Results (
Thai) 1:
[Copy]Copied!
An essential part of an expert- nding task, such as matchingreviewers to submitted papers, is the ability to model the ex-pertise of a person based on documents. We evaluate severalmeasures of the association between a document to be re-viewed and an author, represented by their previous papers.We compare language-model-based approaches with a noveltopic model, Author-Persona-Topic (APT). In this model,each author can write under one or more personas," whichare represented as independent distributions over hiddentopics. Examples of previous papers written by prospectivereviewers are gathered from the Rexa database, which ex-tracts and disambiguates author mentions from documentsgathered from the web. We evaluate the models using a re-viewer matching task based on human relevance judgmentsdetermining how well the expertise of proposed reviewersmatches a submission. We nd that the APT topic modeloutperforms the other models.
Being translated, please wait..
