Experiments of Opinion Analysis On Two Corpora MPQA and NTCIR-6.pdf

Experiments of Opinion Analysis On Two Corpora MPQA and NTCIR-6 Yaoyong Li, Kalina Bontcheva and Hamish Cunningham Department of Computer Science, The University of Sheffield 211 Portobello Street, Sheffield, S1 4DP, UK {yaoyong, kalina, hamish}@dcs.shef.ac.uk Abstract This paper describes the algorithms and linguistic features used in our participating system for the opin- ion analysis pilot task at NTCIR-6. It presents and dis- cusses the results of our system on the opinion analysis task. It also presents our experiments of opinion anal- ysis on the two corpora MPQA and NTCIR-6, by us- ing our learning based system. Our system was base on the SVM learning. It achieved state of the art re- sults on the MPQA corpus for the two problems, opin- ionated sentence recognition and opinion holder ex- traction. The results using the NTCIR-6 English cor- pus for both training and testing are certainly among the first ones. Our results on the opinionated sen- tence recognition sub-task of the NTCIR-6 were en- couraging. The results on the English evaluation of the NTCIR-6 opinion analysis task were obtained from the models learned from the MPQA corpus. The lower results on the NTCIR-6 opinion holder extraction sub- task, in comparison with those using each corpus for both training and testing, may possibly show that there exist substantial differences between the MPQA cor- pus and the NTCIR-6 English corpus. 1 Introduction With the rapid expansion of internet into almost ev- ery area of the society, we are able to express opinions about many interesting and/or useful topics (such as commercial products and cultural and political topics) through internet. Actually there are a huge amount of opinions available on-line (such as product reviews from customers, personal comments of on-line news, and on-line community forum and Blogs), and is in- creasing rapidly. Most of the on-line opinions are the free texts in different languages. Opinion analysis or mining would be very helpf