Natural language processing

A branch of AI that uses a combination of methods, including computational linguistics, to interpret human-generated text or audio data.

Source: AIEOG AI Lexicon (Feb 2026), adapted from NIST AI 100-1 and NSCAI Technical Glossary

Natural language processing (NLP) is the field of AI focused on enabling computers to understand, interpret, and generate human language. NLP bridges the gap between how humans communicate (unstructured text and speech) and how computers process information (structured data).

NLP encompasses a wide range of tasks: text classification, sentiment analysis, named entity recognition, language translation, text summarization, question answering, speech recognition, and text generation.

In financial services, NLP has become one of the most widely adopted AI technologies. Institutions use NLP to process regulatory documents, analyze customer communications, extract information from contracts, classify complaints, summarize meeting notes, and power chatbots.

NLP applications touch regulated processes, making governance essential:

• Regulatory text analysis. NLP tools that interpret regulations must be accurate. Errors could lead to compliance gaps.
• Customer communications. Chatbots must comply with disclosure requirements, UDAP/UDAAP standards, and fairness expectations.
• Complaint management. NLP-based complaint classification affects how complaints are categorized, prioritized, and resolved.
• Document review. NLP used for contract analysis or SAR narrative review must be reliable and validated.

1. Validate NLP accuracy. Test NLP systems for accuracy in your specific domain, including financial and regulatory terminology.
2. Assess bias in language processing. NLP models can exhibit bias based on language patterns in their training data.
3. Review customer-facing NLP. Chatbots and automated communications should be reviewed for compliance with disclosure and fairness requirements.
4. Include in AI governance. NLP deployments should be inventoried, risk-assessed, and monitored.
5. Maintain human oversight. For high-stakes NLP applications, maintain human review processes.
6. Monitor performance. Track NLP accuracy metrics and establish thresholds for investigation.

Large language model, Deep learning, Generative AI, Machine learning, Computer vision, Text/word embedding