Improving Postgres keyword search to avoid zero results(PgvectorKeywordRetriever) #255
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
|
thanks @maylad31 for the notebook! I will review it this week |
|
If anything else is needed from my side, please let me know. But I guess this could be a helpful notebook for people. Many use postgres and when all other things of the pipeline are working and so easy to set up, zero results for keyword search should not be a blocker. By the way it is a nice library and your cookbooks are amazing, would love to contribute more in future 🙂 |
| @@ -0,0 +1,576 @@ | |||
| { | |||
There was a problem hiding this comment.
You can just keep the title and the goal here. Also, give yourself credit for the work. Here's an example:
Improving PostgreSQL Keyword Search to Avoid Zero Results
Notebook by <your linkedin or github profile>
In this notebook, you'll learn how to extend the PgvectorDocumentStore implementation to better transform your query and avoid getting zero results.
Reply via ReviewNB
| @@ -0,0 +1,576 @@ | |||
| { | |||
There was a problem hiding this comment.
| @@ -0,0 +1,576 @@ | |||
| { | |||
There was a problem hiding this comment.
| @@ -0,0 +1,576 @@ | |||
| { | |||
There was a problem hiding this comment.
bilgeyucel
left a comment
bilgeyucel
left a comment
There was a problem hiding this comment.
Thanks for the notebook @maylad31, It looks great! ❤️ I left some comments. After you resolved them, I'll add my commits (mainly around formatting) to this PR, if you dont mind :)
Co-authored-by: Bilge Yücel <[email protected]>
|
Hi, I’ve pushed the changes you suggested. I hope everything looks good now. |
2 participants
As per haystack documents, PgvectorKeywordRetriever does not apply fuzzy search out of the box, so it's necessary to carefully formulate the query in order to avoid getting zero results. This tutorial will help people to understand how to improve the default implementation and how to better transform their query to avoid getting zero results.