Open Domain Question Answering System Using Wikipedia
Author(s):
Ria Singh, Preetam B. A., Prince Zalavadia, Dr. Sharvari Govilkar
Keywords:
Question Answering System, Open Domain, Wikipedia, Deep Learning, Natural Language Processing, Document Retrieval, Document Reader, Answer Ranker.
Abstract
The open-domain question answering task has recently been addressed using unstructured data such as websites and online encyclopedias. Here, open-domain questions are answered by making full use of knowledge sources of Wikipedia via its API for many types of questions, it is critical to analyze user questions in terms of the nature of the answers being sought. The analyzed result of a question has three components: Answer Format, Answer Theme and Question Target (question analysis). The next step involves finding the most relevant documents or passages related to the question using either word embedding distances or Deep Learning (document retrieval). Finally, the answers are extracted from the passage (machine comprehension). PageRank technique can be used while computing the "document score" to assess relevance of a document to a query. BERT or ALBERT architecture (document reader trained on SQuAD 2.0 dataset) can be used which will dramatically improve performance for machine comprehension. The Answer Ranker (SoftMax function) extracts the 1-2 lines of answer for the query. The question provided and the answer generated are in audio format using the SpeechRecognition library and gTTS API.
Article Details
Unique Paper ID: 151185

Publication Volume & Issue: Volume 7, Issue 12

Page(s): 165 - 173
Article Preview & Download


Share This Article

Conference Alert

ICM - STEP

International conference on Management, Science, Technology, Engineering, Pharmact and Humanities.

Latest Publication

Go To Issue



Call For Paper

Volume 7 Issue 9

Last Date 25 February 2020

About Us

IJIRT.org enables door in research by providing high quality research articles in open access market.

Send us any query related to your research on editor@ijirt.org

Social Media

Google Verified Reviews

Contact Details

Telephone:6351679790
Email: editor@ijirt.org
Website: ijirt.org

Policies