Datasets:
id stringlengths 33 33 | source_dataset stringclasses 3
values | source_id stringlengths 1 5 | topic stringclasses 15
values | topic_zh stringclasses 15
values | license stringclasses 3
values | messages listlengths 69 125 | planned_dialogue_acts listlengths 73 93 |
|---|---|---|---|---|---|---|---|
CounselTrajD_seed_04537__idx04536 | SmileChat | 34977 | academic_school | 学业校园 | CC0-1.0 | [
{
"role": "assistant",
"dialogue_act": "Greeting",
"content": "你好,今天想先从哪里说起?如果一时不知道怎么写,也可以慢慢来。"
},
{
"role": "user",
"dialogue_act": "Greeting",
"content": "你好……我最近状态有点乱,主要是睡觉的问题。"
},
{
"role": "assistant",
"dialogue_act": "Paraphrase",
"content": "你想先聊聊最近睡眠不好带来的困扰。"
},... | [
{
"role": "assistant",
"dialogue_act": "Greeting"
},
{
"role": "user",
"dialogue_act": "Greeting"
},
{
"role": "assistant",
"dialogue_act": "Paraphrase"
},
{
"role": "user",
"dialogue_act": "Hesitation"
},
{
"role": "assistant",
"dialogue_act": "Affirm"
},
... |
CounselTrajD_seed_04538__idx04537 | SmileChat | 27484 | academic_school | 学业校园 | CC0-1.0 | [
{
"role": "assistant",
"dialogue_act": "Greeting",
"content": "你好,今天想先从哪里说起?如果一时有点乱,也可以慢慢写。"
},
{
"role": "user",
"dialogue_act": "Greeting",
"content": "你好。我是想问一个同学的事,也可能是我自己想太多了。"
},
{
"role": "assistant",
"dialogue_act": "Affirm",
"content": "你能把“也可能是我想太多”这一点说出来,说明你不是只... | [
{
"role": "assistant",
"dialogue_act": "Greeting"
},
{
"role": "user",
"dialogue_act": "Greeting"
},
{
"role": "assistant",
"dialogue_act": "Affirm"
},
{
"role": "user",
"dialogue_act": "Hesitation"
},
{
"role": "assistant",
"dialogue_act": "Paraphrase"
},
... |
CounselTrajD_seed_04539__idx04538 | SmileChat | 32994 | academic_school | 学业校园 | CC0-1.0 | [
{
"role": "assistant",
"dialogue_act": "Greeting",
"content": "你好,今天想先从哪里说起?如果一时有点乱,也可以慢慢写。"
},
{
"role": "user",
"dialogue_act": "Greeting",
"content": "你好。其实我有点不知道怎么讲,就是最近总觉得自己不太配得上现在的成绩和资源。"
},
{
"role": "assistant",
"dialogue_act": "Affirm",
"content": "你愿意把这种“配不上”的感觉... | [
{
"role": "assistant",
"dialogue_act": "Greeting"
},
{
"role": "user",
"dialogue_act": "Greeting"
},
{
"role": "assistant",
"dialogue_act": "Affirm"
},
{
"role": "user",
"dialogue_act": "DetailedDescription"
},
{
"role": "assistant",
"dialogue_act": "Paraphras... |
CounselTrajD_seed_04540__idx04539 | SmileChat | 1688 | academic_school | 学业校园 | CC0-1.0 | [
{
"role": "assistant",
"dialogue_act": "Greeting",
"content": "你好,今天想先从哪里说起?如果一时组织不好,也可以慢慢写。"
},
{
"role": "user",
"dialogue_act": "Greeting",
"content": "你好。我是大一学生,最近有点不知道怎么办,感觉每天都不想学习。"
},
{
"role": "assistant",
"dialogue_act": "Affirm",
"content": "你能在刚开始觉得乱的时候就来谈这件事,说... | [
{
"role": "assistant",
"dialogue_act": "Greeting"
},
{
"role": "user",
"dialogue_act": "Greeting"
},
{
"role": "assistant",
"dialogue_act": "Affirm"
},
{
"role": "user",
"dialogue_act": "DetailedDescription"
},
{
"role": "assistant",
"dialogue_act": "Paraphras... |
CounselTrajD_seed_04541__idx04540 | SmileChat | 47136 | academic_school | 学业校园 | CC0-1.0 | [
{
"role": "assistant",
"dialogue_act": "Greeting",
"content": "你好,今天想先从哪里说起?如果是关于心理咨询本身的疑问,也可以直接问。"
},
{
"role": "user",
"dialogue_act": "Greeting",
"content": "你好。我其实是想问,怎么挑选心理咨询师。"
},
{
"role": "assistant",
"dialogue_act": "Inform",
"content": "可以的。挑选咨询师确实不只是“看谁名气大”或者“谁... | [
{
"role": "assistant",
"dialogue_act": "Greeting"
},
{
"role": "user",
"dialogue_act": "Greeting"
},
{
