Classification of short text comments by sentiment and actionability for voice your view

William Simm, Maria Angela Ferrario, Scott Piao, Jon Whittle, Paul Rayson

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearchpeer-review

12 Citations (Scopus)

Abstract

Much has been documented in the literature on sentiment analysis and document summarisation. Much of this applies to long structured text in the form of documents and blog posts. With a shift in social media towards short commentary (see Facebook status updates and twitter tweets), the difference in comment structure may affect the accuracy of sentiment analysis techniques. From our VoiceYourView trial, we collected over 2000 individual short comments on the topic of library refurbishment, many of which are transcribed spoken comments. We have shown success in determining the theme of comments by looking for the first noun and using a semantic tag set to categorise this noun and hence the comment for short comments. Sentiment is a measure of how positive or negative a comment is, and the actionability metric is a measure of how actionable the comment is, i.e. how useful it is. This paper looks towards applying methods from the literature to our dataset with the aim of evaluating methods of automatic sentiment and actionability analysis for our VoiceyourView application data and has relevance to data from other applications, e.g. those from the social media. With many social media commentary applications moving to add speech platforms, VoiceYourView data may be representative of the type of free-form spoken text input to be expected in such platforms.

Original languageEnglish
Title of host publicationProceedings - SocialCom 2010: 2nd IEEE International Conference on Social Computing, PASSAT 2010: 2nd IEEE International Conference on Privacy, Security, Risk and Trust
Pages552-557
Number of pages6
DOIs
Publication statusPublished - 2010
Externally publishedYes
Event2nd IEEE International Conference on Social Computing, SocialCom 2010, 2nd IEEE International Conference on Privacy, Security, Risk and Trust, PASSAT 2010 - Minneapolis, MN, United States of America
Duration: 20 Aug 201022 Aug 2010

Conference

Conference2nd IEEE International Conference on Social Computing, SocialCom 2010, 2nd IEEE International Conference on Privacy, Security, Risk and Trust, PASSAT 2010
CountryUnited States of America
CityMinneapolis, MN
Period20/08/1022/08/10

Keywords

  • Actionability analysis
  • Sentiment analysis
  • Text classification

Cite this