similarities/README.md

269 lines
9.6 KiB
Markdown
Raw Normal View History

2022-02-23 19:44:53 +08:00
[![PyPI version](https://badge.fury.io/py/similarities.svg)](https://badge.fury.io/py/similarities)
[![Downloads](https://pepy.tech/badge/similarities)](https://pepy.tech/project/similarities)
[![Contributions welcome](https://img.shields.io/badge/contributions-welcome-brightgreen.svg)](CONTRIBUTING.md)
[![GitHub contributors](https://img.shields.io/github/contributors/shibing624/similarities.svg)](https://github.com/shibing624/similarities/graphs/contributors)
[![License Apache 2.0](https://img.shields.io/badge/license-Apache%202.0-blue.svg)](LICENSE)
[![python_version](https://img.shields.io/badge/Python-3.5%2B-green.svg)](requirements.txt)
[![GitHub issues](https://img.shields.io/github/issues/shibing624/similarities.svg)](https://github.com/shibing624/similarities/issues)
[![Wechat Group](http://vlog.sfyc.ltd/wechat_everyday/wxgroup_logo.png?imageView2/0/w/60/h/20)](#Contact)
# Similarities
2022-03-08 20:20:41 +08:00
Similarities is a toolkit for similarity calculation and semantic search, supports text and image.
2022-02-23 19:44:53 +08:00
2022-03-05 23:48:17 +08:00
similarities相似度计算、语义匹配搜索工具包。
2022-02-23 19:44:53 +08:00
2022-03-08 19:51:28 +08:00
**similarities** 实现了多种相似度计算、匹配搜索算法支持文本、图像python3开发pip安装开箱即用。
2022-02-23 19:44:53 +08:00
**Guide**
2022-03-08 20:20:41 +08:00
2022-02-23 19:44:53 +08:00
- [Feature](#Feature)
- [Install](#install)
- [Usage](#usage)
- [Contact](#Contact)
- [Citation](#Citation)
- [Reference](#reference)
# Feature
### 文本相似度比较方法
- 余弦相似Cosine Similarity两向量求余弦
- 点积Dot Product两向量归一化后求内积
2022-03-07 01:14:37 +08:00
- [RankBM25](similarities/literalsim.py)BM25的变种算法对query和文档之间的相似度打分得到docs的rank排序
2022-03-08 20:20:41 +08:00
- [SemanticSearch](https://github.com/shibing624/similarities/blob/main/similarities/similarity.py#L99)向量相似检索使用Cosine
Similarty + topk高效计算比一对一暴力计算快一个数量级
2022-02-23 19:44:53 +08:00
# Demo
Official Demo: http://42.193.145.218/product/short_text_sim/
2022-03-05 15:54:52 +08:00
HuggingFace Demo: https://huggingface.co/spaces/shibing624/text2vec
2022-02-23 19:44:53 +08:00
![](docs/hf.png)
# Install
2022-03-08 20:20:41 +08:00
2022-02-23 19:44:53 +08:00
```
2022-02-24 14:05:54 +08:00
pip3 install torch # conda install pytorch
2022-02-23 19:44:53 +08:00
pip3 install -U similarities
```
or
```
git clone https://github.com/shibing624/similarities.git
cd similarities
python3 setup.py install
```
# Usage
2022-03-08 20:15:31 +08:00
### 1. 文本语义相似度计算
2022-02-23 19:44:53 +08:00
2022-03-08 20:20:41 +08:00
```python
from similarities import Similarity
2022-02-23 19:44:53 +08:00
2022-03-08 20:20:41 +08:00
m = Similarity("shibing624/text2vec-base-chinese")
r = m.similarity('如何更换花呗绑定银行卡', '花呗更改绑定银行卡')
print(f"similarity score: {r:.4f}") # similarity score: 0.8551
2022-03-07 01:14:37 +08:00
```
2022-02-23 19:44:53 +08:00
2022-03-08 19:51:28 +08:00
> 余弦值`score`范围是[-1, 1],值越大越相似。
2022-02-23 19:44:53 +08:00
2022-03-08 20:15:31 +08:00
### 2. 文本语义匹配搜索
2022-02-23 19:44:53 +08:00
一般在文档候选集中找与query最相似的文本常用于QA场景的问句相似匹配、文本相似检索等任务。
2022-03-08 20:20:41 +08:00
example: [examples/base_demo.py](./examples/base_demo.py)
2022-03-07 01:14:37 +08:00
```python
2022-03-08 20:15:31 +08:00
import sys
sys.path.append('..')
