Experts' Retrieval with Multiword-Enhanced Author Topic Model
Nikhil Johri, Dan Roth, Yuancheng Tu · 2010
In this paper, we propose a multiwordenhanced author topic model that clusters authors with similar interests and expertise, and apply it to an information retrieval system that returns a ranked list of authors related to a keyword. For example, we can retrieve Eugene Charniak via search for statistical parsing. The existing works on author topic modeling assume a “bag-of-words ” representation. However, many semantic atomic concepts are represented by multiwords in text documents. This paper presents a pre-computation step as a way to discover these multiwords in the corpus automatically and tags them in the termdocument matrix. The key advantage of this method is that it retains the simplicity and the computational efficiency of the unigram model. In addition to a qualitative evaluation, we evaluate the results by using the topic models as a component in a search engine. We exhibit improved retrieval scores when the documents are represented via sets of latent topics and authors. 1