MultiEL v0.5

MultiEL

Multilingual Entity Linking model by BELA model

This project want to create easy-to-use Multilingual Entity Linking model by BELA model.

Origin Project

• Bi-encoder Entity Linking Architecture (BELA): https://github.com/facebookresearch/BELA

Install

pip install multiel

Usage

from multiel import BELA

bela_run = BELA(device="cuda")

print(bela_run.process_batch(["นายกประยุทธ์ประกาศจัดการเลือกตั้ง"]))

output:

[{'offsets': [0], 'lengths': [12], 'entities': ['Q2108126'], 'md_scores': [0.22365164756774902], 'el_scores': [0.6967974901199341]}]

API

from multiel import BELA

BELA(
 md_threshold:float=0.2,
 el_threshold:float=0.4, 
 checkpoint_name: str="wiki", 
 device: str="cuda:0",
 config_name:str="joint_el_mel_new",
 repo:str="wannaphong/BELA"
)

• md_threshold: md threshold
• el_threshold: Entity Linking threshold
• checkpoint_name: checkpoint name (wiki, aida, mewsli, and e2e) or your file name with extension
• device: device
• config_name: config name (in the BELA project)
• repo: Huggingface Hub repo (Default wannaphong/BELA)

Predict

BELA.process_batch([str, str])

License

MIT license and the model is MIT license. (BELA is MIT licensed)

MultiEL v0.4

MultiEL

Multilingual Entity Linking model by BELA model

This project want to create easy-to-use Multilingual Entity Linking model by BELA model.

Origin Project

• Bi-encoder Entity Linking Architecture (BELA): https://github.com/facebookresearch/BELA

Install

pip install multiel

Usage

from multiel import BELA

bela_run = BELA(device="cuda")

print(bela_run.process_batch(["นายกประยุทธ์ประกาศจัดการเลือกตั้ง"]))

Full Changelog: 0.3...0.4

MultiEL v0.3

• Try to reduce memory

Full Changelog: 0.2...0.3

MultiEL v0.2

Full Changelog: 0.1...0.2

MultiEL v0.1

Full Changelog: https://github.com/wannaphong/MultiEL/commits/0.1