Textual context analysis for information retrieval
Mark A. Stairmand · 1997
We &. 9dx? b WPbUkSU3ofQUESCOT, a program whkh analyacs and quantifies textual amtexta in doaunenta with reference to the WordNet databae.and hence awemina the dominance oftopics inadocument.0uranaly3ia ia baaedon previous work in lexical cobeaim, a feature of texts which contributes to theif functioning aa a cuhaent unit.The applications arc diverse, but all pertain to infurmatian retrieval.Whilatour resulta auggeat that QuEXO'ria notwellauitedto word aenae diaambiguatkm and text segmentation, our ~~~_u@3Q~~=~* Lxqonent produces premising results.We also used QUESM representationsto automaticallygenerate a resourceto aupplememt WerdNet, baaed on collocatkal relations between conceptsin a document collection.We conclude that QLrEsxrr is auitcd to applications based on document-level descriptions, where the degreeof granularityallowsinaccuraciesto be smoothedout.