You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

885 lines
31 KiB
Plaintext

.. Copyright (C) 2001-2020 NLTK Project
.. For license information, see LICENSE.TXT
=========
Parsing
=========
Unit tests for the Context Free Grammar class
---------------------------------------------
>>> from nltk import Nonterminal, nonterminals, Production, CFG
>>> nt1 = Nonterminal('NP')
>>> nt2 = Nonterminal('VP')
>>> nt1.symbol()
'NP'
>>> nt1 == Nonterminal('NP')
True
>>> nt1 == nt2
False
>>> S, NP, VP, PP = nonterminals('S, NP, VP, PP')
>>> N, V, P, DT = nonterminals('N, V, P, DT')
>>> prod1 = Production(S, [NP, VP])
>>> prod2 = Production(NP, [DT, NP])
>>> prod1.lhs()
S
>>> prod1.rhs()
(NP, VP)
>>> prod1 == Production(S, [NP, VP])
True
>>> prod1 == prod2
False
>>> grammar = CFG.fromstring("""
... S -> NP VP
... PP -> P NP
... NP -> 'the' N | N PP | 'the' N PP
... VP -> V NP | V PP | V NP PP
... N -> 'cat'
... N -> 'dog'
... N -> 'rug'
... V -> 'chased'
... V -> 'sat'
... P -> 'in'
... P -> 'on'
... """)
Unit tests for the rd (Recursive Descent Parser) class
------------------------------------------------------
Create and run a recursive descent parser over both a syntactically ambiguous
and unambiguous sentence.
>>> from nltk.parse import RecursiveDescentParser
>>> rd = RecursiveDescentParser(grammar)
>>> sentence1 = 'the cat chased the dog'.split()
>>> sentence2 = 'the cat chased the dog on the rug'.split()
>>> for t in rd.parse(sentence1):
... print(t)
(S (NP the (N cat)) (VP (V chased) (NP the (N dog))))
>>> for t in rd.parse(sentence2):
... print(t)
(S
(NP the (N cat))
(VP (V chased) (NP the (N dog) (PP (P on) (NP the (N rug))))))
(S
(NP the (N cat))
(VP (V chased) (NP the (N dog)) (PP (P on) (NP the (N rug)))))
(dolist (expr doctest-font-lock-keywords)
(add-to-list 'font-lock-keywords expr))
font-lock-keywords
(add-to-list 'font-lock-keywords
(car doctest-font-lock-keywords))
Unit tests for the sr (Shift Reduce Parser) class
-------------------------------------------------
Create and run a shift reduce parser over both a syntactically ambiguous
and unambiguous sentence. Note that unlike the recursive descent parser, one
and only one parse is ever returned.
>>> from nltk.parse import ShiftReduceParser
>>> sr = ShiftReduceParser(grammar)
>>> sentence1 = 'the cat chased the dog'.split()
>>> sentence2 = 'the cat chased the dog on the rug'.split()
>>> for t in sr.parse(sentence1):
... print(t)
(S (NP the (N cat)) (VP (V chased) (NP the (N dog))))
The shift reduce parser uses heuristics to decide what to do when there are
multiple possible shift or reduce operations available - for the supplied
grammar clearly the wrong operation is selected.
>>> for t in sr.parse(sentence2):
... print(t)
Unit tests for the Chart Parser class
-------------------------------------
We use the demo() function for testing.
We must turn off showing of times.
