "4" |
| }, |
| { |
| "text": "inventator(1.1) instance_hyponym James_Watt(x) James_Watt(x) instance_hypernym inginer(1.1) inginer(1.1) hyponym inginer_software(1) inginer_software(1) domain_member_TOPIC \u015ftiin\u0163a_calculatoarelor(x) \u015ftiin\u0163a_calculatoarelor(x) domain_TOPIC pro- grama(3) programa(3) hyponym crea_mental(1) crea_mental(1) hypernym inventa(1)", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Derivational Relations", |
| "sec_num": "4" |
| }, |
| { |
| "text": "The strangeness of this example results from the intricate path from inventator to inventa, uncommon for whatever speaker of Romanian: inventator -James Wattinginer -engineer\u2016inginer software -software engineer\u2016programa -to program\u2016crea mental -to create by mental act\u2016inventa.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Derivational Relations", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Faced with a number of such cases, we decided to implement derivational relations into our wordnet.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Derivational Relations", |
| "sec_num": "4" |
| }, |
| { |
| "text": "This type of relations exists in other wordnets as well: the Turkish WordNet (Bilgin et al., 2004) , PWN (Fellbaum et al., 2007) , the Czech WordNet (Pala and Hlav\u00e1\u010dkov\u00e1, 2007) , the Polish WordNet (Piasecki et al., 2012) , the Estonian one (Kahusk, et al., 2010) . Given the language-specific character of such relations, each team undertook their own strategy for finding the relations in their wordnet. However, there are teams that transferred the derivational relations in PWN and then validated them: this is the case for the Bulgarian WordNet (Koeva, 2008) , the Serbian and the Finnish one (Lind\u00e9n and Niemi, 2013) .", |
| "cite_spans": [ |
| { |
| "start": 77, |
| "end": 98, |
| "text": "(Bilgin et al., 2004)", |
| "ref_id": "BIBREF0" |
| }, |
| { |
| "start": 105, |
| "end": 128, |
| "text": "(Fellbaum et al., 2007)", |
| "ref_id": "BIBREF2" |
| }, |
| { |
| "start": 149, |
| "end": 176, |
| "text": "(Pala and Hlav\u00e1\u010dkov\u00e1, 2007)", |
| "ref_id": "BIBREF11" |
| }, |
| { |
| "start": 198, |
| "end": 221, |
| "text": "(Piasecki et al., 2012)", |
| "ref_id": "BIBREF13" |
| }, |
| { |
| "start": 241, |
| "end": 263, |
| "text": "(Kahusk, et al., 2010)", |
| "ref_id": "BIBREF6" |
| }, |
| { |
| "start": 550, |
| "end": 563, |
| "text": "(Koeva, 2008)", |
| "ref_id": "BIBREF7" |
| }, |
| { |
| "start": 598, |
| "end": 622, |
| "text": "(Lind\u00e9n and Niemi, 2013)", |
| "ref_id": "BIBREF10" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Derivational Relations", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Whereas most of the undertakings above aimed at expanding the network with new synsets derivationally linked with the literals already in the wordnet, we were interested in adding such relations between literals that are in the synsets. No extension was intended, at least for the moment.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Derivational Relations", |
| "sec_num": "4" |
| }, |
| { |
| "text": "We discuss below some theoretical aspects of derivational relations and the significance of their representation in a wordnet and then present the methodology we adopted for identifying and annotating them in RoWN.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Derivational Relations", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Derivation is one means of creating new words in a language from existing morphemes, i.e. the smallest units of a language that have their own meaning. It ensures both formal and semantic relatedness between the root and the derived word: the formal relatedness is ensured by the fact that the root and the derived word contain (almost) the same string of letters that represent the root, while the semantic relatedness is ensured by the compositionality of meaning of the derived word: its meaning is a sum of the meaning of the root and the meaning of the affix(es). Thus, the Romanian words alerga -run\u2016 and alerg\u0103tor -runner\u2016 are derivationally related: the latter is obtained from the former by adding the suffix -\u0103tor (after removing -a, the infinitive suffix) and it means -the one who runs\u2016. However, derivational relations cannot be established for all meanings of these words: when considered with their proper meaning, they are related, but when alerga is considered with its figurative meaning -to try hard to get something\u2016, it does not establish a derivational relation with alerg\u0103tor, as it has not developed any related figurative meaning.