Base_JSON/prefixR/json/R09/R09-1049.json

{
    "paper_id": "R09-1049",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T15:00:20.550923Z"
    },
    "title": "A morphological and syntactic wide-coverage lexicon for Spanish: The Leffe",
    "authors": [
        {
            "first": "Miguel",
            "middle": [
                "A"
            ],
            "last": "Molinero",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Univ. of A Coru\u00f1a A Coru\u00f1a",
                "location": {
                    "country": "Spain"
                }
            },
            "email": "mmolinero@udc.es"
        },
        {
            "first": "Grupo",
            "middle": [],
            "last": "Lys",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Univ. of A Coru\u00f1a A Coru\u00f1a",
                "location": {
                    "country": "Spain"
                }
            },
            "email": ""
        },
        {
            "first": "Beno\u00eet",
            "middle": [],
            "last": "Sagot",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "INRIA Paris",
                "location": {
                    "country": "France"
                }
            },
            "email": "benoit.sagot@inria.fr"
        },
        {
            "first": "Lionel",
            "middle": [],
            "last": "Nicola\u015b",
            "suffix": "",
            "affiliation": {
                "laboratory": "Laboratoire I3S Sophia Antipolis",
                "institution": "",
                "location": {
                    "country": "France"
                }
            },
            "email": "lnicolas@i3s.unice.fr"
        },
        {
            "first": "Equipe",
            "middle": [],
            "last": "Rl",
            "suffix": "",
            "affiliation": {
                "laboratory": "Laboratoire I3S Sophia Antipolis",
                "institution": "",
                "location": {
                    "country": "France"
                }
            },
            "email": ""
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "In this paper, we introduce the L\u00e9xico de Formas Flexionadas del Espa\u00f1ol (Leffe), a wide-coverage morphological and syntactic Spanish lexicon based on the Alexina lexical framework. We explain how the Leffe has been created by merging together several heterogeneous lexicons and how the Alexina lexical framework has been applied to Spanish. We also introduce a semi-automatic technique based on a tagger to detect the lexicon's deficiencies. A preliminary evaluation shows the potential of the Leffe and the relevance of both creation and extension processes.",
    "pdf_parse": {
        "paper_id": "R09-1049",
        "_pdf_hash": "",
        "abstract": [
            {
                "text": "In this paper, we introduce the L\u00e9xico de Formas Flexionadas del Espa\u00f1ol (Leffe), a wide-coverage morphological and syntactic Spanish lexicon based on the Alexina lexical framework. We explain how the Leffe has been created by merging together several heterogeneous lexicons and how the Alexina lexical framework has been applied to Spanish. We also introduce a semi-automatic technique based on a tagger to detect the lexicon's deficiencies. A preliminary evaluation shows the potential of the Leffe and the relevance of both creation and extension processes.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Abstract",
                "sec_num": null
            }
        ],
        "body_text": [
            {
                "text": "High-level Natural Language Processing (NLP) tools require reliable linguistic resources, such as lexicons and grammars. Nowadays, such relevant resources exist for English, but are often absent or incomplete for other languages, even major ones. For example, some lexical resources exist for Spanish, but none of them combines satisfactorily the following properties:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "\u2022 coverage: all words, including rare ones, in all categories should be included;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "\u2022 quality: manually and automatically developed resources contain various errors;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "\u2022 richness: applications such as (deep) parsing require at least morphological and syntactic information, including subcategorization frames.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "The Leffe 1 is a wide-coverage morphological and syntactic lexicon based on the Alexina framework [13, 15, 1] .",
                "cite_spans": [
                    {
                        "start": 98,
                        "end": 102,
                        "text": "[13,",
                        "ref_id": "BIBREF12"
                    },
                    {
                        "start": 103,
                        "end": 106,
                        "text": "15,",
                        "ref_id": "BIBREF14"
                    },
                    {
                        "start": 107,
                        "end": 109,
                        "text": "1]",
                        "ref_id": "BIBREF0"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "This lexicon follows the linguistic criteria applied on the French lexicon Lefff 2 taking advantage of the linguistic proximity between Spanish and French as Romance languages.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "The main contributions of this piece of research are the following:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "\u2022 we present a morphological and syntactic wide coverage lexicon for Spanish;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "\u2022 we describe an enhanced available lexical framework, 1 L\u00e9xico de formas flexionadas del espa\u00f1ol -Lexicon of Spanish inflected forms 2 Lexique des formes fl\u00e9chies du fran\u00e7ais -Lexicon of French inflected forms",
                "cite_spans": [
                    {
                        "start": 55,
                        "end": 56,
                        "text": "1",
                        "ref_id": "BIBREF0"
                    },
                    {
                        "start": 134,
                        "end": 135,
                        "text": "2",
                        "ref_id": "BIBREF1"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "\u2022 we expose a simple semi-automatic PoS tagger-based approach to detect numerous missing entries in a lexicon (including homonyms).",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "The work described here is one of the starting points of the recently created Victoria project. This project aims at developing techniques and tools for an efficient acquisition and correction of the linguistic resources necessary to symbolic syntactic parsers. The first phase of the project focuses on Spanish, Galician 3 and French.",
                "cite_spans": [
                    {
                        "start": 322,
                        "end": 323,
                        "text": "3",
                        "ref_id": "BIBREF2"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "This paper is organized as follows: we present first a brief description of Spanish language in Section 2. In Section 3 we describe the lexical framework used to formalize the linguistic information. Then, we briefly describe how some available resources were used to develop the Leffe. In Section 5 we present a semi-automatic technique to correct and extend the lexicon. Finally in Section 6 we show preliminary evaluations of the lexicon and present our conclusions in Section 7.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "Spanish is a Romance Language, just like Italian, French, Portuguese, and many others. Despite being spoken as a mother tongue by more than 400 million people, this language is little formalized within the framework of NLP when compared with English.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Spanish language in brief",
                "sec_num": "2"
            },
            {
                "text": "Spanish is an inflected language, with a two-gender system, about fifty conjugated forms per verb, but limited inflections for nouns, adjectives, and determiners. It is morphologically characterized with the Latin alphabet plus the letter\u00f1 and the digraphs ch, ll and rr. Apart from this, the acute accents are commonly used and they enable homophones to be distinguished: e.g., te ('you', object pronoun) and t\u00e9 ('tea').",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Spanish language in brief",
