How Skimle analyzes your documents

Skimle leverages the power of LLMs to analyze your documents. But instead of a sophisticated system prompt, it relies on a multi-step process that iterates through your data to create a structured analysis. This is how the process works:

Step 1: Extracting insights from your documents

Skimle utilizes LLM to analyze your data roughly one page at a time, identifying desired content along with direct quotes from the text.

Step 2: Combine and verify quotes within each document

Skimle then takes the results of the previous step and combines similar quotes within each document to create a coherent set of insights. At this stage, the accuracy of all quotes are verified in relation to the original document contents. If LLM-extracted quotes does not match original document contents, Skimle reanalyzes the document to find a verbatim quote.

Step 3: Form categories based on the extracted insights

Skimle uses LLMs to form categories based on the extracted insights, taking into account the original analysis choices. Then each insight is assigned to the most appropriate category.

Step 4: Form subcategories based on the extracted insights

Skimle uses LLMs to form subcategories based on the extracted insights, taking into account the original analysis choices. Then each insight is assigned to the most appropriate subcategory.

Step 5: Write category summaries based on the extracted insights

Finally, Skimle writes a summary for each category. Please note that the project spreadsheet is accessible already before the creation of summaries has been completed. However, you will only receive the email when the summaries are ready.

Here are the details for each specific form of analysis:

• Automatic thematic analysis
• Predefined categories
• Inductive analysis

Once the analysis is complete, you can explore the results through the spreadsheet view, categories view, or document view. See working with insights to learn how to edit and reorganise the extracted insights.