"role": "assistant",
"dialogue_act": "Inform"
},
{
"role": "user",
"dialogue_act": "DetailedDescription"
},
{
"role": "assistant",
"dialogue_act": "Paraphras... |
CounselTrajD_seed_04542__idx04541 | SmileChat | 4208 | academic_school | 学业校园 | CC0-1.0 | [
{
"role": "assistant",
"dialogue_act": "Greeting",
"content": "你好,今天想先从哪里说起?如果一时说不清,也可以慢慢写。"
},
{
"role": "user",
"dialogue_act": "Greeting",
"content": "你好。我最近有点纠结,主要是关于一个家教的事。我做完决定以后很快就后悔了,不知道是不是我太犹豫。"
},
{
"role": "assistant",
"dialogue_act": "Paraphrase",
"content": "... | [
{
"role": "assistant",
"dialogue_act": "Greeting"
},
{
"role": "user",
"dialogue_act": "Greeting"
},
{
"role": "assistant",
"dialogue_act": "Paraphrase"
},
{
"role": "user",
"dialogue_act": "Hesitation"
},
{
"role": "assistant",
"dialogue_act": "Affirm"
},
... |
CounselTrajD_seed_04543__idx04542 | SmileChat | 3307 | academic_school | 学业校园 | CC0-1.0 | [
{
"role": "assistant",
"dialogue_act": "Greeting",
"content": "你好,今天想先从哪里说起?如果一时不知道怎么组织,也可以慢慢写。"
},
{
"role": "user",
"dialogue_act": "Greeting",
"content": "你好。我其实犹豫了挺久的,就是有个事困扰我很多年了,跟初中的班主任有关。"
},
{
"role": "assistant",
"dialogue_act": "Paraphrase",
"content": "你想聊的是一段已... | [
{
"role": "assistant",
"dialogue_act": "Greeting"
},
{
"role": "user",
"dialogue_act": "Greeting"
},
{
"role": "assistant",
"dialogue_act": "Paraphrase"
},
{
"role": "user",
"dialogue_act": "Hesitation"
},
{
"role": "assistant",
"dialogue_act": "Affirm"
},
... |
CounselTrajD_seed_04544__idx04543 | SmileChat | 45384 | academic_school | 学业校园 | CC0-1.0 | [
{
"role": "assistant",
"dialogue_act": "Greeting",
"content": "你好,今天想先从哪里说起?如果一时很乱,也可以慢慢写。"
},
{
"role": "user",
"dialogue_act": "Greeting",
"content": "你好。我想聊一下宿舍关系吧……其实我憋了挺久的,感觉自己太容易被别人影响了。"
},
{
"role": "assistant",
"dialogue_act": "Paraphrase",
"content": "你现在主要困扰的是,同... | [
{
"role": "assistant",
"dialogue_act": "Greeting"
},
{
"role": "user",
"dialogue_act": "Greeting"
},
{
"role": "assistant",
"dialogue_act": "Paraphrase"
},
{
"role": "user",
"dialogue_act": "Hesitation"
},
{
"role": "assistant",
"dialogue_act": "Affirm"
},
... |
CounselTrajD_seed_04545__idx04544 | SmileChat | 40793 | academic_school | 学业校园 | CC0-1.0 | [
{
"role": "assistant",
"dialogue_act": "Greeting",
"content": "你好,今天想先从哪里说起都可以,如果一时说不清,也可以慢慢写。"
},
{
"role": "user",
"dialogue_act": "Greeting",
"content": "你好。其实我有点不知道怎么说,就是最近总觉得自己做什么都不太对。"
},
{
"role": "assistant",
"dialogue_act": "Affirm",
"content": "你能把这种“总觉得不对”的感觉说出... | [
{
"role": "assistant",
"dialogue_act": "Greeting"
},
{
"role": "user",
"dialogue_act": "Greeting"
},
{
"role": "assistant",
"dialogue_act": "Affirm"
},
{
"role": "user",
"dialogue_act": "Hesitation"
},
{
"role": "assistant",
"dialogue_act": "Paraphrase"
},
... |
CounselTrajD_seed_04547__idx04546 | SmileChat | 35844 | academic_school | 学业校园 | CC0-1.0 | [{"role":"assistant","dialogue_act":"Greeting","content":"你好,今天想先从哪里说起?如(...TRUNCATED) | [{"role":"assistant","dialogue_act":"Greeting"},{"role":"user","dialogue_act":"Greeting"},{"role":"a(...TRUNCATED) |
CounselTrajD
📄 Paper: CounselTraj (arXiv:2611.24168)
CounselTrajD contains 10,964 synthetic Chinese counseling dialogues, totaling 958,791 utterances across 15 topics. The mean length is 87.45 individual utterances per dialogue. These are generated expansions conditioned on seed conversations, rather than newly collected human counseling sessions.