                "sec_num": "2"
            },
            {
                "text": "Regarding syntax and grammar, it is right-branching, uses prepositions, and usually, though not always, places adjectives after nouns. Its syntax is generally Subject Verb Object, though variations are valid and very common. The subject is usually omitted but appears in an implicit fashion. Contrary to English, but similarly to other Romance languages, it is verb-framed, i.e., many Spanish verbs directly encode motion path, and may leave out the manner of motion or express it in a complement of manner: e.g., entrar (go in), salir (go out), subir (go up). It also use a noticeable range of pronominal verbs.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Spanish language in brief",
                "sec_num": "2"
            },
            {
                "text": "A corner stone of this work relies on the fact that Spanish is similar to other Romance Languages in many ways. Indeed, a network of correspondances can easily be established between their features.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Spanish language in brief",
                "sec_num": "2"
            },
            {
                "text": "This led us to consider the benefits of reusing resources describing related languages when building the Leffe. Such approach presents many advantages :",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Spanish language in brief",
                "sec_num": "2"
            },
            {
                "text": "\u2022 a more flexible and complete formalism could be found to develop the Leffe,",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Spanish language in brief",
                "sec_num": "2"
            },
            {
                "text": "\u2022 establishing interlingual links between resources written with a common formalism results easier,",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Spanish language in brief",
                "sec_num": "2"
            },
            {
                "text": "\u2022 the data contained in resources describing other related languages can be more easily acquired.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Spanish language in brief",
                "sec_num": "2"
            },
            {
                "text": "According to these statements, we identified the Lefff, an enhanced morphological and syntactic wide-coverage French lexicon based on the Alexina format (See section 3), as the best candidate.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Spanish language in brief",
                "sec_num": "2"
            },
            {
                "text": "A detailed lexical description of all words (or as many as possible) belonging to a language is needed in order to perform high-level NLP tasks such as deep parsing. This information is usually compiled into a lexicon, which could be defined as a list of lexical forms associated with morphological and syntactic information. Alexina is a framework that represents lexical information in a complete, efficient and readable way [11, 1] , and is compatible with the LMF 4 [2] standard. The flexibility and completeness of the Alexina format allow a straightforward integration with deep grammatical formalisms (LFG, LTAG) which require detailed syntactic data for all forms, and allow to model lexical information for diverse languages. It is indeed the lexical framework of the Lefff, a large-coverage morphological and syntactic lexicon for French, but also that of other lexical resources for languages such as Polish, Slovak, and soon English.",
                "cite_spans": [
                    {
                        "start": 427,
                        "end": 431,
                        "text": "[11,",
                        "ref_id": "BIBREF10"
                    },
                    {
                        "start": 432,
                        "end": 434,
                        "text": "1]",
                        "ref_id": "BIBREF0"
                    },
                    {
                        "start": 468,
                        "end": 469,
                        "text": "4",
                        "ref_id": "BIBREF3"
                    },
                    {
                        "start": 470,
                        "end": 473,
                        "text": "[2]",
                        "ref_id": "BIBREF1"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Alexina framework",
                "sec_num": "3"
            },
            {
                "text": "The Alexina model is based on a two-level representation, detailed below, that separates the description of a lexicon from its use:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Alexina framework",
                "sec_num": "3"
            },
            {
                "text": "\u2022 The intensional lexicon factorizes the lexical information by associating each lemma with a morphological class and deep syntactic information; it is used for lexical resource development",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Alexina framework",
                "sec_num": "3"
            },
            {
                "text": "\u2022 The extensional lexicon, which is generated automatically by compiling the intensional lexicon, associates each inflected form with a detailed structure that represents all its morphological and syntactic information; it is directly used by NLP tools such as parsers.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Alexina framework",
                "sec_num": "3"
            },
            {
                "text": "The first task achieved by the compilation process, which turns an intensional lexicon (an .ilex file) into an extensional lexicon (a .lex file), is to inflect lemmas according to their morphological class. Morphological classes are defined in a formalized morphological description described in [11, 12] . In case a lemma inflects in a very specific way, and/or if a lemma has additional inflected forms apart from those generated by its morphological class, these forms are \"manually\" listed in an additional file (the corresponding .mf file).",
                "cite_spans": [
                    {
                        "start": 296,
                        "end": 300,
                        "text": "[11,",
                        "ref_id": "BIBREF10"
                    },
                    {
                        "start": 301,
                        "end": 304,
                        "text": "12]",
                        "ref_id": "BIBREF11"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Alexina framework",
                "sec_num": "3"
            },
            {
                "text": "As sketched above, the compilation process also maps deep syntactic information into surface syntactic information. Deep syntactic information (deep subcategorization frames and other syntactic information) is common to all redistributions, whereas each redistribution corresponds to different surface syntactic information, and therefore to different extensional entries.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Alexina framework",
                "sec_num": "3"
            },
            {
                "text": "Each entry in the intensional lexicon is usually defined by a lemma and a POS. Nevertheless, it is possible to find several entries with same lemma and POS but differing in the morphological and syntactic information. This allows to split one lemma into different semantic meanings which implies different syntactic constructions.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "An intensional entry details the following information:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "\u2022 a morphological class, which defines the patterns that build all inflected forms of the lemma [12] ;",
                "cite_spans": [
                    {
                        "start": 96,
                        "end": 100,
                        "text": "[12]",
