Third DIHARD Challenge Development

<h3>Introduction</h3>
<p>Third DIHARD Challenge Development was developed by Linguistic Data Consortium (LDC) and contains approximately 34 hours of English and Chinese speech data along with corresponding annotations used in support of the <a href="https://dihardchallenge.github.io/dihard3">Third DIHARD Challenge</a>.</p>
<p>The DIHARD Challenges are a set of shared tasks on diarization focusing on "hard" diarization; that is, speech diarization for challenging corpora where there was an expectation that existing state-of-the-art systems would fare poorly. As with the&nbsp;<a href="https://dihardchallenge.github.io/dihard1/">first</a>&nbsp;and&nbsp;<a href="https://dihardchallenge.github.io/dihard2/">second</a> challenges, the third development and evaluation sets were drawn from a diverse sampling of sources including monologues, map task dialogues, broadcast interviews, sociolinguistic interviews, meeting speech, speech in restaurants, clinical recordings, and amateur web videos.</p>
<h3>Data</h3>
<p>Data sources in this release are as follows (all sources are in English unless otherwise indicated):</p>
<ul>
<li>Autism Diagnosis Observation Schedule (ADOS) interviews</li>
<li>Conversations in Restaurants</li>
<li><a href="../../../LDC96S38">DCIEM/HCRC map task (LDC96S38)</a></li>
<li>Conversational telephone speech collected by LDC for the Fisher project</li>
<li>Audiobook recordings from&nbsp;<a href="https://librivox.org/">LibriVox</a></li>
<li>Meeting speech from&nbsp;<a href="../../../LDC2007S11">2004 Spring NIST Rich Transcription (RT-04S) Development (LDC2007S11)</a>&nbsp;and&nbsp;<a href="../../../LDC2007S12">Evaluation (LDC2007S12)</a>&nbsp;releases</li>
<li>2001 U.S. Supreme Court oral arguments</li>
<li>Sociolinguistic interviews from&nbsp;<a href="../../../LDC2003T15">SLX Corpus of Classic Sociolinguistic Interviews (LDC2003T15)</a></li>
<li><a href="../../../LDC2013S03">Mixer 6 Speech (LDC2013S03)</a></li>
<li>English and Chinese video collected by LDC as part of the Video Annotation for Speech Technologies (VAST) project</li>
<li>YouthPoint radio interviews</li>
</ul>
<p>&nbsp;</p>
<p>All audio is provided in the form of 16 kHz, 16-bit, mono-channel FLAC files. The diarization for each recording is stored as a NIST Rich Transcription Time Marked (RTTM) file. RTTM files are space-separated text files containing one turn per line. Segmentation files are stored as HTK label files. Each of these files contains one speech segment per line. Scoring regions for each recording are specific by un-partitioned evaluation map (UEM) files. All annotation file types are encoded as UTF-8. More information about the file formats and data sources and domains are in the included documentation.</p>
<h3>Samples</h3>
<p>Please view these samples:</p>
<ul>
<li><a href="desc/addenda/LDC2021S10.flac">Audio Sample (FLAC)</a></li>
<li><a href="desc/addenda/LDC2021S10.lab">Label Sample (TXT)</a></li>
<li><a href="desc/addenda/LDC2021S10.rttm">RTTM Sample (TXT)</a></li>
</ul>
<h3>Updates</h3>
<p>None at this time.</p>

### 引言
第三届DIHARD挑战赛开发集由语言数据联盟（Linguistic Data Consortium, LDC）构建，包含约34小时的英语与汉语语音数据及对应标注文件，用于支撑第三届DIHARD挑战赛。

DIHARD挑战赛系列是一组聚焦于"困难场景下说话人diarization"的共享任务，即针对挑战性语料的语音diarization任务——这类语料被认为会让现有主流系统表现不佳。与第一届和第二届挑战赛一致，本届开发集与评测集的数据来源多样，涵盖独白、地图任务对话、广播访谈、社会语言学访谈、会议语音、餐厅场景语音、临床录音以及业余网络视频等。

### 数据概况
本次发布的数据来源如下（若无特别说明，所有数据均为英语）：
- 自闭症诊断观察量表（Autism Diagnosis Observation Schedule, ADOS）访谈录音
- 餐厅对话录音
- DCIEM/HCRC地图任务（LDC96S38）
- 语言数据联盟（LDC）为Fisher项目采集的会话电话语音
- LibriVox平台的有声书录音
- 2004年春季NIST高质量语音转录（RT-04S）开发集（LDC2007S11）与评测集（LDC2007S12）中的会议语音
- 2001年美国最高法院口头辩论录音
- 经典社会语言学访谈SLX语料库（LDC2003T15）中的社会语言学访谈录音
- Mixer 6语音数据集（LDC2013S03）
- 语言数据联盟（LDC）为语音技术视频标注（Video Annotation for Speech Technologies, VAST）项目采集的英语与汉语视频语音数据
- YouthPoint电台访谈录音

所有音频均采用16 kHz、16位单声道FLAC格式提供。每份录音的说话人diarization标注存储为NIST高质量语音转录时间标记（NIST Rich Transcription Time Marked, RTTM）文件。RTTM文件为空格分隔的文本文件，每行对应一个语音轮次。分割文件采用HTK标注文件格式存储，每份文件每行对应一个语音片段。各录音的评分区域由无分区评测映射（un-partitioned evaluation map, UEM）文件指定。所有标注文件均采用UTF-8编码。更多关于文件格式、数据来源与领域的细节，请参见随附的文档。

### 示例
以下为示例文件：
- 音频示例（FLAC）
- 标注示例（TXT）
- RTTM示例（TXT）

### 更新说明
暂无更新。

搜集汇总

背景与挑战

背景概述

该数据集包含34小时的英语和汉语语音数据及注释，用于支持第三届DIHARD挑战赛，专注于挑战性语音分割任务。数据来源多样，涵盖多种语音场景和领域。