Metadata-Version: 2.1
Name: texta-bert-tagger
Version: 3.0.1
Summary: texta-bert-tagger
Home-page: https://git.texta.ee/texta/texta-bert-tagger-python
Author: TEXTA
Author-email: info@texta.ee
License: GPLv3
Description-Content-Type: text/markdown
License-File: LICENSE

# TEXTA Bert Tagger

![Py3.8](https://img.shields.io/badge/python-3.8-green.svg)
![Py3.9](https://img.shields.io/badge/python-3.9-green.svg)

## Installation

##### Using built package
`pip install texta-bert-tagger`

##### Using Git
`pip install git+https://git.texta.ee/texta/texta-bert-tagger-python.git`

### Testing

`python -m pytest -v tests`

### Documentation

Documentation for version 1.* is available [here](https://git.texta.ee/texta/texta-bert-tagger-python/-/wikis/Documentation-v1.*).

Documentation for version 2.* is available [here](https://git.texta.ee/texta/texta-bert-tagger-python/-/wikis/Documentation-v2.*).

Documentation for version 3.* is available [here](https://git.texta.ee/texta/texta-bert-tagger-python/-/wikis/Documentation-v3.*).

## Usage (for versions >=3.*.*)

### Fine-tune BERT model

```python
from texta_bert_tagger.tagger import BertTagger
bert_tagger = BertTagger()

data_sample = {"good": ["It was a nice day.", "All was well."], "bad": ["It was horrible.", "What a disaster."]}

# Train a model

# pos_label - used in metrics (precision, recall, f1-score etc) calculations as true label
bert_tagger.train(data_sample, pos_label="bad", n_epochs=2)

# Predict
result = bert_tagger.tag_text("How awful!")
print(result)
```

#### Output

```
[{"prediction": "bad", "probability": 0.55200404, "attributions": []}]
```