                        "ref_id": "BIBREF11"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "\u2022 a category (or part-of-speech, often written POS), that is taken from the chosen tagset -Leffe uses the Multext (Parole) tagset; categories can be divided in two types: open (productive) categories (adjectives, adverbs, verbs, nouns) and closed (grammatical) categories;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "\u2022 a (deep-syntax) subcategorization frame, that explicits how the lemma might be used in valid syntactic constructions: it lists the canonical syntactic functions of the lemma's possible arguments, 5 and the possible realizations of each of these functions; 6, 7 \u2022 additional syntactic information (control, raising, attributes. . . );",
                "cite_spans": [
                    {
                        "start": 198,
                        "end": 199,
                        "text": "5",
                        "ref_id": "BIBREF4"
                    },
                    {
                        "start": 258,
                        "end": 260,
                        "text": "6,",
                        "ref_id": "BIBREF5"
                    },
                    {
                        "start": 261,
                        "end": 262,
                        "text": "7",
                        "ref_id": "BIBREF6"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "\u2022 possible (re)distributions, that define how the deepsyntax subcategorization frame is to be transformed so as to build extensional surface-syntax subcategorization frames (usual (re)distributions are %actif, %passif, %se moyen). 8",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "For example, here is the intensional (slightly simplified 9 for clarity reasons) entry in the Leffe for the Spanish lemma 5 The Leffe uses the following syntactic functions: Suj for subjects,",
                "cite_spans": [
                    {
                        "start": 122,
                        "end": 123,
                        "text": "5",
                        "ref_id": "BIBREF4"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "Obj for direct objects that can be cliticized into an accusative clitic pronoun, Obja for indirect objects introduced by the preposition a, Loc and Dloc for locative and delocative arguments, Att for (subject, object or a-object) attributes, and Obl (and Obl2) for other (noncliticizable) arguments. More detailed defining criteria for their French counterparts in the Lefff can be found in [14] . 6 Possible realizations are threefold:",
                "cite_spans": [
                    {
                        "start": 391,
                        "end": 395,
                        "text": "[14]",
                        "ref_id": "BIBREF13"
                    },
                    {
                        "start": 398,
                        "end": 399,
                        "text": "6",
                        "ref_id": "BIBREF5"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "-clitic pronouns: cln (nominative clitic), cla (accusative clitic), cld (dative clitic), serefl (reflexive se); -direct phrases: sn (noun phrase), sa (adjectival phrase), sinf (infinitive clause), scompl (completive clause), qcompl (interrogative clause); -prepositional phrases: a direct phrase introduced by a preposition (e.g., a-sn) 7 Note that realizations have the same (French) names as their French counterparts in the Lefff. This should change in the next version of the Leffe. 8 As for realizations, redistributions have the same (French) names as their French counterparts in the Lefff. This should also change in the next version of the Leffe. 9 In particular, additional syntactic features such as control information are not shown. It describes a transitive entry with the following informations:",
                "cite_spans": [
                    {
                        "start": 337,
                        "end": 338,
                        "text": "7",
                        "ref_id": "BIBREF6"
                    },
                    {
                        "start": 487,
                        "end": 488,
                        "text": "8",
                        "ref_id": "BIBREF7"
                    },
                    {
                        "start": 656,
                        "end": 657,
                        "text": "9",
                        "ref_id": "BIBREF8"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "\u2022 its morphological class is V4, the class of the firstconjugation verbs (ending -ar) whose stem changes for present subjunctive(c changes to que);",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "\u2022 its semantic predicate can be represented by the Lemma as is, i.e., diagnosticar;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "\u2022 its category is verb (v);",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "\u2022 it has two arguments canonically realized by the syntactic functions Suj (subject) and Obj (direct object). Each syntactic function is associated with a list of possible realizations. ;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "\u2022 it allows for two different redistributions: active (%actif) and passive (%passif).",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Intensional format",
                "sec_num": "3.1"
            },
            {
                "text": "The compilation process builds one extensional entry for each inflected form and each compatible redistribution, by applying formalized definitions of these redistributions. For example, the only inflected forms of diagnosticar that are compatible with the passive redistribution are the past participle forms. The (simplified) extensional passive entry for diagnosticados (diagnosed) is the following (MP00SM is the morphological tag for past participle masculine plural forms):",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Extensional format",
                "sec_num": "3.2"
            },
            {
                "text": "diagnosticados v [pred='diagnosticar1<arg1:Suj:cln|scompl|sn, arg0:Obl2:(por-sn)>',@passive,@pers,@MP00PM]; %passif",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Extensional format",
                "sec_num": "3.2"
            },
            {
                "text": "As can be seen the original direct object (Obj) has been transformed into the passive Subject and an optional Agent (Obl2) realized by a noun phrase preceded by a preposition (por-sn) was added.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The Extensional format",
                "sec_num": "3.2"
            },
            {
                "text": "In order to create a first version of the Leffe, the first step was of course to reuse available Spanish lexical resources.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "Reusing available linguistic resources is a handy way to start developing new ones. However, it requires to interpret all input resources even though their lexical models are partially incompatible, convert them into a common model and format, and finally merge the converted lexicons. None of these three steps is trivial.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "Indeed, available resources might describe a given language from different points of view and/or using different linguistic criteria. This can be used to acquire information covering different aspects of a language. When considering whether a resource was worth using or not for this task, we payed more attention to quality or richness than coverage. After all, combining several resources shall lead to a good coverage that will generally be wider than the largest of them. Thus, we ensured as more important the reliability of the information put into the new resource. The application of the technique described in section 5 allowed us later to regain more coverage.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "As stated in the introduction, several resources are available for Spanish, but none of them fulfilled our requirements:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "\u2022 wide coverage, good precision and satisfying richness,",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "\u2022 complete separation between lexical and grammatical information, i.e., independence from the grammatical formalism it is going to be used with,",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "\u2022 clear and compact format easily readable by humans,",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "\u2022 free availability in terms of access, modification and distribution;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "\u2022 easily linkable with resources in other languages.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "Nevertheless, in order to create a first version of the Leffe, we reused the following resources:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "Multext is an international project [6] which aims, among other things, at developing standards and specifications for the encoding and processing of linguistic corpora.",
