[4Q2-GS-9] Natural language processing, information retrieval: Q&A system

[4Q2-GS-9-04] Comparison and consideration of natural language SQL query conversion technology and question understanding technology

〇Hitomi Misumi¹, Hiroki Unno¹ (1. INET)

Keywords:Natural Language, seq2SQL, Tableau, SQL, Japanese

In recent years, creating SQL queries using natural language questions has attracted a great deal of interest.Seq2SQL uses machine learning to generate SQL queries corresponding to questions based on questions in a natural language and database schema information.SQLNet eliminates the need to use reinforcement learning algorithms,The accuracy of the WHERE clause was 9% to 13% better than that of the conventional technology.In this paper, we created a model that generates a conditional part interactively in order to Identify problems when applying the conditional part generation model to Japanese question sentences.From the results of this research, it became clear that the length of the question sentence was important as one of the factors affecting the accuracy.