2022-03-07 01:14:37 +08:00
from similarities import Similarity
2022-03-08 20:15:31 +08:00
# 1.Compute cosine similarity between two sentences.
sentences = ['如何更换花呗绑定银行卡',
'花呗更改绑定银行卡']
corpus = [
'花呗更改绑定银行卡',
'我什么时候开通了花呗',
'俄罗斯警告乌克兰反对欧盟协议',
'暴风雨掩埋了东北部新泽西16英寸的降雪',
'中央情报局局长访问以色列叙利亚会谈',
'人在巴基斯坦基地的炸弹袭击中丧生',
]
model = Similarity("shibing624/text2vec-base-chinese")
print(model)
similarity_score = model.similarity(sentences[0], sentences[1])
print(f"{sentences[0]} vs {sentences[1]}, score: {float(similarity_score):.4f}")
# 2.Compute similarity between two list
similarity_scores = model.similarity(sentences, corpus)
print(similarity_scores.numpy())
for i in range(len(sentences)):
for j in range(len(corpus)):
print(f"{sentences[i]} vs {corpus[j]}, score: {similarity_scores.numpy()[i][j]:.4f}")
# 3.Semantic Search
model.add_corpus(corpus)
q = '如何更换花呗绑定银行卡'
print("query:", q)
for i in model.most_similar(q, topn=5):
print('\t', i)
2022-03-07 01:14:37 +08:00
```
output:
2022-03-08 20:20:41 +08:00
2022-03-07 01:14:37 +08:00
```shell
如何更换花呗绑定银行卡 vs 花呗更改绑定银行卡, score: 0.8551
...
如何更换花呗绑定银行卡 vs 花呗更改绑定银行卡, score: 0.8551
如何更换花呗绑定银行卡 vs 我什么时候开通了花呗, score: 0.7212
如何更换花呗绑定银行卡 vs 俄罗斯警告乌克兰反对欧盟协议, score: 0.1450
如何更换花呗绑定银行卡 vs 暴风雨掩埋了东北部新泽西16英寸的降雪, score: 0.2167
如何更换花呗绑定银行卡 vs 中央情报局局长访问以色列叙利亚会谈, score: 0.2517
如何更换花呗绑定银行卡 vs 人在巴基斯坦基地的炸弹袭击中丧生, score: 0.0809
花呗更改绑定银行卡 vs 花呗更改绑定银行卡, score: 1.0000
花呗更改绑定银行卡 vs 我什么时候开通了花呗, score: 0.6807
花呗更改绑定银行卡 vs 俄罗斯警告乌克兰反对欧盟协议, score: 0.1714
花呗更改绑定银行卡 vs 暴风雨掩埋了东北部新泽西16英寸的降雪, score: 0.2162
花呗更改绑定银行卡 vs 中央情报局局长访问以色列叙利亚会谈, score: 0.2728
花呗更改绑定银行卡 vs 人在巴基斯坦基地的炸弹袭击中丧生, score: 0.1279
query: 如何更换花呗绑定银行卡
(0, '花呗更改绑定银行卡', 0.8551459908485413)
(1, '我什么时候开通了花呗', 0.721195638179779)
(4, '中央情报局局长访问以色列叙利亚会谈', 0.2517135739326477)
(3, '暴风雨掩埋了东北部新泽西16英寸的降雪', 0.21666759252548218)
(2, '俄罗斯警告乌克兰反对欧盟协议', 0.1450251191854477)
```
2022-03-08 20:20:41 +08:00
2022-03-08 19:51:28 +08:00
> 余弦`score`的值范围[-1, 1]值越大表示该query与corpus的文本越相似。
2022-03-07 01:14:37 +08:00
2022-03-08 20:20:41 +08:00
#### 英文语义相似度计算和匹配搜索
2022-03-07 01:14:37 +08:00
2022-03-08 20:20:41 +08:00
example: [examples/base_english_demo.py](./examples/base_english_demo.py)
2022-03-07 01:14:37 +08:00
2022-03-08 20:15:31 +08:00
### 3. 快速近似语义匹配搜索
2022-02-23 19:44:53 +08:00
2022-03-08 20:15:31 +08:00
支持Annoy、Hnswlib的近似语义匹配搜索常用于百万数据集的匹配搜索任务。
2022-02-23 19:44:53 +08:00
2022-03-08 20:20:41 +08:00
example: [examples/fast_sim_demo.py](./examples/fast_sim_demo.py)
2022-03-07 01:14:37 +08:00
2022-03-08 20:15:31 +08:00
### 4. 基于字面的文本相似度计算和匹配搜索
2022-03-07 01:14:37 +08:00
2022-03-08 20:15:31 +08:00
支持同义词词林Cilin、知网Hownet、词向量WordEmbedding、Tfidf、SimHash、BM25等算法的相似度计算和字面匹配搜索常用于文本匹配冷启动。
2022-03-07 01:14:37 +08:00
2022-03-08 20:20:41 +08:00
example: [examples/literal_sim_demo.py](./examples/literal_sim_demo.py)
2022-03-07 01:14:37 +08:00
```python
2022-03-08 20:15:31 +08:00
from similarities.literalsim import SimHashSimilarity, TfidfSimilarity, BM25Similarity, \
2022-03-07 01:23:24 +08:00
WordEmbeddingSimilarity, CilinSimilarity, HownetSimilarity
2022-03-07 01:14:37 +08:00
text1 = "如何更换花呗绑定银行卡"
text2 = "花呗更改绑定银行卡"
m = TfidfSimilarity()
print(text1, text2, ' sim score: ', m.similarity(text1, text2))
2022-03-08 20:15:31 +08:00
2022-03-07 01:14:37 +08:00
zh_list = ['刘若英是个演员', '他唱歌很好听', 'women喜欢这首歌', '我不是演员吗']
m.add_corpus(zh_list)
print(m.most_similar('刘若英是演员'))
```
output:
2022-03-08 20:20:41 +08:00
2022-03-07 01:14:37 +08:00
```shell
如何更换花呗绑定银行卡 花呗更改绑定银行卡 sim score: 0.8203384355246909
[(0, '刘若英是个演员', 0.9847577834309504), (3, '我不是演员吗', 0.7056381915655814), (1, '他唱歌很好听', 0.5), (2, 'women喜欢这首歌', 0.5)]
```
2022-02-23 19:44:53 +08:00
2022-03-08 20:15:31 +08:00
### 5. 图像相似度计算和匹配搜索
支持[CLIP](similarities/imagesim.py)、pHash、SIFT等算法的图像相似度计算和匹配搜索。
2022-03-08 20:20:41 +08:00
example: [examples/image_demo.py](./examples/image_demo.py)
2022-03-08 20:15:31 +08:00
```python
import sys
import glob
sys.path.append('..')
