AIbase
EN
Home
Byt5 Small English
B

Byt5 Small English

Developed by hmbyt5
Historical multilingual and monolingual ByT5 base model, current version focuses on English text processing.
Large Language Model EnglishOpen Source License:MIT#Multilingual Named Entity Recognition#Historical Text Processing#Character-level Pretraining
Downloads 30
Release Time : 4/8/2023

Model Overview

A base language model based on the ByT5 architecture, specifically pretrained for English text, suitable for various natural language processing tasks.

Model Features

Historical Text Optimization
Trained on the British Library book corpus, particularly suitable for processing historical documents and book texts.
Multi-task Adaptation
Excellent performance on downstream tasks such as named entity recognition, with an average F1 score above 85.
Efficient Training
Pretrained using a single v3-8 TPU, ensuring high training efficiency.

Model Capabilities

English text understanding
Named entity recognition
Historical document processing

Use Cases

Academic Research
Historical Document Analysis
Named entity recognition and information extraction from British Library historical books
Achieved an F1 score of 85.65 on the English AjMC dataset
Information Extraction
Multilingual Entity Recognition
Handling named entity recognition tasks in multiple languages such as English, German, and French
F1 score of 87.27 on German AjMC and 84.44 on French AjMC
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
English简体中文繁體中文にほんご
© 2025AIbase