Skip to main content
Generating an answer to your question is being split up into two major steps:
1

Document (or Web) Search

Documents are generally divided in text segments. Each text segment will receive a value based on it’s content (semantic vector). This allows Genow to perform semantic searches based on the user input. As result of this search, Genow will hand over all relevant text segments to the LLM which will generate the answer in the second step.
2

Answer Generation

The LLM will receive the user input and all relevant text snippets and will then formulate an answer. For each statement, Genow will cite the given text snippets, if possible.

​
Remarks on Sub-Task-Splitting

It is often the case that a user enquiry can be divided into sub-tasks. In this case, Genow handles these sub-tasks separately (regarding search and the handover of relevant text passages) combines their results at the end to optimise the answer to the question. These sub-tasks can be displayed to the user by hovering the mouse pointer over the question. It should be noted that sometimes not all of the results of sub-tasks are included in the final answer, for example if they are discarded due to low relevance.