from similarities.imagesim import ImageHashSimilarity, SiftSimilarity, ClipSimilarity
image_fp1 = 'data/image1.png'
image_fp2 = 'data/image12-like-image1.png'
m = ClipSimilarity()
print(m)
print(m.similarity(image_fp1, image_fp2))
# add corpus
m.add_corpus(glob.glob('data/*.jpg') + glob.glob('data/*.png'))
r = m.most_similar(image_fp1)
print(r)
```
output:
2022-03-08 20:20:41 +08:00
2022-03-08 20:15:31 +08:00
```shell
0.9579
[(6, 'data/image1.png', 1.0), (0, 'data/image12-like-image1.png', 0.9579654335975647), (4, 'data/image8-like-image1.png', 0.9326782822608948), ... ]
```
2022-03-08 20:20:41 +08:00
2022-03-08 20:15:31 +08:00
![image_sim](docs/image_sim.png)
2022-02-23 19:44:53 +08:00
# Contact
2022-03-08 20:20:41 +08:00
- Issue(建议)
[![GitHub issues](https://img.shields.io/github/issues/shibing624/similarities.svg)](https://github.com/shibing624/similarities/issues)
2022-02-23 19:44:53 +08:00
- 邮件我xuming: xuming624@qq.com
2022-03-08 20:20:41 +08:00
- 微信我: 加我*微信号xuming624, 备注:姓名-公司-NLP* 进NLP交流群。
2022-02-23 19:44:53 +08:00
<img src="docs/wechat.jpeg" width="200" />
# Citation
如果你在研究中使用了similarities请按如下格式引用
2022-03-05 15:54:52 +08:00
APA:
2022-03-08 20:20:41 +08:00
2022-03-05 15:54:52 +08:00
```
Xu, M. Similarities: Compute similarity score for humans (Version 0.0.4) [Computer software]. https://github.com/shibing624/similarities
```
BibTeX:
2022-03-08 20:20:41 +08:00
2022-03-05 15:54:52 +08:00
```
@software{Xu_Similarities_Compute_similarity,
author = {Xu, Ming},
title = {Similarities: similarity calculation and semantic search toolkit},
url = {https://github.com/shibing624/similarities},
version = {0.0.4}
2022-02-23 19:44:53 +08:00
}
```
# License
授权协议为 [The Apache License 2.0](/LICENSE)可免费用做商业用途。请在产品说明中附加similarities的链接和授权协议。
# Contribute
2022-03-08 20:20:41 +08:00
2022-02-23 19:44:53 +08:00
项目代码还很粗糙,如果大家对代码有所改进,欢迎提交回本项目,在提交之前,注意以下两点:
2022-03-08 20:20:41 +08:00
- 在`tests`添加相应的单元测试
- 使用`python setup.py test`来运行所有单元测试,确保所有单测都是通过的
2022-02-23 19:44:53 +08:00
之后即可提交PR。
# Reference
2022-03-08 20:20:41 +08:00
2022-02-23 19:44:53 +08:00
- [A Simple but Tough-to-Beat Baseline for Sentence Embeddings[Sanjeev Arora and Yingyu Liang and Tengyu Ma, 2017]](https://openreview.net/forum?id=SyK00v5xx)
2022-03-05 15:54:52 +08:00
- [liuhuanyong/SentenceSimilarity](https://github.com/liuhuanyong/SentenceSimilarity)
2022-03-08 19:51:28 +08:00
- [shibing624/text2vec](https://github.com/shibing624/text2vec)
- [qwertyforce/image_search](https://github.com/qwertyforce/image_search)
- [ImageHash - Official Github repository](https://github.com/JohannesBuchner/imagehash)