The Impact of Multiple Passages on Machine Comprehension
Loading...
Links to Files
Permanent Link
Author/Creator
Author/Creator ORCID
Date
2019-01-01
Type of Work
Department
Computer Science and Electrical Engineering
Program
Computer Science
Citation of Original Publication
Rights
Distribution Rights granted to UMBC by the author.
Access limited to the UMBC community. Item may possibly be obtained via Interlibrary Loan thorugh a local library, pending author/copyright holder's permission.
This item is likely protected under Title 17 of the U.S. Copyright Law. Unless on a Creative Commons license, for uses protected by Copyright Law, contact the copyright holder or the author.
Access limited to the UMBC community. Item may possibly be obtained via Interlibrary Loan thorugh a local library, pending author/copyright holder's permission.
This item is likely protected under Title 17 of the U.S. Copyright Law. Unless on a Creative Commons license, for uses protected by Copyright Law, contact the copyright holder or the author.
Subjects
Abstract
Machine Reading Comprehension has made rapid progress in recent times with datasets like SQuAD. Most of the existing question-answering models consider single paragraph for answering a query. This theses topic aims to perform a comprehensive analysis of the results obtained from a machine comprehension model and understand how well it is able to comprehend text and answer different types of questions.This includes experiments such as measuring how the increase in the number of passages with very similar information could affect the performance of a Question-Answering model. In this theses, a well-known Machine Comprehension model like BiDaF is considered and an exhaustive set of experiments are performed to measure the performance variation. Experiments show that the model's accuracy drops considerably when testing on a dataset that has multiple similar passages as compared to a dataset with single passage for each question. The model's results were analysed using the official evaluator from SQuAD. We believe that this analysis leads to a new research direction - to build QA systems with an ability to answer a question accurately irrespective of the length of the passages .