Command Me Specification Shawn Mathew, Anthony Tan, Amy Wong
Background Information - Speech Recognition Automatic speech recognition systems takes a speech signal and converts into words for natural processing language Process: Digital Sampling Acoustic Signal Processing Recognition of Words (Hidden Markov Modeling) hidden Markov model is a tool for representing probability distributions over sequences of observations
Background Information - Natural Language Processing What is natural language? Natural language is stated to be any language that are formed by humans and has evolved naturally through human use. It can be categorized as speech, writing, or signing (sign language). What is natural language processing? The study of this language and its interaction with computers.
Researched uses for Natural Language Processing Algorithms These are some of the tasks that have been researched for Natural Language Processing Summarizing chunks of text Translating one human language into another Converting information that is understandable to computers into readable language for humans Determining the part of speech of a word Converting images of text into machine understandable text
Project Idea Our project idea revolves around taking a verified human speech and converting that into an action that will be performed by a computer. Step 1: Provide computer with a sound clip from a person speaking. Step 2: Convert that sound representation into text. Step 3: The computer will perform a certain task/action based on the textual representation that was derived from human speech For now, we are focusing on commands for a computer such opening a certain program or searching what the weather is like through the web Over time, we may improve through machine learning such that the computer will learn new commands that are given to it instead of choosing from a set of predetermined tasks.
Project Implementations Take a saying and the speaker’s voice from microphone Turn the saying into text Text needs to be tokenized Tokenized text should be passed into machine learning algorithm or set of if statements Algorithm will decide what should be executed
Parsing Text
Algorithm After Tokenizer Machine Learning? Train the machine If Statements? Key words connected with commands
Additional Implementation Add voice recognition for user Commands may change according to user
Libraries Bob.spear (Speaker recognition Python toolkit) CMU Sphinx (Voice to Text) Natural Language Toolkit (Tokenizer) Keras (Machine Learning) Chatterbox (take text and provide response)
Sources http://www.ll.mit.edu/mission/cybersec/publications/publication- files/full_papers/020513_Reynolds.pdf https://www.ll.mit.edu/publications/journal/pdf/vol03_no1/3.1.3.speechrecognitio n.pdf http://mi.eng.cam.ac.uk/~mjfg/mjfg_NOW.pdf http://blog.algorithmia.com/introduction-natural-language-processing-nlp/ https://en.wikipedia.org/wiki/Natural_language_processing https://en.wikipedia.org/wiki/Speech_recognition