Bleurt Base 128
BLEURT is a text generation evaluation metric developed by Google Research for automatically assessing the quality of generated text.
Downloads 143
Release Time : 3/2/2022
Model Overview
BLEURT is an automatic evaluation metric based on pre-trained language models, specifically designed to assess the quality of machine-generated text. It learns from human ratings to predict similarity scores between generated text and reference text.
Model Features
Robust Evaluation
Specifically designed to handle various challenges in text generation evaluation, such as semantic similarity and fluency assessment
Pre-trained Model Based
Leverages the powerful representation capabilities of pre-trained language models like BERT
Human Rating Alignment
Fine-tuned with human rating data to make automatic scoring more consistent with human judgments
Model Capabilities
Text Quality Assessment
Machine Translation Evaluation
Text Summarization Evaluation
Dialogue System Evaluation
Use Cases
Natural Language Processing
Machine Translation Quality Assessment
Automatically evaluates the quality difference between machine translation results and reference translations
Provides similarity scores between 0-1
Text Summarization Evaluation
Assesses the semantic similarity between generated summaries and reference summaries
Quantifies summary quality metrics
Dialogue Systems
Dialogue Response Quality Assessment
Evaluates the relevance and fluency of responses generated by dialogue systems
Provides comparable scoring metrics
Featured Recommended AI Models