An Iterative Similarity based Adaptation Technique for Cross-domain Text Classification

Proceedings of the 19th Conference on Computational Language Learning, pages 52–61, Beijing, China, July 30-31, 2015. c⃝2015 Association for Computational Linguistics An Iterative Similarity based Adaptation Technique for Cross Domain Text Classification Himanshu S. Bhatt Deepali Semwal Xerox Research Center India, Bengaluru, INDIA {Himanshu.Bhatt,Deepali.Semwal,Shourya.Roy}@xerox.com Shourya Roy Abstract Supervised machine learning classifica- tion algorithms assume both train and test data are sampled from the same domain or distribution. However, performance of the algorithms degrade for test data from different domain. Such cross do- main classification is arduous as features in the test domain may be different and absence of labeled data could further ex- acerbate the problem. This paper proposes an algorithm to adapt classification model by iteratively learning domain specific fea- tures from the unlabeled test data. More- over, this adaptation transpires in a simi- larity aware manner by integrating similar- ity between domains in the adaptation set- ting. Cross-domain classification exper- iments on different datasets, including a real world dataset, demonstrate efficacy of the proposed algorithm over state-of-the- art. 1 Introduction A fundamental assumption in supervised statis- tical learning is that training and test data are independently and identically distributed (i.i.d.) samples drawn from a distribution. Otherwise, good performance on test data cannot be guar- anteed even if the training error is low. In real life applications such as business process automa- tion, this assumption is often violated. While re- searchers develop new techniques and models for machine learning based automation of one or a handful business processes, large scale adoption is hindered owing to poor generalized performance. In our interactions with analytics software devel- opment teams, we noticed such pervasive diver- sity of learning tasks and associated inefficiency. Novel predictive analytics techniques on standard datasets (or limited client data) did not general- ize across different domains ( new products & ser- vices) and has limited applicability. Training mod- els from scratch for every new domain requires hu- man annotated labeled data which is expensive and time consuming, hence, not pragmatic. On the other hand, transfer learning techniques allow domains, tasks, and distributions used in training and testing to be different, but related. It works in contrast to traditional supervised tech- niques on the principle of transferring learned knowledge across domains. While transfer learn- ing has generally proved useful in reducing the labelled data requirement, brute force techniques suffer from the problem of negative transfer (Pan and Yang, 2010a). One cannot use transfer learn- ing as the proverbial hammer, but needs to gauge when to transfer and also how much to transfer. To address these issues, this paper proposes a domain adaptation technique for cross-domain text classification. In our setting for cross-domain classification, a classifier trained on one domain with sufficient labelled training data is applied to a different test domain with no labelled data. As shown in Figure 1, this paper proposes an iterative similarity based adaptation algorithm which starts with a shared feature representation of source and target domains. To adapt, it iteratively learns do- main specific …

Content truncated to 500 words