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Pre-requisites", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "In the derivation process only one affix of a type is added. So, a prefix and a suffix can be added to a root in the same derivation step, but never two suffixes or/and two prefixes. If a word contains more than two affixes of the same type, then they were attached in different steps in the derivation.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Pre-requisites", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "Having available a list of (492) Romanian affixes and the list of (31872) simple literals in RoWN, we searched for pairs of literals (literal 1 and literal 2 ) such that literal 1 +/-affix(es) = literal 2 . The -+\u2016 version covers progressive derivation, while the --\u2016 version covers backformation. We allowed for at most 2 affixes, but of different types, as discussed above. The results are presented in The percents are reasonable: it is a wellknown fact that prefixation is weakly productive in Romanian, unlike suffixation.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Identifying derivational relations between literals in RoWN", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "We subjected the found pairs to an automatic and then a manual validation. For the former, we enriched the list of affixes with information about the part of speech of the words to which they can attach and of the words they help create. The list is available at www.racai.ro/~vergi under Research. For example, the suffix -a can be attached to nouns or to adjectives to create verbs:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Identifying derivational relations between literals in RoWN", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "-a n>v a>v Examples include: buton (-button\u2016) + -a > butona (-to channel-surf\u2016), scurt (-short\u2016) + -a > scurta (-to shorten\u2016).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Identifying derivational relations between literals in RoWN", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "Afterwards we proceeded to a manual validation of the whole number of pairs. The results are presented in Table 4 : for each type of derivation (DT) (prefixation P or suffixation S), from the found pairs (column 2) we present the number of those passing the automatic validation (AV) in column 3 and then of those that passed the manual validation (MV) in column 4; the last column presents the percent of manually validated pairs for each derivation type. Examples of pairs that passed the automatic validation but not the manual one include: prinde -to catch\u2016surprinde -to surprise\u2016, abate -to deviate\u2016abator -slaughter house\u2016.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 106, |
| "end": 113, |
| "text": "Table 4", |
| "ref_id": "TABREF3" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Identifying derivational relations between literals in RoWN", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "Having already established that derivational relations need to be marked at the word sense level, not for all senses of the words in a pair, the next necessary step is to calculate the Cartesian product of the sets of synsets in which the members of the validated pairs occur. Thus, for the 10442 pairs of literals resulted after manual validation, we calculated the Cartesian product for each pair, obtaining a total of 101729 pairs of synsets. They display formal relatedness and, in order to mark a derivational relation for them, it is also necessary to subject them to a semantic evaluation. A linguist goes through them and whenever semantic similarity is noticed, the pair is labeled with one of the 57 semantic labels we established: 16 for prefixed words (together, subsumption, opposition, mero, eliminate, iterative, through, repeat, imply, similitude, instead, aug, before, anti, out, back) and 41 for suffixed ones (subsumption, member_holo, member_mero, substance_holo, substance_mero, ingredient_holo, holonym, part, agent, result, location, of_origin, job, state, period, undergoer, instrument, sound, cause, container, vehicle, body_part, material, destination, gender, wife, dim, aug, object_made_by, subject_to, by_means_of, clothes, event, abstract, colour, tax, make_become, make_acquire, manner, similitude, related) .", |
| "cite_spans": [ |
