QaNER
Unofficial implementation of QaNER: Prompting Question Answering Models for Few-shot Named Entity Recognition.
You can adopt this pipeline for arbitrary BIO-markup data.
Installation
pip install qaner
CoNLL-2003
Pipeline results on CoNLL-2003 dataset:
How to use
Training
Script for training QaNER model:
qaner-train \
--bert_model_name 'bert-base-uncased' \
--path_to_prompt_mapper 'data/conll2003/prompt_mapper.json' \
--path_to_train_data 'data/conll2003/train.bio' \
--path_to_test_data 'data/conll2003/test.bio' \
--path_to_save_model 'dayyass/qaner-conll-bert-base-uncased' \
--n_epochs 2 \
--batch_size 128 \
--learning_rate 1e-5 \
--seed 42 \
--log_dir 'runs/qaner'
Required arguments:
- --bert_model_name - base bert model for QaNER fine-tuning
- --path_to_prompt_mapper - path to prompt mapper json file
- --path_to_train_data - path to train data (BIO-markup)
- --path_to_test_data - path to test data (BIO-markup)
- --path_to_save_model - path to save trained QaNER model
- --n_epochs - number of epochs to fine-tune
- --batch_size - batch size
- --learning_rate - learning rate
Optional arguments:
- --seed - random seed for reproducibility (default: 42)
- --log_dir - tensorboard log_dir (default: 'runs/qaner')
Infrerence
Script for inference trained QaNER model:
qaner-inference \
--context 'EU rejects German call to boycott British lamb .' \
--question 'What is the organization?' \
--path_to_prompt_mapper 'data/conll2003/prompt_mapper.json' \
--path_to_trained_model 'dayyass/qaner-conll-bert-base-uncased' \
--n_best_size 1 \
--max_answer_length 100 \
--seed 42
Result:
question: What is the organization?
context: EU rejects German call to boycott British lamb .
answer: [Span(token='EU', label='ORG', start_context_char_pos=0, end_context_char_pos=2)]
Required arguments:
- --context - sentence to extract entities from
- --question - question prompt with entity name to extract (examples below)
- --path_to_prompt_mapper - path to prompt mapper json file
- --path_to_trained_model - path to trained QaNER model
- --n_best_size - number of best QA answers to consider
Optional arguments:
- --max_answer_length - entity max length to eliminate very long entities (default: 100)
- --seed - random seed for reproducibility (default: 42)
Possible inference questions for CoNLL-2003:
- What is the location? (LOC)
- What is the person? (PER)
- What is the organization? (ORG)
- What is the miscellaneous entity? (MISC)
Requirements
Python >= 3.7
Citation
@misc{liu2022qaner,
title = {QaNER: Prompting Question Answering Models for Few-shot Named Entity Recognition},
author = {Andy T. Liu and Wei Xiao and Henghui Zhu and Dejiao Zhang and Shang-Wen Li and Andrew Arnold},
year = {2022},
eprint = {2203.01543},
archivePrefix = {arXiv},
primaryClass = {cs.LG}
}
Metadata
Release files for qaner 0.1.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| qaner-0.1.1.tar.gz | 11.4 kB | Without provenance | Details for qaner-0.1.1.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| qaner-0.1.1-py3-none-any.whl | Python 3 | none | any | Details for qaner-0.1.1-py3-none-any.whl |
Total release size: 25.6 kB
Release files / qaner-0.1.1.tar.gz
| Download URL | qaner-0.1.1.tar.gz |
|---|---|
| Size | 11.4 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
d8e4f0f95e40b5041dd86eee2bb29d144a7cdafbb4556d859a77431cb67c538b
|
|
BLAKE2b-256 checksum How to use checksums |
452cc6ed17d088dd7860aa33fe671dadfbe102c47eaf4ef578b65159df396a55
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/4.0.1 CPython/3.7.5
|
Release files / qaner-0.1.1-py3-none-any.whl
| Download URL | qaner-0.1.1-py3-none-any.whl |
|---|---|
| Size | 14.2 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
c7a107579016e40b2289daa7b3c7df963963107b343e233085e91dfe410556fb
|
|
BLAKE2b-256 checksum How to use checksums |
333a92f2b5e5327327c7e8d90a3fcf34eb2278ae692b24e17640ed24b85d4b55
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/4.0.1 CPython/3.7.5
|
Security for qaner 0.1.1
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.