Résumé de séminaire

Résumé :

All academic fields of research share a method of recording and disseminating knowledge that allows researchers to review this knowledge and further its development. To enable efficient retrieval of this knowledge, the documents containing this knowledge are usually indexed with both a subject index and a citation index. A citation index consists of entries of authors' works together with corresponding lists of bibliographic descriptions of citing works.

Each citation link may serve one of many functions. For example, some citations give credit to related work, while others criticize previous work. Citation index searches have low precision partly because the bibliographic description of each citing document does not include the function (or purpose) of the citation.

Classifying a citation's function is a first step to a full interpretation of a citation. We present some new work that continues the process of developing an automatic citation function classifier. We have built on previous work whose success has been limited by the small number of classification categories and the impoverished linguistic information available to the automated classifier. This talk describes a more comprehensive classification scheme and an automated citation function classifier that uses techniques motivated by semantic grammars.