# Decent success for sentence sentiment embedding/querying with all-MiniLM-L6-v2 and rolling window

**URL:** https://community.pinecone.io/t/decent-success-for-sentence-sentiment-embedding-querying-with-all-minilm-l6-v2-and-rolling-window/1896
**Category:** General
**Tags:** semantic-search
**Created:** [May 17, 2023, 11:34pm UTC](https://community.pinecone.io/t/decent-success-for-sentence-sentiment-embedding-querying-with-all-minilm-l6-v2-and-rolling-window/1896 "2023-05-17T23:34:14Z")
**Posts on this page:** 1
**Page:** 1

<div class="post-metadata">

### Author: ![Sean](https://sea2.discourse-cdn.com/flex020/user_avatar/community.pinecone.io/sean/32/543_2.png) [@Sean](https://community.pinecone.io/u/Sean)
#### Post date: [May 17, 2023, 11:34pm UTC](https://community.pinecone.io/t/decent-success-for-sentence-sentiment-embedding-querying-with-all-minilm-l6-v2-and-rolling-window/1896/1 "2023-05-17T23:34:14Z")

</div>

Hope this information might be useful for people hoping to use Pinecone for text chat/NLP:

For context, I’m using the [https://github.com/Azure-Samples/azure-search-openai-demo/data](https://github.com/Azure-Samples/azure-search-openai-demo/data) PDF documents as my data repository.

I used PDFPlumber to extract text from each PDF page (code can be found in the above repo). Unfortunately, the query results were less than optimal - often query results were returned with no relevance to the original query (for instance, there is a demo “Northwind Health Plus Benefits Details” document - asking “are glasses covered under my healthplan?” would return a paragraph about prosthetics).

I tried several approaches (using the all-MiniLM-L6-v2 model):

1. page with 256 max\_seq\_length
2. page with 512 max\_seq\_length
3. chunk of 4 sentences of a page with 256 or 512 max\_seq\_length.
4. single sentence embedding.
5. rolling window paragraph

I finally settled on a hybrid of embedding each sentence, as well as embedding a rolling window of the sentence PLUS 4 subsequent sentences. Using this method, I was able to achieve reasonable results.

The reason I chose this approach is because occasionally, the initial sentence has a high cosine similarity, however additional data in the following sentences improve the intended result of the query. I batched these up into vectors of 700 calls (Pinecone has a limit of 1000 vector calls or 2MB per request).

The balance seems to provide a good balance of relevant information to then push the data to a chatGPT prompt for summarization of the top XX results.

Please comment if you’ve found other approaches!