| { |
| "start": 928, |
| "end": 1338, |
| "text": "(subsumption, member_holo, member_mero, substance_holo, substance_mero, ingredient_holo, holonym, part, agent, result, location, of_origin, job, state, period, undergoer, instrument, sound, cause, container, vehicle, body_part, material, destination, gender, wife, dim, aug, object_made_by, subject_to, by_means_of, clothes, event, abstract, colour, tax, make_become, make_acquire, manner, similitude, related)", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Sense level annotation", |
| "sec_num": "4.3" |
| }, |
| { |
| "text": "The most frequently attached semantic labels are: for prefixed words: opposition (neesen\u0163ial -unessential\u2016essential -essential\u2016) (792), subsumption (subclas\u0103 -subclass\u2016 -clas\u0103 -class\u2016) (363), repeat (reaprinde -reignite\u2016aprinde -ignite\u2016) (305); for suffixed words: related (c\u0103lduros -warm\u2016 -c\u0103ldur\u0103 -warmth\u2016) (1294), event (\u00eemp\u0103rt\u0103\u015fanie -communion\u2016 -\u00eemp\u0103rt\u0103\u015fi -commune\u2016) (699), abstract (cerin\u0163\u0103 -require-ment\u2016cere -require\u2016) (490), manner (primejdios -dangerous\u2016 -primejdie -danger\u2016) (436), agent (lingu\u015fitor -adulator\u2016 -lingu\u015fi -ad-ulate\u2016) (394). At the end of the article, in the Annex, containing Table7 and Table 8 , we present the semantic labels and their frequencies for prefixed and, respectively, suffixed words, accompanied by examples.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 612, |
| "end": 619, |
| "text": "Table 8", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Sense level annotation", |
| "sec_num": "4.3" |
| }, |
| { |
| "text": "Going through 55849 such pairs of synsets, we obtained the results in Table 5 . Table 5 . Semantically annotated pairs.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 70, |
| "end": 77, |
| "text": "Table 5", |
| "ref_id": null |
| }, |
| { |
| "start": 80, |
| "end": 87, |
| "text": "Table 5", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Statistics about derivational relations", |
| "sec_num": "4.4" |
| }, |
| { |
| "text": "The aim of marking these derivational relations was to increase the number of links between synsets, especially between synsets of different parts of speech. For the validated pairs we included in Table 6 statistics about the derivational relations involving words of the same and of different part of speech. It is obvious that, on the whole, adding derivational relations to a wordnet increases the number of cross-part of speech (PoS) relations. Table 6 . Distribution of derivational relation on PoS.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 197, |
| "end": 204, |
| "text": "Table 6", |
| "ref_id": null |
| }, |
| { |
| "start": 449, |
| "end": 456, |
| "text": "Table 6", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Prefixed", |
| "sec_num": null |
| }, |
| { |
| "text": "We have built an Application Programming Interface (API) for RoWN, called RoWordNetLib, meant as a tool to aid quick implementations of RoWN into both research-oriented and industry applications. When designing it, we envisaged a tool that should be easy to use, easy to extend and that would offer a sufficiently large array of functionalities. The chosen programming language is Java. The main functionalities that RoWordNetLib provides are: \uf0b7 Similarity Metrics (both distance-based and semantic).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "RoWordNetLib", |
| "sec_num": "5" |
| }, |
| { |
| "text": "\uf0b7 Input/", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "RoWordNetLib", |
| "sec_num": "5" |
| }, |
| { |
| "text": "The API's uses can be classified as (1) internalit helps to facilitate the continuous work of enriching RoWN and (2) externalto quicken the development of Romanian-enabled smart applications. By providing set operations like difference, intersection or reunion on RoWordNet objects, more people can work in parallel on RoWN and then easily join their versions into a single wordnet, thus easing its development. Externally, wordnets are successfully used to perform word sense disambiguation, information retrieval, information extraction, machine translation, automatic text classification and summarization. RoWordNetLib is structured into several packages, each with its assigned functionality. The main packages are: 'data', 'io', 'op' and 'wsd'.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "RoWordNetLib", |
