TermoPL - a Flexible Tool for Terminology Extraction

TermoPL — a Flexible Tool for Terminology Extraction Małgorzata Marciniak, Agnieszka Mykowiecka, Piotr Rychlik Institute of Computer Science PAS ul. Jana Kazimierza 5 01-248 Warsaw, Poland Malgorzata.Marciniak, Agnieszka.Mykowiecka, Piotr.Rychlik@ipipan.waw.pl Abstract The purpose of this paper is to introduce the TermoPL tool created to extract terminology from domain corpora in Polish. The program extracts noun phrases, term candidates, with the help of a simple grammar that can be adapted for user’s needs. It applies the C-value method to rank term candidates being either the longest identified nominal phrases or their nested subphrases. The method operates on simplified base forms in order to unify morphological variants of terms and to recognize their contexts. We support the recognition of nested terms by word connection strength which allows us to eliminate truncated phrases from the top part of the term list. The program has an option to convert simplified forms of phrases into correct phrases in the nominal case. TermoPL accepts as input morphologically annotated and disambiguated domain texts and creates a list of terms, the top part of which comprises domain termi- nology. It can also compare two candidate term lists using three different coefficients showing asymmetry of term occurrences in this data. Keywords: terminology extraction, domain corpora, Polish, C-value 1. Introduction Every domain of language communication has its own spe- cial terminology — words and multi-word expressions. They are listed in domain glossaries and terminology dic- tionaries which are used both by people and computer ap- plications. Terminology resources are useful for annotating document repositories, for better text classification and au- tomatic translation. For newly created and rapidly chang- ing domains, automatic terminology extraction (TE) helps in building new domain lexicons and facilitates keeping the existing ones up to dates TE is typically performed in three steps. The first one con- sists in a preselection of term candidates. In the second step, the candidates are sorted according to an accepted term quality measure. Finally, general terms, i.e. terms which are used in many different domains, are filtered out by comparison of phrases obtained from domain and non- domain corpora. Various solutions for these three prob- lems were already introduced in the relevant literature. The most interesting ones were presented in (Pazienza et al., 2005). Candidate preselection is almost always done by using simple shallow grammars which recognize the most common syntactic term structure. Term ordering is done using various strategies which can take into account: phrase frequency, number of contexts in which a phrase occurs, phrase length, and phrase elements’ association measures (Pecina and Schlesinger, 2006) such as mutual information, statistical tests of independence and likelihood. Many terminology extraction tools have already been im- plemented. Tools available on the Internet are adapted to process the following languages: • English, French, Spanish, Portuguese and Italian by TermoStat (Drouin, 2003). It uses statistical and lin- guistic methods to identify candidate terms. • English, Italian, French by Terminology Extrac- tion (of Translated Labs). It uses Poisson statis- tics, the Maximum Likelihood Estimation and Inverse Document …

Content truncated to 500 words