ANNLOR: A Naïve Notation-system for Lexical Outputs Ranking

First Joint Conference on Lexical and Computational Semantics (*SEM), pages 487–492, Montr´eal, Canada, June 7-8, 2012. c⃝2012 Association for Computational Linguistics ANNLOR: A Na¨ıve Notation-system for Lexical Outputs Ranking Anne-Laure Ligozat LIMSI-CNRS/ENSIIE rue John von Neumann 91400 Orsay, France annlor@limsi.fr Cyril Grouin LIMSI-CNRS rue John von Neumann 91400 Orsay, France cyril.grouin@limsi.fr Anne Garcia-Fernandez CEA-LIST NANO INNOV, Bt. 861 91191 Gif-sur-Yvette cedex, France anne.garcia-fernandez@cea.fr Delphine Bernhard LiLPa, Universit´e de Strasbourg 22 rue Ren´e Descartes, BP 80010 67084 Strasbourg cedex, France dbernhard@unistra.fr Abstract This paper presents the systems we developed while participating in the first task (English Lexical Simplification) of SemEval 2012. Our first system relies on n-grams frequencies computed from the Simple English Wikipedia version, ranking each substitution term by de- creasing frequency of use. We experimented with several other systems, based on term fre- quencies, or taking into account the context in which each substitution term occurs. On the evaluation corpus, we achieved a 0.465 score with the first system. 1 Introduction In this paper, we present the methods we used while participating to the Lexical Simplification task at Se- mEval 2012 (Specia et al., 2012). We experimented with several methods: • using word frequencies or other statistical fig- ures from the BNC corpus, Google Books NGrams, the Simple English Wikipedia, and results from the Bing search engine (with/with- out lemmatization); • using association measures for a word and its context based on language models (with/with- out inflection); • making a combination of previous methods with SVMRank. Depending on the results obtained on the training corpus, we chose the methods that seemed to best fit the data. 2 Task description 2.1 Presentation The Lexical Simplification task aimed at determin- ing the degree of simplicity of words. The inputs given were a short text, in which a target word was chosen, and several substitutes for the target word that fit the context. An example of a short text follows; the target word is “outdoor”, and other words of this text will be considered as the context of this target word. <i n s t a n c e id =” 270 ”> <c o n t e x t>With the growing demand f o r t h e s e f i n e garden f u r n i s h i n g s , they found i t n e c e s s a r y to d e d i c a t e a p o r t i o n of t h e i r b u s i n e s s to <head>outdoor</ head> l i v i n g and p a t i o f u r n i s h i n g s .</ c o n t e x t> </ i n s t a n c e> The substitutes given for this target word were the following: “alfresco;outside;open-air;outdoor;”. The objective was to order these words by descend- ing simplicity. 2.2 Corpora Two corpora were …

Content truncated to 500 words