Natural Language Processing
By
Dr. Parminder Kaur
What is NLP?
• Natural Language Processing (NLP)
– Computers use (analyze, understand,
generate) natural language
– A somewhat applied field
• Computational Linguistics (CL)
– Computational aspects of the human
language faculty
– More theoretical
Goals of NLP
• Scientific Goal
– Identify the computational machinery
needed for an agent to exhibit various
forms of linguistic behavior
• Engineering Goal
– Design, implement, and test systems
that process natural languages for
practical applications
Applications
• speech processing: get flight information or book
a hotel over the phone
• information extraction: discover names of people
and events they participate in, from a document
• machine translation: translate a document from
one human language into another
• question answering: find answers to natural
language questions in a text collection or
database
• summarization: generate a short biography of
Noam Chomsky from one or more news articles
General Themes
• Ambiguity of Language
• Language as a formal system
• Computation with human language
• Rule-based vs. Statistical Methods
• The need for efficiency
Topic Ideas
1.Text to Speech – artificial voices
2.Speech Recognition - understanding
3.Textual Analysis – readability
4.Plagiarism Detection – candidate selection
5.Intelligent Agents – machine interaction
Text to Speech – artificial voice
• Text Input
• Break text into phonemes
– Match phonemes to voice elements
– Concatenate voice elements
– Manipulate pitch and spacing
• Output results
• Research question: How can a human voice be
used to produce an artificial voice?
• Model Talker - opportunities for active, hands-on
research
Speech Recognition
• Spoken Input
• Identify words and phonemes in speech
– Generate text for recognized word parts
– Concatenate text elements
– Perform spelling, grammar and context checking
• Output results
• Research question: How can speech recognition
assist a deaf student taking notes in class?
• VUST – Villanova University Speech Transcriber
Textual Analysis - Readability
• Text Input
• Analyze text & estimate “readability”
– Grade level of writing
– Consistency of writing
– Appropriateness for certain educ. level
• Output results
• Research question: How can computer
analyze text and measure readability?
• Opportunities for hands-on research
Plagiarism Detection
• Text Input
• Analyze text & locate “candidates”
– Find one or more passages that might be plagiarized
– Algorithm tries to do what a teacher does
– Search on Internet for candidate matches
• Output results
• Research question: What algorithms work like
humans when finding plagiarism?
• Experimental CS research
Intelligent Agents
• Example: ELIZA
• AIML: Artificial Intelligence Modeling Lang.
• Human types something
• Computer parses, “understands”, and generates
response
• Response is viewed by human
• Research question: How can computers
“understand” and “generate” human writing?
• Also good area for experimentation
Digital Image Processing
• The images in previous slides are digital
(now), but they are NOT the result of DIP
• Digital Image Processing is
– Processing digital images by a digital
computer
• DIP requires a digital computer and other
supporting technologies (e.g., data
storage, display and transmission)
Photography
Motion Pictures
Law Enhancement and Biometrics
Remote Sensing
Hurricane Andrew America at night
taken by NOAA GEOS (Nov. 27, 2000)
Thermal Images
Operate in infrared frequency
Human body disperses Different colors indicate
heat (red pixels) varying temperatures
Medical Diagnostics
Operate in X-ray frequency
chest head
PET and Astronomy
Operate in gamma-ray frequency
Cygnus Loop in the
constellation of Cygnus
Positron Emission Tomography
Cartoon Pictures (Non-photorealistic)
Synthetic Images in Gaming
Age of Empire III by Ensemble Studios
Virtual Reality (Photorealistic)
Speech Recognition in AI
• AI-based speech recognition has made it
possible for computers to understand and
recognize human speech, enabling
frictionless interaction between humans
and machines.
How does Speech
Recognition in AI Work?
• Recording: The voice recorder that is built into the gadget is used to carry out the first
stage. The user's voice is kept as an audio signal after being recorded.
• Sampling: As you are aware, computers and other electronic gadgets use data in their
discrete form. By basic physics, it is known that a sound wave is continuous. Therefore,
for the system to understand and process it, it is converted to discrete values. This
conversion from continuous to discrete is done at a particular frequency.
• Transforming to Frequency Domain: The audio signal's time domain is changed to its
frequency domain in this stage. This stage is very important because the frequency
domain may be used to examine a lot of audio information. Time domain refers to the
analysis of mathematical functions, physical signals, or time series of economic or
environmental data, concerning time. Similarly, the frequency domain refers to the
analysis of mathematical functions or signals concerning frequency, rather than time.
• Information Extraction from Audio: Each voice recognition system's foundation is at
this stage. At this phase, the audio is transformed into a vector format that may be used.
For this conversion, many extraction methods, including PLP, MFCC, etc., are applied.
• Recognition of Extracted Information: The idea of pattern matching is applied in this
step. Recognition is performed by taking the extracted data and comparing it to some pre-
defined data. Pattern matching is used to accomplish this comparing and matching. One
of the most popular pieces of software for this is Google Speech API.
techniques used in AI for
speech recognition are:
• Hidden Markov Models (HMMs): HMMs are statistical models that
are widely used in speech recognition AI. HMMs work by modelling
the probability distribution of speech sounds, and then using these
models to match input speech to the most likely sequence of sounds.
• Deep Neural Networks (DNNs): DNNs are a type of machine
learning model that is used extensively in speech recognition AI.
DNNs work by using a hierarchy of layers to model complex
relationships between the input speech and the corresponding text
output.
• Convolutional Neural Networks (CNNs): CNNs are a type of
machine learning model that is commonly used in image recognition,
but have also been applied to speech recognition AI. CNNs work by
applying filters to input speech signals to identify relevant features.
Some recent advancements in
speech recognition AI include:
• Transformer-based models: Transformer-based models, such as BERT
and GPT, have been highly successful in natural language processing
tasks, and are now being applied to speech recognition AI.
• End-to-end models: End-to-end models are designed to directly map
speech signals to text, without the need for intermediate steps. These
models have shown promise in improving the accuracy and efficiency of
speech recognition AI.
• Multimodal models: Multimodal models combine speech recognition AI
with other modalities, such as vision or touch, to enable more natural and
intuitive interactions between humans and machines.
• Data augmentation: Data augmentation techniques, such as adding
background noise or changing the speaking rate, can be used to generate
more training data for speech recognition AI models, improving their
accuracy and robustness.
