Artificial Intelligence

George F Luger
ARTIFICIAL INTELLIGENCE 6th edition
Structures and Strategies for Complex Problem Solving
Understanding Natural Language
Luger: Artificial Intelligence, 6th edition. © Pearson Education Limited, 2009
15.0 Role of Knowledge in Language
Understanding
15.1 Deconstructing Language: A
Symbolic Analysis
15.2 Syntax
15.3 Syntax and Knowledge with ATN
Parsers
15.4 Stochastic Tools for Language Analysis
15.5 Natural Language Applications
15.6 Epilogue and References
15.7 Exercises
1

Fig 15.1 A blocks world, adapted from Winograd (1972).
2

To manage this complexity, linguists have defined different levels of analysis for
natural language:
3

Fig 15.2 Stages in producing an internal representation of a sentence.
4

Fig 15.3 Parse tree for the sentence “The man bites the dog.”
5

Sentence −−> . Noun Verb predict: Noun followed by a Verb
Noun −−> . mary predict: mary
Noun --> mary . scanned: mary
Sentence --> Noun . Verb completed: Noun; predict: Verb
Verb --> . runs predict: runs
Verb --> runs . scanned: runs
Sentence --> Noun Verb . completed: Verb,
completed: sentence
Luger: Artificial Intelligence, 6th edition. © Pearson Education Limited, 2009 6

Figure 15.4 The relationship of dotted r ules to the generation of a parse tree.
Noun .Verb Noun Verb .
Noun
mary .
mar y
runs .
runs
Sentence
.Noun Verb
.
..
.
.
Verb

The chart for mary runs, with three state lists, is:
S0: [($ --> . S), start
(S --> . Noun Verb)] predictor
S1: [(Noun --> mary .), scanner
(S --> Noun . Verb)] completer
S2: [(Verb --> runs .)] scanner
(S --> Noun Verb .), completer
($ --> S .)] completer

function EARLEY-PARSE(words, grammar) returns chart
begin
chart := empty
ADDTOCHART(($ Æ . S, [0, 0]), chart[0]) % dummy start state
for i from 0 to LENGTH(words) do
for each state in chart[i] do
if rule_rhs(state) = … . A … and A is not a part of speech
then PREDICTOR(state)
else if rule_rhs(state) = … . L … % L is part of speech
then SCANNER(state)
else COMLETER(state) % rule_rhs = RHS
end

procedure PREDICTOR((A Æ … . B …, [i, j]))
begin
for each (B Æ RHS) in grammar do
ADDTOCHART((B Æ . RHS, [j, j]), chart[j])
end
procedure SCANNER((A Æ … . L …, [i, j]))
begin
if (L Æ word[j]) is_in grammar
then ADDTOCHART((L Æ word[j] ., [j, j + 1]), chart[j + 1])
end
procedure COMPLETER((B Æ … ., [j, k]))
begin
for each (A Æ … . B …, [i, j]) in chart[j] do
ADDTOCHART((A Æ … B . …, [i, k]), chart[k])
end
procedure ADDTOCHART(state, state-list)
begin
if state is not in state-list
then ADDTOEND(state, state-list)
end

Fig 15.5 Transition network definition of a simple English grammar.
11

Pseudo-code for a transition network parser appears on the following two slides. It
is defined using two mutually recursive functions, parse and transition.
continued…
12

continued from previous slide
13

Fig 15.6 Trace of a transition network parse of the sentence “Dog bites.”
14

Fig 15.7 Structures representing the sentence, noun phrase, and verb phrase
nonterminals of the grammar.
15

Fig 15.8 Dictionary entries for a simple ATN
16

Fig 15.9 An ATN grammar that checks number agreement and builds a parse tree.
continued…
17

Fig 15.9 continued from previous slide.
continued…
18

continued…
19

20

Fig 10 Parse tree for the sentence “The dog likes a man” returned by an ATN
parser.
21

Fig 15.11 Type hierarchy used in “dogs world” example.
22

Fig 15.12 Case frames for the verbs “like” and “bite.”
23

Rules for our example are described as pseudo-code procedures. In each
procedure, if a specified join or other test fails, that interpretation is rejected as
semantically incorrect.
continued…
24

continued from previous slide.
25

Fig 15.13 Construction of a semantic representation from the parse tree of Figure
15.10.
26

Fig 15.14 Two different parses of “Print the file on the printer.”
27

Fig 15.15 Conceptual graph for the question “Who loves Jane?”
28

Fig 15.16 Two relations in an employee database.
29

Fig 15.17 Entity-relationship diagrams of the manager_of_hire and
employee_salary relations.
30

Fig 15.18 Knowledge base entry for “hire” queries.
31

Fig 15.19 Development of a database query from the graph of a natural language
input..
32

Fig 15.20 Sample text, template summary, and information extraction for computer
science advertisement.

Fig 15.21 An architecture for information extraction, from Cardie (1997).
34

Artificial Intelligence

More Related Content

What's hot (6)

Viewers also liked (17)

More from Muhammad Ahad (20)

Recently uploaded (20)

Artificial Intelligence