Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multiple Models for Recommending Temporal Aspects of Entities

Tu Nguyen, Nattiya Kanhabua, Wolfgang Nejdl

TL;DR

This work tackles temporal entity aspect recommendation by explicitly modeling event-driven dynamics through time and event type. It introduces an event-centric ensemble ranking framework that combines multiple time- and type-dependent models, an aspect-extraction pipeline from co-click query-URL data, and a cascaded time/type identification module, all guided by a unified loss. The approach leverages both salience and short-term features drawn from Wikipedia and query logs, and validates with real AOL/Google logs and Wikipedia alignment, showing robust improvements over baselines in both classification and ranking tasks. The results indicate that balancing long-term salience with short-term recency is crucial for accurately surfacing temporally relevant aspects of event-driven entities, with potential impact on search experience and relevance modeling.

Abstract

Entity aspect recommendation is an emerging task in semantic search that helps users discover serendipitous and prominent information with respect to an entity, of which salience (e.g., popularity) is the most important factor in previous work. However, entity aspects are temporally dynamic and often driven by events happening over time. For such cases, aspect suggestion based solely on salience features can give unsatisfactory results, for two reasons. First, salience is often accumulated over a long time period and does not account for recency. Second, many aspects related to an event entity are strongly time-dependent. In this paper, we study the task of temporal aspect recommendation for a given entity, which aims at recommending the most relevant aspects and takes into account time in order to improve search experience. We propose a novel event-centric ensemble ranking method that learns from multiple time and type-dependent models and dynamically trades off salience and recency characteristics. Through extensive experiments on real-world query logs, we demonstrate that our method is robust and achieves better effectiveness than competitive baselines.

Multiple Models for Recommending Temporal Aspects of Entities

TL;DR

This work tackles temporal entity aspect recommendation by explicitly modeling event-driven dynamics through time and event type. It introduces an event-centric ensemble ranking framework that combines multiple time- and type-dependent models, an aspect-extraction pipeline from co-click query-URL data, and a cascaded time/type identification module, all guided by a unified loss. The approach leverages both salience and short-term features drawn from Wikipedia and query logs, and validates with real AOL/Google logs and Wikipedia alignment, showing robust improvements over baselines in both classification and ranking tasks. The results indicate that balancing long-term salience with short-term recency is crucial for accurately surfacing temporally relevant aspects of event-driven entities, with potential impact on search experience and relevance modeling.

Abstract

Entity aspect recommendation is an emerging task in semantic search that helps users discover serendipitous and prominent information with respect to an entity, of which salience (e.g., popularity) is the most important factor in previous work. However, entity aspects are temporally dynamic and often driven by events happening over time. For such cases, aspect suggestion based solely on salience features can give unsatisfactory results, for two reasons. First, salience is often accumulated over a long time period and does not account for recency. Second, many aspects related to an event entity are strongly time-dependent. In this paper, we study the task of temporal aspect recommendation for a given entity, which aims at recommending the most relevant aspects and takes into account time in order to improve search experience. We propose a novel event-centric ensemble ranking method that learns from multiple time and type-dependent models and dynamically trades off salience and recency characteristics. Through extensive experiments on real-world query logs, we demonstrate that our method is robust and achieves better effectiveness than competitive baselines.

Paper Structure

This paper contains 17 sections, 2 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Background and Problem Statement
  4. Preliminaries
  5. Problem Definitions
  6. Our Approach
  7. Aspect Extraction
  8. Time and Type Identification
  9. Time and Type-Dependent Ranking Models
  10. Ranking Features
  11. Salience features
  12. Short-term interest features
  13. Evaluation
  14. Experimental Setting
  15. Cascaded Classification Evaluation
  16. ...and 2 more sections

Figures (4)

  • Figure 1: [Screenshot] Recommendation generated by a commercial search engine for academy awards 2017 and australia open 2017, submitted on March $31^{th}$, 2017, on a clean history browser.
  • Figure 2: Dynamic aspect behaviors for entity ncaa in AOL and Google.
  • Figure 3: Learning time and type-specific ranking models.
  • Figure 4: Performance of different models for event entities of different types.