The Role of Artificial Neural Network in Word Sense Disambiguation (WSD)—A Survey

Abstract

The chore of reducing ambiguity in distinct sense of words is known as word sense disambiguation (WSD). It’s a key area of research in computational linguistics to deal with the senses being assigned automatically to the words in a particular circumstances (Yuan et al. in Semi-supervised word sense disambiguation with neural models [1]). Human are naturally excellent at WSD and can tell the difference between senses utilized in the vocabulary through verbal language. On the contrary, computers have a hard time distinguishing between proper and incorrect meanings of words. Knowledge-based, Supervised, Semi-Supervised, and Unsupervised techniques have all been used to make progress in the problem of disambiguation. A better knowledge of human language will aid to computer performance in a variety of applications, including search and retrieval. The major goal of the paper is to describe a supervised neural network model that uses multiple strategies to maximize sense detection accuracy. The neural network’s input layer will be made up of binary valued nodes based on whether or not frequently recurring context words connected to the ambiguous phrases are present. Amount of nodes in the outer layer will be equal to the amount of senses of the ambiguous word.