[Announcements]
[General Information]
[Syllabus]
[Top]
Description
Computatioal Models of Social Meaning is a seminar in Natural Language Processing, focusing on computational methods for extracting social
and interactional meaning, mainly from conversational text (and speech). Topics include detection of speaker's sentiments, emotions, opinions and beliefs,
sarcasm, deception, persuation, perspective, power and influence, politeness, and personality. Analysis of meaning-bearing characteristics of the
speaker and topic, including text, discourse, prosodic and other cues.
Prerequisites
- COMS 3133/4/7/9 (Data Structures) or equivalent programming ability in at least one systems or scripting language (C++, Java, Python)
- smara [who is at] ccls [dot] columbia [dot] edu
- Office Hours: Thursdays 3:00pm-4:00pm
Office: Center for Computational Learning System, Interchurch Center (Room 850)
TA: TBD
- Email: TBD
- Office Hours: TBD
Lectures
- Thursday 4:10-6:00, MUD 644
The class consists of lectures and discussion of research papers led by students. To facilitate the discussion of research articles
on a particular topic each week, the topic is introduced the previous week. Thus, the class structure will be as follows: lecture on new topic and
discuss of papers on topic introduced in the previous week.
Grade Breakdown
- 10% Data Analysis Assignments
There will be 3 data analysis assignments that will require relatively small amount of work to learn
about a data source. You will be asked to answer one or two questions and to notice some interesting
points about the data source. The grading for this will be Excellent/Good/Insufficient. The data
analysis assignments will be due before class, on Courseworks site. Late assignments will not be accepted, except emergency situations.
- 30% Discussion of Papers
Each student will do a critical discussion of ONE of the research articles proposed for reading.
Students will prepare a brief presentation of the paper (Input? Output? What does it do? How is evaluated? How does it relate to other work in the field) followed by leading a critical
discussion on key positive and negative aspects, availability of code and datasets. You can consider as inspiration guidelines given to reviewers of conference papers Review Form ). But all the time explain your ratings.
NOTE: Every week every student needs to write a short paragraph about each of the papers given for discussion (there will be 2-3 papers), and write what you liked and what you did not like about the paper (can be one positive aspect and one negative). This will allow a good discussion in class. These summaries are not graded BUT they are
REQUIRED. Missing a significant number of these write-up will result in a lower grade!!
- 60% Final Project
Design and implement a software
product of appropriate scope and complexity given the time
constraints. Project should be related to one topic discussed in class or related issues.
Prepare a literature review (by mid semester), give a class presentation and a paper write-up
describing the methodology and results. Groups of 2-3 are welcome as
is individual work (complexity expectations being tuned accordingly). Split up of grades will be
10% for literature review, 5% class presentations, 45% for project and final paper.
[Top]
Schedule is tentative and highly subject to change.
Date |
Topics |
Readings |
Handouts & Due dates |
1/22 |
Class Introduction |
|
|
1/29 |
Sentiment Analysis: Sentiment Lexicons |
General reading for lecture
|
|
2/5 |
Sentiment Analysis |
Discussion papers on Sentiment Lexicons:
|
Data Analysis Assignment 1 Out |
2/12 |
Emotion/Mood Detection |
Discussion papers on Sentiment Analysis:
|
Data Analysis Assignment 1 due |
2/19 |
Belief Analysis and Hedging |
Suggested General Readings relevant to guest lecture:
Discussion papers on Emotion/Mood Detection:
|
Project: Project ideas due Feb 17, 11:59pm (note unusual time) |
2/26 |
Irony and Sarcasm Detection |
Discussion papers on Belief Analysis and Hedging:
|
|
3/5 |
Agreement/Disagreement
|
Discussion papers on Sarcasm/Irony Detection:
- [SI1]Riloff, Ellen, Qadir, Ashequl, Surve, Prafulla, Silva, Lalindra De, Gilbert, Nathan and Huang, Ruihong. Sarcasm as Contrast between a Positive Sentiment and Negative Situation Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 704–714. [Keerthana Kumar]
- [SI2] Tepperman, Joseph, David Traum, and Shrikanth Narayanan. 2006. Yeah right: Sarcasm recognition for spoken dialogue systems . Proceedings of InterSpeech 2006. Harish Viswanathan
|
Project Proposal due |
3/12 |
Perspective/Ideology |
Discussion Papers on Agreement/Disagreement:
- [AD1] Murakami, Akiko and Rudy Raymond. 2010. Support or oppose?: Classifying positions in online debates from reply activities and opinion expressions. Proceedings of the 23rd International Conference on Computational Linguistics: Posters, pp. 869--875.
- [AD2] Bender, Emily M, Jonathan T Morgan, Meghan Oxley, Mark Zachry, Brian Hutchinson, Alex Marin, Bin Zhang, and Mari Ostendorf. 2011. Annotating social acts: Authority claims and alignment moves in Wikipedia talk pages. Proceedings of the ACL-HLT Workshop on Language in Social Media, pp. 48--57. [Rishina Tah]
|
|
3/19 |
No Lecture: Spring Break |
3/26 |
Social Power
|
Discussion Papers on Perspective/Ideology:
|
|
4/2 |
Deception |
Discussion Papers on Social Power:
|
Project: Literature Review Due |
4/9 |
Extracting Social Networks |
Discussion Papers on Deception:
|
|
4/16 |
Personality and Interpersonal Stance |
Discussion Papers on Extracting Social Networks:
|
|
4/23 |
TBD |
Discussion Papers on Personality and Interpersonal Stance:
|
|
4/30 |
Final Project Presentations (start time 4:00pm) |
Project: Presentations due;
Project: Final paper due May 8 11:59pm
|
smara [who is at] ccls [dot] columbia [dot] edu
Design adapted from David Elson's site design
|