The three compressed JSONL files and their manifest are identical to the GitHub data release. They are separated by provenance to preserve attribution and make source-specific licensing explicit.
Load the dataset
pip install datasets
from datasets import load_dataset
dataset = load_dataset("UEC-InabaLab/CounselTrajD", split="train")
print(len(dataset)) # 10964
messages = dataset[0]["messages"]
The train split contains the entire public release; no predefined evaluation
split is supplied. The default configuration includes all three sources. To load
a single source or stream the data:
smilechat = load_dataset("UEC-InabaLab/CounselTrajD", "smilechat", split="train")
psydt = load_dataset("UEC-InabaLab/CounselTrajD", "psydtcorpus", split="train")
cpsy = load_dataset("UEC-InabaLab/CounselTrajD", "cpsycound", split="train")
stream = load_dataset("UEC-InabaLab/CounselTrajD", split="train", streaming=True)
first_dialogue = next(iter(stream))
After downloading this repository, the standard-library loader can verify the file checksums, unique IDs, dialogue counts, and utterance counts:
python3 scripts/load_counseltraj.py
The full-record comparison and configuration loading checks are documented in validation.json. SHA256SUMS covers the release files.
Sources and licenses
Licenses apply by source, not as interchangeable license options for the whole
dataset. Each record includes its applicable license identifier. Please
retain attribution and notices as required by the respective source terms.
| Source of seed conversations | Configuration | Dialogues | License | Paper |
|---|---|---|---|---|
| SmileChat | smilechat |
7,672 | CC0-1.0 | SMILE (Qiu et al., 2024) |
| PsyDTCorpus | psydtcorpus |
2,595 | Apache-2.0 | PsyDT (Xie et al., 2025) |
| CPsyCounD | cpsycound |
697 | CC-BY-SA-4.0 | CPsyCoun (Zhang et al., 2024) |
| Total | default |
10,964 | Source-specific terms |
See LICENSING.md, THIRD_PARTY_NOTICES.md, and the complete license texts in LICENSES. The data do not inherit a blanket code license.
Statistics
| Statistic | Overall | Client | Counselor |
|---|---|---|---|
| Dialogues | 10,964 | — | — |
| Utterances | 958,791 | 473,910 | 484,881 |
| Mean utterances per dialogue | 87.45 | 43.22 | 44.22 |
| Characters | 42,197,155 | 18,321,830 | 23,875,325 |
| Topics | 15 | — | — |
Utterances are individual messages. Character counts exclude speaker and tag markup. Counts, topic frequencies, and source-file SHA256 checksums are recorded in data/manifest.json.
Schema
Each JSONL record contains:
| Field | Meaning |
|---|---|
id |
Stable unique CounselTrajD identifier |
source_dataset, source_id |
Source of the seed conversation and original record identifier |
topic, topic_zh |
Topic labels |
license |
Source-specific release license identifier |
messages |
Ordered generated messages with role, dialogue_act, and content |
planned_dialogue_acts |
Intended trajectory entries with role and dialogue_act |
user denotes the client, and assistant denotes the counselor. A dialogue may
start with a counselor greeting; strict speaker alternation is not enforced.
Dialogue-act fields
The message-level dialogue_act values preserve tags emitted during generation.
They are stored separately from the planned trajectory and are not guaranteed
to align with it one-to-one. In this release, 114 dialogues have different
planned and generated sequence lengths. These records are retained without
truncation or silent repair. Label definitions are documented in the
project README.
The export excludes raw seed dialogue text, API wrappers, request IDs, billing, usage metadata, and internal file paths. See the data card for additional schema details.
Construction and use
CounselTraj learns session-level interaction patterns from KokoroChat, plans dialogue-act trajectories from brief dialogues, and uses these plans to guide dialogue expansion. The CounselTrajD release was generated using GPT-5.5. The construction overview and DA annotation prompts are available in the project repository. The CounselTrajLM model versions are fine-tuned on this dataset for next counselor response generation.
This dataset is intended for counseling dialogue research. Synthetic generation and automated pattern screening do not establish complete anonymity, clinical safety, or freedom from bias. The data are not a substitute for professional diagnosis or treatment.
Citation
@inproceedings{qi-inaba-2027-counseltraj,
title = "{CounselTraj}: Synthesizing Process-Aware Psychological Counseling Dialogues via Dialogue-Act Trajectory Planning",
author = "Qi, Zhiyang and Inaba, Michimasa",
booktitle = "Proceedings of the 20th Conference of the European Chapter of the Association for Computational Linguistics",
year = "2027",
url = "https://arxiv.org/abs/2611.24168",
}
- Downloads last month
- -