Response Text Generation on StyleTalk (test)
4BLEUText-LLM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Text-LLMStyle Labeling Strategy=ground truth (upper bound)2024.02 | 4 | 17.9 | 19.6 | 76.3 | |
| Spoken-LLMSpeech Embedding Granularity=chunk-wise2024.02 | 4 | 17.8 | 19.4 | 76.3 | |
| Text-LLMStyle Labeling Strategy=cascaded2024.02 | 3.2 | 17.3 | 19.1 | 76 | |
| Text-LLMStyle Labeling Strategy=text-only2024.02 | 3.1 | 16.2 | 17.4 | 75.3 | |
| ParalinGPTSpeech Embedding Granularity=utterance-wise2024.02 | 3.1 | 16.8 | 18.5 | 75.9 | |
| ParalinGPTSpeech Embedding Granularity=chunk-wise2024.02 | 3.1 | 16.5 | 18.2 | 75.8 | |
| Spoken-LLMSpeech Embedding Granularity=utterance-wise2024.02 | 2.8 | 16.6 | 20.2 | 75.8 |