                "cite_spans": [
                    {
                        "start": 36,
                        "end": 39,
                        "text": "[6]",
                        "ref_id": "BIBREF5"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "The USC lexicon is a large morphological Spanish lexicon [16] , created for PoS tagging tasks in the research group Gram\u00e1tica del Espa\u00f1ol of the University of Santiago de Compostela (Spain).",
                "cite_spans": [
                    {
                        "start": 57,
                        "end": 61,
                        "text": "[16]",
                        "ref_id": "BIBREF15"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "ADESSE is a database of Spanish verbs developed at the University of Vigo (Spain) [3] with syntactic and some semantic information. It is a high quality work which includes subcategorization frames for more than 4,000 verbs. However, it is restricted to verbs and does not include morphological information;",
                "cite_spans": [
                    {
                        "start": 82,
                        "end": 85,
                        "text": "[3]",
                        "ref_id": "BIBREF2"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "The Spanish Resource Grammar (SRG) is an opensource multi-purpose large-coverage and precise grammar for Spanish [7] grounded in the theoretical framework of Head-driven Phrase Structure Grammar (HPSG). It includes a lexicon describing syntactic information for Spanish in a well organized hierarchy of syntactic classes. However, it is not easily readable, and specific to the HPSG formalism.",
                "cite_spans": [
                    {
                        "start": 113,
                        "end": 116,
                        "text": "[7]",
                        "ref_id": "BIBREF6"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "In order to merge these resources, we followed a process described in details in [9] . We briefly remind it here.",
                "cite_spans": [
                    {
                        "start": 81,
                        "end": 84,
                        "text": "[9]",
                        "ref_id": "BIBREF8"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "As mentioned in Section 2, Multext and USC lexicons only include morphological information, whereas the SRG Lexicon and ADESSE include syntactic information. Therefore, we proceeded in the following way:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "1. we built a morphological baseline lexicon by converting the Multext lexicon into the Alexina format and added some Alexina-specific entries (prefixes, suffixes, named entities, punctuation signs);",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "2. we converted the USC Lexicon into the Alexina format and merged it with the baseline lexicon extracted from Multext, so as to obtain the morphological base of the Leffe;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "3. we converted the syntactic information from ADESSE and the SRG lexicon into the Alexina format;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "4. we merged the morphological Leffe from step 2 and both verbal syntactic lexicons built during step 3; the result was the Leffe beta.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "The final result is a morphological and syntactic lexicon with an important coverage in terms of morphological information but a more restricted one in terms of syntactic information. Indeed, for morphological entries 10 for which no syntactic information could be found, we added default syntactic features corresponding to the most common ones among entries with the same PoS. For example, all verbal lemmas that were not covered by ADESSE or SRG received the following subcategorization frame: <Suj:sn|cln,Obj:(sn|cla)> (transitive verb with optional direct object). However, the application of semiautomatic techniques to extend and correct a lexicon, as described in [10] , should help us fixing this aspect.",
                "cite_spans": [
                    {
                        "start": 218,
                        "end": 220,
                        "text": "10",
                        "ref_id": "BIBREF9"
                    },
                    {
                        "start": 672,
                        "end": 676,
                        "text": "[10]",
                        "ref_id": "BIBREF9"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Reusing other lexical resources",
                "sec_num": "4"
            },
            {
                "text": "The next step after obtaining a first version of the Leffe was to continue upgrading it by adding missing entries. Usually, this task is manually performed and thus, is a time-costly process. We now present a simple but effective semi-automatic technique which greatly eases the process by identifying possible missing entries. We distinguish two types of missing entries:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Tagger-based identification of missing entries",
                "sec_num": "5"
            },
            {
                "text": "1. totally non referenced forms, 2. missing homonyms of forms referenced in the lexicon, i.e., forms non associated to a different Part-of-Speech (PoS).",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Tagger-based identification of missing entries",
                "sec_num": "5"
            },
            {
                "text": "In order to detect missing entries, a PoS tagger [5, 8] might be used to discover new PoS tags thanks to its ability to guess PoS tags for unknown words. The tagger we use is trained with a Spanish training corpus of approx. 500,000 words extracted from the Ancora 11 corpora and Leffe as an external lexicon. According to the kind of missing entries we are trying to identify, the tagger is used in two different ways.",
                "cite_spans": [
                    {
                        "start": 49,
                        "end": 52,
                        "text": "[5,",
                        "ref_id": "BIBREF4"
                    },
                    {
                        "start": 53,
                        "end": 55,
                        "text": "8]",
                        "ref_id": "BIBREF7"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Tagger-based identification of missing entries",
                "sec_num": "5"
            },
            {
                "text": "When looking for non referenced forms, we simply rely on the tagger's ability to guess tags for unknown words.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Tagger-based identification of missing entries",
                "sec_num": "5"
            },
            {
                "text": "When looking for missing homonyms, we allow the tagger to assign new tags to known forms that are different from those included in its lexicon by forcing it to consider known forms as unknown. Indeed, the default strategy for most taggers when facing a form included in their internal lexicon is to consider as candidate tags only the ones associated there. Thus, when facing a missing homonym of a form, the tagger will never consider as a potential candidate the correct missing tag. In order to obtain such behavior, we simply bypass the internal lexicon. Thus, the tagger guesses new tags basing itself on the morphology of the form and its local context.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Tagger-based identification of missing entries",
                "sec_num": "5"
            },
            {
