gest-IT corpus

The gest-IT corpus is an Italian multimodal corpus consisting of maximally ecological interactions, transcribed on three separate layers aligned to each other:

• an orthographic transcription;
• a prosodic transcription;
• a gestural transcription.

The resource is concerned with sighted and visually impaired L1 speakers in dialogical situations. In total, the corpus contains recording of 6 blind and 8 sighted participants, recruited on a voluntary basis and through a protocol that has been evaluated as compliant with GDPR and ethical requirements (Positive evaluation of the Bioethics Committee of the University of Bologna n. 0020349, 24/01/2024.).

All participants were paired and later involved in 30-minutes seated conversations, to elicit samples of spontaneous speech. Pairs and interaction settings were built accordind to two parameters:

• speakers could belong to the same category of participant (both blind, both sighted) or different categories. We coded these two situations as S (same, blind-blind or sighted-sighted conversation) or D (different, blind-sighted conversation);
• speakers could be facing each other or be seated back-to-back, to ensure that participants could not perceive the other's nonverbal communication. We coded these two situations as M (masked, back-to-back situation) or U (unmasked, facing situation).

Overall, 13 conversations were recorded, for a total of roughly 7 hours.

Progress

Name	Length	Metadata	Video Editing	Automatic Transcription	Anonimization	Ortographic	Prosodic	Gesture
DMC31051430	32.57	✓	✓	✓	✓	0:65.605	0:65.605	0:65.605
DMC31051630	35.35	✓	✓	✓
DUC22051430	32.00	✓	✓	✓
DUL31051430	35.21	✓	✓	✓
DUL31051630	34.30	✓	✓
SMC22051530	30.18	✓	✓
SMC31051730	36.15	✓	✓
DML22051630	33.28	✓	✓	✓
SML22051730	30.57	✓	✓
SUC22051630	23.33	✓	✓
SUC31051530	35.55	✓	✓	✓
SUL31051530	34.52	✓	✓	✓
SUL31051730	35.14	✓	✓

Issues:

• decide if any standardization is needed for the "notes" field
• add right license

Changelog

v 0.0.1

v 0.0.2

• added "student" and "retired" cases to classification of professions

v 0.0.3

• continuous integrarion added

v 0.1.0

• first release, one minute pilot transcription

v 0.1.1

• update metadata for S005