Mining Interesting Trivia for Entities from Wikipedia

TL;DR

This paper presents a system that automatically mines and ranks interesting trivia about entities from Wikipedia pages to enhance user engagement, using a learned interestingness model trained on web data.

Contribution

It introduces a novel approach for extracting and ranking trivia from Wikipedia using a learned interestingness model based on linguistic and entity features.

Findings

System outperforms baselines in movies and celebrity domains

Rich feature set effectively surfaces interesting trivia

Qualitative analysis confirms feature importance in ranking

Abstract

Trivia is any fact about an entity, which is interesting due to any of the following characteristics - unusualness, uniqueness, unexpectedness or weirdness. Such interesting facts are provided in 'Did You Know?' section at many places. Although trivia are facts of little importance to be known, but we have presented their usage in user engagement purpose. Such fun facts generally spark intrigue and draws user to engage more with the entity, thereby promoting repeated engagement. The thesis has cited some case studies, which show the significant impact of using trivia for increasing user engagement or for wide publicity of the product/service. In this thesis, we propose a novel approach for mining entity trivia from their Wikipedia pages. Given an entity, our system extracts relevant sentences from its Wikipedia page and produces a list of sentences ranked based on their interestingness as trivia. At the heart of our system lies an interestingness ranker which learns the notion of interestingness, through a rich set of domain-independent linguistic and entity based features. Our ranking model is trained by leveraging existing user-generated trivia data available on the Web instead of creating new labeled data for movie domain. For other domains like sports, celebrities, countries etc. labeled data would have to be created as described in thesis. We evaluated our system on movies domain and celebrity domain, and observed that the system performs significantly better than the defined baselines. A thorough qualitative analysis of the results revealed that our engineered rich set of features indeed help in surfacing interesting trivia in the top ranks.