Azure Cognitive Search Indexer: How to recognize graphs and colors in PDF pie charts?

Question

I build a Azure skillset of an Indexer with the following skills:

1. "@odata.type": "#Microsoft.Skills.Util.DocumentExtractionSkill", to extract information from PDF;
2. "@odata.type": "#Microsoft.Skills.Vision.OcrSkill", to read .jpeg image stamped in a PDF;
3. "@odata.type": "#Microsoft.Skills.Text.MergeSkill", to merge content with the OcrSkill output;
4. "@odata.type": "#Microsoft.Skills.Text.SplitSkill", to splits the merge text into overlapping segments for vectorization
5. "@odata.type": "#Microsoft.Skills.Text.AzureOpenAIEmbeddingSkill", to vectorize the output of Text.MergeSkill;
6. "@odata.type": "#Microsoft.Skills.Text.KeyPhraseExtractionSkill". (for semantic search) And I've built a Multi-Round Q&A on Your Data Azure promptflow for querying the documents in the Index, using gpt-4 and Ada.

Issue: As is show in the below picture, the promptflow doesn't recognize graphs or colors in pie charts.

Questions:

• How can I solve this problem?
• Can be enough add #Microsoft.Skills.Vision.ImageAnalysisSkill?

More general questions:

• What is the best strategy (i.e., skillset) to interpret and understand various types of graphs, such as histograms, line charts, plots, and other similar visualizations?

Every comment is welcome and encourage.