Simulation of Within-Session Query Variations Using a Text Segmentation Approach

Abstract

Abstract. We propose a generative model for automatic query reformulations from an initial query using the underlying subtopic structure of top ranked retrieved documents. We address three types of query reformulations a) specialization; b) generalization; and c) drift. To test our model we generate the three reformulation variants starting with selected fields from the TREC-8 topics as the initial queries. We use manual judgments from multiple assessors to calculate the accuracy of the reformulated query variants and observe accuracies of 65%, 82% and 69% respectively for specialization, generalization and drift reformulations.

CLEF 15th Birthday

“…Language Tools and Resources continued the CLEF interest in multilinguality by dealing with tools, algorithm, and resources for multiple languages such as: lemmatizers, decompounders and normalizers for underrepresented resources using statistical approaches; named entity extraction, linking and clustering in cross-lingual settings; exploitation of multiple translation resources; and language-independent generation of document snippets [24,37,52,93,139,144,151] Tools, Systems, and Applications covered the design and development of various kinds of algorithms, systems, and applications focused on multilinguality and specialised domains such as: semantic discovery of resources in cloud-based systems; Arabic question answering; cross-language similarity search using thesauri; automatic annotation of bibliographic references; exploitation of visual context in multimedia translation; sub-topic mining in Web documents; exploiting relevance feedback for building tag-clouds in image search; query expansion for image retrieval; and transcript-based video retrieval [34,61,63,75,92,97,106,108,111,116,132,133,138,147,148,230,233,268,[271][272][273]275];…”

Utilizing sub-topical structure of documents for information retrieval

“…Experiments with TREC 8 topic titles as initial queries involving manual assessments to judge the quality of the reformulations, show that both specific and general queries can be generated with reasonable accuracy, with a greater inter-assessor agreement for general reformulations [4].…”