Article Structure
Abstract
This paper presents an unsupervised opinion analysis method for debate-side classification, i.e., recognizing which stance a person is taking in an online debate.
Introduction
This paper presents a method for debate-side classification, i.e., recognizing which stance a person is taking in an online debate posting.
The Debate Genre
In this section, we describe our debate data, and elaborate on characteristic ways of expressing opinions in this genre.
Method
We propose an unsupervised approach to classifying the stance of a post in a dual-topic debate.
Experiments
On http://www.convinceme.net, the html page for each debate contains side information for each post (sidel is blue in color and sideg is green).
Discussion
In this section, we discuss the results from the previous section and describe the sources of errors.
Related Work
Several researchers have worked on similar tasks.
Conclusions
This paper addresses challenges faced by opinion analysis in the debate genre.
Topics
F-measure
- #Correct’ Recall m and F-measure #guessed #relevantPage 6, “Experiments”
- Finally, both of the OpPr systems are better than both baselines in Accuracy as well as F-measure for all four debates.Page 6, “Experiments”
- The F-measure improves, on average, by 25 percentage points over the OpTopic system, and by 17 percentage points over the OpPMI system.Page 7, “Experiments”
- On average, there is a 3 percentage point improvement in Accuracy, 5 percentage point improvement in Precision and 5 percentage point improvement in F-measure due to the added concession information.Page 7, “Experiments”
See all papers in Proc. ACL 2009 that mention F-measure.
See all papers in Proc. ACL that mention F-measure.
Back to top.
Linear Programming
- We combine this knowledge with discourse information, and formulate the debate side classification task as an Integer Linear Programming problem.Page 1, “Abstract”
- This information is employed, in conjunction with discourse information, in an Integer Linear Programming (ILP) framework.Page 1, “Introduction”
- We formulate the problem of finding the overall side of the post as an Integer Linear Programming (ILP) problem.Page 5, “Method”
See all papers in Proc. ACL 2009 that mention Linear Programming.
See all papers in Proc. ACL that mention Linear Programming.
Back to top.
Semantic Relatedness
- We find semantic relatedness of each noun in the post with the two main topics of the debate by calculating the Pointwise Mutual Information (PMI) between the term and each topic over the entire web corpus.Page 6, “Experiments”
- We use the API provided by the Measures of Semantic Relatedness (MSR)4 engine for this purpose.Page 6, “Experiments”
- All of them are indeed semantically related to the domain.Page 7, “Discussion”
See all papers in Proc. ACL 2009 that mention Semantic Relatedness.
See all papers in Proc. ACL that mention Semantic Relatedness.
Back to top.
