Automatic Assamese Text Categorization Using WordNet

The increasing rate of Assamese text contents in digital format encourages us to generate a system that automatically categorizes them. This paper discusses a system that will perform the categorization of texts automatically based on the knowledge from Assamese WordNet. In WordNet, synset correspond to the words which implies the same concept and words having more than one sense in a particular text content is disambiguated in this approach. This approach extracts words occurred in the document and uses them to create a synset vector with union to its corresponding synsets from WordNet. To increase our performance, we present a process where it increases the weight of not only the terms but also that of the synsets corresponding to the terms. We later count the occurrences of the senses that help in disambiguation tasks by propagating the relationship between synsets. The proposed method outcomes with a reasonable state of art accuracy when measured with Precision and Recall.