Multiple inconsistencies in dataset annotation, file naming, and question_id sequences
Hi team,
I noticed several critical inconsistencies in the dataset that may affect its usability, and list the details below:
- Deceptive/Truthful count mismatch (claimed vs actual)
The dataset claims each subject has 9 deceptive and 15 truthful samples, but multiple subjects deviate from this rule:
045: deceptive=6, truthful=18
056: deceptive=6, truthful=18
059: deceptive=8, truthful=16
068: deceptive=6, truthful=18
074: deceptive=6, truthful=18
105: deceptive=8, truthful=16
145: deceptive=6, truthful=18
150: deceptive=6, truthful=18
160: deceptive=6, truthful=18 - Invalid question_id sequences (not 1-24 unique)
Two subjects have duplicate question_ids (missing unique values from 1-24):
109: question_id list = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 12, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24]
(duplicate 12, missing 13)
131: question_id list = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 14, 15, 16, 17, 18, 19, 19, 20, 21, 22, 23, 24]
(duplicate 19, missing 12) - Inconsistent file naming & ambiguous subject mapping
a. labels/084.csv contains entries like "084-1-1" and "084-22-5";
the features folder has "084-1-1" and "084-22-1"
b. Feature files use mixed naming conventions (e.g., "069-1-2" vs "069_1_2")
Could you please verify and correct these issues?
Thanks!
Thank you for your thorough review and for bringing these inconsistencies to our attention. We truly appreciate the time and effort you took to examine the dataset in such detail.
We have re-verified the issues you pointed out and fixed these errors.
To apply the corrections, you can simply:
Re-download deception/labels.zip to get the updated label files.
Rename the file '084-22-1' to '084-22-5' in the /deception/ folder. （including raw_audio, raw_video and deception_features）
Of course you can download the full updated dataset.
Thanks again for your valuable feedback—it helps improve the dataset for everyone. Please let us know if you encounter any further issues.