Incollection,

Incorporating User Feedback into Name Disambiguation of Scientific Cooperation Network

, , , , and .
Web-Age Information Management, volume 6897 of Lecture Notes in Computer Science, Springer, Berlin/Heidelberg, (2011)
DOI: 10.1007/978-3-642-23535-1_39

Abstract

In scientific cooperation network, ambiguous author names may occur due to the existence of multiple authors with the same name. Users of these networks usually want to know the exact author of a paper, whereas we do not have any unique identifier to distinguish them. In this paper, we focus ourselves on such problem, we propose a new method that incorporates user feedback into the model for name disambiguation of scientific cooperation network. Perceptron is used as the classifier. Two features and a constraint drawn from user feedback are incorporated into the perceptron to enhance the performance of name disambiguation. Specifically, we construct user feedback as a training stream, and refine the perceptron continuously. Experimental results show that the proposed algorithm can learn continuously and significantly outperforms the previous methods without introducing user interactions.

Tags

Users

  • @jaeschke
  • @dblp

Comments and Reviewsshow / hide

  • @jaeschke
    12 years ago
    The paper presents author name disambiguation as a classification task that is solved using a Perceptron. Several features are used as input for the Perceptron: organization, title similarity, homepage, etc. The User feedback is incorporated as constraint in a continuous training stream that adapts the weights of the Perceptron. On a comparison on 20 (ambigue) author names the method outperforms a baseline method based on graph clustering. The idea is interesting and well presented. Howevery, the language of the paper makes it wearisome to read and its related work section only mentions some related papers without any discussion about their relevance or a comparison to the presented approach. In particular, the topic of real-time learning or learning from streams is not tackled, although the approach falls into that category.
Please log in to take part in the discussion (add own reviews or comments).