>>> import nltk
First we test tracing with a short sentence
>>> nltk.parse.chart.demo(2, print_times=False, trace=1,
... sent='I saw a dog', numparses=1)
* Sentence:
I saw a dog
['I', 'saw', 'a', 'dog']
<BLANKLINE>
* Strategy: Bottom-up
<BLANKLINE>
|. I . saw . a . dog .|
|[---------] . . .| [0:1] 'I'
|. [---------] . .| [1:2] 'saw'
|. . [---------] .| [2:3] 'a'
|. . . [---------]| [3:4] 'dog'
|> . . . .| [0:0] NP -> * 'I'
|[---------] . . .| [0:1] NP -> 'I' *
|> . . . .| [0:0] S -> * NP VP
|> . . . .| [0:0] NP -> * NP PP
|[---------> . . .| [0:1] S -> NP * VP
|[---------> . . .| [0:1] NP -> NP * PP
|. > . . .| [1:1] Verb -> * 'saw'
|. [---------] . .| [1:2] Verb -> 'saw' *
|. > . . .| [1:1] VP -> * Verb NP
|. > . . .| [1:1] VP -> * Verb
|. [---------> . .| [1:2] VP -> Verb * NP
|. [---------] . .| [1:2] VP -> Verb *
|. > . . .| [1:1] VP -> * VP PP
|[-------------------] . .| [0:2] S -> NP VP *
|. [---------> . .| [1:2] VP -> VP * PP
|. . > . .| [2:2] Det -> * 'a'
|. . [---------] .| [2:3] Det -> 'a' *
|. . > . .| [2:2] NP -> * Det Noun
|. . [---------> .| [2:3] NP -> Det * Noun
|. . . > .| [3:3] Noun -> * 'dog'
|. . . [---------]| [3:4] Noun -> 'dog' *
|. . [-------------------]| [2:4] NP -> Det Noun *
|. . > . .| [2:2] S -> * NP VP
|. . > . .| [2:2] NP -> * NP PP
|. [-----------------------------]| [1:4] VP -> Verb NP *
|. . [------------------->| [2:4] S -> NP * VP
|. . [------------------->| [2:4] NP -> NP * PP
|[=======================================]| [0:4] S -> NP VP *
|. [----------------------------->| [1:4] VP -> VP * PP
Nr edges in chart: 33
(S (NP I) (VP (Verb saw) (NP (Det a) (Noun dog))))
<BLANKLINE>
Then we test the different parsing Strategies.
Note that the number of edges differ between the strategies.
Top-down
>>> nltk.parse.chart.demo(1, print_times=False, trace=0,
... sent='I saw John with a dog', numparses=2)
* Sentence:
I saw John with a dog
['I', 'saw', 'John', 'with', 'a', 'dog']
<BLANKLINE>
* Strategy: Top-down
<BLANKLINE>
Nr edges in chart: 48
(S
(NP I)
(VP (Verb saw) (NP (NP John) (PP with (NP (Det a) (Noun dog))))))
(S
(NP I)
(VP (VP (Verb saw) (NP John)) (PP with (NP (Det a) (Noun dog)))))
<BLANKLINE>
Bottom-up
>>> nltk.parse.chart.demo(2, print_times=False, trace=0,
... sent='I saw John with a dog', numparses=2)
* Sentence:
I saw John with a dog
['I', 'saw', 'John', 'with', 'a', 'dog']
<BLANKLINE>
* Strategy: Bottom-up
<BLANKLINE>
Nr edges in chart: 53
(S
(NP I)
(VP (VP (Verb saw) (NP John)) (PP with (NP (Det a) (Noun dog)))))
(S
(NP I)
(VP (Verb saw) (NP (NP John) (PP with (NP (Det a) (Noun dog))))))
<BLANKLINE>
Bottom-up Left-Corner
>>> nltk.parse.chart.demo(3, print_times=False, trace=0,
... sent='I saw John with a dog', numparses=2)
* Sentence:
I saw John with a dog
['I', 'saw', 'John', 'with', 'a', 'dog']
<BLANKLINE>
* Strategy: Bottom-up left-corner
<BLANKLINE>
Nr edges in chart: 36
(S
(NP I)
(VP (VP (Verb saw) (NP John)) (PP with (NP (Det a) (Noun dog)))))
(S
(NP I)
(VP (Verb saw) (NP (NP John) (PP with (NP (Det a) (Noun dog))))))
<BLANKLINE>
Left-Corner with Bottom-Up Filter
>>> nltk.parse.chart.demo(4, print_times=False, trace=0,
... sent='I saw John with a dog', numparses=2)
* Sentence:
I saw John with a dog
['I', 'saw', 'John', 'with', 'a', 'dog']
<BLANKLINE>
* Strategy: Filtered left-corner
<BLANKLINE>
Nr edges in chart: 28
(S
(NP I)
(VP (VP (Verb saw) (NP John)) (PP with (NP (Det a) (Noun dog)))))
(S
(NP I)
(VP (Verb saw) (NP (NP John) (PP with (NP (Det a) (Noun dog))))))
<BLANKLINE>
The stepping chart parser
