Using Noun Phrases for Local Analysis in Automatic Query Expansion

João Marcelo Azevedo Arcoverde, Maria das Graças Volpe Nunes, Wendel Scardua
2006 Conference and Labs of the Evaluation Forum  
Blind relevance feedback constitutes a widely used technique to improve the performance of IR systems. Its result is directly influenced by the correct choice of the potentially qualified features used to expand the initial query. It is well known the power of noun phrases in the role of descriptors with high discriminatory and informative potential. This work presents a local analysis technique to automatic query expansion through the use of noun phrases extracted from the pseudo-relevant set,
more » ... for the Ad-Hoc, monolingual (Portuguese) track. Even though the technique is language independent, specific resources for Portuguese were used to the noun phrases extraction through the use of Machine Learning techniques. In our experiments with a specific IR system, it has been observed an improvement in the results obtained over 38% of the topics, and also its depreciation over some others topics. However, this fact constitutes a clear evidence of how the use of NLP techniques can influence such processes, showing real possibilities for improving the presented technique.
dblp:conf/clef/ArcoverdeNS06 fatcat:rhskj4mr5nf7tnfqmkr6ozpnbq