Table 3 Dataset content.

GESRes_dataset.csv

Annotation data, including the 16 gestural attributes described above as well as the

following information:

A unique identifier number

Annotation start and end times in milliseconds and hh:mm:ss format

Clip duration in seconds

The video ID corresponding to the full video clip

The video ID corresponding to the annotation clip

The speaker group (Lecturer, Politician, Clinician)

The speaker ID (e.g., Politician1, Politician2 etc.)

The longer utterance spoken just before, during, and after the annotation utterance

GESRes_codebook.xlsx

A code book detailing the information contained in each column of the annotation data.

GESRes_dataset.json

A JSON file containing the full dataset, metadata including video information, and the full codebook.

lexeme_descriptions.csv

A file containing the names and detailed descriptions of each lexeme.

gesture_annotation_template.etf

An ELAN template file. This can be used to import tier structure into ELAN for annotation.

GESRes_annotation_manual.pdf

An annotation manual detailing exact annotation procedures.

Licensing_information.pdf

Contains detailed information on the license types for each video and the availability of non-licensed videos.

01Gesture_videos

Contains individual video clips for each hand gesture (licensed videos only) including audio.

02Full_videos

Contains full videos where licensing allowed. This allows for other researchers to add

their own annotations or try out our annotation approach.

03Transcripts

Contains Transcripts for each video.

04Tracking_data

Hand tracking data in 3D for 33 pose and 21 hand landmarks.

05Code

Folder containing all code used to produce and evaluate the dataset.

See ReadMe.rtf file for description of scripts.

ReadMe.rtf

A file listing all files and folders in the directory, including listing individual analysis scripts.