>>> nltk.parse.chart.demo(5, print_times=False, trace=1,
... sent='I saw John with a dog', numparses=2)
* Sentence:
I saw John with a dog
['I', 'saw', 'John', 'with', 'a', 'dog']
<BLANKLINE>
* Strategy: Stepping (top-down vs bottom-up)
<BLANKLINE>
*** SWITCH TO TOP DOWN
|[------] . . . . .| [0:1] 'I'
|. [------] . . . .| [1:2] 'saw'
|. . [------] . . .| [2:3] 'John'
|. . . [------] . .| [3:4] 'with'
|. . . . [------] .| [4:5] 'a'
|. . . . . [------]| [5:6] 'dog'
|> . . . . . .| [0:0] S -> * NP VP
|> . . . . . .| [0:0] NP -> * NP PP
|> . . . . . .| [0:0] NP -> * Det Noun
|> . . . . . .| [0:0] NP -> * 'I'
|[------] . . . . .| [0:1] NP -> 'I' *
|[------> . . . . .| [0:1] S -> NP * VP
|[------> . . . . .| [0:1] NP -> NP * PP
|. > . . . . .| [1:1] VP -> * VP PP
|. > . . . . .| [1:1] VP -> * Verb NP
|. > . . . . .| [1:1] VP -> * Verb
|. > . . . . .| [1:1] Verb -> * 'saw'
|. [------] . . . .| [1:2] Verb -> 'saw' *
|. [------> . . . .| [1:2] VP -> Verb * NP
|. [------] . . . .| [1:2] VP -> Verb *
|[-------------] . . . .| [0:2] S -> NP VP *
|. [------> . . . .| [1:2] VP -> VP * PP
*** SWITCH TO BOTTOM UP
|. . > . . . .| [2:2] NP -> * 'John'
|. . . > . . .| [3:3] PP -> * 'with' NP
|. . . > . . .| [3:3] Prep -> * 'with'
|. . . . > . .| [4:4] Det -> * 'a'
|. . . . . > .| [5:5] Noun -> * 'dog'
|. . [------] . . .| [2:3] NP -> 'John' *
|. . . [------> . .| [3:4] PP -> 'with' * NP
|. . . [------] . .| [3:4] Prep -> 'with' *
|. . . . [------] .| [4:5] Det -> 'a' *
|. . . . . [------]| [5:6] Noun -> 'dog' *
|. [-------------] . . .| [1:3] VP -> Verb NP *
|[--------------------] . . .| [0:3] S -> NP VP *
|. [-------------> . . .| [1:3] VP -> VP * PP
|. . > . . . .| [2:2] S -> * NP VP
|. . > . . . .| [2:2] NP -> * NP PP
|. . . . > . .| [4:4] NP -> * Det Noun
|. . [------> . . .| [2:3] S -> NP * VP
|. . [------> . . .| [2:3] NP -> NP * PP
|. . . . [------> .| [4:5] NP -> Det * Noun
|. . . . [-------------]| [4:6] NP -> Det Noun *
|. . . [--------------------]| [3:6] PP -> 'with' NP *
|. [----------------------------------]| [1:6] VP -> VP PP *
*** SWITCH TO TOP DOWN
|. . > . . . .| [2:2] NP -> * Det Noun
|. . . . > . .| [4:4] NP -> * NP PP
|. . . > . . .| [3:3] VP -> * VP PP
|. . . > . . .| [3:3] VP -> * Verb NP
|. . . > . . .| [3:3] VP -> * Verb
|[=========================================]| [0:6] S -> NP VP *
|. [---------------------------------->| [1:6] VP -> VP * PP
|. . [---------------------------]| [2:6] NP -> NP PP *
|. . . . [------------->| [4:6] NP -> NP * PP
|. [----------------------------------]| [1:6] VP -> Verb NP *
|. . [--------------------------->| [2:6] S -> NP * VP
|. . [--------------------------->| [2:6] NP -> NP * PP
|[=========================================]| [0:6] S -> NP VP *
|. [---------------------------------->| [1:6] VP -> VP * PP
|. . . . . . >| [6:6] VP -> * VP PP
|. . . . . . >| [6:6] VP -> * Verb NP
|. . . . . . >| [6:6] VP -> * Verb
*** SWITCH TO BOTTOM UP
|. . . . > . .| [4:4] S -> * NP VP
|. . . . [------------->| [4:6] S -> NP * VP
*** SWITCH TO TOP DOWN
*** SWITCH TO BOTTOM UP
*** SWITCH TO TOP DOWN
*** SWITCH TO BOTTOM UP
*** SWITCH TO TOP DOWN
*** SWITCH TO BOTTOM UP
Nr edges in chart: 61
(S
(NP I)
(VP (VP (Verb saw) (NP John)) (PP with (NP (Det a) (Noun dog)))))
(S
(NP I)
(VP (Verb saw) (NP (NP John) (PP with (NP (Det a) (Noun dog))))))
<BLANKLINE>
Unit tests for the Incremental Chart Parser class
-------------------------------------------------
The incremental chart parsers are defined in earleychart.py.
