https://github.com/strangetom/averaged-perceptron

https://github.com/strangetom/averaged-perceptron

Last synced: about 1 month ago
JSON representation

• Host: GitHub
• URL: https://github.com/strangetom/averaged-perceptron
• Owner: strangetom
• License: mit
• Created: 2024-08-10T18:35:13.000Z (3 months ago)
• Default Branch: main
• Last Pushed: 2024-08-25T09:55:37.000Z (3 months ago)
• Last Synced: 2024-08-25T10:51:05.512Z (3 months ago)
• Language: Python
• Size: 33.2 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Ingredient Parser: Averaged Perceptron

This repository contains a prototype implementation of a pure python Averaged Perceptron sequence tagger, intended as a replacement for the Conditional Random Fields model used by [ingredient-parser](https://github.com/strangetom/ingredient-parser).

The implementation is based on a post by Matthew Honnibal: [A Good Part-of-Speech Tagger in about 200 Lines of Python](https://explosion.ai/blog/part-of-speech-pos-tagger-in-python).

## Set up

This repository relies on some of the files in the [ingredient-parser](https://github.com/strangetom/ingredient-parser) repository, so both must be downloaded to be able to use the implementation here:

```bash

# Clone repos

$ git clones https://github.com/strangetom/ingredient-parser

$ git clone https://github.com/strangetom/averaged-perceptron

# Set up venv

$ cd averaged-perceptron

$ python3 -m venv venv

$ source venv/bin/activate

# Train model, outputting html results, detailed results and confusion matrix

$ python train.py train --database train/data/training.sqlite3 --html --detailed --confusion

```

## Performance

The performance of this implementation is show below.

```bash

Sentence-level results:

	Accuracy: 94.45%

Word-level results:

	Accuracy 98.06%

	Precision (micro) 98.04%

	Recall (micro) 98.06%

	F1 score (micro) 98.05%

```

This compares favourably with current state of the art performance for the [ingredient-parser](https://github.com/strangetom/ingredient-parser) library, although the performance is not quite a good.