the installation of an electronic dictionary of French computer science terms (compound words), with an aim of analysing, automatically indexing texts.
Proceedings of the 10th WSEAS International Conference on COMPUTERS, Vouliagmeni, Athens, Greece, July 13-15, 2006 (pp458-462)
An Electronic Dictionary of Computer Science Terminology Farida AOUGHLIS (1), Elisabeth METAIS (2) (1) Université Mouloud Mammeri, 15000 - Tizi Ouzou, Algérie (2) CEDRIC / CNAM: 292 rue Saint Martin - 75003 Paris, France
Abstract: Automatic text analysis systems can lexically recognize a word only if it already exists in the electronic dictionary. The same thing is true for the NOOJ system analysis programs. One understands here by electronic dictionaries the lexical databases where all information is explicit because they are intended for computer programs use. These bases aim at the modelling of the language, which distinguishes them from the electronic lexicons created for particular applications needs. For technical languages or of speciality, work remains to be made to build dictionaries. Our work applies to the NOOJ system, with for immediate objective the installation of an electronic dictionary of French computer science terms (compound words), with an aim of analysing, automatically indexing texts. The linguistic aspects of the terminology are retained. Key-Words: Terminology, terminology extraction, electronic dictionary, compound words, NOOJ.
1 Introduction The exchanges internationalization, the always increasing production of documents under electronic format, the stakes related Internet and networks Intranet development create needs increasingly significant and varied in terminological resources. They can be [5] simple lists of terms that are more or less structured (structured indices, thesaurus, lexical networks) used by automatic indexing systems or for information retrieval or more documented terminological reference frames. [5] The construction of a terminology [2] depends on the application in which one wants to use it. The selected terms and their degree of description are different according to whether one wants to build a reference terminology for a drafting assistance system or a lexical network to improve
information retrieval on Internet. The computer science compound words dictionary INFO_COMP that we build will allow the analysis, the automatic computer science texts indexing, using NOOJ system developed by Silberztein M. For the construction of our dictionary for NOOJ, the extraction of terminology is manual and we also extract terms automatically with NOOJ. This paper presents INFO_COMP, the electronic computer science dictionary of terminology for NOOJ. In section 2, we present the context of the works. Section 3 describes related works about terminology extraction. The section 4 presents the INFO_COMP dictionary, its entries, the inflexional descriptions, the NOOJ grammars and an example of automatic extraction of terms from a text, with
NOOJ. We find our remarks and future works in section 5. An extract of the dictionary is given in Annex 1 for the term card.
2 Context of the works The environment of development is NOOJ [21] which is a new linguistic development environment, issued of author Silberztein M. 12 years experience, as the INTEX user and designer at the LADL [18]. The LADL was the principal research center on the electronic dictionaries of French. NOOJ is free and can be downloaded at [21].
3 Related Works about terminology extraction One will find in [14] works an examination of the techniques of extraction of the terminological data and their impact on the terminologist work. A census of the various existing systems is made in [10].They make it possible to extract new terms starting from texts or corpus. Here also the various linguistic, statistical or mixed methods exist and are used to develop the tools for automatic extraction. The majority of the authors [10] consider it essential to maintain a human activity in the acquisition systems, with an acceptance or a refusal of the results but also as far as being the acquisition centre,
\
Proceedings of the 10th WSEAS International Conference on COMPUTERS, Vouliagmeni, Athens, Greece, July 13-15, 2006 (pp458-462)
the computer processing will thus be reduced to a presentation and data-recording tool.
3.1 Tools using the linguistic methods They are also called symbolic methods. They are based on a syntactic analysis of the texts. A preliminary morpho-syntactic tagging locates the noun syntagms, to analyze them in order to extract the terms which are candidates. The candidate terms will be submitted to the expert to retain the compound words that are terms. Four systems are listed in [5], we have TERMINO [9], LEXTER [4], XTERM by Cerbah, F. developed in 1999, and FASTER [13]. The system LEXPRO [17] uses dictionaries DELAS and DELAC of English as well as system INTEX [18] to acquire English computer science terms. With NOOJ [21] we can extract terms.
3.2 Statistical methods They make it possible to extract candidate terms without preliminary linguistic analysis. System ANA [11] automatically extracts concepts from texts to produce a semantic network. The system MANTEX developed by Oueslati R. [15] is founded on the repeated segments principle. Rochibeau B. [16] extracts collocations using statistical methods. In his doctoral thesis, Balvet A. [3] evaluates the statistical approaches capital in locating thematic signatures (complex lexical units).
3.3 Mixed methods System like ACABIT, developed by Daille B. [8] allows to cure the problems of noise which arises in the statistical methods with the introduction of linguistic owners and the use of statistical filters, with XTRACT, Smadja F. [22] gives a generic tool for location of collocations and not only of the terms.
4 The electronic dictionary of computer science terms: INF0_COMP for NOOJ NOOJ developed by Silberztein M. is a development environment [21] [20] used to construct large-coverage formalized descriptions of natural languages, and apply them to large corpora, in real time.
