Use of topical and temporal profiles and their hybridisation for content-based recommendation

TL;DR

The paper addresses building stronger content-based recommender systems by fusing temporality and topicality into item/user profiles. It proposes two hybrid strategies—TopTemp and TempTop—and evaluates them on publication venue recommendation using time-sliced topical subprofiles produced by LDA within temporal partitions, with a Lucene Language Model and rank fusion. Results show that temporally partitioned, locally topic-discovered profiles (TempTop) outperform baselines, with the best setting at $k=110$ topics and a $2Sqrt$ decay, highlighting the value of time-aware topical modelling. The findings suggest that hybridising topical and temporal signals can substantially improve recommendation quality and offer directions for temporal topic models and explainability in CBRS.

Abstract

In the context of content-based recommender systems, the aim of this paper is to determine how better profiles can be built and how these affect the recommendation process based on the incorporation of temporality, i.e. the inclusion of time in the recommendation process, and topicality, i.e. the representation of texts associated with users and items using topics and their combination. The main contribution of the paper is to present two different ways of hybridising these two dimensions and to evaluate and compare them with other alternatives.

Figures (2)

• Figure 1: Building monolithic and atomic (sub)profiles
• Figure 2: Building topical profiles