VHash: Vectorizing Hash Tables
If you have documents with class labels, and you want to create numeric representations of those documents that maximize inter-class differences, then this package is for you. This package provides vectorizing hash tables that quickly transform your text, optimizing for the maximum distance between document-class vectors.
This project has a C++ backend with a python interface, allowing for maximum speed and maximum interopability.
Installation
To install, use pip:
pip install vhashMinimum Working Example
This package follows the conventions of scikit-learn to provide an
intuitive and familiar interface.
from typing import Any
from nptyping import NDArray
from vhash import VHash
# sample documents: each item in the list is its own document
docs = [
'hi, my name is Mike',
'hi, my name is George',
'hello, my name is Mike',
]
# class labels for each document
labels = [1, 0, 1]
# create & train model
vhash = VHash().fit(docs, labels)
# create numeric representation (2D float array)
numeric: NDArray[(Any, Any), float] = vhash.transform(docs)Metrics
To see how this text transformer compares to BERT, check out the
sample notebook,
where we show how vhash outperforms sBERT on a sentiment analysis task.
Documentation
To get started, check out the docs!
If you will be contributing to this repo, checkout the developer guide.
1 maintainer
Metadata
1 maintainer
Release files for vhash 0.0.27
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| vhash-0.0.27.tar.gz | 14.2 kB | Details |
Release files / vhash-0.0.27.tar.gz
| Download URL | vhash-0.0.27.tar.gz |
|---|---|
| Size | 14.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
5c67fd99bd9e96cc8b164558649d6281dbe4ee41b3cd19b00d91ac000afa88fd
|
|
BLAKE2b-256 checksum How to use checksums |
0e4dfb436cd775d3a9920d21cb300c9b7fe168e9a6aca903268471ce7f97293c
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/3.8.0 pkginfo/1.8.2 readme-renderer/32.0 requests/2.27.1 requests-toolbelt/0.9.1 urllib3/1.26.8 tqdm/4.62.3 importlib-metadata/4.11.1 keyring/23.5.0 rfc3986/2.0.0 colorama/0.4.4 CPython/3.9.13
|