4.1 The compound words, the term Our aim is to build an electronic dictionary of computer terms for NOOJ. A term can be simple if it contains one word or compound if it contains more than one. A compound word is built starting from simple words. Silberztein M. defines a compound noun as a consecutive sequence of at least two simple forms and blocks of separators. A simple form is a nonempty consecutive sequence of characters of the alphabet appearing between two separators. A simple word is a simple form that constitutes an input of dictionary. We will use indifferently term or compound noun to indicate the same concept within the selected technical language (computer science). We studied the shape of an input and the compound nouns gender and number, the possible determinants, and the inflection of the compound nouns. In [18], we find the plural of the compound nouns for the principal classes. In [1] we inventoried the different classes of compounds.
4.2 NOOJ dictionaries The description of natural languages are formalized as electronic dictionaries, and grammars represented by organized sets of graphs. NOOJ dictionaries [20] are a great enhancement over INTEX DELA-type dictionaries [6] as well as lexicon grammars. NOOJ dictionaries are similar to DELAS-DELAC dictionaries and can represent spelling and terminological variants. In INTEX [18], DELAS is the electronic dictionary of simple words DELAC [7] is the dictionary of the compound nouns. In NOOJ [20], the INTEX dictionaries are represented in one unique format the full description of the inflexion and derivation is encoded inside NOOJ dictionaries for the entries. NOOJ dictionaries are used to represent, describe and recognize simple and compound words. Dictionaries are .nod files that are compiled from editable .dic source files.
4.3 Entry format of the dictionary INFO_COMP The dictionary contains all the lemmas of the language and with the lemma we have a morphosyntactical code, possible semantic and syntactic codes, inflectional and derivational paradigms. The fig.1 shows some entries. For the term algorithme de gestion mémoire : • The term: algorithme de gestion mémoire; • The category: N+NPNPN; • Info: informatique;
Proceedings of the 10th WSEAS International Conference on COMPUTERS, Vouliagmeni, Athens, Greece, July 13-15, 2006 (pp458-462)
• •
FLX: algogemem is the flexional model name for this term; Hild98: text source of extraction.
4.5 NOOJ grammars A NOOJ local grammar (with file extension .nog) makes it possible to gather the terms by family, a grammar can contain several graphs.
Fig.1 Extract of the compound nouns dictionary
4.4 Inflection of the compound words NOOJ’s inflexion module is triggered by adding the special property “+FLX” to a lexical entry. NOOJ provides two equivalent tools to describe the inflexional paradigms: - Inflection description files with .flx extension are set of rewriting rules, - Inflectional grammars with .nof extension are special kinds of morphological grammars.
Fig.3 A NOOJ grammar for the term "registre"
4.6 Automatic extraction of compound terms with NOOJ After the linguistic analysis, we can use “locate a pattern” to extract compound nouns from texts or corpus, here we want to locate de terms in the text:
Fig.4 A sample of text Fig. 2 Extract of the inflexion dictionary
The concordance located with the de pattern for the text infotest.not is in the Fig.5:
Proceedings of the 10th WSEAS International Conference on COMPUTERS, Vouliagmeni, Athens, Greece, July 13-15, 2006 (pp458-462)
Acknowledgements: I would like to thank Max Silberztein, from the LASELDI laboratory from the Franche-Comté University for his help and all his remarks.
Fig.5
Some new terms extracted by NOOJ
5 Conclusion and Future Works These newspapers like "Le Monde Informatique" made it possible to collect manually terms but not much. Specialized books of computer science, dictionaries were taken as corpus for manual extracting and the number of terms clearly increased with the Hildebert [12] dictionary. Actually, more than 10 000 terms were extracted and listed manually, and codified and added as entries of the to INFO_COMP dictionary. Others informations (conceptual, syntactic, synonymous, English translation) can be added at the entry of the dictionary according to the use that one will make of it. 30 000 terms and more are found and will be codified. The collection of the terms is manual; it is long and needs to read of big corpus. We are building a big corpus of computer science with different texts of computer science from PDF, HTML, WORD and PS files. The corpus is in the Unicode text format . We meet difficulties to find French electronic corpus of computer science terminology. NOOJ can analyze the corpus and we can automatically locate new terms. The elaborate dictionary INFO_COMP will make it possible to analyze computer science corpus, texts. With NOOJ, one will be able to treat computer science technical texts and to use them in various applicability such are the automatic indexing, the information retrieval, the automatic analysis of texts, and machine translation. It remains to finalize the coding of others terms of the dictionary, to set up all the other grammars. Tests are then designed to analyze the big corpus elaborated and computer science texts. We plan to compare NOOJ terms acquisition with statistic and mixed methods.
