ACCESS TO RELATIONAL DATABASES USING INTERROGATIVE SENTENCES IN INDONESIAN LANGUAGE

• Main Author: Ghassani , Hishshah
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/27881
• Institution: Institut Teknologi Bandung

One of way to access relational databases is using Natural Language Interface to Databases or NLIDB, which is accessing relational databases using natural language. Generally, NLIDB developed for domain specific databases. NLIDB in Indonesian language that can be used in any domain already developed by Fibrian in 2016. However, this NLIDB system only accept imperative sentences, while interrogative sentences was not handled yet. Therefore, this thesis was carried out to design the translation method from interrogative sentences in Indonesian language to SQL query. The translation method is carried out in three steps. First step, interrogative sentences that user input is analyzed using syntactic parser, then the result is mapped to SQL parts. Second step, in order to NLIDB that can be used in any domain, ontology is build automatically from relational databases. In the ontology the list of words associated with the ontology object is stored, which is build using WordNet bahasa. Then after we get the ontology object, last step is to build whole SQL query corresponding to the query type. Interrogative sentences type that can be accepted by this application is ordinary interrogative sentence and clarification/confirmation interrogative sentence. Sentences can be translated is depending on parser used, and sentences with object sentences with desired objects from the database are explicitly mentioned. NLIDB application development was carried out to test the translation method from interrogative sentences in Indonesian language to SQL query. NLIDB in this thesis is using PC-PATR syntactic parser with production rules and lexicon from Joice in 2005. Application implementation is using Java programming language and MySQL databases. This application successfully translates various types of Indonesian interrogative sentences into SQL queries. This application can also build ontology automatically from relational databases in any domain. Some development that can be made on this thesis for future works are identification method of attribute’s semantic type in ontology, handling SQL query with HAVING, handling input sentence which has implicit ontology object, and searching method for ontology relations between classes. <br />