ASL and Technology
ASL stands for American Sign Language, as it refers to the complete, and natural language conveyed through gestures such as the movement of hands and face, instead of through the spoken or written word. Having the same properties with different grammar and syntaxes as spoken languages, ASL is the primary language of several people in North America who belong to the deaf community. As no universal sign language exists, ASL exists uniquely in North America, as people use different sign languages in different countries. ASL is a distinctly proper language of its own, possessing all the fundamental features of language such as grammar rules for word formation, syntax, and word order.
Just as AI is revolutionizing human engagement in language, it has tremendous possibilities to offer in the realm of ASL too. If spoken languages can be parsed by AI for different purposes, it is possible to do the same with ASL too, for example for purposes of live speech-to-text transcription and translation. This is achieved through feeding software through NLP algorithms that convert languages from one text to another. While of course with spoken languages, the live translation is made possible with the help of special mics embedded with special software, mics are of no use for ASL. How then do we go about live translating ASL? Here, cameras come to the rescue, as cameras serve as visual input devices that allow computers to parse what is being communicated in real-time.
These computers make use of Convolution Neural Nets (CNN) a supremely helpful way for computers to classify images on the basis of their available data of thousands of photos to be able to sort them in the right category and thus choose the correct translation. But this is a deeply convoluted task. This is because while the human mind can easily identify and differentiate between two entities such as cats and dogs, computers are unable to do this. This is because they are unable to truly “see” images, instead parsing them through a series of numbers arranged in an array. This is why bot detectors on websites ask humans to do simple visual tests to prove that the user is in fact a human. Further, computers also lack inductive reasoning that allows them to look at an animal and say that it is in fact a dog.
Here, technological progress and innovation have taken up the mantle of honing the abilities of AI to enable real-time translation of speech-to-text. It is precisely through working in fields like these that make lives easier for people where the emancipatory potential of AI lies.
Sources:
https://towardsdatascience.com/using-ai-to-translate-sign-language-in-real-time-96fe8c8223ed