References: [1] F. Aoughlis, «Building an Electronic Dictionary of Computer Science", in 8th INTEX/NOOJ workshop, Besançon, may 30june 1, 2005. http://www.nooj4nlp.net. [2] G. Aussenac-Gilles, A. Condamines, S. Szulman, "Prise en compte de l’application dans la constitution de produits terminologiques " in Actes des 2èmes assises nationales du GDR13, Toulouse Cepaduès ed., Nancy 2002, pp. 289-302. [3] A. Balvet, "Approches catégoriques et non catégoriques en linguistique des corpus spécialisés, application à un système de filtrage d'information", thèse de doctorat, université Paris X, Nanterre, 2002. [4] D. Bourigault, 1994, "LEXTER, un logiciel d’extraction de terminologie. Application à l’extraction de connaissances à partir de textes", Thèse de doctorat, Ecole des hautes Etudes en Sciences Sociales, Paris, 1994. [5] “Constitution de corpus à partir du Web pour l’acquisition automatique de terminologie : une expérience ". http: //www.poleia.lip6.fr/~slodzian/sberland/ [6] B. Courtois, M. Silberztein, "Dictionnaires électroniques du français ", in Langue Française n°87, Larousse, Paris, 1990. [7] B. Courtois, Garrigues M., Gross M., Gross M., Rung M., Mathieu-Colas, Silberztein M., Vives M., "Dictionnaire électronique des noms composés DELAC : les composants NA et NN", Technical report, LADL, N° 55, 1997. [8] Daille B., Approche mixte pour l’extraction automatique de terminologie : statistique lexicale et filtres linguistiques, Thèse de doctorat en informatique fondamentale, université de Paris7, 1994. [9] S. David, P. Plante, "De la nécessité d’une approche morphosyntaxique en analyse de textes", in OICO, 2(3) pp. 140-155, Québec, 1990. [10] G. De Chalendar, "STEVLAN : un système de structuration du lexique guidé par la détermination automatique du contexte thématique ".Thèse de docteur de l’université de Paris XI, Orsay, décembre 2001.
Proceedings of the 10th WSEAS International Conference on COMPUTERS, Vouliagmeni, Athens, Greece, July 13-15, 2006 (pp458-462)
[11] C. Enguehard, "Acquisition naturelle automatique d’un réseau sémantique", thèse de doctorat, Compiègne, 1992. [12] J. Hildebert, " Dictionnaire des Technologies de l’Informatique ", volume II, français/anglais, éditions La maison du dictionnaire, Hippocrene books inc, Paris, 1998. [13] C. Jacquemin, "Variation terminologique et acquisition automatique de termes et leurs variantes en corpus", Habilitation à diriger des recherches en informatique", IRIN, université de Nantes, 1997. [14] M.C. L'homme, "Nouvelles technologies et recherche terminologique. Techniques d'extraction des données terminologiques et leur impact sur le travail du terminographe. L'Impact des nouvelles technologies sur la gestion terminologique", Montreal, 2001. http://www.ling.umontreal.ca/lhomme/cgi-bin. [15] R. Oueslati, "Aide à l'acquisition de connaissances à partir de corpus". Thèse de doctorat, Université Louis Pasteur, Strasbourg, 1999. [16] B. Rochibeau, "Extraction de collocations fondée sur les méthodes statistiques", in Ateliers sur les corpus spécialisés en terminologie, 2003. http: //www.ling.umontreal.ca/lhomme/cgibin/BR.zip [17] A. Savary, "Recensement et description des mots composés- Méthodes et applications", thèse de doctorat, université de Paris7, 2000. [18] M. Silberztein, "Dictionnaires électroniques et analyse automatique de textes : le système INTEX, Edition Masson, Paris, 1993. [19] M. Silberztein, 1990, "Le dictionnaire électronique des mots composés ", in Langue Française, N° 87, septembre 1990. [20] M. Silberztein, "NOOJ’s dictionaries», in the Proceedings of the LTC, 2005. [21] M. Silberztein, "NOOJ’ manual», 2006. http ://www.nooj4nlp.net [22] F. Smadja, "XTRACT: an overview", in Computers and Humanities, 26, pp.399-413, Kluwer Academic Publishers, Netherlands, 1996.
Annexe.1 carte à 80 colonnes carte à 90 colonnes carte à bande carte à carte carte de chargement carte à fenêtre carte à mémoire flash
carte à mémoire carte accélératrice carte binaire carte binaire par colonne carte binaire par rangée carte chercheuse carte circuit carte circuit imprimé carte courte carte Datavoice So carte de chargement carte de contrôle carte de décompression carte de fin carte de jeu électronique carte de lancement carte de police de caractères carte de télécopie carte demi-format carte d'expansion mémoire carte d'extension carte d'extension carte d'extension carte d'interface carte dorsale carte fille carte graphique 3D carte graphique Hercules carte graphique carte intelligente carte local bus PCI carte magnétique carte mère carte mère Mercury PCI carte mère NA H83 carte mère Pentium PCI carte mère Pentium PRO carte multi-fonctions carte non peuplée carte nue carte PCMCIA carte Pentium carte perforée carte perforée carte pilote carte programme carte RAM carte ROM carte Scii Telecom Expres SOISA carte son H carte système carte video carte video graphique carte video local bus PCI carte vierge
