Chain-based recommendations
| dc.availability.bitstream | openaccess | |
| dc.contributor.advisor | Bridge, Derek G. | en |
| dc.contributor.advisor | Provan, Gregory | en |
| dc.contributor.author | Rana, Arpit | |
| dc.contributor.funder | Science Foundation Ireland | en |
| dc.date.accessioned | 2020-03-10T13:02:43Z | |
| dc.date.available | 2020-03-10T13:02:43Z | |
| dc.date.issued | 2019-10-14 | |
| dc.date.submitted | 2019-10-14 | |
| dc.description.abstract | Recommender systems are discovery tools. Typically, they infer a user's preferences from her behaviour and make personalized suggestions. They are one response to the overwhelming choices that the Web affords its users. Recent studies have shown that a user of a recommender system is more likely to be satisfied by the recommendations if the system provides explanations that allow the user to understand their rationale, and if the system allows the user to provide feedback on the recommendations to improve the next round of recommendations so that they take account of the user's ephemeral needs. The goal of this dissertation is to introduce a new recommendation framework that offers a better user experience, while giving quality recommendations. It works on content-based principles and addresses both the issues identified in the previous paragraph, i.e.\ explanations and recommendation feedback. We instantiate our framework to produce two recommendation engines, each focusing on one of the themes: (i) the role of explanations in producing recommendations, and (ii) helping users to articulate their ephemeral needs. For the first theme, we show how to unify recommendation and explanation to a greater degree than has been achieved hitherto. This results in an approach that enables the system to find relevant recommendations with explanations that have a high degree of both fidelity and interpretability. For the second theme, we show how to allow users to steer the recommendation process using a conversational recommender system. Our approach allows the user to reveal her short-term preferences and have them taken into account by the system and thus assists her in making a good decision efficiently. Early work on conversational recommender systems considers the case where the candidate items have structured descriptions (e.g.\ sets of attribute-value pairs). Our new approach works in the case where items have unstructured descriptions (e.g.\ sets of genres or tags). For each of the two themes, we describe the problem settings, the state-of-the-art, our system design and our experiment design. We evaluate each system using both offline analyses as well as user trials in a movie recommendation domain. We find that the proposed systems provide relevant recommendations that also have a high degree of serendipity, low popularity-bias and high diversity. | en |
| dc.description.status | Not peer reviewed | en |
| dc.format.mimetype | application/pdf | en |
| dc.identifier.citation | Raca, D. 2019. Chain-based recommendations. PhD Thesis, University College Cork. | en |
| dc.identifier.endpage | 181 | en |
| dc.identifier.uri | https://hdl.handle.net/10468/9743 | |
| dc.language.iso | en | 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 |
| dc.rights | © 2019, Arpit Rana. | en |
| dc.rights.uri | https://creativecommons.org/licenses/by-nc-nd/4.0/ | en |
| dc.subject | Recommendations | en |
| dc.subject | Explanations | en |
| dc.subject | Conversational recommender | en |
| dc.subject | User trials | en |
| dc.title | Chain-based recommendations | en |
| dc.type | Doctoral thesis | en |
| dc.type.qualificationlevel | Doctoral | en |
| dc.type.qualificationname | PhD - Doctor of Philosophy | en |