                "text": "Obviously, such a process introduces ambiguity on purpose. In order to keep it beyond limits, we only force one form in a sentence at a time to be considered as unknown. Thus, to guess PoS tags for all words in the sentence, the sentence is entirely tagged several times.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Tagger-based identification of missing entries",
                "sec_num": "5"
            },
            {
                "text": "Since forms belonging to closed categories 12 are generally well described (and their homonyms correctly included too), only forms belonging to open categories 13 are forced as unknown.",
                "cite_spans": [
                    {
                        "start": 160,
                        "end": 162,
                        "text": "13",
                        "ref_id": "BIBREF12"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Tagger-based identification of missing entries",
                "sec_num": "5"
            },
            {
                "text": "Of course, taggers make mistakes, particularly when dealing with unknown (forced or not) forms. A well-known situation for a tagger is to consider an unknown proper noun as a common noun. However, the scope of the process span an entire corpora and not only one sentence. Thus, considering a large amount of text allows us to compute a statistical ranking of the suspected missing forms which balance the false positives produced by tagging errors. This ranking takes into account the precision rate prec t for a tag t, as evaluated relatively to the training corpus, and n wt and the number of occurrences of the form w tagged as t. More precisely, we assign to each couple form w and tag t a score S sc (w, t) defined as follows:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Tagger-based identification of missing entries",
                "sec_num": "5"
            },
            {
                "text": "EQUATION",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [
                    {
                        "start": 0,
                        "end": 8,
                        "text": "EQUATION",
                        "ref_id": "EQREF",
                        "raw_str": "S sc (w, t) = prec t \u2022 log(n w/t )",
                        "eq_num": "(1)"
                    }
                ],
                "section": "Tagger-based identification of missing entries",
                "sec_num": "5"
            },
            {
                "text": "Thanks to this ranking, we are able to generate an ordered list of candidate pairs (form,PoS) which minimizes the appearance of false positives. As we will see in section 6, this list was good enough to be manually reviewed in a short amount of time.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Tagger-based identification of missing entries",
                "sec_num": "5"
            },
            {
                "text": "In order to evaluate the quality of Leffe, currently in beta version, we performed the following tests: on the one hand, we compared Leffe with other known Spanish lexicons in terms of coverage; on the other hand, we measured the improvement achieved on the baseline lexicon after adding the information extracted from all other sources.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "Regarding coverage, the Leffe contains more than 165,000 unique (lemma,PoS) pairs, which correspond to approx. 1,590,000 extensional entries that associate a form with both morphological and syntactic information (approx. 680,000 unique (form,PoS) pairs). We computed the following properties for the other lexicons:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "\u2022 SRG: 76,000 unique (lemma,PoS) pairs 14 (53.9% fewer than Leffe), some of them associated with syntactic information;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "\u2022 Multext: 510,710 unique (form,PoS) pairs 15 (24.9% fewer than Leffe), and no syntactic information is provided;",
                "cite_spans": [
                    {
                        "start": 43,
                        "end": 45,
                        "text": "15",
                        "ref_id": "BIBREF14"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "\u2022 Spanish gilcUB-M Dictionary: 70,000 lemmas 15 (57.6% fewer than Leffe), and no syntactic information is provided;",
                "cite_spans": [
                    {
                        "start": 45,
                        "end": 47,
                        "text": "15",
                        "ref_id": "BIBREF14"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "\u2022 USC Lexicon: 490,000 unique (form,PoS) pairs (27.95% fewer than Leffe), and no syntactic information is provided.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "We also tested the morphological coverage of our lexicon in the context of a real application: a morphological pre-processor [4] developed by the COLE 16 and LYS 17 groups. We first performed a test with our baseline lexicon and another one with the Leffe.",
                "cite_spans": [
                    {
                        "start": 125,
                        "end": 128,
                        "text": "[4]",
                        "ref_id": "BIBREF3"
                    },
                    {
                        "start": 151,
                        "end": 153,
                        "text": "16",
                        "ref_id": "BIBREF15"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "The corpus of raw text we used as input for these tests was obtained from Wikipedia Sources 18 . It includes more than 4,322,000 words after clearing Wikipedia references and foreign expressions. The evaluation took into account how many words were not tagged by the pre-processor and thus remained unknown. It is worth noting that unknown words are the main cause of PoS-tagging errors. Such problems can be tackled by relying on (very) large coverage lexicons.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "As can be observed in Table 1 , the process has noticeable benefits. The Leffe has beaten other large lexicons in the morphological preprocessing task. Even if the difference is slight, this demonstrates the advantage of merging existing resources to create an enhanced one.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 22,
                        "end": 29,
                        "text": "Table 1",
                        "ref_id": "TABREF2"
                    }
                ],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "In order to measure the syntactic coverage of the lexicons at all stages of the merging process, we used the notion of expanded intensional entry [9] which is just a defactorized Alexina subcategorization frame. Thus, each expanded intensional entry describes one fully-specified syntactic behaviour.",
                "cite_spans": [
                    {
                        "start": 146,
                        "end": 149,
                        "text": "[9]",
                        "ref_id": "BIBREF8"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "The expanded intensional lexicon acquired from SRG contains 42,689 unique entries, i.e., fully-specified subcategorization frames, while the one obtained from the ADESSE contains 39,040 entries. After merging these lexicons, the number of such unique entries jumps to 66,028. Finally, the Leffe, which associates default syntactic information with all verbs not covered by the result of this merging, contains 91,507 unique expanded entries. After factorization, the Leffe contains 16,311 verbal entries.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "Once the first version of Leffe was built, we used the technique described in section 5 to upgrade its coverage. We used a corpus built from a subset of the Spanish part of the Europarl 19 containing approx. 6 million words. The only restriction applied to this corpus was to avoid the inclusion of sentences containing foreign words, since they would lead to false positives.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "A ranking of suspected missing pairs (form, tag) was obtained. The quality of this list was not exceptional since it included many false positives, but, thanks to this list, we did include in the Leffe at a very small cost (it was manually done by one person in two days) nothing less than 1,800 lemmas. We must note that the original coverage of the Leffe was very high and thus it is reasonable to think that the proportion of false positive would have been reduced when dealing with lexicons with a smaller coverage. Table 2 shows the number of lemmas added to the Leffe classified by categories. The great majority were proper nouns, since they were very incomplete in Leffe up to this point. The approx. 1,800 intensional entries added to the Leffe correspond to more than 3,700 inflected forms in the extensional lexicon. For example, we added the verbs it abstraer (to abstract) and documentar (to document), the adjective francoespa\u00f1ol (Franco-Spanish), the common noun biocarburante (biofuel), the adverb precipitadamente (hastily) and the proper noun Niza (Nice).",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 520,
