A Reinforcement Learning-driven Translation Model for Search-Oriented\n Conversational Systems

Search-oriented conversational systems rely on information needs expressed in\nnatural language (NL). We focus here on the understanding of NL expressions for\nbuilding keyword-based queries. We propose a reinforcement-learning-driven\ntranslation model framework able to 1) learn the translation from NL\nexpressions to queries in a supervised way, and, 2) to overcome the lack of\nlarge-scale dataset by framing the translation model as a word selection\napproach and injecting relevance feedback in the learning process. Experiments\nare carried out on two TREC datasets and outline the effectiveness of our\napproach.\n