Similar Segments in Social Speech

Documentation and Downloads

Items in bold are suggested places to start.

Documentation

• Task Overview Paper: The Similar Segments in Social Speech Task (pdf)
• Technical Report: Data Collection for the Similar Segments in Social Speech Task (pdf)

Corpus Collection Documentation

• Data-Collection Protocol
• Recruitment Flier
• Procedure and Consent Form
• Participant Verbal Instructions

Corpus Documentation

• Per-dialog Metadata
• Recording Notes

Annotation Documentation

• Annotator Training Overview
• Annotators Guide
• How To Tag Social Speech using Elan
• Annotation Notes

• Evaluation Code, Version 5 (old version 4) (old version 3)
• Guide to the Metrics
• Baselines
• Sample Queries
• Sample Results (search-system-generated answerfile)
• Sample Reference ("correct") Similiarity Sets
• Sample Eval-code Output
• All Possible Queries over the Training Set
• Testset Queries

Similarity Sets

• Similarity Sets (July 19), (old release), (older release)
• same similarity sets, but one file per annotator: a1, a2 , a3, a4, a5, a6

Video (mpeg-4)

• obtainable from any of the task organizers

Training-Set Transcripts

• ASR transcripts for dialogs 000-018 and 021 (better transcripts are available on request)

Prosodic Features

• Descriptions
• F₀ Values
• Contextual: 0-9 left; 10-18,21 left; 0-9 right; 10-18, 21 right
• Dimensions: left tracks, right tracks

Other

• Buglist and Todo List
• Similarity Annotations (collection 1) in Elan format, which might be useful if listening to the data

New

• Mandarin annotations/simsets (for LDC CallHome Mandarin)

For access to the training data and the test data, please contact Nigel Ward.

Back up to the Similar Segments in Social Speech main page