Rich-context: an unsupervised context-driven recommender system based on user reviews

Show simple item record

dc.contributor.advisor Bridge, Derek G. en
dc.contributor.author Peña, Francisco J.
dc.date.accessioned 2019-05-20T15:29:39Z
dc.date.available 2019-05-20T15:29:39Z
dc.date.issued 2019
dc.date.submitted 2019
dc.identifier.citation Peña, F. J. 2019. Rich-context: an unsupervised context-driven recommender system based on user reviews. PhD Thesis, University College Cork. en
dc.identifier.endpage 162 en
dc.identifier.uri http://hdl.handle.net/10468/7942
dc.description.abstract In the digital era, users have, more than at any point in history, a large amount of products or services to choose from. Recommender systems help to overcome this problem by suggesting which products or services to consume based on the users’ past behavior along with additional information about users, products and services. For the most part, however, they have done this in a context-insensitive way. Yet it is clear that a knowledge of the context in which a user intends to consume a product or service is desirable to ensure that the recommended products and services match the intended context. Context-Aware Recommender Systems incorporate contextual information in order to make recommendations that take into account both the users’ preferences and his/her current contextual situation. However, there are problems in building such recommender systems: most of the time, contextual information is not available; when it is available, it is limited to a very small number of predefined variables; human intervention is required to define what contextual variables there will be; and many other possible contextual variables are left out. Given that users sometimes express their context while writing reviews about consumption of a product or service, user-generated reviews present themselves as a source to extract contextual information. In this thesis, we research the problem of how to make contextual recommendations without the need to pre-define what context is. We mine the contextual information from user-generated reviews in an unsupervised way. We present Rich-Context, an unsupervised context-driven recommender system that extracts contextual information out of reviews in order to make recommendations. By using natural language processing techniques such as part-of-speech tagging, text classification and topic modeling, Rich-Context is able to successfully extract the contextual information out of the reviews. Experimental results on multiple publicly-available, sparse, real-world datasets from different domains show that Rich-Context has better performance in both rating and ranking prediction tasks compared to several state-of-the-art algorithms, including six Context-Aware Recommender Systems, with the advantage that no contextual keywords or other variables need to be pre-defined. Additionally, since contextual recommendations are often cold-start recommendations, we performed experiments with users that had no previous ratings, again outperforming all of the state-of-the-art recommenders. en
dc.format.mimetype application/pdf en
dc.language.iso en en
dc.publisher University College Cork en
dc.rights © 2019, Francisco J. Peña. en
dc.rights.uri http://creativecommons.org/licenses/by-nc-nd/3.0/ en
dc.subject Recommender systems en
dc.subject Machine learning en
dc.subject Natural language processing en
dc.subject Unsupervised learning en
dc.subject Topic modeling en
dc.title Rich-context: an unsupervised context-driven recommender system based on user reviews en
dc.type Doctoral thesis en
dc.type.qualificationlevel Doctoral en
dc.type.qualificationname PhD en
dc.internal.availability Full text available en
dc.check.info Not applicable en
dc.description.version Accepted Version
dc.contributor.funder Science Foundation Ireland en
dc.description.status Not peer reviewed en
dc.internal.school Computer Science and Information Technology en
dc.check.type No Embargo Required
dc.check.reason Not applicable en
dc.check.opt-out Not applicable en
dc.thesis.opt-out false
dc.check.embargoformat Embargo not applicable (If you have not submitted an e-thesis or do not want to request an embargo) en
ucc.workflow.supervisor d.bridge@ucc.ie
dc.internal.conferring Summer 2019 en
dc.internal.ricu Insight - Centre for Data Analytics en
dc.relation.project info:eu-repo/grantAgreement/SFI/SFI Research Centres/12/RC/2289/IE/INSIGHT - Irelands Big Data and Analytics Research Centre/ en


Files in this item

This item appears in the following Collection(s)

Show simple item record

© 2019, Francisco J. Peña. Except where otherwise noted, this item's license is described as © 2019, Francisco J. Peña.
This website uses cookies. By using this website, you consent to the use of cookies in accordance with the UCC Privacy and Cookies Statement. For more information about cookies and how you can disable them, visit our Privacy and Cookies statement