A Swedish Sign Language Database of Video-Recorded Sign-Pseudosign Pairs With Matching Neighborhood Density and...

Area of Science:

• Linguistics
• Psycholinguistics
• Sign Language Studies

Background:

• Understanding the phonotactic structure of sign languages is crucial for linguistic analysis.
• Lexical metrics like neighborhood density (ND) and phonotactic probability (PP) are vital for sign language research.
• Developing standardized datasets for sign languages like Swedish Sign Language (STS) is essential for advancing the field.

Purpose of the Study:

• To create a video-recorded dataset of real sign-pseudosign pairs for Swedish Sign Language (STS).
• To ensure pseudosigns are phonotactically similar to real signs, matched for neighborhood density (ND) and phonotactic probability (PP).
• To develop algorithms for calculating ND and PP in STS and investigate their correlations.

Main Methods:

• Utilized publicly available STS data to generate phonotactically feasible pseudosign candidates.
• Developed and applied automatic algorithms for calculating ND and PP for both real signs and pseudosigns.
• Conducted manual evaluation of pseudosign candidates and video-recorded 466 matched sign-pseudosign pairs with four actors.

Main Results:

• Successfully created 466 video-recorded sign-pseudosign pairs for STS, matched for ND and PP.
• Ensured similar distributions of ND, PP, and duration between real signs and pseudosigns.
• Confirmed that the overall distributions of ND, PP, and sign frequency (SF) in the dataset mirror those of STS.

Conclusions:

• The developed dataset is publicly available and suitable for psycholinguistic experiments in STS.
• This study establishes a new standard for estimating lexical metrics in sign languages, applicable to other sign languages.
• The findings contribute to a deeper understanding of sign language phonotactics and lexical structure.