Natural Language Processing

Key Points

• What & why NLP?

• Steps involves in language processing:
Morphological, syntactic, semantic, pragmatic,
discourse integration
• Parsing: bottom up vs. top down
• Parser Implementation: ATN
• Why semantic is important?
• CFG/CSG/TGG
Natural Language Processing (NLP)
• Natural Language?
– Refers to the language spoken/written by people, e.g. English,
Japanese, Bengali, as opposed to artificial languages, like C++, Java,
LISP, Prolog, etc.
• Natural Language Processing
– Applications that deal with natural language in a way or another
– Computers use (analyze, understand, generate) natural language
– A somewhat applied field

 Computational Linguistics
– Doing linguistics on computers
– More on the linguistic side than NLP, but closely related
– More theoretical
 Why NLP?
• Language is meant for
communicating about the
world
– NLP offers insights into
language
• Language is the medium
of the web
• Help in communication
Kismet: MIT Media Lab
– With computers
– With other humans (MT)
– HCI/HRI
 Why NLP?
• Classify text into categories
• Index and search large texts
Applications for • Automatic translation
processing large • Speech understanding
amounts of texts – Understand phone conversations
-require NLP • Information extraction
expertise – Extract useful information from resumes
• Automatic summarization
– Condense 1 book into 1 page
• Question answering
• Knowledge acquisition
• Text generations / dialogues
 NLP Involves ….
• Two tasks:
Processing written text
-Lexical
-Syntactic
-Semantic
-Pragmatic
Processing spoken language,
-Lexical, Syntactic, Semantic, Pragmatic+++
- Phonology
 NLP Involves ….
Natural Language Understanding (NLU): understanding & reasoning
while the input is NL
-Internal structure of the input NL
Natural Language Generation (NLG): generate other language

Natural Language Processing

NL Input Machine NL Output

Understanding Generation
 Linguistics and Language Processing

High-level

Low-level
 Steps in the NLP
• Morphological/Lexical Analysis: Individual words are
analyzed into their components
-non-word tokens such as punctuation are separated from
the words.
-Divide the text into paragraphs, sentences & words

• Syntactic Analysis: Linear sequences of words are

transformed into structures that show how the words
relate to each other.
-Boy the go to store (rejected due to violate the rules)
 Steps in the NLP
• Semantic Analysis: The structures created by the syntactic analyzer
are assigned meanings.
-Manush gass khai (syntactically correct but semantically not)

• Discourse integration: The meaning of an individual sentence may

depend on the sentences that precede it and may influence the
meanings of the sentences that follow it.
-you wanted it (it depends on the prior discourse context)

• Pragmatic Analysis: The structure representing what was said is

reinterpreted to determine what was actually meant.
-commonsense knowledge
-Do you know what is it? (request)
 Morphological Analysis
• Morphology is the analysis of words into
morphemes & conversely the synthesis of
words from morphemes
• Morphemes
-a smallest meaningful unit in the grammar
-a smallest linguistic unit that has semantic meaning
-a unit of language immediately below the ‘word level’
-a smallest part of a word that can carry a discrete meaning
 Morphological Analysis
Example:
The word “unbreakable” has 03 morphemes
1. un- a bound morpheme
2. -break- a free morpheme
3. -able- a bound morpheme
Also, un: prefix, able-suffix--affixes

How many morphemes?

“aagniporikhha” “unreliable”,
 Types of Morphemes
 Free Morphemes:
-can appear stand alone/free
town, dog
-with other lexemes
town hall, dog house
 Bound Morphemes
-appear only together with other morphemes to form a lexeme
un-: may be prefix/suffix
 Inflectional Morphemes
-modify a word’s tense, number, aspects etc
dog morpheme with plural marker
morpheme s becomes dogs
 Types of Morphemes
 Derivational Morphemes
-can be added to a word to derive another word
Addition of “-nees” to “happy”: “happiness”
 Root Morphemes
-primary lexical unit of a word. Roots can be either free or bound
morphemes
chatters has the inflectional root chatter but the lexical root chat
 Null Morphemes
-if is an invisible affix, also called zero morpheme represented as either
the figure zero (0), the empty set symbol , or its variant 
-Adding a null morpheme is called null affixation/null derivation/zero
derivation;
cat = cat + -0 = ROOT (“cat”) + singular
cats = cat + - s = ROOT (“cat”) + plural
 Syntactic Analysis
• Syntax
-Syntax is the structure of language.
-It is the grammatical arrangement of the words in a sentence
to show its relationship to one another in a sentence
-Syntax is a finite set of rules that specify a language
-Rules: Phase Structure Rules
Context-Free or Context-Sensitive
-Syntax rules govern proper sentence structure
-Syntax is represented by Parse tree:
-a way to show the structure of a language
fragment or by a list
 Syntactic Analysis
He ate the pizza
PS Rules
Input String S NP VP
NP  PRO
NP  ART N
VP  V NP

