Minimally Supervised Instance Matching: An Alternate Approach

Conference paper

DOI: 10.1007/978-3-319-25639-9_14

Part of the Lecture Notes in Computer Science book series (LNCS, volume 9341)
Cite this paper as:
Kejriwal M., Miranker D.P. (2015) Minimally Supervised Instance Matching: An Alternate Approach. In: Gandon F., Guéret C., Villata S., Breslin J., Faron-Zucker C., Zimmermann A. (eds) The Semantic Web: ESWC 2015 Satellite Events. ESWC 2015. Lecture Notes in Computer Science, vol 9341. Springer, Cham

Abstract

Instance matching concerns identifying pairs of instances that refer to the same underlying entity. Current state-of-the-art instance matchers use machine learning methods. Supervised learning systems achieve good performance by training on significant amounts of manually labeled samples. To alleviate the labeling effort, this poster (The work presented herein is also being published as a full conference paper at ESWC 2015. This poster provides a more high-level overview and discusses supplemental experimental findings beyond the scope of the material in the full paper.) presents a minimally supervised instance matching approach that is able to deliver competitive performance using only 2 % training data. As a first step, a committee of base classifiers is trained in an ensemble setting using boosting. Iterative semi-supervised learning is used to improve the performance of the ensemble classifier even further, by self-training it on the most confident samples labeled in the current iteration. Empirical evaluations on real-world data show that, using a multilayer perceptron as base classifier, the system is able to achieve an average F-Measure that is within 2.5 % of that of state-of-the-art supervised systems.

Keywords

Instance Matching Boosting Semi-supervision Self-training 

Copyright information

© Springer International Publishing Switzerland 2015

Authors and Affiliations

  1. 1.University of Texas at AustinAustinUSA

Personalised recommendations