In this paper we describe our efforts for TREC Contextual Suggestion task. Our goal of this year is to evaluate the effectiveness of (1) an opinion-based method to model user profiles and rank candidate suggestions; and (2) a template-based summarization method that leverages the information from multiple resources to generate the description of a candidate suggestion. Existing approaches often uses the description or category information of the suggestions to estimate the user profile. However, one limitation of such approaches is that it may not be generalized well to other suggestions. To overcome this limitation, we propose to estimate the user profile based on the reviews of the candidate suggestions. Our preliminary results show that the proposed new method can improve the performance. Moreover, we explore a new way of generating summaries for the candidate suggestions. The main idea is that we want to leverage the information from multiple sources to generate a more comprehensive summary for a candidate suggestion. In particular, the summary includes the category information of the suggestion, meta-description of the website, reviews of the suggestion and the similar example suggestions that the user liked.
[1]
Hui Fang,et al.
An Exploraton of Ranking-Based Strategy for Contextual Suggestion
,
2012,
TREC.
[2]
ChengXiang Zhai,et al.
An exploration of axiomatic approaches to information retrieval
,
2005,
SIGIR '05.
[3]
Dan Klein,et al.
Feature-Rich Part-of-Speech Tagging with a Cyclic Dependency Network
,
2003,
NAACL.
[4]
Hui Fang,et al.
Opinion-based User Profile Modeling for Contextual Suggestions
,
2013,
ICTIR.
[5]
Jiawei Han,et al.
Opinosis: A Graph Based Approach to Abstractive Summarization of Highly Redundant Opinions
,
2010,
COLING.