Citation Contexts

Citation contexts of a reference have many extensive applications in the area of computational linguistics and information retrieval. To use the citation contexts, it is common practice to employ the complete sentences in which the reference appears or the text in fixed window around the reference. Most of the time, the whole citation context does not particulary speaks about the target reference but only there is a sentence fragment which is the cause of making a reference. For instance, the citation causes like acknowledgement, explanation of a concept, critics, extension of a work or presence of another reference, all need a small amout of text in a sentence. In using the citation contexts, this observation needs to be taken into account as including unneeded text effects the performance of the application. In this paper, we attempt to identify only the text which actually describes a citation in a sentence long citation context. Particulary, different from other works, we use attention model based on deep learning approach for the solution of our problem which in not considered before for this type of work. Our experiments show that using limited amount of data, our results are comparable with existing techniques but can be improved more by using more data.