Revett: Academic Papers

Collection of academic papers I have authored.

View the Project on GitHub revett/Academic-Papers

An Investigation into Noisy Language in YouTube Comments

2014
Supervisor: Dr. Paul Rayson
Grade: 1st (100%)

As the popularity of YouTube, a social video sharing platform, increases so does the quantity of comments a video creator receives from their audience. This influx in viewership has seen a need for natural language processing (NLP) tools to better extract value from comments, however the effectiveness of these tools is bounded by the high level of variance found in YouTube comments (YTC). This paper presents an in-depth linguistic examination of the noise found within YTC, partnered with a novel application that allows for unsupervised normalisation tailored specifically to the unique set of requirements present in YTC. It sets out to prove the level of variance in comments is higher than that found in other forms of short unstructured text, such as Twitter or SMS, and through the use of normalisation can improve the accuracy of NLP techniques on a corpus of YTC.

Keywords: Natural Language Processing, Normalisation, Social Media, YouTube.

12,703 words.

Automatic Text Summarization of Emails on Mobile Platforms

2013
Supervisor: Dr. Paul Rayson
Grade: 1st (80%)

This paper presents a novel approach for email management on a mobile device, using automatic text summarization (ATS) techniques tailored to the unique set of requirements presented by summarising emails and the limitations of working on mobile platforms. It sets out to prove that the addition of email mobile email client (app) can improve user productivity.

Keywords: Natural Language Processing, Summarization, Email, YouTube.

6,990 words.