Yes.It’s an example of a simple use case. You can use more complex libraries too to achieve the result. What use case are you struggling with llamaindex?
Sorry! It seems there is some issue with the voice! My voice has been cut off so often, and you hear some words are NOT properly articulated. It must have happened during the editing. We'll see if we can re-upload the video. :)
I have one question! What if the content of a page continues onto another page? For example, there is a clause or point called "2. Interpretation" and its elaboration is from page 2 to page 3. How do u deal with this?
That's a good question. One solution would be to have page overlap. I mean when summarizing each page (e.g. page3) you can add a few sentences or a paragraph from previous page (i.e. page2) to this current page and then summarize it. Nevertheless, even if there is no overlap, like in this video, I don't think that would significantly impact the result as the goal is to simply get a gist of each page. :)
You guys are great. Subscribed. Thanks for all you do! Why would you do this? These summaries could be much better keys to store in things like PrivateGPT than the discrete sections of text that it does by default. I think creating embeddings from these summaries would result in much better performance from PrivateGPT.
Thank you! I understand. As I said, the idea came from some folks asking about page by page summarization (that was their need). Otherwise there are several other approaches to summarizing a long document and further processing like embedding, etc and using for downstream tasks. :)
Based on the title, I thought this video was going to show me something new that solves the same typical uses cases that llamaindex solves. I would never think that summarizing a pdf page is an appropriate use case for using llamaindex.
The intention was to show that for many use cases there is not always a need to big libraries like llamaindex as they add plenty of dependencies to your code.
Cool Guys love your Content.
Yes.It’s an example of a simple use case. You can use more complex libraries too to achieve the result. What use case are you struggling with llamaindex?
Sorry! It seems there is some issue with the voice! My voice has been cut off so often, and you hear some words are NOT properly articulated. It must have happened during the editing. We'll see if we can re-upload the video. :)
I have one question! What if the content of a page continues onto another page? For example, there is a clause or point called "2. Interpretation" and its elaboration is from page 2 to page 3. How do u deal with this?
That's a good question. One solution would be to have page overlap. I mean when summarizing each page (e.g. page3) you can add a few sentences or a paragraph from previous page (i.e. page2) to this current page and then summarize it. Nevertheless, even if there is no overlap, like in this video, I don't think that would significantly impact the result as the goal is to simply get a gist of each page. :)
You guys are great. Subscribed. Thanks for all you do! Why would you do this? These summaries could be much better keys to store in things like PrivateGPT than the discrete sections of text that it does by default. I think creating embeddings from these summaries would result in much better performance from PrivateGPT.
Thank you! I understand. As I said, the idea came from some folks asking about page by page summarization (that was their need). Otherwise there are several other approaches to summarizing a long document and further processing like embedding, etc and using for downstream tasks. :)
Based on the title, I thought this video was going to show me something new that solves the same typical uses cases that llamaindex solves. I would never think that summarizing a pdf page is an appropriate use case for using llamaindex.
The intention was to show that for many use cases there is not always a need to big libraries like llamaindex as they add plenty of dependencies to your code.
Exactly my thought. The title was clickbaity. This video had nothing to do with langchain and llamaindex