JacobLinCool/ami-multiscale

---
dataset_info:
- config_name: window_10min
features:
- name: meeting_id
dtype: string
- name: begin_time
dtype: float32
- name: end_time
dtype: float32
- name: duration
dtype: float32
- name: audio
dtype:
audio:
sampling_rate: 16000
- name: transcription
struct:
- name: speaker_id
list: string
- name: text
list: string
- name: begin_time
list: float32
- name: end_time
list: float32
- name: previous_context
struct:
- name: speaker_id
list: string
- name: text
list: string
- name: begin_time
list: float32
- name: end_time
list: float32
splits:
- name: train
num_bytes: 9771197297
num_examples: 593
- name: validation
num_bytes: 1155657785
num_examples: 71
- name: test
num_bytes: 1101296865
num_examples: 67
download_size: 9742665407
dataset_size: 12028151947
- config_name: window_15min
features:
- name: meeting_id
dtype: string
- name: begin_time
dtype: float32
- name: end_time
dtype: float32
- name: duration
dtype: float32
- name: audio
dtype:
audio:
sampling_rate: 16000
- name: transcription
struct:
- name: speaker_id
list: string
- name: text
list: string
- name: begin_time
list: float32
- name: end_time
list: float32
- name: previous_context
struct:
- name: speaker_id
list: string
- name: text
list: string
- name: begin_time
list: float32
- name: end_time
list: float32
splits:
- name: train
num_bytes: 9517182000
num_examples: 402
- name: validation
num_bytes: 1121687203
num_examples: 49
- name: test
num_bytes: 1066944509
num_examples: 46
download_size: 9473332900
dataset_size: 11705813712
- config_name: window_20min
features:
- name: meeting_id
dtype: string
- name: begin_time
dtype: float32
- name: end_time
dtype: float32
- name: duration
dtype: float32
- name: audio
dtype:
audio:
sampling_rate: 16000
- name: transcription
struct:
- name: speaker_id
list: string
- name: text
list: string
- name: begin_time
list: float32
- name: end_time
list: float32
- name: previous_context
struct:
- name: speaker_id
list: string
- name: text
list: string
- name: begin_time
list: float32
- name: end_time
list: float32
splits:
- name: train
num_bytes: 9353704457
num_examples: 310
- name: validation
num_bytes: 1104971468
num_examples: 39
- name: test
num_bytes: 1046304744
num_examples: 34
download_size: 9302129635
dataset_size: 11504980669
- config_name: window_25min
features:
- name: meeting_id
dtype: string
- name: begin_time
dtype: float32
- name: end_time
dtype: float32
- name: duration
dtype: float32
- name: audio
dtype:
audio:
sampling_rate: 16000
- name: transcription
struct:
- name: speaker_id
list: string
- name: text
list: string
- name: begin_time
list: float32
- name: end_time
list: float32
- name: previous_context
struct:
- name: speaker_id
list: string
- name: text
list: string
- name: begin_time
list: float32
- name: end_time
list: float32
splits:
- name: train
num_bytes: 9272977513
num_examples: 262
- name: validation
num_bytes: 1092752457
num_examples: 33
- name: test
num_bytes: 1038577383
num_examples: 30
download_size: 9224680581
dataset_size: 11404307353
- config_name: window_30min
features:
- name: meeting_id
dtype: string
- name: begin_time
dtype: float32
- name: end_time
dtype: float32
- name: duration
dtype: float32
- name: audio
dtype:
audio:
sampling_rate: 16000
- name: transcription
struct:
- name: speaker_id
list: string
- name: text
list: string
- name: begin_time
list: float32
- name: end_time
list: float32
- name: previous_context
struct:
- name: speaker_id
list: string
- name: text
list: string
- name: begin_time
list: float32
- name: end_time
list: float32
splits:
- name: train
num_bytes: 9236661820
num_examples: 240
- name: validation
num_bytes: 1084865634
num_examples: 28
- name: test
num_bytes: 1035711555
num_examples: 29
download_size: 9186170106
dataset_size: 11357239009
- config_name: window_5min
features:
- name: meeting_id
dtype: string
- name: begin_time
dtype: float32
- name: end_time
dtype: float32
- name: duration
dtype: float32
- name: audio
dtype:
audio:
sampling_rate: 16000
- name: transcription
struct:
- name: speaker_id
list: string
- name: text
list: string
- name: begin_time
list: float32
- name: end_time
list: float32
- name: previous_context
struct:
- name: speaker_id
list: string
- name: text
list: string
- name: begin_time
list: float32
- name: end_time
list: float32
splits:
- name: train
num_bytes: 10937114925
num_examples: 1236
- name: validation
num_bytes: 1289114263
num_examples: 145
- name: test
num_bytes: 1228214982
num_examples: 140
download_size: 10898379925
dataset_size: 13454444170
configs:
- config_name: window_10min
data_files:
- split: train
path: window_10min/train-*
- split: validation
path: window_10min/validation-*
- split: test
path: window_10min/test-*
- config_name: window_15min
data_files:
- split: train
path: window_15min/train-*
- split: validation
path: window_15min/validation-*
- split: test
path: window_15min/test-*
- config_name: window_20min
data_files:
- split: train
path: window_20min/train-*
- split: validation
path: window_20min/validation-*
- split: test
path: window_20min/test-*
- config_name: window_25min
data_files:
- split: train
path: window_25min/train-*
- split: validation
path: window_25min/validation-*
- split: test
path: window_25min/test-*
- config_name: window_30min
data_files:
- split: train
path: window_30min/train-*
- split: validation
path: window_30min/validation-*
- split: test
path: window_30min/test-*
- config_name: window_5min
data_files:
- split: train
path: window_5min/train-*
- split: validation
path: window_5min/validation-*
- split: test
path: window_5min/test-*
---

