o Method:
Computer Processing of Tone with same frequency as vocal
Human Language chords
Formant synthesis – emphasizing
harmonics corresponding to the
COMPUTATIONAL LINGUISTICS
formants required for a certain
– subfield of linguistics and computer science that is sound OR
concerned with the interactions of human language Concatenative synthesis – recorded
and computers units phones, syllables, words,
– Includes: phrases, and sentences; what is
o analysis of written texts and spoken generally used today
discourse o Computers still have a machinelike accent
o translation of text and speech – Text-to-Speech
o use of human languages for communication o Computer program which converts written
between computers and people texts to basic units for synthesizer
o modeling and testing of linguistic theories Formant synthesizers: translates
into phonetic representation; text
Computational phonetics and phonology to speech happens precedes
– Processing of speech electronic conversion to sound
– Converting speech to text and vice versa Concatenative synthesis: translates
– Speech recognition input depending on what
o analyzing speech into phones and sounds/phones/words are joined
phonemes → producing a phonetic together
transcription → converting to ordinary text o Problems:
o Computers find difficulty in segmenting Words spelled alike but read
speech b/c differently (e.g. read – computer
its original design is to detect sound must have structural knowledge)
patterns of only certain Inconsistent spelling (e.g. tough,
words/phrases bough, cough, dough
it recognizes speech of only one Computational morphology
person
of coarticulation - different ways to - Processing of word structures for the computer to
pronounce a word/phrase be able to understand and use the word
of noise – computers do not have - Interwoven rules, exceptions, and word/morpheme
the ability to zero in on the voice of forms
the speaker (cocktail party effect) - Methods:
o Computers use statistics or pre- o Computer looks for roots and affixes
programmed words/phrases to decide o Stemming- affixes are repeatedly stripped
– Speech synthesis off
o creating electronic signals that stimulate
Computational syntax
phones and other features of speech →
assemble into words and phrases for output - Parsing (input)
to electronic speaker o Parser- computer program which tries to
o Many have attempted to create “talking replicate our “mental parser” – both use
machines” by using different materials to grammar to assign a phrase structure to a
imitate the mouth but now, we can produce string of words
these sounds electronically o 2 approaches:
Top-down – looks at rules first then
tries to apply the sentences
Bottom-up – it looks at the Computational Sign Language
sentence first, identifies its parts,
- Make use of a camera and visual processing
then looks for the rules
- Two purposes
o Problems
o Video dictionary
words with different syntactic
o Enable a computer to search through ASL
categories (e.g. orange)
videos for a particular sign
different ways to read sentences
- Generation (output)
o Assign words to the ideas or concepts
o Words must fit into phrases or sentences
which must comply with the syntax of the
language
o 2 approaches:
Top-down- highest-lowest level
categories
Bottom-up- words → combine
words for higher level categories
o Transition network- visualization of
grammar; uses nodes (circles) and arcs
(arrows)
Figure 1. Transition network for S -> NP VP
Computational semantics
– representation of the meaning of words and
morphemes in the computer, as well as the
meanings derived from their combinations
– Higher levels of linguistic processing
– 2 concerns:
o Produce semantic representation in the
computer
o Take semantic representation to produce
natural language output which has meaning
– Computer determines the concept it wants to
express → assembles components of meaning →
syntactic rules applied to produce grammatical
sentences
Computational pragmatics
- Pragmatics as the interaction between the real
world and the language system (e.g the switch v a
wire)
- Also used to determine meaning when two
expressions refer to the same object
