Sunteți pe pagina 1din 29

I.

Introduction

2. NLP and computational


linguistics
Natural Language Processing

Neurology & brain physiology

NLP Psycholinguistics
Biolinguistics
Language learning

Artificial intelligence
Damage to Broca's Area
Damage to Broca's area is called Broca's
asphasia.
A person suffering damage will likely
exhibit problems like words not coming
out properly formed, slowed or slurred
speech, prevention of producing words,
and possibility of not being able to
comprehend language.
Damage to Wernicke's area
known as Wenicke's aphasia would
result in people suffering from the loss
of ability to comprehend language, and
also the ability to speak words clearly;
the inability to properly put them
together.
the words would be out of order and
would make no sense.
NLP and AI
In computer science, NLP is an
interdisciplinary field;
The goal of this new field is to get computers
to perform useful tasks involving human
language, tasks like enabling human-machine
communication, improving human-human
communication, or simply doing useful
processing of text or speech.
(Daniel Jurafsky and James Martin, 2000)
NLP and AI
Computer tasks involving NLP:

automatic speech recognition and natural language


understanding (voice activated controls, digital dictation,
transcription of speech for the hearing-impaired, etc.);
machine translation;
information extraction (web-based question answering,
computer-aided summarization);
spelling correction;
grammar and style checking;
conversational agents (programs able to understand and
produce natural language speech recognition and speech
synthesis - with a high degree of autonomy and fluency).
Human NLP & computer NLP
Language = natural mechanism of complex
symbol systems
Computers = artificial mechanical symbol
systems
Human NLP = modular computational process
Computer NLP = modular computational
process
Both human & computer NLP improve
performance by integration of rules +
statistical analyses
Computer NLP
NLP in AI makes use of:

programming knowledge;
knowledge about how natural
languages work (human NLP);
knowledge about language itself;
Computational linguistics
a new branch of linguistics (language
study);
aims to start from the study of natural
languages in order to create the
machinery needed for a non-human
agent to exhibit various forms of
linguistic behaviour;
Conversational agent?

Are you sure you want to send


lesson_1 to the Recycle Bin?

Yes No Cancel
Conversational agents?
Commander Data, engage!
Whats going on in Datas mind?

Speech recognition in a noise sequence


Segmentation of noise continuum
Identification of parts structure
Identification of combinatorial principles
.
.
.
The role of computational
linguistics

Human NLP Computer NLP


NLP step 1
Tokenization

I/ can / fish/.
Tagging
PVN
P aux V
NLP - step 2
Morphological analysis: stem +
affixes/inflection
Stored forms
{fish, fishes, fished}
Rule + irregular forms
NLP step 3
Syntactic analysis parsing
Application of linguistic rules for how words combine

for a given language (tagging + parsing)


Generation of single/multiple structures (parse trees);

Choosing the right tree through various methods

(disambiguation based on statistical analysis, semantic


rules, discourse principles, etc.)
Context - free grammars

Systems of rules for braking a stretch of


language into constituent elements:
sentences, phrases, words.
Context-free grammars
Components:
S = SENTENCE
VP = VERB PHRASE
NP = NOUN PHRASE
Context-free grammars
Composition rules (non-terminals):
S [NP VB]

VP [VERB]
VP [AUXILIARY VERB]
VP [VERB NP]
VP [AUXILIARY VERB NP]

NP [DETERMINER NOUN]
NP [NOUN]
NP [PRONOUN]
Context-free grammars
Composition rules (terminals):

V {sleep, write, can, fish, love}


AUXILIARY {can, have, didnt}
N {door, fish, head}
DETERMINER {the, this, blue}
PRONOUN {I, we, you, she, he, it}
Parse tree 1

NP VP

Pronoun Auxiliary Verb

I can fish
Parse tree 2

NP VB

Pronoun Verb NP

I can fish
Disambiguation
Unification = adding rules to the syntactic
parser; unification-based grammar;
Learning patterns; Corpus
analysis
Statistical analysis; tools

Semantic disambiguation;
Discourse disambiguation (register analysis)
NLP Step 4
Semantic analysis:

Assigning meaning to components;


Applying compositionality;
Applying other semantic rules.
Semantic rules

Compositionality
VP [Aux V]

VP.meaning = Apply (Aux.meaning,

V.meaning)
Syntactic + semantic +???
I can fish = Eu stiu sa pescuiesc

NP VP

Pronoun Auxiliary Verb

I can fish
Linguistic knowledge involved
in NLP

Phonetics structure, quality/quantity of sounds


Morphology word components and their function;
Syntax structure of phrases/sentences and
functional relationships of sentence components;
Semantics word and sentence meaning;
Pragmatics f(x)= compositionality + context
Discourse f(x) = language + use
Key concepts in NLP
Tokenization
Tagging
Morphological analysis
Parsing (syntactic analysis)
Parse trees
Disambiguation (unification rules, statistical
data, etc.)
Semantic analysis
Compositionality

S-ar putea să vă placă și