ReSyf: a French lexicon with ranked synonyms

Proceedings of the 27th International Conference on Computational Linguistics, pages 2570–2581 Santa Fe, New Mexico, USA, August 20-26, 2018. 2570 ReSyf: a French lexicon with ranked synonyms Billami, Mokhtar B. Aix-Marseille University LIS UMR 7020 Marseille, France mokhtar.billami@univ-amu.fr Franc¸ois, Thomas Catholic University of Louvain FNRS-CENTAL/IL&C Louvain-la-Neuve, Belgium thomas.francois@uclouvain.be Gala, N´uria Aix-Marseille University LPL UMR 7309 Aix-en-Provence, France nuria.gala@univ-amu.fr Abstract In this article, we present ReSyf, a lexical resource of monolingual synonyms ranked according to their difficulty to be read and understood by native learners of French. The synonyms come from an existing lexical network and they have been semantically disambiguated and refined. A ranking algorithm, based on a wide range of linguistic features and validated through an eva- luation campaign with human annotators, automatically sorts the synonyms corresponding to a given word sense by reading difficulty. ReSyf is freely available and will be integrated into a web platform for reading assistance. It can also be applied to perform lexical simplification of French texts. 1 Introduction With the availability of very large corpora and the growing maturity of corpus linguistics and NLP techniques, quantitative descriptions of the lexicon have made noteworthy progress. The coverage of the resources has increased, tokenizing and part-of-speech tagging have enabled a better annotation of the lexical units and statistical models have yielded more accurate descriptions of the lexicon (in terms of frequencies, n-gram models, etc.). However, to the best of our knowledge, on-line lexical resources with a ranking of the lexical units as regards to their difficulty to be read and understood are scarse. Such information is highly relevant in communicative and educational contexts (e.g. clear and efficient writing, calibration of teaching materials, etc.) given that complex synonyms are identified and their simpler equivalents proposed. In this paper, we present and describe such a resource for French. In ReSyf, synonyms are automati- cally disambiguated and ranked according to their complexity for L1 schoolchildren readers. The lexicon has been developed as part of the project ALECTOR1 (Reading Aids to leverage Document Accessibility for Children with Dyslexia) that aims to support poor readers and dyslexic children to acquire French vocabulary and improve their reading skills in French through practise. More specifically, ALECTOR addresses the challenge of automatic text simplification for this population and illustrates one of the possible uses of ReSyf: it can be a knowledge database to carry out the substitution of complex words present in reading materials. The evaluation of the proposed ranking algorithm of the synonyms shows that, in 91% of the cases, the ranks in ReSyf correspond to the ranks given by human annotators. These results are encouraging taking into account the difficulty of the task (i.e. subjectivity, inter-annotator agreement). They are also important because the simplest synonym is always identified, which implies that ReSyf can indeed be used for reading assistance and can also be integrated into a model for automatic lexical simplification. The paper is organized as follows. After reviewing related work at Section 2, we present the metho- dology applied to …

Content truncated to 500 words