<h3>Introduction</h3> <p>Second DIHARD Challenge Evaluation - SEEDLinGS was developed by Duke University and LDC and contains approximately two hours of English child language recordings along with corresponding annotations used in support of the <a href="https://dihardchallenge.github.io/dihard2">Second 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 <a href="https://dihardchallenge.github.io/dihard1/">first challenge</a>, the second 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, extended child language acquisition recordings, and YouTube videos.</p> <h3>Data</h3> <p>Source data is from the <a href="https://homebank.talkbank.org/access/Password/Bergelson.html">SEEDLingS</a> (The Study of Environmental Effects on Developing Linguistic Skills) corpus, designed to investigate how infants' early linguistic and environmental input plays a role in their learning. Recordings were generated in the home environment of infants in the Rochester, New York area. A subset of that data was annotated by LDC for use in the First and Second DIHARD Challenges.</p> <p>This release, when combined with Second DIHARD Challenge Evaluation - Eleven Sources (<a href="../../../LDC2022S06">LDC2022S06</a>), contains the evaluation set audio data and annotation, except for CHiME-5 audio files which must be obtained from the University of Sheffield.</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 file formats, data sources and domains is contained in the included documentation.</p> <h3>Updates</h3> <p>None at this time.</p>

Creator(s)

Neville Ryant

Mark Liberman

James Fiumara

Christopher Cieri

Distributor(s)

Linguistic Data Consortium

Right Holder(s)

Portions © 2019, 2022 Duke University, © 2019, 2022 Trustees of the University of Pennsylvania