bbm25-haystack 0.1.0

Haystack 2.x In-memory Document Store with Enhanced Efficiency

Better BM25 In-Memory Document Store

An in-memory document store is a great starting point for prototyping and debugging before migrating to production-grade stores like Elasticsearch. However, the original implementation of BM25 retrieval recreates an inverse index for the entire document store on every new search. Furthermore, the tokenization method is primitive, only permitting splitters based on regular expressions, making localization and domain adaptation challenging. Therefore, this implementation is a slight upgrade to the default BM25 in-memory document store by implementing incremental index update and incorporation of SentencePiece statistical sub-word tokenization.

Installation

This package has not yet been published to PyPI. Please install the package directly from the main branch using:

pip install git+https://github.com/Guest400123064/bbm25-haystack.git@main

Usage

The initializer takes three BM25+ hyperparameters, namely k1, b, and delta, and one path to a trained SentencePiece tokenizer .model file. All parameters are optional. The default tokenizer is directly copied from this SentencePiece test tokenizer with a vocab size of 1000.

from haystack import Document
from bbm25_haystack import BetterBM25DocumentStore, BetterBM25Retriever


document_store = BetterBM25DocumentStore()
document_store.write_documents([
   Document(content="There are over 7,000 languages spoken around the world today."),
   Document(content="Elephants have been observed to behave in a way that indicates a high level of self-awareness, such as recognizing themselves in mirrors."),
   Document(content="In certain parts of the world, like the Maldives, Puerto Rico, and San Diego, you can witness the phenomenon of bioluminescent waves.")
])

retriever = BetterBM25Retriever(document_store)
retriever.run(query="How many languages are spoken around the world today?")

Filtering Logic and Caveats

The filtering logic is slightly different from the default implementation shipped with Haystack, but this logic may be subject to changes, and I am open to different suggestions. Please find comments and implementation details in filters.py. TL;DR:

• Comparison with None, i.e., missing values, involved will always return False, no matter the document attribute value or filter value.
• Comparison with DataFrame is always prohibited to reduce surprises.
• No implicit datetime conversion from string values.

These differences lead to a few caveats. Firstly, some test cases are overridden to take into account the different expectations. However, this means that passed, non-overridden tests may not be faithful, i.e., the filters behave in the same way as the old implementation while different behaviors are expected. Further, the negation logic needs to be considered again because False can now issue from both input check and the actual comparisons. But I think having input processing and comparisons separated makes the filtering behavior more transparent.

License

bbm25-haystack is distributed under the terms of the Apache-2.0 license.