The workshop lasts for two days: 1st day, 2nd day.
Monday March 12th 2012
Location: The Interchurch Center (corner of 120th and Claremont Ave) Room C & D on first floor. You will have to sign in at the front desk. Your names will be there.
8:30 - 9:00am Breakfast and Webex set up9:00 - 9:30am Introductions and Overarching goals of workshop (Mona & Eneko)
9:30 - 10:30am Discussion of What is STS? [Item A]
- 1. STS granularity (document, paragraph, sentence, phrase, word, subword)
- 2. Similarity between gradeable and binary characterizations
- 3. How do we characterize textual similarity? (lexical, syntactic, semantic, pragmatic levels of representation)
- 4. What are the different dimensions of semantic similarity?
- 5. How is semantic similarity different from semantic relatedness/inference?
- 6. How is STS different from textual entailment?
- 7. Desiderata for an STS system
- 8. Current approaches to textual similarity
11:00 - 11:30am Semeval STS task (Eneko & Dan)
- 1. Task design
- 2. Data sets
- 3. Amazon Mechanical Turk Experiments
- 4. Metrics and Initial evaluation
12:00 - 1:00pm Discussion of annotations
1:00 - 2:00pm Lunch
2:00 - 2:30pm Evaluation of STS (Mona & Eneko) [Item B]
- 1. intrinsic vs extrinsic considerations
- 2. Metrics
- 1. MT
- 2. MT evaluation
- 3. Summarization
- 4. Machine Reading
- 5. Watson Jeopardy
- 6. Distillation
- 7. Generation
- 8. Opinion Mining
- 9. Social Media Mining (trending)
- 10. Inference
4:30 - 5:30 How to create an STS blackbox? (Discussion, please send us your thoughts ahead of the workshop) [Item D]
- 1. What semantic components contribute to STS?
- 2. Component interface issues
Tuesday March 13th
Location: (Change of Location from previous day) Room 750, Interschool lab, CEPSR building on campus, entrance on 120th St, between Broadway and Amsterdam Ave.
8:30-9:00am Breakfast and webex set up9:00 - 9:30am Review of day 1 discussions (Mona & Eneko)
9:30 - 10:30 Discussion on how to create an STS system [Item E]
- 1. What components exist that are relevant for the task
- 2. what desired components are missing that would complete the STS pipeline?
11:00 - 12:30 Infrastructure desiderata [Item F]
- 1. Interoperability between components
- 2. What kind of platform would be of interest: UIMA, webservices, distributed architecture?
1:30-3:00 Discussion of Open issues [Item G]
- 1. Evaluation Revisited
- 2. issues of interpretability
- 3. towards a multilingual STS
- 4. Possibility of an empirical semantic framework
3:30-4:30pm Next steps and Wrapping up
- 1. Shared Task
- 2. Committee formation
- 3. Funding opportunities
- 4. Other issues