Semantically-Sensitive NLP Tasks

Term projects in Ling 575 A/B/D (Spring 2016) should each concern a semantically-sensitive NLP task, which furthermore has the following properties:

• Standard train/test datasets are available
• Evaluation metrics and scripts that implement them are available
• The language in question has a sufficiently large DELPH-IN grammar (English, or depending on the task German, Japanese, Spanish)
• There exists a baseline system to which you can add MRS-based features

Some inspiration can be found on the DelphinApplications page of the DELPH-IN wiki, but note that not all tasks approached there meet all of the criteria above.

This page collects links to tasks which I believe (are likely to) meet these criteria, but other tasks are also welcome!

• SemEval
  • SemEval-2015
  • SemEval-2014
  • SemEval-2013
  • SemEval-2012
• Biomedical
  • SemEval 2015 Task 6
  • i2b2 2014
  • BioNLP-ST 2013
  • i2b2 2012
  • i2b2 2010
  • CoNLL-2010 Shared Task
• Other
  • Shallow Discourse Parsing (CoNLL 2015 shard task) (NB: Also 2016, but the timing is bad for us.)
  • Coreference resolution (CoNLL 2011 shared task)
  • Summarization (TAC 2011)
  • STS: Semantic Textual Similarity
  • Recognizing Textual Entailment (RTE)