Skip to content

Problems

Word similarity

Given a pair of words, give a score to them that reflects semantic similarity. The higher the score, the more related are the words in meaning.

Models for this task are evaluated by computing the scores for some human-annotaded dataset. Here are some of those:

  • WordSim-353
  • SCWS (Huang, 2012)
  • RW (Luong, 2013)

Word Analogy

Consists in answer questions such as:

a is to be as c is to ______

For example, "Athens is to Greece as Berlin is to _____"

Named Entity Recognition

Consists in labeling words that correspond to some given entities. For example, label the words in the text that refers to: person, location and organization.

Datasets:

  • CoNLL-03
  • ACE Phase 2
  • ACE-2003