We use the demo() function for testing. We must turn off showing of times.
>>> import nltk
Earley Chart Parser
>>> nltk.parse.earleychart.demo(print_times=False, trace=1,
... sent='I saw John with a dog', numparses=2)
* Sentence:
I saw John with a dog
['I', 'saw', 'John', 'with', 'a', 'dog']
<BLANKLINE>
|. I . saw . John . with . a . dog .|
|[------] . . . . .| [0:1] 'I'
|. [------] . . . .| [1:2] 'saw'
|. . [------] . . .| [2:3] 'John'
|. . . [------] . .| [3:4] 'with'
|. . . . [------] .| [4:5] 'a'
|. . . . . [------]| [5:6] 'dog'
|> . . . . . .| [0:0] S -> * NP VP
|> . . . . . .| [0:0] NP -> * NP PP
|> . . . . . .| [0:0] NP -> * Det Noun
|> . . . . . .| [0:0] NP -> * 'I'
|[------] . . . . .| [0:1] NP -> 'I' *
|[------> . . . . .| [0:1] S -> NP * VP
|[------> . . . . .| [0:1] NP -> NP * PP
|. > . . . . .| [1:1] VP -> * VP PP
|. > . . . . .| [1:1] VP -> * Verb NP
|. > . . . . .| [1:1] VP -> * Verb
|. > . . . . .| [1:1] Verb -> * 'saw'
|. [------] . . . .| [1:2] Verb -> 'saw' *
|. [------> . . . .| [1:2] VP -> Verb * NP
|. [------] . . . .| [1:2] VP -> Verb *
|[-------------] . . . .| [0:2] S -> NP VP *
|. [------> . . . .| [1:2] VP -> VP * PP
|. . > . . . .| [2:2] NP -> * NP PP
|. . > . . . .| [2:2] NP -> * Det Noun
|. . > . . . .| [2:2] NP -> * 'John'
|. . [------] . . .| [2:3] NP -> 'John' *
|. [-------------] . . .| [1:3] VP -> Verb NP *
|. . [------> . . .| [2:3] NP -> NP * PP
|. . . > . . .| [3:3] PP -> * 'with' NP
|[--------------------] . . .| [0:3] S -> NP VP *
|. [-------------> . . .| [1:3] VP -> VP * PP
|. . . [------> . .| [3:4] PP -> 'with' * NP
|. . . . > . .| [4:4] NP -> * NP PP
|. . . . > . .| [4:4] NP -> * Det Noun
|. . . . > . .| [4:4] Det -> * 'a'
|. . . . [------] .| [4:5] Det -> 'a' *
|. . . . [------> .| [4:5] NP -> Det * Noun
|. . . . . > .| [5:5] Noun -> * 'dog'
|. . . . . [------]| [5:6] Noun -> 'dog' *
|. . . . [-------------]| [4:6] NP -> Det Noun *
|. . . [--------------------]| [3:6] PP -> 'with' NP *
|. . . . [------------->| [4:6] NP -> NP * PP
|. . [---------------------------]| [2:6] NP -> NP PP *
|. [----------------------------------]| [1:6] VP -> VP PP *
|[=========================================]| [0:6] S -> NP VP *
|. [---------------------------------->| [1:6] VP -> VP * PP
|. [----------------------------------]| [1:6] VP -> Verb NP *
|. . [--------------------------->| [2:6] NP -> NP * PP
|[=========================================]| [0:6] S -> NP VP *
|. [---------------------------------->| [1:6] VP -> VP * PP
(S
(NP I)
(VP (VP (Verb saw) (NP John)) (PP with (NP (Det a) (Noun dog)))))
(S
(NP I)
(VP (Verb saw) (NP (NP John) (PP with (NP (Det a) (Noun dog))))))
Unit tests for LARGE context-free grammars
------------------------------------------
Reading the ATIS grammar.
