Unsupervised, Efficient and Semantic Expertise Retrieval

TL;DR

This paper presents an unsupervised, efficient, and semantic model for expert retrieval in online collections, matching supervised methods' performance without external evidence or relevance feedback.

Contribution

It introduces a novel unsupervised discriminative log-linear model that achieves high retrieval accuracy with low inference cost, outperforming traditional statistical and probabilistic models.

Findings

Achieves state-of-the-art performance in expert retrieval

Matches supervised methods using only textual evidence

Demonstrates semantic matching capabilities of the model

Abstract

We introduce an unsupervised discriminative model for the task of retrieving experts in online document collections. We exclusively employ textual evidence and avoid explicit feature engineering by learning distributed word representations in an unsupervised way. We compare our model to state-of-the-art unsupervised statistical vector space and probabilistic generative approaches. Our proposed log-linear model achieves the retrieval performance levels of state-of-the-art document-centric methods with the low inference cost of so-called profile-centric approaches. It yields a statistically significant improved ranking over vector space and generative models in most cases, matching the performance of supervised methods on various benchmarks. That is, by using solely text we can do as well as methods that work with external evidence and/or relevance feedback. A contrastive analysis of rankings produced by discriminative and generative approaches shows that they have complementary strengths due to the ability of the unsupervised discriminative model to perform semantic matching.