Design and Development of Amharic-English Cross Language Question Answering System
No Thumbnail Available
Date
10/6/2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Most of web documents are published in English. Due to this reason, different information retrieval systems that retrieve English documents are developed by different researchers. And, Web users who do not know English are unable to access those systems in their native language. To enable Web users to access web documents written in a different language, a cross-language information retrieval is one of the solutions. So that, a question answering system is one of a practical application of information retrieval. Developing a cross lingual question answering system is essential to enable user to ask a question in a language different than the language in which documents are written. We used a statistical machine translation-based approach for Amharic-English cross language question answering.
In this research work, we have designed architecture of Amharic-English cross language question answering. In doing so, we have developed a statistical machine translation system for translating user question. Also, to make the system more effective, we developed an Amharic named entity tagger with an algorithm of maximum entropy. And by using this system we make a semantic based indexing; it is a great technique to increase precision of the system and pinpointing an answer.
To evaluate the overall system, precision and recall metrics are used. The Experimental result obtained shows that for Amharic-English cross lingual retrieval we got a result of 72% of precision and recall of 79% and for English-Amharic cross lingual retrieval got a result of 65% of precision and recall of 70%.
Description
Keywords
Cross-Lingual Information Retrieval, Sematic Based Indexing and Statistical Machine Translation