I am not working on the field of natural language processing (NLP). I am not sure if what I wrote in this report is new or promising. What I wrote here is more on the output of NLP technology that I perceive would help me and probably other people.
I. The NLP technology that I propose is a cheap, compact-but-comprehensive, flexible and offline multilingual speech-to-text recorder that I can bring anywhere.
Problem statement: "Usually I have a good idea popping-up in my mind when I am jogging, driving or biking, without any pen and paper that I can use to record the new idea. Hence, I forget the new ideas often, resulting in lost opportunities. I wish to have a voice-to-text machine that is flexible to accommodate any language, that is, I can switch language (e.g., Tagalog-English-Japanese) without clicking a button (because clicking buttons just to switch languages interrupts my thinking activity during jogging, driving or biking)."
I see a lot of potential in making flexible multilingual voice-to-text app with automated language/dialect detection. Many people are like me --- multilingual. This app can help researchers in science and writers/creative artists. It can also help make note-taking during meetings easier (e.g., transcribing minutes of the meeting). Currently, there is an existing free app in android or iphone, called the Dragon Dictation, but this app is not multilingual (that is, you can choose only one language at a time).
One of the additional features of this recorder could be automatic grammar correction and automatic style (e.g., from active to passive voice) conversion.
II. Aside from speech-to-text recorder, it is also good to extend this recorder to a translator. That is, a multilingual text/speech to a single-language text/speech. Currently, Google translate only translates text from one language to another language. What I envision is automatically translating a mix-source text/speech (e.g., multilingual) to another language.
III. Language is not only through speech and text. Language is also communicated through actions and emotions. Another possible NLP technology is interpreting, recording or translating actions and emotions (such as translating emotions to emoticons).
IV. Combining speech (words, tone, pitch, etc.), text, actions, emotions and other forms of language medium could be a frontier innovation. For example, one of the NLP + robotics machine that would definitely change our current technology is creating “EX MACHINA”.
See the trailer of EX MACHINA here:
V. Creating robots with good NLP capability is really a dream. One challenge for NLP is building database, algorithms and machines that can be stored and implemented in small structures (e.g., nano-computers). These NLP machines could also be efficient, fast, energy-saving and cheap so that it can be easily marketed to the public.
Quantum computing can help NLP. Error-correcting codes/language could also improve NLP. Formulating machine learning techniques that can generate additional words from small database could also be built. For example, if the database only has the word “walk” then “walking” could be added automatically as a new word.
VI. Personalized NLP can also be built. That is, some people or group of people do not always use major languages (English, Chinese, Latin, etc.). A good NLP product could be a machine that can be easily personalized by any person. This personalized NLP is not only for text/speech but also for actions and emotions. For example, certain people have mannerisms that a personalized NLP can detect and interpret.