>>> grammar = nltk.data.load('grammars/large_grammars/atis.cfg')
>>> grammar
<Grammar with 5517 productions>
Reading the test sentences.
>>> sentences = nltk.data.load('grammars/large_grammars/atis_sentences.txt')
>>> sentences = nltk.parse.util.extract_test_sentences(sentences)
>>> len(sentences)
98
>>> testsentence = sentences[22]
>>> testsentence[0]
['show', 'me', 'northwest', 'flights', 'to', 'detroit', '.']
>>> testsentence[1]
17
>>> sentence = testsentence[0]
Now we test all different parsing strategies.
Note that the number of edges differ between the strategies.
Bottom-up parsing.
>>> parser = nltk.parse.BottomUpChartParser(grammar)
>>> chart = parser.chart_parse(sentence)
>>> print((chart.num_edges()))
7661
>>> print((len(list(chart.parses(grammar.start())))))
17
Bottom-up Left-corner parsing.
>>> parser = nltk.parse.BottomUpLeftCornerChartParser(grammar)
>>> chart = parser.chart_parse(sentence)
>>> print((chart.num_edges()))
4986
>>> print((len(list(chart.parses(grammar.start())))))
17
Left-corner parsing with bottom-up filter.
>>> parser = nltk.parse.LeftCornerChartParser(grammar)
>>> chart = parser.chart_parse(sentence)
>>> print((chart.num_edges()))
1342
>>> print((len(list(chart.parses(grammar.start())))))
17
Top-down parsing.
>>> parser = nltk.parse.TopDownChartParser(grammar)
>>> chart = parser.chart_parse(sentence)
>>> print((chart.num_edges()))
28352
>>> print((len(list(chart.parses(grammar.start())))))
17
Incremental Bottom-up parsing.
>>> parser = nltk.parse.IncrementalBottomUpChartParser(grammar)
>>> chart = parser.chart_parse(sentence)
>>> print((chart.num_edges()))
7661
>>> print((len(list(chart.parses(grammar.start())))))
17
Incremental Bottom-up Left-corner parsing.
>>> parser = nltk.parse.IncrementalBottomUpLeftCornerChartParser(grammar)
>>> chart = parser.chart_parse(sentence)
>>> print((chart.num_edges()))
4986
>>> print((len(list(chart.parses(grammar.start())))))
17
Incremental Left-corner parsing with bottom-up filter.
>>> parser = nltk.parse.IncrementalLeftCornerChartParser(grammar)
>>> chart = parser.chart_parse(sentence)
>>> print((chart.num_edges()))
1342
>>> print((len(list(chart.parses(grammar.start())))))
17
Incremental Top-down parsing.
>>> parser = nltk.parse.IncrementalTopDownChartParser(grammar)
>>> chart = parser.chart_parse(sentence)
>>> print((chart.num_edges()))
28352
>>> print((len(list(chart.parses(grammar.start())))))
17
Earley parsing. This is similar to the incremental top-down algorithm.
>>> parser = nltk.parse.EarleyChartParser(grammar)
>>> chart = parser.chart_parse(sentence)
>>> print((chart.num_edges()))
28352
>>> print((len(list(chart.parses(grammar.start())))))
17
Unit tests for the Probabilistic CFG class
------------------------------------------
>>> from nltk.corpus import treebank
>>> from itertools import islice
>>> from nltk.grammar import PCFG, induce_pcfg, toy_pcfg1, toy_pcfg2
Create a set of PCFG productions.
>>> grammar = PCFG.fromstring("""
... A -> B B [.3] | C B C [.7]
... B -> B D [.5] | C [.5]
... C -> 'a' [.1] | 'b' [0.9]
... D -> 'b' [1.0]
... """)
>>> prod = grammar.productions()[0]
>>> prod
A -> B B [0.3]
>>> prod.lhs()
A
>>> prod.rhs()
(B, B)
>>> print((prod.prob()))
0.3
>>> grammar.start()
A
>>> grammar.productions()
[A -> B B [0.3], A -> C B C [0.7], B -> B D [0.5], B -> C [0.5], C -> 'a' [0.1], C -> 'b' [0.9], D -> 'b' [1.0]]
Induce some productions using parsed Treebank data.