| "sec_num": "5" |
| }, |
| { |
| "text": "The 'data' package contains the data structures RoWordNetLib uses internally. Its structure is simple, following the way the data is naturally structured in a wordnet: a RoWordNet object contains an array of Synset objects which are indexed by the synset ID for retrieval speed. Each Synset object contains a number of primitive types as well as an array of Literal objects and an array of Relation objects. A Literal object contains a word and an associated sense. A Relation object contains a relation (string) that points to a target synset (defined as an ID), as well as optionally having a source and target literal for cases where the relation is not between synsets but between two synsets' particular literals.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "RoWordNetLib", |
| "sec_num": "5" |
| }, |
| { |
| "text": "The 'io' package provides input and output functions. The most important I/O function is reading and writing RoWordNet objects in their native XML format.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "RoWordNetLib", |
| "sec_num": "5" |
| }, |
| { |
| "text": "The 'op' package provides different operational tools: (1) set operation methods for joining, intersecting, complementing, etc., multiple RoWordNet objects; (2) through the BFWalk class, the ability to perform a breadth-first walk through the RoWN semantic network; (3) a number of distance-based and semantic similarity measures (Resnik, 1995) for measuring the closeness of concepts (lexicalized by literals in synsets).", |
| "cite_spans": [ |
| { |
| "start": 330, |
| "end": 344, |
| "text": "(Resnik, 1995)", |
| "ref_id": "BIBREF14" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "RoWordNetLib", |
| "sec_num": "5" |
| }, |
| { |
| "text": "The 'wsd' package implements two Word Sense Disambiguation algorithms: Lesk (1986) and an adapted version of Lesk. They are used to obtain information content values for synsets in RoWN given an arbitrary Romanian text as the input corpus, which is further used to enable the semantic similarity measures.", |
| "cite_spans": [ |
| { |
| "start": 71, |
| "end": 82, |
| "text": "Lesk (1986)", |
| "ref_id": "BIBREF9" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "RoWordNetLib", |
| "sec_num": "5" |
| }, |
| { |
| "text": "RoWN is a valuable resource for the Romanian language and the NLP group of RACAI uses it in most of their applications. We presented here our latest qualitative and quantitative achievements.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusions and Further Work", |
| "sec_num": "6" |
| }, |
| { |
| "text": "Further enrichment of RoWN is a constant preoccupation of our team. It follows all the time the other interests of the group. For instance, the last set of implemented synsets was made up of verbs exclusively, given our present interest to cover VerbNet 3.1, with the prospect of creating a parser for Romanian.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusions and Further Work", |
| "sec_num": "6" |
| }, |
| { |
| "text": "Increasing the density of relations between synsets in order to make RoWN more effective in applications was obtained by adding derivational relations. Although they are relations between literals, the semantic labels we attached to them can be viewed as a link between the synsets to which the respective literals belong. After finishing the semantic annotation of the derivative pairs, we could try to expand the network with automatically derived words. For Romanian an experiment of automatically deriving words is reported by Petic (2011) , who used very productive and reliable affixes. With the list of affixes and their combination possibilities (available at www.racai.ro/~vergi under Research) that we have created, we can dare test new cases of automatic derivation for Romanian.", |
| "cite_spans": [ |
| { |
| "start": 531, |
| "end": 543, |
| "text": "Petic (2011)", |
| "ref_id": "BIBREF12" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusions and Further Work", |
| "sec_num": "6" |
| }, |
| { |
| "text": "http://ws.racai.ro:9191/repository/", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [ |
| { |
| "text": "Occur ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Label", |
| "sec_num": null |
| } |
| ], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "Morphosemantic relations in and across wordnets: A study based on Turkish", |
| "authors": [ |
| { |
| "first": "Orhan", |
| "middle": [], |
| "last": "Bilgin", |
