Automatic topic labelling and opinion summarisation
Saved in:
| Title: | Automatic topic labelling and opinion summarisation |
|---|---|
| Authors: | Barawi, Mohamad Hardyman |
| Committee Members: | Lin, Chenghua; Siddharthan, Advaith |
| Summary: | With the global increase in online tools such as online reviews and social media platforms, individuals all around the globe have changed their way of making a decision, interacting and sharing information. This change has led researchers to explore various interests in these invaluable sources of information by using a set of statistical methods such as the topic models to discover the hidden thematic structure in a large collection of documents. As an illustration, these models learn sets of topics from words frequency that co-occurs in the document collection automatically. Topics discovered are associated with relevant documents and often represent abstract themes, i.e. Politics or Sports. As a result, these characteristics make topic models a useful tool to extract interesting topics automatically from a mass amount of data such as reviews, online expressions and ratings. The main aim of this thesis is to focus on some fundamental challenges in inter-preting topic models, making them more useful and comprehensible to humans. First, we look at the problem of labelling the topics discovered by topic models. We propose novel methods for labelling the sentiment-bearing topics automatically and show that our approaches work better than previously proposed methods. Next, we propose methods for summarising opinions in a large document collection. Our opinion summarisation approach is scalable and also provides diverse and general summaries. Finally, we look at the problem of organising large collections of opinionated articles to visualising relevant information in articles. We develop a browsing system that allows users to navigate and identify relevant information in article collections by using the topics discovered by topic models as keywords. We also propose approaches to visualise topics discovered in a quantitative way, such as the heat maps. We also show the topics visualisation discovered in the different region using the reverse geo-coding approach. |
| URL: | https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.794105 |
| Database: | OpenDissertations |
| FullText | Text: Availability: 0 |
|---|---|
| Header | DbId: ddu DbLabel: OpenDissertations An: ddu.oai.ethos.bl.uk.794105 AccessLevel: 6 PubType: Dissertation/ Thesis PubTypeId: dissertation PreciseRelevancyScore: 0 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: Automatic topic labelling and opinion summarisation – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Barawi%2C+Mohamad+Hardyman%22">Barawi, Mohamad Hardyman</searchLink> – Name: Author Label: Committee Members Group: Au Data: <searchLink fieldCode="CO" term="%22Lin%2C+Chenghua%22">Lin, Chenghua</searchLink>; <searchLink fieldCode="CO" term="%22Siddharthan%2C+Advaith%22">Siddharthan, Advaith</searchLink> – Name: Abstract Label: Summary Group: Ab Data: With the global increase in online tools such as online reviews and social media platforms, individuals all around the globe have changed their way of making a decision, interacting and sharing information. This change has led researchers to explore various interests in these invaluable sources of information by using a set of statistical methods such as the topic models to discover the hidden thematic structure in a large collection of documents. As an illustration, these models learn sets of topics from words frequency that co-occurs in the document collection automatically. Topics discovered are associated with relevant documents and often represent abstract themes, i.e. Politics or Sports. As a result, these characteristics make topic models a useful tool to extract interesting topics automatically from a mass amount of data such as reviews, online expressions and ratings. The main aim of this thesis is to focus on some fundamental challenges in inter-preting topic models, making them more useful and comprehensible to humans. First, we look at the problem of labelling the topics discovered by topic models. We propose novel methods for labelling the sentiment-bearing topics automatically and show that our approaches work better than previously proposed methods. Next, we propose methods for summarising opinions in a large document collection. Our opinion summarisation approach is scalable and also provides diverse and general summaries. Finally, we look at the problem of organising large collections of opinionated articles to visualising relevant information in articles. We develop a browsing system that allows users to navigate and identify relevant information in article collections by using the topics discovered by topic models as keywords. We also propose approaches to visualise topics discovered in a quantitative way, such as the heat maps. We also show the topics visualisation discovered in the different region using the reverse geo-coding approach. – Name: URL Label: URL Group: URL Data: <link linkTarget="URL" linkTerm="https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.794105" linkWindow="_blank">https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.794105</link> |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=ddu&AN=ddu.oai.ethos.bl.uk.794105 |
| RecordInfo | BibRecord: BibEntity: Languages: – Code: eng Text: English Subjects: – SubjectFull: 004 Type: general – SubjectFull: Text data mining ; Probabilities Type: general Titles: – TitleFull: Automatic topic labelling and opinion summarisation Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Barawi, Mohamad Hardyman IsPartOfRelationships: – BibEntity: Dates: – D: 01 M: 01 Type: published Y: 2019 |
| ResultId | 1 |
