Data analysis examines, transforms, and interprets data to discover patterns, answer questions, and derive insights. With large language models, it can connect raw data with natural-language instructions and analytical outputs. section 1.0 section 1.1
Use data analysis across several data types and tasks:
- Classify, extract information from, and group text documents by related topics. section 1.2
- Ask questions about images, identify people from profiles, and analyze video content. section 1.2
- Transcribe or translate audio and query information through spoken questions. section 1.2
- Generate code that performs computations and analysis on tables and other structured data. section 1.2
Language models support two main workflows. Direct analysis places the data and task instructions in the prompt. Indirect analysis asks the model to translate an analysis request into code or another formal language, then delegates processing to specialized tools. section 1.4 section 2.0
A multimodal starting point
A short natural-language request can ask a model to connect data analysis with large language models by generating a poem and a corresponding picture. The example shows how instructions can produce coherent text and images, while also illustrating that data analysis can combine data, code, and meaning. section 1.1
Figure 1.1 Illustration by GPT-4o, connecting the topics “data analysis” and “large language models”
View the illustration connecting data analysis and large language models. section 1.1