                        "end": 527,
                        "text": "Table 2",
                        "ref_id": "TABREF3"
                    }
                ],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "Apart from the correct entries, the list allowed us to detect some systematic deficiencies, such as 16 http://www.grupocole.org, April 2009 17 http://www.grupolys.org, April 2009 18 http://download.wikimedia.org, January 2009 19 A parallel corpus from the European Parliament proceedings diminutives/augmentatives and adverbs ending in -mente. In a near future, they will be automatically generated after updating the morphological rules used to obtain the extensional lexicon from the intensional one (see sect.3).",
                "cite_spans": [
                    {
                        "start": 100,
                        "end": 102,
                        "text": "16",
                        "ref_id": "BIBREF15"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preliminary Evaluation",
                "sec_num": "6"
            },
            {
                "text": "In this work we have presented a morphological and syntactic wide-coverage lexicon for Spanish built by taking advantage of existing lexical resources in Spanish and French. Nowadays, for many languages, several scattered linguistic resources exist, but usually none of them is satisfying in terms of coverage, richness or precision. Nevertheless, the amount of work invested in their development should not be ignored. In fact, we believe reusing already formalized knowledge is a handy and productive way to build and/or upgrade other linguistic resources and it will be the usual strategy in the near future.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusion",
                "sec_num": "7"
            },
            {
                "text": "We also described a tagger-based approach to detect missing entries in a lexicon. Even when applied to a quite exhaustive lexicon, such as Leffe, this simple approach has allowed us to add more than 3,700 lexical forms in a very short amount of time. 20 The resulting lexicon, the Leffe, is currently in beta version and will soon be distributed under a LGPL-LR license 21 . Although it is still far from perfect, we have shown that the Leffe has already overtaken other wellknown Spanish lexicons in terms of morphological and syntactic coverage.",
                "cite_spans": [
                    {
                        "start": 251,
                        "end": 253,
                        "text": "20",
                        "ref_id": null
                    },
                    {
                        "start": 370,
                        "end": 372,
                        "text": "21",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusion",
                "sec_num": "7"
            },
            {
                "text": "In the near future, we plan to further evaluate the Leffe by comparing the coverage and precision of different deep parsers that rely on the same grammar but using different morphological and syntactic lexicons such as the Leffe. ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusion",
                "sec_num": "7"
            },
            {
                "text": "A co-official language in north-west Spain.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "Lexical Markup Framework, the ISO/TC37 standard for NLP lexicons.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "The condition to add an entry to the Leffe was to acquire at least its morphological information. 11 http://clic.ub.edu/ancora/index.php, July 2009.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "Such as prepositions, pronouns and determiners.13 Adverbs, common nouns, proper nouns, verbs, adjectives.14 As provided by Freeling (http://garraf.epsevg.upc.es/ freeling/) in a version from April 2008.15 According to ELRA webpage http://catalog.elra.info, April 2009.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "The use of a tagger with a lower error rate, in particular on words unknown to the tagger's training corpus, should allow to scale up the efficiency of this approach. Such a tagger is under development in the team of one of the authors.21 As explained in this paper, the construction of the Leffe involved the Spanish morphological lexicon developed within the Multext project, which is freely available for research. The Leffe beta is the result of the research work described here. It merges lexical information coming from various resources, most of them with a coverage that is larger than the Spanish Multext lexicon. For this reason, we consider it appropriate to publish the Leffe beta under the LGPL-LR.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [
            {
                "text": "This work was supported in part by the Ministerio de Educaci\u00f3n y Ciencia of Spain and FEDER (HUM2007-66607-C04-02), the Xunta de Galicia (INCITE08PXIB302179PR, INCITE08E1R104022ES, PGIDIT07SIN005206PR) and the \"Galician Network for Language Processing and Information Retrieval\" (2006)(2007)(2008)(2009). We would like also to thank the group Gram\u00e1tica del Espa\u00f1ol from USC, and especially to Guillermo Rojo, M. Paula Santalla and Susana Sotelo, for granting us access to their lexicon.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Acknowledgements",
                "sec_num": null
            }
        ],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "Constructions pronominales dans dicovalence et le lexique-grammaireint\u00e9gration dans le Lefff",
                "authors": [
                    {
                        "first": "L",
                        "middle": [],
                        "last": "Danlos",
                        "suffix": ""
                    },
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Sagot",
                        "suffix": ""
                    }
                ],
                "year": 2008,
                "venue": "Proceedings of the 27th Lexicon-Grammar Conference",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "L. Danlos and B. Sagot. Constructions pronom- inales dans dicovalence et le lexique-grammaire - int\u00e9gration dans le Lefff. In Proceedings of the 27th Lexicon-Grammar Conference, L'Aquila, Italy, 2008.",
                "links": null
            },
            "BIBREF1": {
                "ref_id": "b1",
                "title": "Lexical Markup Framework (LMF)",
                "authors": [
                    {
                        "first": "G",
                        "middle": [],
                        "last": "Francopoulo",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "George",
                        "suffix": ""
                    },
                    {
                        "first": "N",
                        "middle": [],
                        "last": "Calzolari",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Monachini",
                        "suffix": ""
                    },
                    {
                        "first": "N",
                        "middle": [],
                        "last": "Bel",
                        "suffix": ""
                    },
                    {
                        "first": "C",
                        "middle": [],
                        "last": "Pet",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Soria",
                        "suffix": ""
                    }
                ],
                "year": 2006,
