Introduction

Expertise matching, aiming to find the alignment between experts and queries, is a common problem in many applications such as conference paper-reviewer assignment, product-reviewer alignment, and product-endorser matching. Most of existing methods for this problem focus on independently finding a set of experts for each query by using, e.g., an information retrieval method. However, in real-world systems, various constraints usually should be considered.  Straightforward, we hope that the human experts who are assigned to answer a question have the specific expertise related to the question. But it is obviously insufficient. An ideal matching system should also consider various constraints in the real world, for example, an expert can only answer a certain number of questions (load balance); as the authoritative degree of different experts would vary largely, it is desirable that each question can be answered/reviewed by at least one senior expert (authority balance); the combined expertise of all assigned experts would cover all aspects of questions (topic coverage). In this work, we explore such an approach by formulating the expertise matching problem in a constrain-based optimization framework. The proposed approach has been evaluated on two different genres of expertise matching problem.

Data Sets

Conference Paper-Reviewer Assignment Data Set

The paper-reviewer data set consists of 338 papers and 354 reviewers. The reviewers are program committee members of KDD'09 and the 338 papers are those published on KDD'08, KDD'09, and ICDM'09. 

The data can be downloaded here.

Course-Teacher Assignment Data Set

In the course-teacher assignment, we manually crawled graduate courses from the department of Computer Science (CS) of four top universities, namely CMU, UIUC, Stanford and MIT. In total, there are 609 graduate courses from fall semester in 2008 to 2010 spring, and course is instructed by 1 to 3 teachers.

The data can be downloaded here.