IBM Watson NLU is one of the specialized NLP (Natural Language Processing) APIs offered by major cloud providers, alongside Google Cloud Natural Language API and Amazon Comprehend. It provides various modules tailored for semantic analysis tasks. Its core modules include Entity Identification and Entity Sentiment, which inspect text for known entities and combine entity analysis with sentiment analysis to determine the positive or negative sentiment expressed about those entities within the text.
Another primary module is Sentiment Analysis, which analyzes text to identify the dominant emotional opinion, determining if the text’s attitude is generally positive, negative, or neutral. IBM Watson NLU also features modules for finding relationships (Relation Mapping) between entities and for emotion detection, which identifies specific emotions like joy, sadness, fear, anger, and disgust expressed in the text. The API uses Watson’s NLU capabilities for precise sentiment scoring across different use cases.
Additional capabilities provided by IBM Watson NLU include key phrases detection (keyword extraction), concepts extraction, text classification, metadata extraction, and semantic roles (advanced syntax analysis). When analyzing sentiment, IBM Watson NLU can indicate whether a text contains ‘mixed sentiment’. In comparative analyses of sentiment classification, IBM Watson NLU was found to be more precise on neutral samples, although with smaller recall, compared to AWS and Google Cloud NLP.