Reuse and Adaptation for Entity Resolution through Transfer Learning
This addresses the high human effort in feature engineering and training data creation for entity resolution in data integration, though it is incremental as it builds on existing transfer learning concepts.
The paper tackles the problem of training machine learning classifiers for entity resolution with limited or no training data by reusing and adapting training data from related domains, showing that their algorithms provide superior performance for a fixed training data size in experiments on 12 datasets from 5 domains.
Entity resolution (ER) is one of the fundamental problems in data integration, where machine learning (ML) based classifiers often provide the state-of-the-art results. Considerable human effort goes into feature engineering and training data creation. In this paper, we investigate a new problem: Given a dataset D_T for ER with limited or no training data, is it possible to train a good ML classifier on D_T by reusing and adapting the training data of dataset D_S from same or related domain? Our major contributions include (1) a distributed representation based approach to encode each tuple from diverse datasets into a standard feature space; (2) identification of common scenarios where the reuse of training data can be beneficial; and (3) five algorithms for handling each of the aforementioned scenarios. We have performed comprehensive experiments on 12 datasets from 5 different domains (publications, movies, songs, restaurants, and books). Our experiments show that our algorithms provide significant benefits such as providing superior performance for a fixed training data size.