NATURAL LANGUAGE INTERFACE TO DATABASE (NLIDB) FOR INTERROGATIVE SENTENCES WITH TEMPORAL ELEMENTS IN INDONESIAN LANGUAGE

• Main Author: Grady Daniel Thamrin, Hansel
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/69203
• Institution: Institut Teknologi Bandung

NLIDB (Natural Language Interface to Database) is a solution that can allow users to retrieve data from databases in natural language. There are various studies that have been conducted on NLIDB, such as the Poetra’s research (2019) related to NLIDB with imperative sentences for temporal data. However, the NLIDB system built by Poetra has not been able to facilitate input in the form of interrogative sentences. The purpose of this Final Project is to develop an NLIDB that is able to translate interrogative sentences that have temporal elements into SQL queries that can be executed and the results of the execution are displayed. The translation process is carried out by first masking the input sentence and decomposing the sentence into a collection of syntax trees. The resulting set of syntax trees will be selected with tree similarity to produce the tree that is most similar to the tree that has been generated previously. The result of parsing this sentence will be analyzed per component based on the nodes generated in the tree. Furthermore, the entity of the search results for question sentences will be sought in ontology and produce a list of tables and relationships involved in the formation of SQL queries. SQL queries are built based on the results of analysis of syntax tree components and mapping of entities to database objects. The SQL query will be executed and the user can provide feedback regarding the tree selected by the system. The user-selected tree will be stored by the system in a collection that stores trees that have been correctly generated before. The implementation is done with MySQL database and PC-PATR as interrogative sentence syntax parser. From the tests carried out, the built NLIDB succeeded in translating interrogative sentences with the question words what, who, when, where, how, what, and how much. For sentences with the question word how, the translation is only done for historical queries on temporal data.