该数据集包含6种配置，分别对应不同的滑动窗口时长：5分钟、10分钟、15分钟、20分钟、25分钟、30分钟。每种配置的详细信息如下：
1. 配置名称：window_10min
特征字段包括：
- meeting_id：数据类型为字符串（string），用于标识唯一会议
- begin_time：数据类型为32位浮点型（float32），记录当前片段的起始时间
- end_time：数据类型为32位浮点型（float32），记录当前片段的结束时间
- duration：数据类型为32位浮点型（float32），记录当前片段的时长
- audio：音频数据，其数据格式为采样率16000赫兹的音频
- transcription：转录结果结构体（struct），包含4个子字段：
* speaker_id：字符串列表，存储各发言段落的说话人ID
* text：字符串列表，存储各发言段落的转录文本
* begin_time：32位浮点型列表，存储各发言段落的起始时间
* end_time：32位浮点型列表，存储各发言段落的结束时间
- previous_context：前文上下文结构体（struct），结构与transcription一致，存储当前片段之前的会话上下文，包含speaker_id、text、begin_time、end_time四个列表类型子字段
数据划分如下：
- 训练集（train）：字节数9771197297，样本量593
- 验证集（validation）：字节数1155657785，样本量71
- 测试集（test）：字节数1101296865，样本量67
该配置的下载大小为9742665407字节，数据集总大小为12028151947字节

2. 配置名称：window_15min
特征字段与上述window_10min配置完全一致
数据划分：
- 训练集：字节数9517182000，样本量402
- 验证集：字节数1121687203，样本量49
- 测试集：字节数1066944509，样本量46
下载大小9473332900字节，数据集总大小11705813712字节

3. 配置名称：window_20min
特征字段同其他配置一致
数据划分：
- 训练集：字节数9353704457，样本量310
- 验证集：字节数1104971468，样本量39
- 测试集：字节数1046304744，样本量34
下载大小9302129635字节，数据集总大小11504980669字节

4. 配置名称：window_25min
数据划分：
- 训练集：字节数9272977513，样本量262
- 验证集：字节数1092752457，样本量33
- 测试集：字节数1038577383，样本量30
下载大小9224680581字节，数据集总大小11404307353字节

5. 配置名称：window_30min
数据划分：
- 训练集：字节数9236661820，样本量240
- 验证集：字节数1084865634，样本量28
- 测试集：字节数1035711555，样本量29
下载大小9186170106字节，数据集总大小11357239009字节

6. 配置名称：window_5min
数据划分：
- 训练集：字节数10937114925，样本量1236
- 验证集：字节数1289114263，样本量145
- 测试集：字节数1228214982，样本量140
下载大小10898379925字节，数据集总大小13454444170字节

数据集配置与对应数据文件路径如下：
- 配置名称window_10min：训练集数据路径为`window_10min/train-*`，验证集路径为`window_10min/validation-*`，测试集路径为`window_10min/test-*`
- 配置名称window_15min：训练集路径`window_15min/train-*`，验证集路径`window_15min/validation-*`，测试集路径`window_15min/test-*`
- 配置名称window_20min：训练集路径`window_20min/train-*`，验证集路径`window_20min/validation-*`，测试集路径`window_20min/test-*`
- 配置名称window_25min：训练集路径`window_25min/train-*`，验证集路径`window_25min/validation-*`，测试集路径`window_25min/test-*`
- 配置名称window_30min：训练集路径`window_30min/train-*`，验证集路径`window_30min/validation-*`，测试集路径`window_30min/test-*`
- 配置名称window_5min：训练集路径`window_5min/train-*`，验证集路径`window_5min/validation-*`，测试集路径`window_5min/test-*`