This paper presents Strathclyde iSchool's (SiS) participation in the Technological Assisted Reviews in Empirical Medicine Task. For the ranking task, we explored two ways in which assistance to reviewers could be provided during the assessment process: (i) topic models, where we use Latent Dirichlet Allocation to identify topics within the set of retrieved documents, ranking documents by the topic most likely to be relevant and (ii) relevance feedback, where we use Rocchio's algorithm to update the query model for subsequent rounds of interaction. A third approach combines the topic and relevance feedback to quickly identify the relevant abstracts. For the thresholding task, we apply a score threshold, and exclude documents which did not exceed the threshold given BM25.
|Number of pages||5|
|Journal||CEUR Workshop Proceedings|
|Publication status||Published - 11 Sep 2017|
- Technological Assisted Reviews in Empirical Medicine Task
- relevance feedback
- information retrieval