Word Sense Disambiguation (WSD)

Word Sense Disambiguation (WSD) is the task in Natural Language Processing (NLP) of determining which sense (meaning) of a word is used in a given context when the word has multiple meanings (i.e., it is polysemous).

Example:

Consider the word “bank” in the following sentences:

1. She sat by the bank of the river.

2. He went to the bank to deposit a check.

Here, “bank” has different meanings:

• In (1), it refers to the side of a river.

• In (2), it refers to a financial institution.

WSD helps algorithms decide the correct meaning based on surrounding words (context).

Why WSD is Important:

• Machine Translation: Translating ambiguous words correctly.

• Information Retrieval: Understanding query intent.

• Chatbots / Virtual Assistants: Providing accurate responses.

• Text Mining / Knowledge Extraction: Linking correct entities or facts.

Common Approaches to WSD:

1. Knowledge-based Methods:

   • Use lexical resources like WordNet.

   • Algorithms: Lesk algorithm, semantic similarity measures.

2. Supervised Learning:

   • Uses labeled corpora (sense-annotated text).

   • Algorithms: Decision Trees, SVMs, Neural Networks.

   • Limitation: Requires large, annotated datasets.

3. Unsupervised Learning:

   • Clusters word occurrences into different senses without labeled data.

   • Uses context similarity, co-occurrence statistics.

4. Neural / Deep Learning Methods:

   • Contextual embeddings (e.g., BERT, ELMo) have significantly improved WSD performance.

   • Fine-tuning models on WSD tasks gives state-of-the-art results.