Ambiguity Diagnosis for Terms in Digital Humanities

Ambiguity Diagnosis for Terms in Digital Humanities B´eatrice Daille1, Evelyne Jacquey2, Ga¨el Lejeune3, Luis Felipe Melo4, Yannick Toussaint4 1LINA UMR 6241, University of Nantes; 2UMR 7118 ATILF CNRS Universit´e de Lorraine; 3GREYC UMR 6072, Normandy University; 4INRIA Nancy Grand-Est & LORIA CNRS Universit´e de Lorraine. beatrice.daille@univ-nantes.fr, evelyne.jacquey@atilf.fr, gael.lejeune@univ-nantes.fr, luisfe.melo@gmail.com, yannick.toussaint@loria.fr Abstract Among all researches dedicating to terminology and word sense disambiguation, little attention has been devoted to the ambiguity of term occurrences. If a lexical unit is indeed a term of the domain, it is not true, even in a specialised corpus, that all its occurrences are terminological. Some occurrences are terminological and other are not. Thus, a global decision at the corpus level about the terminological status of all occurrences of a lexical unit would then be erroneous. In this paper, we propose three original methods to characterise the ambiguity of term occurrences in the domain of social sciences for French. These methods differently model the context of the term occurrences: one is relying on text mining, the second is based on textometry, and the last one focuses on text genre properties. The experimental results show the potential of the proposed approaches and give an opportunity to discuss about their hybridisation. Keywords: Terminology, Disambiguation, Ambiguity, Pol- ysemy, Text Mining, Textometry, Salience 1. Introduction This paper introduces and compares three original methods for ambiguity diagnosis. The objective is to decide for any occurrence of a term candidate (TC) if it is terminological (TO) or not (NTO). This ambiguity diagnosis is useful for information retrieval, keyphrase extraction, and also for text summarization. While lexical disambiguation is a very productive issue (Navigli, 2009), research on term disam- biguation remains surprisingly unexplored. Nevertheless, in any domain, TCs may be ambiguous (L’Homme, 2004), having TOs and NTOs as well. As an illustration, let us consider two occurrences of aspect in a research paper belonging to the linguistic domain in French : (I) L’ aspect est une cat´egorie qui refl`ete le d´eroulement interne d’un proc`es ’Aspect is a cat- egory which expresses the internal sequence of a process’ (Cothire-Robert, 2007) (II) Ce dernier aspect est primordial ’This last aspect is primordial’ (El-Khoury, 2007) In the first example, aspect is a term, while it is not in the second example. Ambiguity occurs also for multi- word terms when they are submitted to lexical reduc- tion in discourse. The reduced form is often ambigu- ous: the occurrence of the TC analysis could refer to syntactic analysis, semantic analysis or to a non-terminological sens of analysis. Term disambiguation, as in most works on word sense dis- ambiguation, is considered here as a classification problem. We propose three supervised learning methods which ex- plore different modelizations of TCs context. They are tested on a manually annotated corpus in the broad domain of social sciences in French. This paper is organised as follows: first we give some in- sights about related work (section 2.), then we present the dataset (section 3.) and the proposed methods (section 4.). Finally, we evaluate …

Content truncated to 500 words