hugging-mapper 1.0.4

A lightweight python tool for effortless text similarity scoring using Hugging Face models

Hugging-Mapper

A lightweight python tool for easy text similarity scoring using Hugging Face models

Table of Contents

• Installation
• Features
• Usage
• Documentation
• License

Installation

pip install hugging-mapper

Features

• Easily compare how similar two pieces of text are
• Customizable model selection at initialization
• Works with Hugging Face models that create sentence embeddings
• Batch scoring for lists of sentence pairs

Usage

Embedding text using huggingface models

from hugger.mapper import HuggingMapper

# init
# default model_name is 'cambridgeltl/SapBERT-from-PubMedBERT-fulltext'
mapper = HuggingMapper(
    model_name="sentence-transformers/all-MiniLM-L6-v2"
)

# generate embedding
embedding = mapper.embed_text("I hope you'll find this helpful.")

Similarity search of given data

from hugger.mapper import NodeMapper
import pandas as pd

# demo data
data = pd.DataFrame({
    "id": ["node1", "node2", "node3"], 
    "text": ["Disease", "Gene", "Drug"]
})

# generate embeddings for data using (default) huggingface model
node_mapper = NodeMapper(data)

# get most similar 
# threshold 0 returns all data sorted by similarity to the given term
most_similar = node_mapper.get_similar("protein", threshold=0)

# get matching node
node_id, metadata = node_mapper.get_match("genetics", threshold=0.7)

Documentation

Tutorials and documentation are available on Read the Docs :)

License

This project is licensed under the MIT License.