A computational approach to support the creation of terminological neologisms in sign languages.

This work presents a novel approach to support the creation of terminological neologisms (technical signs) for Science, Technology, Engineering, and Mathematics (STEM) dictionaries in sign languages. We propose the use of computational methods to assist the creation of terminological neologisms that will constitute new STEM dictionaries. More specifically, we propose the use of computer vision algorithms for shape extraction and description from the visual representation (illustration) of a signless technical concept, as well as natural language processing methods for determining the stem of its corresponding oral term. By comparing those pieces of information of the signless technical concept with the corresponding ones of general-purpose signs in a repository, our approach creates a meaningful new sign from the existing ones with similar features. As far as we know, this is the first computer-aided proposal to create new signs and deal with the problem of technical vocabulary absence in sign languages, making the process of sign creation more agile and scalable. To validate the created technical signs, an evaluation committee composed by fluent signers considers aspects such as, the phonetic, phonological, and morphological structures of the sign language. Experiments were performed with technical concepts of the architectural drawing area, for which new signs in the Brazilian Sign Language were created. The results are promising with 60% of the new signs being rated on a five-point scale with a similar satisfaction level of signs of a reference dataset.