Computational Linguistics (CL) is the branch of linguistics in which the concepts and algorithms of computer science are applied in performing analysis and synthesis of language and speech. This field of study focuses on the idea I briefly touched on in my last blog: machines can be taught to understand and output a language. The field deals with both written and spoken forms of the language. The following article from Stanford University beautifully explains the concepts of this field: https://plato.stanford.edu/entries/computational-linguistics/. In this blog, I will touch upon those CL applications that piqued my interest either because they made me think of potential use in breaking communication barriers for neurodivergent or because they reminded me of an everyday application I have come to take for granted.
Speech Processing is the application of Computational Linguistics to process speech. It is widely used in speech recognition (changing spoken audio to text), speech synthesis (changing text to spoken audio), and speech enhancement to improve intelligibility and emotion detection. Speech Processing uses ML methods like deep learning and neural networks to perform feature extraction from spoken audio pattern matching and language modeling.
Natural Language Understanding is one of the most used applications of Computational Linguistics. In this field, algorithms are developed to extract the meaning and intent of spoken speech or text, enabling natural interactions between a human and a computer. Alexa at my home can understand the questions I ask using my voice using Speech Recognition, understand them using Natural Language Processing techniques such as deep learning and semantic parsing, find the answer to my question, and transform it back to an audio response for me using speech synthesis algorithms.
One of the widely used applications of Computational Linguistics is in Translation. My grandparents, who are traveling from India to the USA, are ardent users of translation apps they have downloaded on their mobile phones. They only speak Hindi but can communicate at their transit points in the local language through the app that takes in their audio in Hindi, uses speech recognition software to transform it to text, understands the meaning of what they are saying using natural language processing, performs the Translation into another language and uses text to speech algorithms to output the foreign language in spoken format. While these apps are familiar, their power to break communication barriers is super helpful.
Text generation from unstructured information is an application of computational linguistics we all encounter when asking a question to Google and seeing a summary at the top. In this technique, an AI system uses large language models like chatGPT to process input data and generate output text. Summarization, Paraphrasing for clarity, and chatbots are all various applications of Text Generation we encounter in everyday life. It is a very promising application for me to dig deeper to understand its application for inclusive communication.
My primary goal is to research the applications of this field and explore how it can be harnessed to empower the Neurodivergent community.