Game State Retrieval with Keyword Queries

摘要

There are many databases of game records available online.rnIn order to retrieve a game state from such a database, usersrnusually need to specify the target state in a domain-specifcrnlanguage, which may be difficult to learn for novice users.rnIn this work, we propose a search system that allows usersrnto retrieve game states from a game record database by us-rning keywords. In our approach, we frst train a neural net-rnwork model for symbol grounding using a small number ofrnpairs of a game state and a commentary on it. We thenrnapply it to all the states in the database to associate eachrnof them with characteristic terms and their scores. The en-rnhanced database thus enables users to search for a state us-rning keywords. To evaluate the performance of the proposedrnmethod, we conducted experiments of game state retrievalrnusing game records of Shogi (Japanese chess) with commen-rntaries. The results demonstrate that our approach gives sig-rnnifcantly better results than full-text search and an LSTMrnlanguage model.