Natural Language Processing

It is a field in Computer Science that focuses mainly on the natural language understanding by the computers and human-computer interaction.

Natural Language Processing is applied to various prominent areas such as voice recognition, automatic summarization, machine translation, Optical Character Recognition (OCR), text-to-speech and automated essay scoring.

Voice Recognition

VR can be specialized into two different categories based on its functions:

#Small-vocabulary/many users: used in automated telephone answering, when the computer needs to process many different accents of people, but cannot comprehend much complex words, since very little prebuilt commands are incorporated.

#Large-vocabulary/limited users: used in companies by a few expert users, and are usually highly efficient due to the huge number of command inputs.

How does Voice Recognition work?

When we talk, the sound waves produced are detected by ADC (Analogue to Digital Converter). After filtering the unwanted noise, the digitised sound are separated in terms of its frequency. The signal is divided into numerous small segments and matched with their phonemes. Therefore, the signal is then sent out in the form of text or audio. Weaknesses of the VR include low signal-to-noise ratio, overlapping speech and inability to differentiate homonyms by computers. To improve VR, statistical modelling systems are used.

Automatic Summarization

Process of reducing a text but still retaining the key components of the text, by computers, through extraction and abstraction. While extraction looks for keywords in the article and creates a summary based on it, abstraction has its own semantic representation of symbols and compares it with the main text. Although abstraction is more efficient and produces a highly reduced form of summary, but it has some quality constraints and is very complex.

Text to Speech

It is highly recommended to use when people need to know a lot of information at a very limited time.


# Is text-to-speech, machine translation almost the same?

# How long might it take to have a fully functional Voice Recognition that will translate all kinds of languages spoken?

# If accuracy of Voice Recognition depends on answers from volunteers, how likely is VR going to be error-free?

Useful Links

How Voice Recognition helps values to solve more challenges.

Steps to use google text to speech.

How VR works.