TUW @ TREC Clinical Decision Support Track 2015
نویسندگان
چکیده
In this document, we describe the participation of Vienna University of Technology (TUW in German) in both tasks A and B of the TREC Clinical Decision Support (TREC-CDS) Track 2015. Based on the 2014 data, we concluded that query expansion with PRF resulted in large improvements over a BM25 baseline. Thus, we investigate a manner to add an intermediary layer based on a subset of the concepts annotated by MetaMap. This acts as a way to add weight to relevant concepts in the query and slightly expand the query with the preferred name of relevant concepts, before performing the query expansion with PRF. For TREC-CDS 2014, our method could reach a precision at 10 (P@10) of 0.40, while the best result of that year was 0.39. For 2015, we could reach a P@10 of 0.41 using the intermediary layer proposed, a small improvement over our baseline of P@10 of 0.39 when using only the original query expanded with PRF.
منابع مشابه
TUW @ TREC Clinical Decision Support Track
This document describes the participation of Vienna University of Technology in the TREC Clinical Decision Support Track 2014. Four different search models are investigated, as well as different strategies to index the corpus and to extract the most relevant information from the topics. Our results conclude that BM25 and Vector Space Model had similar performance for P@10 and inferred NDCG.
متن کاملTUW at the First Total Recall Track
Abstract. For the first participation in the TREC Total Recall track, we set out to try some basic changes to the baseline provided by the organisers. Namely, the weighting scheme, the use of stopwords, and the number of learners that contribute to the decision of which documents to ask the virtual assessor to review. We observed that the baseline was extremely strong and none of the runs signi...
متن کاملDUTH at TREC 2015 Clinical Decision Support Track
In this report we give an overview of our participation in the TREC 2015 Clinical Decision Support Track. We present two approaches for pre-processing and indexing of the open-access PubMed articles, and four methods for query construction which are applied to the previous two approaches. Regarding pre-processing, our main assumption is that only particular medical study designs are appropriate...
متن کاملWSU-IR at TREC 2015 Clinical Decision Support Track: Joint Weighting of Explicit and Latent Medical Query Concepts from Diverse Sources
This paper describes participation of WSU-IR group in TREC 2015 Clinical Decision Support (CDS) track. We present a Markov Random Fields-based retrieval model and an optimization method for jointly weighting statistical and semantic unigram, bigram and multi-phrase concepts from the query and PRF documents as well as three specific instantiations of this model that we used to obtain the runs su...
متن کاملEMSE at TREC 2015 Clinical Decision Support Track
This paper describes the participation of the EMSE team at the clinical decision support track of TREC 2015 (Task A). Our team submitted three automatic runs based only on the summary field. The baseline run uses the summary field as a query and the query likelihood retrieval model to match articles. Other runs explore different approaches to expand the summary field: RM3, LSI with pseudo relev...
متن کامل