Department of Computer Engineering
CS 590 SEMINAR
Keyword Mapping for Natural Language Queries in Relational Databases
Computer Engineering Department
Translating Natural Language Queries(NLQ) to Structured Query Language(SQL) is a challenging task that is widely studied in database research recently. Translation process contains multiple steps where each step needs its own solution. However, recent works have mostly focused on the translation step instead of studying on earlier steps. In the translation pipeline, one of the most important problems which are overlooked in the past is keyword mapping; constructing a mapping between tokens in the query and relational database elements(tables, attributes, values etc.). We define the keyword mapping problem as a sequence tagging problem and propose a novel deep learning based approach that utilizes POS(Part of Speech) tags of the NLQ. We introduce a Bi-LSTM language model optimized on three different loss functions simultaneously. We evaluated our approach with IMDB, Yelp and Scholar schemas and report 84%, 89% and 93% accuracy on tokens in the NL queries respectively.
DATE: 02 December 2019, Monday @ 16:40