| "suffix": "" |
| }, |
| { |
| "first": "\u00d6zlem", |
| "middle": [], |
| "last": "\u00c7etinoglu", |
| "suffix": "" |
| }, |
| { |
| "first": "Kemal", |
| "middle": [], |
| "last": "Oflazer", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Orhan Bilgin, \u00d6zlem \u00c7etinoglu, and Kemal Oflazer. 2004. Morphosemantic relations in and across wordnets: A study based on Turkish. P. Sojka, K.", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "Proceedings of GWC", |
| "authors": [ |
| { |
| "first": "P", |
| "middle": [], |
| "last": "Pala", |
| "suffix": "" |
| }, |
| { |
| "first": "C", |
| "middle": [], |
| "last": "Smrz", |
| "suffix": "" |
| } |
| ], |
| "year": null, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Pala, P. Smrz, C. Fellbaum, P. Vossen (Eds.), Pro- ceedings of GWC.", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "Putting Semantics into WordNet's -Morphosemantic\u2016 Links", |
| "authors": [ |
| { |
| "first": "Christiane", |
| "middle": [], |
| "last": "Fellbaum", |
| "suffix": "" |
| }, |
| { |
| "first": "Anne", |
| "middle": [], |
| "last": "Osherson", |
| "suffix": "" |
| }, |
| { |
| "first": "Peter", |
| "middle": [ |
| "E" |
| ], |
| "last": "Clark", |
| "suffix": "" |
| } |
| ], |
| "year": 2007, |
| "venue": "Proceedings of the 3rd Language and Technology Conference", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Christiane Fellbaum, Anne Osherson, and Peter E. Clark. 2007. Putting Semantics into WordNet's -Morphosemantic\u2016 Links. Proceedings of the 3rd Language and Technology Conference.", |
| "links": null |
| }, |
| "BIBREF3": { |
| "ref_id": "b3", |
| "title": "Unsupervised Word Sense Disambiguation with Lexical Chains and Graph-Based Context Formalization. Zygmunt Vetulani", |
| "authors": [ |
| { |
| "first": "Radu", |
| "middle": [], |
| "last": "Ion", |
| "suffix": "" |
| }, |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "\u015etef\u0103nescu", |
| "suffix": "" |
| } |
| ], |
| "year": 2011, |
| "venue": "LTC 2009", |
| "volume": "6562", |
| "issue": "", |
| "pages": "435--443", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Radu Ion and Dan \u015etef\u0103nescu. 2011. Unsupervised Word Sense Disambiguation with Lexical Chains and Graph-Based Context Formalization. Zygmunt Vetulani (ed.): LTC 2009, Lecture Notes in Artifi- cial Intelligence, 6562/2011: 435-443.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "Multilingual versus Monolingual Word Sense Disambiguation", |
| "authors": [ |
| { |
| "first": "Radu", |
| "middle": [], |
| "last": "Ion", |
| "suffix": "" |
| }, |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Tufi\u015f", |
| "suffix": "" |
| } |
| ], |
| "year": 2009, |
| "venue": "International Journal of Speech Technology", |
| "volume": "12", |
| "issue": "2-3", |
| "pages": "113--124", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Radu Ion and Dan Tufi\u015f. 2009. Multilingual versus Monolingual Word Sense Disambiguation. Inter- national Journal of Speech Technology; 12 (2- 3):113-124.", |
| "links": null |
| }, |
| "BIBREF5": { |
| "ref_id": "b5", |
| "title": "RACAI's QA System at the Romanian-Romanian Multiple Language Question Answering (QA@CLEF2008) Main Task", |
| "authors": [ |
| { |
| "first": "Radu", |
| "middle": [], |
| "last": "Ion", |
| "suffix": "" |
| }, |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "\u015etef\u0103nescu", |
| "suffix": "" |
| }, |
| { |
| "first": "Alexandru", |
| "middle": [], |
| "last": "Ceau\u015fu", |
| "suffix": "" |
| }, |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Tufi\u015f", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "Working Notes for the CLEF", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Radu Ion, Dan \u015etef\u0103nescu, Alexandru Ceau\u015fu, and Dan Tufi\u015f. 2008. RACAI's QA System at the Ro- manian-Romanian Multiple Language Question Answering (QA@CLEF2008) Main Task. Carol Peters et al. (eds.) Working Notes for the CLEF 2008 Workshop: 10.", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "Enriching Estonian WordNet with Derivations and Semantic Relations", |
| "authors": [ |
| { |
| "first": "Neeme", |
| "middle": [], |
| "last": "Kahusk", |
| "suffix": "" |
| }, |
| { |
| "first": "Kadri", |
| "middle": [], |
| "last": "Kerner", |
| "suffix": "" |
| }, |
| { |
| "first": "Kadri", |
| "middle": [], |
| "last": "Vider", |
| "suffix": "" |
| } |
| ], |
| "year": 2010, |
| "venue": "Proceeding of the 2010 conference on Human Language Technologies -The", |
| "volume": "", |
