Bleurt Large 512
B
Bleurt Large 512
Developed by Elron
BLEURT is a pre-trained metric model for evaluating the quality of text generation, based on the BERT architecture, capable of automatically scoring the similarity between candidate text and reference text.
Downloads 240
Release Time : 3/2/2022
Model Overview
BLEURT is a text generation evaluation metric developed by Google Research, which scores by learning the semantic similarity between reference text and candidate text. This PyTorch version is converted and implemented by community members.
Model Features
Robust Text Evaluation
Learns text similarity patterns through pre-training, capturing semantic similarity better than traditional metrics (e.g., BLEU).
BERT-based Architecture
Based on the BERT-large model, leveraging its powerful semantic representation capabilities.
End-to-End Scoring
Directly outputs a quality score between 0 and 1 without manual feature engineering.
Model Capabilities
Text Similarity Evaluation
Machine Translation Quality Scoring
Text Generation Quality Evaluation
Use Cases
Natural Language Processing
Machine Translation Evaluation
Automatically evaluates the match quality between machine translation results and reference translations.
Example output shows 'hello world' scored 0.9877 with 'hi universe' and 0.0475 with 'bye world'.
Text Summarization Evaluation
Evaluates the semantic consistency between generated summaries and reference summaries.
Featured Recommended AI Models