https://github.com/pakio/EsBM25SemanticHybridComparison

(Demo) Elasticsearch with ML node and ingest pipeline for hybrid search (Lexical + Semantic)
https://github.com/pakio/EsBM25SemanticHybridComparison

Last synced: about 8 hours ago
JSON representation

(Demo) Elasticsearch with ML node and ingest pipeline for hybrid search (Lexical + Semantic)

• Host: GitHub
• URL: https://github.com/pakio/EsBM25SemanticHybridComparison
• Owner: pakio
• License: other
• Created: 2022-12-03T07:39:36.000Z (almost 2 years ago)
• Default Branch: master
• Last Pushed: 2022-12-04T03:05:21.000Z (almost 2 years ago)
• Last Synced: 2024-08-02T13:17:37.109Z (3 months ago)
• Language: Python
• Homepage:
• Size: 1.58 MB
• Stars: 5
• Watchers: 2
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Elasticsearch BM25 vs KNN vs Hybrid comparison system

This is the ready-to-use demo repository to test Elasticsearch semantic search with embedded transformer using ingest pipeline.

![Demo](https://github.com/pakio/EsBM25SemanticHybridComparison/blob/master/demo.gif?raw=true)

- Data

  - wikimedia enwiki 20221201 dump [url](https://dumps.wikimedia.org/enwiki/20221201/)

- Model

  - sentence-transformers/msmarco-MiniLM-L-12-v3 [link(Hugging Face)](https://huggingface.co/sentence-transformers/msmarco-MiniLM-L-12-v3)

# prerequisites

This repository uses the softwares/tools/frameworkds below.

- docker

- docker-compose

- python (>3.10)

# How to run

## 1. Launch Elasticsearch and upload model

Run `./Es/setup.sh` to launch Elasticsearch, upload model, and configure the ingest pipeline.

## 2. Ingest data

Run `./indexer/setup.sh` to download, and index the data.

If you observe 429 error, reduce the batch size and please retry.

## 3. Launch comparison tool

There is a GUI comparison tool under `./eval` directory.

Go `./eval` directory and run `streamlit run main.py` to launch the comparison tool.

# note

This repository enables Elasticsearch Trial License inorder to use ML node to run embedding transformer model in ingest pipeline.