Question answering enhanced with a window function

Session

Computer Science and Communication Engineering

Description

Question answering (QA) can be generally defined as an automatic process capable to understand questions formulated in a natural language and responding exactly with the requested information [1]. However, this simple definition turns to be very complex when the characteristics and functionalities of the “ideal” QA system are analyzed. The knowledge from several research areas, such as, Natural Language Processing, Information Extraction, Information Retrieval, Knowledge Representation, are required to design the “ideal” QA system. Researchers have experimented with different approaches in every stage of the QA system. Same stadies focus on utilizing question keywords density distribution to enhance retrieval of relevant information snippets [2]. It is evident that the distances between the sequence od words indicating the correct answer and the keywords extracted from the question are especially important for QA systems [3].

This study presents the results from the final stage of the several years developing Question answering system for Macedonian language. A window function is implemented in the process of selecting the correct answer by assigning higher weights to keywords from the question and the evaluated answer that appear densely in a particular window [4]. So, the aim of this study is to assess the influence of keywords proximity in the process of selecting the correct answer for a given multiple-choice question posed in Macedonian language. The preliminary experimental results show that the Hanning window function can be effectively used for that purpose.

Keywords:

Question answering, keyword proximity, density distribution, Hanning window

Session Chair

Felix Breitenecker

Session Co-Chair

Edmond Hajrizi

Proceedings Editor

Edmond Hajrizi

ISBN

978-9951-437-54-7

Location

Durres, Albania

Start Date

28-10-2017 9:00 AM

End Date

28-10-2017 10:30 AM

DOI

10.33107/ubt-ic.2017.101

This document is currently not available here.

Share

COinS
 
Oct 28th, 9:00 AM Oct 28th, 10:30 AM

Question answering enhanced with a window function

Durres, Albania

Question answering (QA) can be generally defined as an automatic process capable to understand questions formulated in a natural language and responding exactly with the requested information [1]. However, this simple definition turns to be very complex when the characteristics and functionalities of the “ideal” QA system are analyzed. The knowledge from several research areas, such as, Natural Language Processing, Information Extraction, Information Retrieval, Knowledge Representation, are required to design the “ideal” QA system. Researchers have experimented with different approaches in every stage of the QA system. Same stadies focus on utilizing question keywords density distribution to enhance retrieval of relevant information snippets [2]. It is evident that the distances between the sequence od words indicating the correct answer and the keywords extracted from the question are especially important for QA systems [3].

This study presents the results from the final stage of the several years developing Question answering system for Macedonian language. A window function is implemented in the process of selecting the correct answer by assigning higher weights to keywords from the question and the evaluated answer that appear densely in a particular window [4]. So, the aim of this study is to assess the influence of keywords proximity in the process of selecting the correct answer for a given multiple-choice question posed in Macedonian language. The preliminary experimental results show that the Hanning window function can be effectively used for that purpose.