Inference and Reconciliation in a Crowdsourced Lexical-Semantic Network

Abstract

Abstract. Lexical-semantic network construction and validation is a major issue in NLP. No matter the construction strategies used, automatically inferring new relations from already existing ones is a way to improve the global quality of the resource by densifying the network. In this context, the purpose of an inference engine is to formulate new conclusions (i.e. relations between terms) from already existing premises (also relations) on the network. In this paper we devise an inference engine for the JeuxDeMots lexical network which contains terms and typed relations between terms. In the JeuxDeMots project, the lexical network is constructed with the help of a game with a purpose and thousands of players. Polysemous terms may be refined in several senses (bank may be a bank-financial institution or a bank-river) but as the network is indefinitely under construction (in the context of a Never Ending Learning approach) some senses may be missing. The approach we propose is based on the triangulation method implementing semantic transitivity with a blocking mechanism for avoiding proposing dubious new relations. Inferred relations are proposed to contributors to be validated. In case of invalidation, a reconciliation strategy is undertaken to identify the cause of the wrong inference : an exception, an error in the premises or a transitivity confusion due to polysemy with the identification of the proper word senses at stake.