| "issue": "", |
| "pages": "195--200", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Neeme Kahusk, Kadri Kerner, and Kadri Vider. 2010. Enriching Estonian WordNet with Derivations and Semantic Relations. Proceeding of the 2010 conference on Human Language Technologies -The Baltic Perspective:195-200.", |
| "links": null |
| }, |
| "BIBREF7": { |
| "ref_id": "b7", |
| "title": "Derivational and Morphosemantic Relations in Bulgarian Wordnet. Intelligent Information Systems", |
| "authors": [ |
| { |
| "first": "Svetla", |
| "middle": [], |
| "last": "Koeva", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "", |
| "volume": "XVI", |
| "issue": "", |
| "pages": "359--369", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Svetla Koeva. 2008. Derivational and Morphoseman- tic Relations in Bulgarian Wordnet. Intelligent In- formation Systems; XVI:359-369.", |
| "links": null |
| }, |
| "BIBREF8": { |
| "ref_id": "b8", |
| "title": "Morpho-semantic Relations in Wordnet -A Case Study for two Slavic Languages", |
| "authors": [ |
| { |
| "first": "Svetla", |
| "middle": [], |
| "last": "Koeva", |
| "suffix": "" |
| }, |
| { |
| "first": "Cvetana", |
| "middle": [], |
| "last": "Krstev", |
| "suffix": "" |
| }, |
| { |
| "first": "Du\u0161ko", |
| "middle": [], |
| "last": "Vitas", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "Proceedings of the Fourth Global WordNet Conference", |
| "volume": "", |
| "issue": "", |
| "pages": "239--254", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Svetla Koeva, Cvetana Krstev, and Du\u0161ko Vitas. 2008. Morpho-semantic Relations in Wordnet -A Case Study for two Slavic Languages. Proceed- ings of the Fourth Global WordNet Confer- ence:239-254.", |
| "links": null |
| }, |
| "BIBREF9": { |
| "ref_id": "b9", |
| "title": "Automatic sense disambiguation using machine readable dictionaries: How to tell a pine cone from an ice cream cone", |
| "authors": [ |
| { |
| "first": "Michael", |
| "middle": [], |
| "last": "Lesk", |
| "suffix": "" |
| } |
| ], |
| "year": 1986, |
| "venue": "5th SIGDOC", |
| "volume": "", |
| "issue": "", |
| "pages": "24--26", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Michael Lesk. 1986. Automatic sense disambiguation using machine readable dictionaries: How to tell a pine cone from an ice cream cone. 5th SIGDOC:24-26.", |
| "links": null |
| }, |
| "BIBREF10": { |
| "ref_id": "b10", |
| "title": "Is it possible to create a very large wordnet in 100 days? An evaluation", |
| "authors": [ |
| { |
| "first": "Krister", |
| "middle": [], |
| "last": "Lind\u00e9n", |
| "suffix": "" |
| }, |
| { |
| "first": "Jyrki", |
| "middle": [], |
| "last": "Niemi", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "Language Resources and Evaluation", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Krister Lind\u00e9n and Jyrki Niemi. 2013. Is it possible to create a very large wordnet in 100 days? An evalu- ation. Language Resources and Evaluation, http://link.springer.com/article/10.1007%2Fs10579 -013-9245-0.", |
| "links": null |
| }, |
| "BIBREF11": { |
| "ref_id": "b11", |
| "title": "Derivational relations in Czech Wordnet", |
| "authors": [ |
| { |
| "first": "Karel", |
| "middle": [], |
| "last": "Pala", |
| "suffix": "" |
| }, |
| { |
| "first": "Dana", |
| "middle": [], |
| "last": "Hlav\u00e1\u010dkov\u00e1", |
| "suffix": "" |
| }, |
| { |
| "first": "D", |
| "middle": [], |
| "last": "", |
| "suffix": "" |
| } |
| ], |
| "year": 2007, |
| "venue": "Proceedings of the Workshop on Balto-Slavonic Natural Language Processing", |
| "volume": "", |
| "issue": "", |
| "pages": "75--81", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Karel Pala and Dana Hlav\u00e1\u010dkov\u00e1, D. 2007. Deriva- tional relations in Czech Wordnet. Proceedings of the Workshop on Balto-Slavonic Natural Lan- guage Processing: 75-81.", |
| "links": null |
| }, |
| "BIBREF12": { |
| "ref_id": "b12", |
| "title": "Generative mechanisms of Romanian derivational morphology", |
| "authors": [ |
| { |
| "first": "Mircea", |
| "middle": [], |
| "last": "Petic", |
| "suffix": "" |
| } |
| ], |
| "year": 2011, |
| "venue": "Memoirs of the Scientific Section of the Romanian Academy. Series IV", |
| "volume": "", |
| "issue": "", |
| "pages": "21--30", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Mircea Petic. 2011. Generative mechanisms of Ro- manian derivational morphology. Memoirs of the Scientific Section of the Romanian Academy. Series IV, Tome XXXIV:21-30.", |
| "links": null |
| }, |
