Advanced Topics in NLP: Opinion Extraction and Sentiment Analysis
CS 3730/ISSP 3120 Natural Language Processing

Schedule    Home   

Course Description:

Opinion Extraction/Sentiment Analysis is one of the fastest growing areas of NLP. Opinion extraction, in its fullest, is an extremely complex task. Among them, the papers we will read will involve semantics, discourse, and pragmatics, as well as many techniques for complex NLP tasks such as topic modeling; joint extraction; integer linear programming (global inference); and various graph-based algorithms. We will read some relevant papers from linguistics and return to some classic AI papers on deep natural language understanding. We will also look at samples of annotated data sets, to get a feeling for what is available. We will cover subjectivity annotations; sense-aware subjectivity and sentiment analysis; product review mining; identifying aspects; identifying holders and targets; domain adaptation for sentiment classification; recognizing stances, arguments, and viewpoints; sentiment summarization; factuality/veridicalty; hedges and uncertainty; recognizing contextual polarity; connotation; compositional polarity calculation; scripts, plot units, and story understanding. Students will lead class discussions, participate in on-line and in-class discussions, and complete a group (2-3 people) course project.


Instructor Jan Wiebe
Office Hours After class or by appointment
Office 5409 Sennott Square
Phone (412) 624-9590


Day Time Place
Tuesday & Thursday 4:00-5:15pm 5313 Sennott Square


Courses in Artificial Intelligence and Natural Language Processing, OR permission of instructor.

Course readings:

A course schedule lists the papers or chapters to be discussed each day.

Grade basis:

Course Project 40% (proposal: 8%, presentation: 8%, report: 24%)

Commentaries on Readings: due by 2pm the day before

Class Presentations of Papers: leading 2 classes 20%, presenting 2 optional papers 10% 30%
Class participation (including exercises, for which there may be preparation) 5%

* Course project information is here.

* Commentaries on Readings: To help presenters direct class discussions, everyone is expected to write a set of short reactions for each assigned paper (in the form of NB annotations). You may find these guidelines (Mitzenmacher and Ramsey 2000) for reading a research paper useful (thanks to Dr. Hwa for the pointer). I also recommend reading "The Task of the Referee" by Alan Jay Smith (IEEE Computer 1990). For our class, the section entitled "Evaluating a research paper" is particularly relevant.

* Leading class discussions: Each student will lead two class discussions. You should prepare slides. Since everyone in the class will have read the assigned papers, there is no need to reiterate everything in the paper. You should present the claims of the paper, and address some interesting issues. What did we learn from the paper? What future work does the research suggest? What are your criticisms of the paper? Please use the commentaries on the readings posted by the other students for inspiration when preparing your presentation. The additional notes about subjective comments from above are relevant here as well.

Feel free to meet with me before your presentation. Bring a sketch of your presentation. I can answer background questions you may have, and help you figure out what to focus on in your presentation.

* Optional paper presentations: To keep the amont of mandatory reading reasonable, we will distribute individual knowledge among the group through supplementary presentations of optional papers per class. Specifically, for two classes you are not leading, you will be expected to read one paper that is mandatory for you but optional for the rest of the class. You should prepare slides. Your presentation should have two parts: an objective overview of the paper and a subjective reaction to the paper.

The objective overview of the paper should describe the following. You cannot include all the details that are in the paper; you will need to abstract away from some details.

The subjective reaction should include the same types of comments as the posted commentaries on the papers. Again, please refer to the additional notes about subjective comments. As above, feel free to meet with me before you presentation.

* Academic Integrity Academic integrity: if you include material or resources from any source in your presentations and/or projects, you must acknowledge it.
* Thanks to Dr. Litman for this idea.