What other inputs are there to Word Sense Disambiguation task?

Question

In Natural Language Processing (NLP), the Word Sense Disambiguation (WSD) task computationally determines the meaning(s) or sense(s) or concept(s) of a polysemous word given a sentence that the word appears in. For example:

• "Some was stupid enough to rob the central bank*."*
• "The river bank is full of stones"

Do anyone know on WSD performed in paragraph or document level?

Other than disambiguate senses/meaning from context words in one sentence, what other input could be introduce to perform WSD task? (I've seen WSD with images before, http://acl.ldc.upenn.edu/W/W03/W03-0601.pdf)

Answer 1

When you start talking about paragraphs or documents, often times rather than word senses you are really interested in topics. You could use a collection of word senses assigned to the words in a paragraph to help you figure out the topic, so the problems are connected, although it's probably better to think in terms of topics rather than senses (which are mostly associated with words or perhaps phrases).

Good luck, Ted