| "BIBREF13": { |
| "ref_id": "b13", |
| "title": "Recognition of Polish Derivational Relations Based on Supervised Learning Scheme", |
| "authors": [ |
| { |
| "first": "Maciej", |
| "middle": [], |
| "last": "Piasecki", |
| "suffix": "" |
| }, |
| { |
| "first": "Radoslaw", |
| "middle": [], |
| "last": "Ramocki", |
| "suffix": "" |
| }, |
| { |
| "first": "Marek", |
| "middle": [], |
| "last": "Maziarz", |
| "suffix": "" |
| } |
| ], |
| "year": 2012, |
| "venue": "Proceedings of LREC 2012", |
| "volume": "", |
| "issue": "", |
| "pages": "916--922", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Maciej Piasecki, Radoslaw Ramocki, and Marek Ma- ziarz. 2012. Recognition of Polish Derivational Re- lations Based on Supervised Learning Scheme. Proceedings of LREC 2012: 916-922.", |
| "links": null |
| }, |
| "BIBREF14": { |
| "ref_id": "b14", |
| "title": "Using information content to evaluate semantic similarity in a taxonomy", |
| "authors": [ |
| { |
| "first": "Philip", |
| "middle": [], |
| "last": "Resnik", |
| "suffix": "" |
| } |
| ], |
| "year": 1995, |
| "venue": "14th International Joint Conference on Artificial Intelligence", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Philip Resnik. 1995. Using information content to evaluate semantic similarity in a taxonomy. 14th International Joint Conference on Artificial Intelligence.", |
| "links": null |
| }, |
| "BIBREF15": { |
| "ref_id": "b15", |
| "title": "Bal-kaNet: Aims, Methods, Results and Perspectives. A General Overview", |
| "authors": [ |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Tufi\u0219", |
| "suffix": "" |
| }, |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Cristea", |
| "suffix": "" |
| }, |
| { |
| "first": "Sofia", |
| "middle": [], |
| "last": "Stamou", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "Romanian Journal on Information Science and Technology", |
| "volume": "7", |
| "issue": "", |
| "pages": "9--34", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Dan Tufi\u0219, Dan Cristea, and Sofia Stamou. 2004. Bal- kaNet: Aims, Methods, Results and Perspectives. A General Overview. Romanian Journal on In- formation Science and Technology; 7:9-34.", |
| "links": null |
| }, |
| "BIBREF16": { |
| "ref_id": "b16", |
| "title": "Radu Ion", |
| "authors": [ |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Tufi\u015f", |
| "suffix": "" |
| }, |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Verginica Barbu Mititelu", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "\u015etef\u0103nescu", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "The Romanian Wordnet in a Nutshell. Language Resources and Evaluation", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Dan Tufi\u015f, Verginica Barbu Mititelu, Dan \u015etef\u0103nescu, Radu Ion. 2013. The Romanian Wordnet in a Nut- shell. Language Resources and Evaluation, http://link.springer.com/article/10.1007%2Fs10579 -013-9230-7.", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "TABREF0": { |
| "text": "Statistics about synsets and literals in RoWN.", |
| "html": null, |
| "content": "<table><tr><td>Relation</td><td>Number</td></tr><tr><td>hypo/hyperonymy</td><td>48316</td></tr><tr><td>instance_hypo/hyperonymy</td><td>3889</td></tr><tr><td>antonym</td><td>4131</td></tr><tr><td>similar_to</td><td>4838</td></tr><tr><td>verb_group</td><td>1530</td></tr><tr><td>member_holonym</td><td>2047</td></tr><tr><td>part_holonym</td><td>5573</td></tr><tr><td>substance_holonym</td><td>410</td></tr><tr><td>also_see</td><td>1333</td></tr><tr><td>attribute</td><td>958</td></tr><tr><td>cause</td><td>196</td></tr><tr><td>entailment</td><td>371</td></tr></table>", |
| "type_str": "table", |
| "num": null |
| }, |
| "TABREF1": { |
| "text": "", |
| "html": null, |
| "content": "<table><tr><td/><td>:</td><td/><td/></tr><tr><td>Derivation</td><td>Number</td><td>of</td><td>Percent</td></tr><tr><td>type</td><td>derived</td><td/><td/></tr><tr><td/><td>words</td><td/><td/></tr><tr><td>Prefixation</td><td colspan=\"2\">2862</td><td>17.43</td></tr><tr><td>Suffixation</td><td colspan=\"2\">13556</td><td>82.57</td></tr><tr><td>TOTAL</td><td colspan=\"2\">16418</td><td/></tr><tr><td colspan=\"4\">Table 3. Derivational relations between simple literals</td></tr><tr><td/><td>in RoWN.</td><td/><td/></tr></table>", |
| "type_str": "table", |
| "num": null |
| }, |
| "TABREF3": { |
| "text": "Validated pairs.", |
| "html": null, |
| "content": "<table/>", |
| "type_str": "table", |
| "num": null |
| } |
| } |
| } |
| } |
