By now, you may have started to play around with AIs. You probably started with ChatGPT-3 or -4, or maybe even Google Bard (now known as Google Gemini). As you progress in skill with prompts and the differences between the AIs, you will notice that each of them has subtle differences in the way they respond to prompts and the way they handle context. The differences are important.
In this article we will begin to look at the reasons for that, and at understanding which of them match the kinds of problems you deal with in your work. As you progress working with AI of different types, you should study the differences and think about which AI is better matched to the kind (or kinds) of work you are doing.
Key tasks: Putting AI to work for you
Artificial intelligence has two sets of abilities: Natural Language Understanding (NLU) and Natural Language Generation (NLG).
Natural Language Understanding
NLU provides a system that enables AI to comprehend and interpret human language. NLU helps a system understand user queries and make accurate responses. NLU extracts meaning from text or speech (or from images in the case of multimodal AI).
The key tasks for NLU include:
- Entity recognition – Identifying specific entities within a sentence (names. dates, locations).
- Intent recognition – Understanding the user’s purpose behind a query.
- Sentiment analysis – Determining the emotional tone expressed in the text (positive, negative, neutral).
- Contextual understanding – Grasping the context in which a statement is made.
Natural Language Generation
The second set of abilities is Natural Language Generation. NLG generates human-like language based on structured data or instructions.
The key tasks in NLG include:
- Text summarization – Condensing lengthy content down to short summaries.
- Content creation – Creating articles, creative writing, and reports.
- Dialog creation – Formulating responses to chatbots or virtual assistants.
- Personalization – Tailoring content to individual users
Use cases for NLU and NLG
ChatGPT-4 and Google Gemini are two well-known AI applications. Each of them can use the NLG and NLU abilities but these abilities have different uses. The different uses are set up in the programming of the applications.
Natural Language Understanding (NLU)
NLU equips AI systems with the capability to comprehend and interpret human language. It involves extracting meaning from text or spoken input. NLU systems often rely on machine learning models, including deep learning architectures like recurrent neural networks (RNNs) and transformer-based models.
Key tasks within NLU include:
- Text Classification: Assigning predefined labels or categories to text (e.g., sentiment analysis, topic categorization).
- Named Entity Recognition (NER): Identifying entities such as names of people, places, organizations, dates, etc.
- Intent Recognition: Determining the user’s intention or purpose behind a query (e.g., understanding that “book a flight” implies a travel-related intent).
- Dependency Parsing: Analyzing the grammatical structure of sentences.
- Coreference Resolution: Resolving pronouns to their corresponding entities.
Natural Language Generation (NLG):
NLG focuses on creating human-like text based on structured data or prompts. NLG models can be rule-based, template-based, or powered by neural networks (such as recurrent neural networks or transformers).
Key tasks within NLG include:
- Text Summarization: Condensing lengthy text into concise summaries.
- Language Translation: Converting text from one language to another.
- Text Generation: Crafting original sentences, paragraphs, or entire documents.
- Dialogue Systems: Generating responses in conversational contexts.
- Storytelling: Creating narratives, poems, or fictional content.
In summary
NLU enables AI to understand human language, while NLG empowers AI to generate meaningful and coherent responses. These two sets of abilities work in tandem to enhance AI’s communication with users.