                "venue": "Proceedings of LREC'06",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "G. Francopoulo, M. George, N. Calzolari, M. Mona- chini, N. Bel, mandy Pet, and C. Soria. Lexi- cal Markup Framework (LMF). In Proceedings of LREC'06, Genoa, Italy, 2006.",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "Verbs, semantic classes and semantic roles in the ADESSE project",
                "authors": [
                    {
                        "first": "J",
                        "middle": [
                            "M"
                        ],
                        "last": "Garc\u00eda-Miguel",
                        "suffix": ""
                    },
                    {
                        "first": "F",
                        "middle": [
                            "J"
                        ],
                        "last": "Albertuz",
                        "suffix": ""
                    }
                ],
                "year": 2005,
                "venue": "Proceedings of the Interdisciplinary Workshop on the Identification and Representation of Verb Features and Verb Classes",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "J. M. Garc\u00eda-Miguel and F. J. Albertuz. Verbs, semantic classes and semantic roles in the ADESSE project. In Proceedings of the Interdisciplinary Workshop on the Identification and Representation of Verb Features and Verb Classes, Saarbr\u00fccken, Germany, 2005.",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "Formal methods of tokenization for part-of-speech tagging",
                "authors": [
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Gra\u00f1a",
                        "suffix": ""
                    },
                    {
                        "first": "F",
                        "middle": [
                            "M"
                        ],
                        "last": "Barcala",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Vilares",
                        "suffix": ""
                    }
                ],
                "year": 2002,
                "venue": "Computational Linguistics and Intelligent Text Processing",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "J. Gra\u00f1a, F. M. Barcala, and J. Vilares. Formal meth- ods of tokenization for part-of-speech tagging. Com- putational Linguistics and Intelligent Text Process- ing, Lecture Notes in Computer Science, 2002.",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "T\u00e9cnicas de An\u00e1lisis Sint\u00e1ctico Robusto para la Etiquetaci\u00f3n del Lenguaje Natural ( robust syntactic analysis methods for natural language tagging). Doctoral thesis",
                "authors": [
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Gra\u00f1a",
                        "suffix": ""
                    }
                ],
                "year": 2000,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "J. Gra\u00f1a. T\u00e9cnicas de An\u00e1lisis Sint\u00e1ctico Robusto para la Etiquetaci\u00f3n del Lenguaje Natural ( robust syntactic analysis methods for natural language tagging). Doctoral thesis, Universidad de La Coru\u00f1a, Spain, 2000.",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "Multext: Multilingual text tools and corpora",
                "authors": [
                    {
                        "first": "N",
                        "middle": [],
                        "last": "Ide",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "V\u00e9ronis",
                        "suffix": ""
                    }
                ],
                "year": 1994,
                "venue": "Proceedings of COLING'94",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "N. Ide and J. V\u00e9ronis. Multext: Multilingual text tools and corpora. In Proceedings of COLING'94, Kyoto, Japan, 1994.",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "title": "An opensource lexicon for Spanish",
                "authors": [
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Marimon",
                        "suffix": ""
                    },
                    {
                        "first": "N",
                        "middle": [],
                        "last": "Seghezzi",
                        "suffix": ""
                    },
                    {
                        "first": "N",
                        "middle": [],
                        "last": "Bel",
                        "suffix": ""
                    }
                ],
                "year": 2007,
                "venue": "Sociedad Espa\u00f1ola para el Procesamiento del Lenguaje Natural",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "M. Marimon, N. Seghezzi, and N. Bel. An open- source lexicon for Spanish. In Sociedad Espa\u00f1ola para el Procesamiento del Lenguaje Natural, n. 39, 2007.",
                "links": null
            },
            "BIBREF7": {
                "ref_id": "b7",
                "title": "Practical application of one-pass viterbi algorithm in tokenization and pos tagging",
                "authors": [
                    {
                        "first": "M",
                        "middle": [
                            "A"
                        ],
                        "last": "Molinero",
                        "suffix": ""
                    },
                    {
                        "first": "F",
                        "middle": [
                            "M"
                        ],
                        "last": "Barcala",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Otero",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Gra\u00f1a",
                        "suffix": ""
                    }
                ],
                "year": 2007,
                "venue": "Recent Advances in Natural Language Processing (RANLP). Proceedings",
                "volume": "",
                "issue": "",
                "pages": "35--40",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "M. A. Molinero, F. M. Barcala, J. Otero, and J. Gra\u00f1a. Practical application of one-pass viterbi algorithm in tokenization and pos tagging. Recent Advances in Natural Language Processing (RANLP). Proceedings, pp. 35-40, 2007.",
                "links": null
            },
            "BIBREF8": {
                "ref_id": "b8",
                "title": "Building a morphological and syntactic lexicon by merging various linguistic resources",
                "authors": [
                    {
                        "first": "M",
                        "middle": [
                            "A"
                        ],
                        "last": "Molinero",
                        "suffix": ""
                    },
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Sagot",
                        "suffix": ""
                    },
                    {
                        "first": "L",
                        "middle": [],
                        "last": "Nicolas",
                        "suffix": ""
                    }
                ],
                "year": 2009,
                "venue": "Proceedings of NODALIDA'09",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "M. A. Molinero, B. Sagot, and L. Nicolas. Building a morphological and syntactic lexicon by merging various linguistic resources. In Proceedings of NODALIDA'09, Odense, Denmark, 2009.",
                "links": null
            },
            "BIBREF9": {
                "ref_id": "b9",
                "title": "Computer aided correction and extension of a syntactic wide-coverage lexicon",
                "authors": [
                    {
                        "first": "L",
                        "middle": [],
                        "last": "Nicolas",
                        "suffix": ""
                    },
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Sagot",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [
                            "A"
                        ],
                        "last": "Molinero",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Farr\u00e9",
                        "suffix": ""
                    },
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Villemonte De La Clergerie",
                        "suffix": ""
                    }
                ],
                "year": 2008,
