Syntax Parsers in Neural Network Architectures for Visual Question Answering

Student: Shaban Makhmud

Supervisor: Alexey Kovalev

Educational Programme: Applied Mathematics and Information Science (Bachelor)

Year of Graduation: 2021

Modern visual question answering systems use various methods to process both visual and linguistic information. One of the promising approaches is the neural-symbolic approach that uses symbolic representations of visual information alongside syntax parsing of the sentence to convert the question to a program that can be executed on a scene representation. While having certain advantages over other approaches, such as interpretability of the answer generation process, neural-symbolic systems lack the ability to generalize to new tasks and datasets. In this thesis, we propose modifications to an existing system, such as multi-head approach and high-dimensional vector scene representations in order to mitigate the described issue and promote the usage of syntax parsing-based approaches.

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses