NLP is used across many industries. Each industry below has 3 typical use cases, and 2 example companies together with the NLP algorithms they rely on.
Industries at a glance
| Industry | Use cases | Companies (NLP algorithms) |
|---|---|---|
| Healthcare | Clinical documentation, diagnostic assistance, drug discovery and development | IBM Watson Health (advanced ML, language models), Epic Systems (text parsing, extraction) |
| Finance and Banking | Financial report automation, fraud detection, customer service chatbots | Kensho (text analytics, sentiment analysis), Yseop (natural language generation, data interpretation) |
| Retail and E-Commerce | Sentiment analysis of product reviews, personalised recommendations, customer service chatbots | Clarabridge (sentiment analysis, deep learning), Algolia (semantic search, user intent understanding) |
| Legal | Document review and analysis, predictive analysis of case outcomes, compliance monitoring | LegalSifter (contract analysis, key term extraction), Lex Machina (text mining, predictive analytics) |
| Automotive | Voice-activated assistants, customer feedback analysis, automated vehicle information support | Nuance Dragon Drive (speech recognition, language understanding), SoundHound's Houndify (voice recognition, NLU) |
| Publishing | Content curation and management, reader engagement analysis, automated content generation | OpenAI's GPT-3 (deep learning, transformer models), Narrativa (natural language generation, data interpretation) |
| Education | Plagiarism detection, content personalisation, language learning tools | Turnitin (text comparison, similarity scoring), Grammarly (grammar checking, style suggestions) |
| Travel and Hospitality | Automated booking assistants, customer feedback analysis, language translation services | Amadeus (semantic search, natural language queries), Booking.com chatbot (chatbot technology, intent recognition) |
| Media and Entertainment | Content recommendation, automated subtitling and dubbing, sentiment analysis of social media | Netflix (content recommendation, user preference analysis), Valossa (video content analysis, metadata generation) |
| Government and Public Sector | Public sentiment analysis, automated public services, document and record management | Palantir Gotham (text extraction, entity recognition), Digital Reasoning (sentiment analysis, entity recognition) |
What to focus on
Know which NLP task serves which use case, for example sentiment analysis for reviews, NER for document records, text generation for reports. The company names are background only.
Common themes
The same 5 themes run through all the industries:
| Theme | What it means | Main industries |
|---|---|---|
| Insight extraction | Gaining insights from data | Healthcare, finance, retail |
| Customer engagement | Better service with chatbots | Banking, retail, hospitality |
| Content personalisation | Tailoring experiences to each user | Media, education, e-commerce |
| Automated compliance | Streamlining document processing | Legal, government |
| Enhanced communication | Better interaction with users | Travel, automotive |
Remembering tip
I C C A E: "I Can Chat And Explain". Insight, Customer engagement, Content personalisation, Automated compliance, Enhanced communication.
Using an LLM for these use cases
A single instruction-tuned LLM can cover several of the use cases above, only by changing the prompt:
| Industry | Prompt idea |
|---|---|
| Healthcare | Extract the key points of a doctor and patient conversation into a markdown table |
| Finance | Summarise the effect of a promotion on sales from a sales table, in bullet points |
| Retail | Suggest how to improve sales from a list of customer reviews |
| Publishing | Write a short introduction to a country in 4 sentences or less |
LLM output is not always accurate, so the results must be checked before use.
The next note looks at a full NLP system in one of these industries, the Fraud Detection Model.