                "venue": "Proceedings of COLING'08",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "L. Nicolas, B. Sagot, M. A. Molinero, J. Farr\u00e9, and E. Villemonte de La Clergerie. Computer aided correction and extension of a syntactic wide-coverage lexicon. In Proceedings of COLING'08, Manchester, United Kingdom, 2008.",
                "links": null
            },
            "BIBREF10": {
                "ref_id": "b10",
                "title": "Automatic acquisition of a Slovak lexicon from a raw corpus",
                "authors": [
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Sagot",
                        "suffix": ""
                    }
                ],
                "year": 2005,
                "venue": "Proceedings of TSD'05",
                "volume": "3658",
                "issue": "",
                "pages": "156--163",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "B. Sagot. Automatic acquisition of a Slovak lexicon from a raw corpus. In Lecture Notes in Artificial Intelligence 3658 ( c Springer-Verlag), Proceedings of TSD'05, pages 156-163, Karlovy Vary, Czech Republic, 2005.",
                "links": null
            },
            "BIBREF11": {
                "ref_id": "b11",
                "title": "Building a morphosyntactic lexicon and a pre-syntactic processing chain for Polish",
                "authors": [
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Sagot",
                        "suffix": ""
                    }
                ],
                "year": 2007,
                "venue": "Proceedings of the 3rd Language & Technology Conference (LTC'05)",
                "volume": "",
                "issue": "",
                "pages": "423--427",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "B. Sagot. Building a morphosyntactic lexicon and a pre-syntactic processing chain for Polish. In Proceedings of the 3rd Language & Technology Conference (LTC'05), pages 423-427, Pozna\u0144, Poland, 2007.",
                "links": null
            },
            "BIBREF12": {
                "ref_id": "b12",
                "title": "The Lefff 2 syntactic lexicon for French: architecture, acquisition, use",
                "authors": [
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Sagot",
                        "suffix": ""
                    },
                    {
                        "first": "L",
                        "middle": [],
                        "last": "Cl\u00e9ment",
                        "suffix": ""
                    },
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Villemonte De La Clergerie",
                        "suffix": ""
                    },
                    {
                        "first": "P",
                        "middle": [],
                        "last": "Boullier",
                        "suffix": ""
                    }
                ],
                "year": 2006,
                "venue": "Proceedings of LREC'06",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "B. Sagot, L. Cl\u00e9ment, E. Villemonte de La Clergerie, and P. Boullier. The Lefff 2 syntactic lexicon for French: architecture, acquisition, use. In Proceedings of LREC'06, 2006.",
                "links": null
            },
            "BIBREF13": {
                "ref_id": "b13",
                "title": "Am\u00e9liorer un lexique syntaxique\u00e0 l'aide des tables du lexique-grammaire -Constructions impersonnelles",
                "authors": [
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Sagot",
                        "suffix": ""
                    },
                    {
                        "first": "L",
                        "middle": [],
                        "last": "Danlos",
                        "suffix": ""
                    }
                ],
                "year": 2007,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "B. Sagot and L. Danlos. Am\u00e9liorer un lexique syntaxique\u00e0 l'aide des tables du lexique-grammaire -Constructions impersonnelles. Cahiers du Cental, 2007.",
                "links": null
            },
            "BIBREF14": {
                "ref_id": "b14",
                "title": "M\u00e9thodologie lexicographique de constitution d'un lexique syntaxique de r\u00e9f\u00e9rence pour le fran\u00e7ais",
                "authors": [
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Sagot",
                        "suffix": ""
                    },
                    {
                        "first": "L",
                        "middle": [],
                        "last": "Danlos",
                        "suffix": ""
                    }
                ],
                "year": 2008,
                "venue": "Proceedings of the workshop \"Lexicographie et informatique : bilan et perspectives",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "B. Sagot and L. Danlos. M\u00e9thodologie lexi- cographique de constitution d'un lexique syntaxique de r\u00e9f\u00e9rence pour le fran\u00e7ais. In Proceedings of the workshop \"Lexicographie et informatique : bilan et perspectives\", Nancy, France, 2008.",
                "links": null
            },
            "BIBREF15": {
                "ref_id": "b15",
                "title": "Avalon, una gram\u00e1tica formal basada en corpus",
                "authors": [
                    {
                        "first": "C",
                        "middle": [],
                        "last": "\u00c1lvarez",
                        "suffix": ""
                    },
                    {
                        "first": "P",
                        "middle": [],
                        "last": "Alvari\u00f1o",
                        "suffix": ""
                    },
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Gil",
                        "suffix": ""
                    },
                    {
                        "first": "T",
                        "middle": [],
                        "last": "Romero",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [
                            "P"
                        ],
                        "last": "Santalla",
                        "suffix": ""
                    },
                    {
                        "first": "S",
                        "middle": [],
                        "last": "Sotelo",
                        "suffix": ""
                    }
                ],
                "year": 1998,
                "venue": "Procesamiento del Lenguaje Natural (Actas del XIV Congreso de la SEPLN)",
                "volume": "",
                "issue": "",
                "pages": "132--139",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "C.\u00c1lvarez, P. Alvari\u00f1o, A. Gil, T. Romero, M. P. Santalla, and S. Sotelo. Avalon, una gram\u00e1tica formal basada en corpus. In Procesamiento del Lenguaje Natural (Actas del XIV Congreso de la SEPLN), pages 132-139, Alicante, Spain, 1998.",
                "links": null
            }
        },
        "ref_entries": {
            "TABREF0": {
                "type_str": "table",
                "num": null,
                "html": null,
                "content": "<table><tr><td>diagnosticar1</td></tr><tr><td>V4</td></tr><tr><td>Lemma;v;</td></tr><tr><td>&lt;arg0:Suj:cln|scompl|sinf|sn,</td></tr><tr><td>arg1:Obj:(cla|scompl|sn)&gt;;</td></tr><tr><td>%actif,%passif</td></tr></table>",
                "text": "diagnosticar 1 , i.e., diagnosticar in the sense of the English to diagnose."
            },
            "TABREF2": {
                "type_str": "table",
                "num": null,
                "html": null,
                "content": "<table><tr><td/><td colspan=\"2\">LEMMAS (INTENSIONAL) INFLECTED FORMS (EXTENSIONAL)</td></tr><tr><td>Adjectives</td><td>88</td><td>298</td></tr><tr><td>Adverbs</td><td>54</td><td>54</td></tr><tr><td>Verbs</td><td>26</td><td>1,693</td></tr><tr><td>Common nouns</td><td>117</td><td>231</td></tr><tr><td>Proper nouns</td><td>1,518</td><td>1,518</td></tr><tr><td>Total</td><td>1,803</td><td>3,740</td></tr></table>",
                "text": "Results obtained by applying the morphological preprocessor with different lexicons."
            },
            "TABREF3": {
                "type_str": "table",
                "num": null,
                "html": null,
                "content": "<table/>",
                "text": "Lemmas acquired using the tagger-based technique."
            }
        }
    }
}