Update README.md
Browse files
README.md
CHANGED
@@ -51,3 +51,84 @@ configs:
|
|
51 |
- split: test
|
52 |
path: data/test-*
|
53 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
51 |
- split: test
|
52 |
path: data/test-*
|
53 |
---
|
54 |
+
|
55 |
+
# StressTest Evaluation Dataset
|
56 |
+
|
57 |
+
This dataset supports the evaluation of models on **Sentence Stress Reasoning (SSR)** and **Sentence Stress Detection (SSD)** tasks, as introduced in our paper:
|
58 |
+
|
59 |
+
**[StressTest: Can YOUR Speech LM Handle the Stress?](https://arxiv.org/abs/2505.22765)**
|
60 |
+
|
61 |
+
📃 [Paper](https://arxiv.org/abs/2505.22765) | 💻 [Code Repository](https://github.com/slp-rl/StressTest) | 🤗 [Model: StresSLM](https://huggingface.co/slprl/StresSLM)
|
62 |
+
|
63 |
+
---
|
64 |
+
|
65 |
+
## 🗂️ Dataset Overview
|
66 |
+
|
67 |
+
This dataset includes **218** evaluation samples (split: `test`) with the following features:
|
68 |
+
|
69 |
+
* `transcription_id`: Identifier for each transcription sample.
|
70 |
+
* `transcription`: The spoken text.
|
71 |
+
* `description`: Description of the interpretation of the stress pattern.
|
72 |
+
* `intonation`: The stressed version of the transcription.
|
73 |
+
* `interpretation_id`: Unique reference to the interpretation imposed by the stress pattern of the sentence.
|
74 |
+
* `audio`: Audio data at 16kHz sampling rate.
|
75 |
+
* `metadata`: Structured metadata including:
|
76 |
+
|
77 |
+
* `gender`: Speaker gender.
|
78 |
+
* `language_code`: Language of the transcription.
|
79 |
+
* `sample_rate_hertz`: Sampling rate in Hz.
|
80 |
+
* `voice_name`: Voice name.
|
81 |
+
* `possible_answers`: List of possible interpretations for SSR.
|
82 |
+
* `label`: Ground truth label for SSR.
|
83 |
+
* `stress_pattern`: Structured stress annotation including:
|
84 |
+
|
85 |
+
* `binary`: Sequence of 0/1 labels marking stressed words.
|
86 |
+
* `indices`: Stressed word positions in the transcription.
|
87 |
+
* `words`: The actual stressed words.
|
88 |
+
* `audio_lm_prompt`: The prompt used for SSR.
|
89 |
+
|
90 |
+
---
|
91 |
+
|
92 |
+
## Evaluate YOUR model
|
93 |
+
|
94 |
+
This dataset is designed for evaluating models following the protocol and scripts in our [StressTest repository](https://github.com/slp-rl/StressTest).
|
95 |
+
|
96 |
+
To evaluate a model, refer to the instructions in the repository. For example:
|
97 |
+
|
98 |
+
```bash
|
99 |
+
python -m stresstest.evaluation.main \
|
100 |
+
--task ssr \
|
101 |
+
--model_to_evaluate stresslm
|
102 |
+
```
|
103 |
+
|
104 |
+
Replace `ssr` with `ssd` for stress detection, and use your model’s name with `--model_to_evaluate`.
|
105 |
+
|
106 |
+
---
|
107 |
+
|
108 |
+
## How to use
|
109 |
+
|
110 |
+
This dataset is formatted for with the HuggingFace Datasets library:
|
111 |
+
|
112 |
+
```python
|
113 |
+
from datasets import load_dataset
|
114 |
+
|
115 |
+
dataset = load_dataset("slprl/StressTest")
|
116 |
+
```
|
117 |
+
|
118 |
+
---
|
119 |
+
|
120 |
+
## 📖 Citation
|
121 |
+
|
122 |
+
If you use this dataset in your work, please cite:
|
123 |
+
|
124 |
+
```bibtex
|
125 |
+
@misc{yosha2025stresstest,
|
126 |
+
title={StressTest: Can YOUR Speech LM Handle the Stress?},
|
127 |
+
author={Iddo Yosha and Gallil Maimon and Yossi Adi},
|
128 |
+
year={2025},
|
129 |
+
eprint={2505.22765},
|
130 |
+
archivePrefix={arXiv},
|
131 |
+
primaryClass={cs.CL},
|
132 |
+
url={https://arxiv.org/abs/2505.22765},
|
133 |
+
}
|
134 |
+
```
|