Development of question answering system for factoid and non-factoid type questions (case: history of khulafaur rasyidin)
Main Author: | Hadzami An Nadisah |
---|---|
Format: | Bachelors |
Terbitan: |
Fakultas Sains dan Teknologi UIN Syarif Hidayatullah Jakarta
|
Daftar Isi:
- Question answering system is one of sub-fields information retrieval that integrates information from multiple sources as document collection whether local or web to provide concise answer to specific question which is entered by the user. This question answering system is about history of khulafaur rasyidin which is expanded from the previous research (Naf?an, 2012) by adding types of question word; such as: apa, mengapa, and bagaimana. So, this question answering system provides question words: siapa, kapan, and dimana as factoid question and apa, mengapa, and bagaimana as non-factoid question. This question answering system consists of several processes: question analyzer, document retrieval, and answer extraction. Question analyzer aims to classify question to obtain answer type, clue word, and query. This query will be used in document retrieval to obtain document candidate which contain exact answer by using search engine which is called Lucene. The sources that are used came from previous research (Naf?an, 2012). It has been reviewed and corrected by the author. Answer extraction aims to obtain exact answer which is desired by the user. Answer extraction for factoid question is performed by calculating distance between answer candidate to all queries from the question in sentence answer and for non-factoid question is performed by sentence pattern which has priority value for each sentence. System testing is performed by giving 30 question words for factoid question and non-factoid question. The evaluation of answer result refers to (Naf?an, 2012) which has 5 categories, namely: benar, tidak tepat, tidak mendukung, tidak terjawab, and salah. From the system evaluation, factoid question has high performance to obtain exact answer and non-factoid question has low performance.