Towards a new perspective on context based citation index of research articles

Citation index measures the impact or quality of a research publication. Currently, all the standard journal citation indices are used to measure the impact of individual research article published in those journals and are based on the citation count, making them a pure quantitative measure. To address this, as our first contribution, we propose to assign weights to the edges of citation network using three context based quality factors: 1. Sentiment analysis of the text surrounding the citation in the citing article, 2. Self-citations, 3. Semantic similarity between citing and cited article. Prior approaches make use of PageRank algorithm to compute the citation scores. This being an iterative process is not essential for acyclic citation networks. As our second contribution, we propose a non-iterative graph traversal based approach, which uses the edge weights and the initial scores of the non-cited nodes to compute the citation indices by visiting the nodes in topologically sorted order. Experimental results depict that rankings of citation indices obtained by our approach are improved over the traditional citation count based ranks. Also, our rankings are similar to that of PageRank based methods; but, our algorithm is simpler and 70 % more efficient. Lastly, we propose a new model for future reference, which computes the citation indices based on solution of system of linear inequalities, in which human-expert’s judgment is modeled by suitable linear constraints.