Improving Document Retrieval for Gene Queries
Author Information
Author(s): Aditya K. Sehgal, Padmini Srinivasan
Primary Institution: The University of Iowa
Hypothesis
Can different ranking strategies improve the retrieval of relevant documents for gene queries in MEDLINE?
Conclusion
The study successfully demonstrates that specific ranking strategies can significantly enhance the relevance of documents retrieved for gene queries, even in the presence of ambiguity.
Supporting Evidence
- Two of the three LocusLink-based strategies offer significant improvements.
- The best ranking strategy improves relevance by 15.9% to 17.7% depending on the baseline.
- 76% of topics have five or fewer relevant documents identified.
- SP and S strategies significantly outperform baseline strategies.
- Improvements are statistically significant at the 0.05 level.
Takeaway
This study shows that using special strategies can help find the right information about genes in a big library of documents, even when the names of the genes can be confusing.
Methodology
The study used five different ranking strategies to evaluate the effectiveness of document retrieval from MEDLINE for gene queries, focusing on the human genome.
Potential Biases
Potential biases may arise from the reliance on specific databases and the ambiguity of gene names.
Limitations
The study's findings may not generalize to all gene queries, especially those without summary information in LocusLink.
Statistical Information
Confidence Interval
95%
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website