%0 Journal Article %1 daud2009scientific %A Daud, Ali %A Shaikh, Muhammad Akram %A Rajpa, Altaf Hussain %D 2009 %J Research Journal of Engineering & Technology %K model paper recommender topic %N 2 %P 253--262 %T Scientific Reference Mining using Semantic Information through Topic Modeling %U http://www.researchgate.net/publication/215904235_Scientific_Reference_Mining_using_Semantic_Information_through_Topic_Modeling/file/1ea53d5db2d0d7e8776c557e49c95b39.pdf?origin=publication_detail %V 28 %X This paper addresses one time-consuming task in preparing scientific paper of finding appropriate citations (references). In the past this task is usually performed manually or by using links information of papers. Link information only used title of paper and ignored the semantic-based text information present in the paper contents. Due to overlaps between different fields e.g. in computer science only title of a paper cannot be real representative of different other hidden topics of that paper. We think it is necessary to model the semantic information present in papers to provide more appropriate citations by capturing hidden topics. In this paper, we address this issue by modeling citations on the basis of latent topics present in the papers. Latent topics can provide us semantic correlations present between the papers. We propose a topic modeling approach in which each citation of a paper is represented as a probability distribution over latent topics, and each latent topic is represented as a probability distribution over words of paper for that topic, which can provide us more appropriate citations for a given paper. Experimental results on citeceer corpus shows the effectiveness of proposed approach and detailed interpretation of results reveals interesting information about scientific recommendation.

@article{daud2009scientific, abstract = {This paper addresses one time-consuming task in preparing scientific paper of finding appropriate citations (references). In the past this task is usually performed manually or by using links information of papers. Link information only used title of paper and ignored the semantic-based text information present in the paper contents. Due to overlaps between different fields e.g. in computer science only title of a paper cannot be real representative of different other hidden topics of that paper. We think it is necessary to model the semantic information present in papers to provide more appropriate citations by capturing hidden topics. In this paper, we address this issue by modeling citations on the basis of latent topics present in the papers. Latent topics can provide us semantic correlations present between the papers. We propose a topic modeling approach in which each citation of a paper is represented as a probability distribution over latent topics, and each latent topic is represented as a probability distribution over words of paper for that topic, which can provide us more appropriate citations for a given paper. Experimental results on citeceer corpus shows the effectiveness of proposed approach and detailed interpretation of results reveals interesting information about scientific recommendation. }, added-at = {2014-01-15T09:23:07.000+0100}, author = {Daud, Ali and Shaikh, Muhammad Akram and Rajpa, Altaf Hussain}, biburl = {https://www.bibsonomy.org/bibtex/2bf3de37589eae72aa8366f5fba005891/jaeschke}, interhash = {95da80aa281679b82ae9d0cce053e4fb}, intrahash = {bf3de37589eae72aa8366f5fba005891}, journal = {Research Journal of Engineering & Technology}, keywords = {model paper recommender topic}, number = 2, pages = {253--262}, timestamp = {2014-07-28T15:57:31.000+0200}, title = {Scientific Reference Mining using Semantic Information through Topic Modeling}, url = {http://www.researchgate.net/publication/215904235_Scientific_Reference_Mining_using_Semantic_Information_through_Topic_Modeling/file/1ea53d5db2d0d7e8776c557e49c95b39.pdf?origin=publication_detail}, volume = 28, year = 2009 }