Currently when we ask the chatbot to say where it found the information it references, and it does so with embedded paths to the source files, like shown in the screenshot below.
It would be nice if we found a way to make it easy for the user to quickly see the relevant chunks in the source files, especially the contextual information surrounding the referenced text. i.e. If it's a transcript of a conversation, it will be important to load the part of the transcript that was relevant in a viewer window. Maybe it should be called the "chunk" window.
If a chunk is interesting, it would be nice to also have a link back to the entire document, so it could be clicked and loaded in another tab.
Pursuant to that we probably need to:
Host the source docs somewhere, and make the path to the source into a hyperlink. Then the user just has to click it to be taken to the source document, at the part of the source document that is most relevant.
Note: To make this work we'd have to make the source documents like the transcripts publicly available, at least the way it is setup now.
Not quite sure how to open the document at the relevant spot.
Add a UI element to the interface that allows user to browse relevant source files like transcripts, webpages, documents, etc.
Option 2 will take some work / mocking up. For example - how to deal with multiple source documents?
Currently when we ask the chatbot to say where it found the information it references, and it does so with embedded paths to the source files, like shown in the screenshot below.
It would be nice if we found a way to make it easy for the user to quickly see the relevant chunks in the source files, especially the contextual information surrounding the referenced text. i.e. If it's a transcript of a conversation, it will be important to load the part of the transcript that was relevant in a viewer window. Maybe it should be called the "chunk" window. If a chunk is interesting, it would be nice to also have a link back to the entire document, so it could be clicked and loaded in another tab. Pursuant to that we probably need to:
Host the source docs somewhere, and make the path to the source into a hyperlink. Then the user just has to click it to be taken to the source document, at the part of the source document that is most relevant.
Add a UI element to the interface that allows user to browse relevant source files like transcripts, webpages, documents, etc.
Option 2 will take some work / mocking up. For example - how to deal with multiple source documents?