As with most machine learing problems, lack of labelled data data is common problem in building disambiguation models. Thus ,we need unsupervised techniques to solve the same problem.
A popular unsupervised algorithm usedfor word sense disambiguation is the Lesk algorithm.
There are various ways in which you can use the lesk algorithm. Apart from what the professor has discussed can just take the defintions corresponding to the different senses of ambiguous word and see which definition overlaps maximum with the neighbouring words of the ambiguous word. The sense which has the maximum overlap with the surroundings word is then chosen as the `correct sense.
Let,s now use lesk algorithm to disambiguate the word bank in a text.
Recall that WordNet has a network of synonyms called synset for individual words.