Parser Lexicon V-ate

PRO-he
N-pizza
ART-the
Parse Tree
 Classification Phrases
• Sentence (S)
• Noun Phrase (NP): Noun/Pronoun as head
-Optionally accompanied by a set of modifiers
Determiners: article (the, a)
Adjectives (the red ball)
the black cat, a cat on the mat
• Verb Phrase (VP): verb as head
“eat cheese”, “jump up and down”
• Adjective Phrase (AP): adjective as head
full of toys, good boy
• Adverbial Phrase (AdvP): adverb as head
very carefully
• Prepositional Phrase (PP): preposition as head
in love, over the rainbow
 Classification Phrases
• Determiner Phrase (DP): determiner as head
a little dog, the little dogs
• In English, determiners are usually placed before the
noun as a noun modifier that includes:
-articles (the, a),
-demonstratives (this, that),
-numerals (two, five, etc.),
-possessives (my, their, etc.),
-quantifiers(some, many, etc.)
 PS Rules
• PS rules are a way to describe language syntax.
• Rules determine what goes into phrase & how its
constituents are ordered.
• They are used to break a sentence down to its constituent
parts namely phrasal categories & lexical categories:
Phrasal category include: noun phrase, verb
phrase, prepositional phrase
Lexical category include: noun, verb,
adjective, adverb, others
 PS Rules
• PS rules are usually of the form A  B C
-meaning “constituent A is separated into sub-constituents
B&C
-Or simply “ A consists of B followed by C”
 S  NP VP Reads: S consists of an NP followed by a VP
-Means a sentence consists of a noun phrase followed by a
verb phrase
 NP  Det N1 Reads: NP consists of an Det followed by N1
-Means a noun phrase consists of a determiner followed by
a noun
PS Rules and Tress for NP
 PS Rules
S NP VP Bill printed the red file
NP  (Det) (ADJS) N S
NP  PRO
NP PN NP VP