>>> productions = []
>>> for fileid in treebank.fileids()[:2]:
... for t in treebank.parsed_sents(fileid):
... productions += t.productions()
>>> grammar = induce_pcfg(S, productions)
>>> grammar
<Grammar with 71 productions>
>>> sorted(grammar.productions(lhs=Nonterminal('PP')))[:2]
[PP -> IN NP [1.0]]
>>> sorted(grammar.productions(lhs=Nonterminal('NNP')))[:2]
[NNP -> 'Agnew' [0.0714286], NNP -> 'Consolidated' [0.0714286]]
>>> sorted(grammar.productions(lhs=Nonterminal('JJ')))[:2]
[JJ -> 'British' [0.142857], JJ -> 'former' [0.142857]]
>>> sorted(grammar.productions(lhs=Nonterminal('NP')))[:2]
[NP -> CD NNS [0.133333], NP -> DT JJ JJ NN [0.0666667]]
Unit tests for the Probabilistic Chart Parse classes
----------------------------------------------------
>>> tokens = "Jack saw Bob with my cookie".split()
>>> grammar = toy_pcfg2
>>> print(grammar)
Grammar with 23 productions (start state = S)
S -> NP VP [1.0]
VP -> V NP [0.59]
VP -> V [0.4]
VP -> VP PP [0.01]
NP -> Det N [0.41]
NP -> Name [0.28]
NP -> NP PP [0.31]
PP -> P NP [1.0]
V -> 'saw' [0.21]
V -> 'ate' [0.51]
V -> 'ran' [0.28]
N -> 'boy' [0.11]
N -> 'cookie' [0.12]
N -> 'table' [0.13]
N -> 'telescope' [0.14]
N -> 'hill' [0.5]
Name -> 'Jack' [0.52]
Name -> 'Bob' [0.48]
P -> 'with' [0.61]
P -> 'under' [0.39]
Det -> 'the' [0.41]
Det -> 'a' [0.31]
Det -> 'my' [0.28]
Create several parsers using different queuing strategies and show the
resulting parses.
>>> from nltk.parse import pchart
>>> parser = pchart.InsideChartParser(grammar)
>>> for t in parser.parse(tokens):
... print(t)
(S
(NP (Name Jack))
(VP
(V saw)
(NP
(NP (Name Bob))
(PP (P with) (NP (Det my) (N cookie)))))) (p=6.31607e-06)
(S
(NP (Name Jack))
(VP
(VP (V saw) (NP (Name Bob)))
(PP (P with) (NP (Det my) (N cookie))))) (p=2.03744e-07)
>>> parser = pchart.RandomChartParser(grammar)
>>> for t in parser.parse(tokens):
... print(t)
(S
(NP (Name Jack))
(VP
(V saw)
(NP
(NP (Name Bob))
(PP (P with) (NP (Det my) (N cookie)))))) (p=6.31607e-06)
(S
(NP (Name Jack))
(VP
(VP (V saw) (NP (Name Bob)))
(PP (P with) (NP (Det my) (N cookie))))) (p=2.03744e-07)
>>> parser = pchart.UnsortedChartParser(grammar)
>>> for t in parser.parse(tokens):
... print(t)
(S
(NP (Name Jack))
(VP
(V saw)
(NP
(NP (Name Bob))
(PP (P with) (NP (Det my) (N cookie)))))) (p=6.31607e-06)
(S
(NP (Name Jack))
(VP
(VP (V saw) (NP (Name Bob)))
(PP (P with) (NP (Det my) (N cookie))))) (p=2.03744e-07)
>>> parser = pchart.LongestChartParser(grammar)
>>> for t in parser.parse(tokens):
... print(t)
(S
(NP (Name Jack))
(VP
(V saw)
(NP
(NP (Name Bob))
(PP (P with) (NP (Det my) (N cookie)))))) (p=6.31607e-06)
(S
(NP (Name Jack))
(VP
(VP (V saw) (NP (Name Bob)))
(PP (P with) (NP (Det my) (N cookie))))) (p=2.03744e-07)
>>> parser = pchart.InsideChartParser(grammar, beam_size = len(tokens)+1)
>>> for t in parser.parse(tokens):
... print(t)
Unit tests for the Viterbi Parse classes
----------------------------------------
>>> from nltk.parse import ViterbiParser
>>> tokens = "Jack saw Bob with my cookie".split()
>>> grammar = toy_pcfg2
Parse the tokenized sentence.
