In general, different levels of knowledge are used for disambiguation. In this paper, word disambiguation is achieved with only two knowledge features (bigram). Bigram features of input (ambiguous words) as well as cases for solving the ambiguity are represented as vector of size 1 x 2. Vector values are taken out from weights assigned for part of speech of the knowledge features. In this paper ten different distance metric functions are empirically analyzed for improving the accuracy performance of word disambiguation with minimal knowledge sources. Four different classifiers such as K-Nearest Neighbouring, Bayes, Hidden Markov Model and Artificial Neural Network are used for extracting correct sense of the ambiguous word from the selected minimal distance cases. In this paper, a long sentence was taken to show the performance of disambiguation process. The projected result shows that Baye's classifier featured with pre-bigram produced 85.71% (recognized twelve ambiguous words out of fourteen) of disambiguation accuracy with spearman function (F8), with comparatively least disambiguation time of 0.071 seconds.

Key words: Word Sense Disambiguation, distance metric functions, bigram, K-Nearest Neighbour, Baye's, Hidden Markov Model and Artificial Neural Network.