ADJS  ADJ |ADJS V NP

PN
VP  V (NP)
N  file|printer Bill printed Det ADJS N
PN  Bill|Rahim|Bird
the ADJ file
PRO  I|he|she
Det  the|a|an red
ADJ  short|long|want|red
V  printed|created|want
 Parser
• A parser is a program, that accepts as input a sequence of words
in a natural language and breaks them up into parts (nouns, verbs,
& their attributes), to be managed by other programming.
• Parsing can be defined as the act of analyzing the grammatically
an utterance according to some specific grammar
• Parsing is the process to check, that a particular sequence of
words in sentence correspond to a language defined by its
grammar
• Parsing means show how we can get from the start symbol of the
grammar to the sentence of words using the production rules
• The output of a parser is a parse tree/Structural Representation
(SR)
 Properties of a Parse Tree
• Parse tree is a way of representing the output of a parser
 Each phrasal constituent found during parsing becomes a
branch node of the parse tree
 Every node corresponds either to an input word or to a non-
terminal
 Each level in the parse tree corresponds to the application
of one PS rule
 The words of the sentence become the leaves of the parse
tree
 There can be more than one parse tree for a single sentence
 Parse tree/Structural Representation (SR)
• For computation, a tree the boy ate a popsicle
must be represented as a S
record, a list or similar data
NP VP
structure.
List: Det N V NP
(S (NP ((Det the)
the boy Det N
(N boy))
ate
(VP (V ate) a popsicle
(NP (Det a)
(N popsicle)))))
 Top-down vs. Bottom-up Parsing
• To parse a sentence, it is necessary to find a
way in which the sentence could have been
generated from the stat symbols
• Two Ways: Top-down and Bottom-up
• The choice between these two approaches is similar
to the choice between forward and backward
reasoning in other problem-solving tasks.
• Sometimes these two approaches are combined to a
single method called “bottom-up parsing with top-
down filtering (hybrid approach)”.
 Top-down vs. Bottom-up Parsing
 Top-down Parsing:
Begin with start symbol and apply the grammar rules
forward until the symbols at the terminals of the tree
correspond to the components of the sentence being
parsed.
 Bottom-up parsing
Begin with the sentence to be parsed and apply the
grammar rules backward until a single tree whose
terminals are the words of the sentence and whose top
node is the start symbol has been produced.
 Top-down vs. bottom-up
• S NP VP •  Kathy jumped the horse
 N VP  N jumped the horse
 Kathy VP  N V the horse
® Kathy V NP ® N V Det horse
® Kathy jumped NP ® N V Det N
® Kathy jumped Det N ® NP V Det N
® Kathy jumped the N ® NP V NP
® Kathy jumped the horse ® NP VP
®S
 Augmented Transition Networks
• ATN: Augmented Transition Network
• Top-down parsing procedure that allows various
kinds of knowledge to be incorporated into the
parsing system so it can operate efficiently
• Similar to a finite state machine in which the
class of labels that can be attached to the arcs
that define transitions between states has been
augmented.
 Semantic Analysis
• Producing a syntactic parse of a sentence is only the
first step toward understanding it.
• We must still produce a representation of the meaning
of the sentence.
• There is no single definitive language in which all
sentence meaning can be described.
• Need to define vocabulary (features)
• The choice of a target language for any particular
natural language understanding program must depend
on what is to be done with the meanings once they
are constructed.
 Semantic Analysis
• Must do two important things
-it must map individual words into appropriate
objects in the knowledge base or database
-it must create the correct structures to
correspond to the way the meaning of the
individual words combine with each other
 Choice of Target Language in Semantic
Analysis
• There are two broad families of target languages that are used
in NL systems, depending on the role that the natural
language system is playing in a larger system:
– When natural language is being considered as a phenomenon on its
own, as for example when one builds a program whose goal is to read
text and then answer questions about it, a target language can be
designed specifically to support language processing.
– When natural language is being used as an interface language to
another program ( such as a db query system or an expert system),
then the target language must be legal input to that other program.
Thus the design of the target language is driven by the backend
program.
 Lexical processing
• The first step in any semantic processing system is to look up
the individual words in a dictionary ( or lexicon) and extract
their meanings.
• Many words have several meanings, and it may not be
possible to choose the correct one just by looking at the word
itself.
• The process of determining the correct meaning of an
individual word is called word sense disambiguation or lexical
disambiguation.
Diamond
-a geometrical shape with 4 equal sides
-a baseball field
-an extremely hard & valuable gemstone
• It is done by associating, with each word in lexicon,
information about the contexts in which each of the word’s
senses may appear.
 Lexical Processing
• Each of the words in a sentence can serve as part of the
context in which the meanings of the other words must be
determined
• Sometimes only very straightforward info about each word
sense is necessary.
• For example, baseball field interpretation of diamond could
be marked as a LOCATION.
• Some useful semantic markers are :
– PHYSICAL-OBJECT
– ANIMATE-OBJECT
– ABSTRACT-OBJECT
 Using these markers, the correct meaning of a word in the
sentence can be computed
 Sentence-Level Processing
• Several approaches to the problem of creating a semantic
representation of a sentence have been developed, including
the following:
– Semantic grammars, which combine syntactic, semantic and
pragmatic knowledge into a single set of rules in the form of grammar.
– Case grammars, in which the structure that is built by the parser
contains some semantic information, although further interpretation
may also be necessary.
– Conceptual parsing in which syntactic and semantic knowledge are
combined into a single interpretation system that is driven by the
semantic knowledge.
– Approximately compositional semantic interpretation, in which
semantic processing is applied to the result of performing a syntactic
parse
 Semantic Grammar
• A semantic grammar is a CFG in which the
choice of non-terminals and production rules is
governed by semantic as well as syntactic
function.
• a semantic action associated with each
grammar rule.
• The result of parsing and applying all the
associated semantic actions is the meaning of
the sentence.
 A semantic grammar
• S-> what is FILE-PROPERTY of FILE?
• { query FILE.FILE-PROPERTY}
• S-> I want to ACTION
• { command ACTION}
• FILE-PROPERTY -> the FILE-PROP
• {FILE-PROP}
• FILE-PROP -> extension | protection | creation date | owner
• {value}
• FILE -> FILE-NAME | FILE1
• {value}
• FILE1 -> USER’s FILE2
• { FILE2.owner: USER }
• FILE1 -> FILE2
• { FILE2}
• FILE2 -> EXT file
• { instance: file-struct extension: EXT}
• EXT -> .init | .txt | .lsp | .for | .ps | .mss
• value
• ACTION -> print FILE
• { instance: printing object : FILE }
• ACTION -> print FILE on PRINTER
• { instance : printing object : FILE printer : PRINTER }
• USER -> Bill | susan
• { value }
 Parsing with Semantic Grammars
• A semantic grammar can be used by a parsing system
Pros:
-when parse tree is complete, the result can be used immediately
without the additional stage processing
-Many ambiguities can be avoided
-syntactic issues that do not affect the semantic issues can be
ignored
Cons
-the number of rules required can become very large
-because the number of grammar rules may be large, the parsing
process may be expensive
I want to print Bill’s.int file
 Case Grammars
• Provide a different approach to the problem of
how syntactic & semantic interpretation can
be combined
• Rules are written to describe syntactic rather
than semantic regularities
• But the structures produce correspond to
semantic relations rather than to strictly
syntactic ones
 An Example

