Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/mrseanryan/entity-classifier
Classify entities into clusters via embedding vectors, using a given list of category names
https://github.com/mrseanryan/entity-classifier
classification classification-algorithm embedding-vectors machine-learning
Last synced: 7 days ago
JSON representation
Classify entities into clusters via embedding vectors, using a given list of category names
- Host: GitHub
- URL: https://github.com/mrseanryan/entity-classifier
- Owner: mrseanryan
- License: mit
- Created: 2024-01-20T11:25:15.000Z (10 months ago)
- Default Branch: master
- Last Pushed: 2024-02-12T15:29:27.000Z (9 months ago)
- Last Synced: 2024-04-18T01:53:37.757Z (7 months ago)
- Topics: classification, classification-algorithm, embedding-vectors, machine-learning
- Language: Python
- Homepage:
- Size: 12.7 KB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# entity-classifier
Classify entities into clusters via a zero-shot approach using embedding vectors, using a given list of category names.
- use an embedding to make vectors of entity names
- use the same embedding to make vectors of category names
- for each embedding, find the category that has a nearest vector
- then can classify the entities, for presentation in logical groups
## Approach
Compare words (labels) by examining how close are their encoded vectors:
- the dot product of 2 normalised vectors = cosine Angle
- cosine distance = 1 - v.w
- smaller means closer
## Dependencies
- Python 3.11
- pyenv - if on Windows use [pyenv-win](https://github.com/pyenv-win/pyenv-win)
## Install
Switch to Python 3.11.6:
```
pyenv install 3.11.6
pyenv local 3.11.6
```
Setup a virtual environment:
```
./create_env.sh
```
Install SBERT and cornsnake via this pip command:
```
pip install -U sentence-transformers==2.2.2 cornsnake==0.0.26
```
## Usage
```
python main.py [threshold (number between 0 and 1)]
```
## Example
To test:
```
./test.sh
```
OUTPUT:
```
CATEGORY: (unknown)
entity ['Aardvark', 'Alpaca', 'Anaconda']
CATEGORY: animal
entity ['Albatross', 'Alligator', 'Ant', 'Zebu']
CATEGORY: country
entity ['Albania', 'Andorra', 'Angola', 'Austria', 'Bangladesh', 'Belgium']
```
The results are not perfect, but not bad considering this is a simple 'out of the box' solution.
## Further improvements
Hierarchy of labels:
- first, classify against a top-level list of labels
- then, for each label, classify against that labels list of sub-labels
Increase accuracy:
- take several embeddings per class and use their average for that class
- try different embeddings, can get better results
- try different distance measures from your library
- consider tuning the embedding (for example, for the domain vocabulary of a particular industry or problem space)
# References
[My Medium article](https://medium.com/@mr.sean.ryan/classify-entities-via-a-zero-shot-approach-using-embedding-encodings-7ee9ee6e6bf2)
[Conference notes from ML Con Berlin 2023](https://github.com/mrseanryan/dev-conferences/blob/master/2023/mlcon-berlin/talk-Embeddings-Intro.md/README.md)
[SBERT: How to Use Sentence Embeddings to Solve Real-World Problems](https://anirbansen2709.medium.com/sbert-how-to-use-sentence-embeddings-to-solve-real-world-problems-f950aa300c72)