>>> parser = ViterbiParser(grammar)
>>> for t in parser.parse(tokens):
... print(t)
(S
(NP (Name Jack))
(VP
(V saw)
(NP
(NP (Name Bob))
(PP (P with) (NP (Det my) (N cookie)))))) (p=6.31607e-06)
Unit tests for the FeatStructNonterminal class
----------------------------------------------
>>> from nltk.grammar import FeatStructNonterminal
>>> FeatStructNonterminal(
... pos='n', agr=FeatStructNonterminal(number='pl', gender='f'))
[agr=[gender='f', number='pl'], pos='n']
>>> FeatStructNonterminal('VP[+fin]/NP[+pl]')
VP[+fin]/NP[+pl]
Tracing the Feature Chart Parser
--------------------------------
We use the featurechart.demo() function for tracing the Feature Chart Parser.
>>> nltk.parse.featurechart.demo(print_times=False,
... print_grammar=True,
... parser=nltk.parse.featurechart.FeatureChartParser,
... sent='I saw John with a dog')
<BLANKLINE>
Grammar with 18 productions (start state = S[])
S[] -> NP[] VP[]
PP[] -> Prep[] NP[]
NP[] -> NP[] PP[]
VP[] -> VP[] PP[]
VP[] -> Verb[] NP[]
VP[] -> Verb[]
NP[] -> Det[pl=?x] Noun[pl=?x]
NP[] -> 'John'
NP[] -> 'I'
Det[] -> 'the'
Det[] -> 'my'
Det[-pl] -> 'a'
Noun[-pl] -> 'dog'
Noun[-pl] -> 'cookie'
Verb[] -> 'ate'
Verb[] -> 'saw'
Prep[] -> 'with'
Prep[] -> 'under'
<BLANKLINE>
* FeatureChartParser
Sentence: I saw John with a dog
|.I.s.J.w.a.d.|
|[-] . . . . .| [0:1] 'I'
|. [-] . . . .| [1:2] 'saw'
|. . [-] . . .| [2:3] 'John'
|. . . [-] . .| [3:4] 'with'
|. . . . [-] .| [4:5] 'a'
|. . . . . [-]| [5:6] 'dog'
|[-] . . . . .| [0:1] NP[] -> 'I' *
|[-> . . . . .| [0:1] S[] -> NP[] * VP[] {}
|[-> . . . . .| [0:1] NP[] -> NP[] * PP[] {}
|. [-] . . . .| [1:2] Verb[] -> 'saw' *
|. [-> . . . .| [1:2] VP[] -> Verb[] * NP[] {}
|. [-] . . . .| [1:2] VP[] -> Verb[] *
|. [-> . . . .| [1:2] VP[] -> VP[] * PP[] {}
|[---] . . . .| [0:2] S[] -> NP[] VP[] *
|. . [-] . . .| [2:3] NP[] -> 'John' *
|. . [-> . . .| [2:3] S[] -> NP[] * VP[] {}
|. . [-> . . .| [2:3] NP[] -> NP[] * PP[] {}
|. [---] . . .| [1:3] VP[] -> Verb[] NP[] *
|. [---> . . .| [1:3] VP[] -> VP[] * PP[] {}
|[-----] . . .| [0:3] S[] -> NP[] VP[] *
|. . . [-] . .| [3:4] Prep[] -> 'with' *
|. . . [-> . .| [3:4] PP[] -> Prep[] * NP[] {}
|. . . . [-] .| [4:5] Det[-pl] -> 'a' *
|. . . . [-> .| [4:5] NP[] -> Det[pl=?x] * Noun[pl=?x] {?x: False}
|. . . . . [-]| [5:6] Noun[-pl] -> 'dog' *
|. . . . [---]| [4:6] NP[] -> Det[-pl] Noun[-pl] *
|. . . . [--->| [4:6] S[] -> NP[] * VP[] {}
|. . . . [--->| [4:6] NP[] -> NP[] * PP[] {}
|. . . [-----]| [3:6] PP[] -> Prep[] NP[] *
|. . [-------]| [2:6] NP[] -> NP[] PP[] *
|. [---------]| [1:6] VP[] -> VP[] PP[] *
|. [--------->| [1:6] VP[] -> VP[] * PP[] {}
|[===========]| [0:6] S[] -> NP[] VP[] *
|. . [------->| [2:6] S[] -> NP[] * VP[] {}
|. . [------->| [2:6] NP[] -> NP[] * PP[] {}
|. [---------]| [1:6] VP[] -> Verb[] NP[] *
|. [--------->| [1:6] VP[] -> VP[] * PP[] {}
|[===========]| [0:6] S[] -> NP[] VP[] *
(S[]
(NP[] I)
(VP[]
(VP[] (Verb[] saw) (NP[] John))
(PP[] (Prep[] with) (NP[] (Det[-pl] a) (Noun[-pl] dog)))))
(S[]
(NP[] I)
(VP[]
(Verb[] saw)
(NP[]
(NP[] John)
(PP[] (Prep[] with) (NP[] (Det[-pl] a) (Noun[-pl] dog))))))
Unit tests for the Feature Chart Parser classes
-----------------------------------------------
The list of parsers we want to test.
