Word Sense Disambiguation

Use vertex centrality measures to disambiguate senses of a word.

Thoughts

• Seems that adding both synset and related synsets' (all possible relationships) words is better

  • maybe it's useful to go deeper (more than one edge) (this seems the correct way)
    • depth 3-4 seems the best in terms of results/computation time

• Maybe working on examples inside words could give us better results

Possible future work

• Tries:

  • Don't create edges between distant words in a sentence, maybe a single sentence is composed by multiple, separate, logical sentences
  • Work on all relationships or weight differently the relationships by words
  • Work on SemCor3
  • Separate results based on NOUNS, VERBS ADJ etc..
  • Try with GLNS solver in fast mode and compare

• // Move all filesystem resources stream to getResourceAsAStream

• // Use indipendent dictionary (add Adapter for BabelNet, for example)