• The semantic roles of “Susan” & “the file” are

identical, but syntactic roles are reversed.

• Case grammars:
(printed (agent Susan)
(object File))
 Example continued

• Syntactic structure are same

• Case1: Mother is the subject of baked
• Case2: the pie is the subject

• But the relationship between Mother and baking is very different

from that between the pie and baking
 Example Continued
• A case grammar analysis reflects this
difference
• (baked (agent Mother)
(timeperiod 3-hours))
(baked (object Pie)
(timeperiod 3-hours))
The semantic roles of “mother” and “the pie”
are explicit
 Deep cases
• (A) Agent-Instigator of the action (animate)
• (I) Instrument- Cause of the event/object used in causing the event (inanimate)
• (D) Dative-Entity affected by the action (animate)
• (F) Factitive-Object/being resulting from the event
• (L) Locative-Place of the event
• (S) Source-Place from which something moves
• (G) Goal-Place to which something moves
• (B) Beneficiary-being on whose behalf the event occurred (animate)
• (T) Time-time at which the event occurred
• (O) Object-Entity that is acted upon/that changes, the most general case
 Parsing process directed by lexical entries
associated with each verb
open [_ _ O (I) (A)] Languages have rules for mapping
The door opened from underlying case structure to
John opened the door surface syntactic forms
The wind opened the door If A is present, it is the subject.
Otherwise, if I is present, it is the
John opened the door with a chisel
subject
Die [_ _ D]
John died
kill [_ _ D (I) (A)]
Bill killed John
Bill killed John with a knife
run [_ _ A]
John ran
want [_ _ A O]
John wanted some ice cream
John wanted Mary to go to the store
 Lexical and Compositional Semantics
• Input sentences are • The meaning of an expression is
derived from the meanings of the
transformed through the
parts of the expression. The
use of domain dependent target knowledge structures
semantic rewrite rules constructed are typically logic
which create the target expressions
knowledge structures. • Sample24 contains silicon
(S DCL
• The boy drunk a soda
(INGEST(ACTION(PP(MAME boy) (CLASS PHYS-OBJ)
(NP (N (Sample24)))
(TYPE ANIMATE) (REF DEF))) (AUX (TENSE (PREENT)))
(OBJECT (PP (NAME soda) (CLASS PHYS-OBJ)
(TYPE INANIMATE) (REF INDEF))) (VP (V (contain))
(TENSE PAST))
(NP (N silicon))))
 Try Yourself
Design a set of PS rules to parse the Bangla
simple sentences
Generate the Parse tree for the following
sentences by using your proposed PS rules:
- Rahim ekti boi porche
- CUET ekti engineering university
- Javed khub bhalo gan gay
- Ek kahana pakhi ahashe urche
 Conclusion
• Why NLP is important?
• Steps involves in
language processing
• Morphological analysis
• Syntactic analysis using
PS rules
• Structural
Representation (SR):
Parse tree Topics Covers:
Artificial Intelligence: Rich & Knight
Artificial Intelligent & Expert System: Patterson