>>> parsers = [nltk.parse.featurechart.FeatureChartParser,
... nltk.parse.featurechart.FeatureTopDownChartParser,
... nltk.parse.featurechart.FeatureBottomUpChartParser,
... nltk.parse.featurechart.FeatureBottomUpLeftCornerChartParser,
... nltk.parse.earleychart.FeatureIncrementalChartParser,
... nltk.parse.earleychart.FeatureEarleyChartParser,
... nltk.parse.earleychart.FeatureIncrementalTopDownChartParser,
... nltk.parse.earleychart.FeatureIncrementalBottomUpChartParser,
... nltk.parse.earleychart.FeatureIncrementalBottomUpLeftCornerChartParser,
... ]
A helper function that tests each parser on the given grammar and sentence.
We check that the number of trees are correct, and that all parsers
return the same trees. Otherwise an error is printed.
>>> def unittest(grammar, sentence, nr_trees):
... sentence = sentence.split()
... trees = None
... for P in parsers:
... result = P(grammar).parse(sentence)
... result = set(tree.freeze() for tree in result)
... if len(result) != nr_trees:
... print("Wrong nr of trees:", len(result))
... elif trees is None:
... trees = result
... elif result != trees:
... print("Trees differ for parser:", P.__name__)
The demo grammar from before, with an ambiguous sentence.
>>> isawjohn = nltk.parse.featurechart.demo_grammar()
>>> unittest(isawjohn, "I saw John with a dog with my cookie", 5)
This grammar tests that variables in different grammar rules are renamed
before unification. (The problematic variable is in this case ?X).
>>> whatwasthat = nltk.grammar.FeatureGrammar.fromstring('''
... S[] -> NP[num=?N] VP[num=?N, slash=?X]
... NP[num=?X] -> "what"
... NP[num=?X] -> "that"
... VP[num=?P, slash=none] -> V[num=?P] NP[]
... V[num=sg] -> "was"
... ''')
>>> unittest(whatwasthat, "what was that", 1)
This grammar tests that the same rule can be used in different places
in another rule, and that the variables are properly renamed.
>>> thislovesthat = nltk.grammar.FeatureGrammar.fromstring('''
... S[] -> NP[case=nom] V[] NP[case=acc]
... NP[case=?X] -> Pron[case=?X]
... Pron[] -> "this"
... Pron[] -> "that"
... V[] -> "loves"
... ''')
>>> unittest(thislovesthat, "this loves that", 1)
Tests for loading feature grammar files
---------------------------------------
Alternative 1: first load the grammar, then create the parser.
>>> fcfg = nltk.data.load('grammars/book_grammars/feat0.fcfg')
>>> fcp1 = nltk.parse.FeatureChartParser(fcfg)
>>> print((type(fcp1)))
<class 'nltk.parse.featurechart.FeatureChartParser'>
Alternative 2: directly load the parser.
>>> fcp2 = nltk.parse.load_parser('grammars/book_grammars/feat0.fcfg')
>>> print((type(fcp2)))
<class 'nltk.parse.featurechart.FeatureChartParser'>