1 # io.py, with sentence positions between word locations (i<k<j)
5 # some of dmv-module bleeding in here... todo: prettier (in inner())
10 def debug(string
, level
='TODO'):
11 '''Easily turn on/off inline debug printouts with this global. There's
12 a lot of cluttering debug statements here, todo: clean up'''
18 '''The PCFG used in the I/O-algorithm.
23 Todo: as of now, this allows duplicate rules... should we check
24 for this? (eg. g = Grammar([x,x],[]) where x.prob == 1 may give
25 inner probabilities of 2.)'''
29 def new_rules(self
, p_rules
):
30 self
.__p
_rules
= p_rules
33 return [rule
for rule
in self
.all_rules() if rule
.LHS() == LHS
]
36 return self
.__head
_nums
38 def sent_nums(self
, sent
):
39 return [self
.tagnum(tag
) for tag
in sent
]
41 def numtag(self
, num
):
42 if num
== ROOTNUM
: # don't want these added to headnums (which we iter through)
45 return self
.__numtag
[num
]
47 def tagnum(self
, tag
):
51 return self
.__tagnum
[tag
]
53 def get_nums_tags(self
):
54 return (self
.__numtag
,self
.__tagnum
)
56 def __init__(self
, numtag
, tagnum
, p_rules
=[], p_terminals
=[]):
57 '''rules and p_terminals should be arrays, where p_terminals are of
58 the form [preterminal, terminal], and rules are CNF_Rule's.'''
59 self
.__numtag
= numtag
60 self
.__tagnum
= tagnum
61 self
.__head
_nums
= [k
for k
in numtag
.iterkeys()]
62 self
.__p
_rules
= p_rules
# todo: could check for summing to 1 (+/- epsilon)
63 self
.p_terminals
= p_terminals
69 '''A single CNF rule in the PCFG, of the form
71 where these are just integers
72 (where do we save the connection between number and symbol?
73 symbols being 'vbd' etc.)'''
74 def __eq__(self
, other
):
75 return self
.LHS() == other
.LHS() and self
.R() == other
.R() and self
.L() == other
.L()
76 def __ne__(self
, other
):
77 return self
.LHS() != other
.LHS() or self
.R() != other
.R() or self
.L() != other
.L()
79 return "%s -> %s %s [%.2f]" % (self
.LHS(), self
.L(), self
.R(), self
.prob
)
80 def __init__(self
, LHS
, L
, R
, prob
):
86 "Return a probability, doesn't care about attachment..."
95 def inner(i
, j
, LHS
, g
, sent
, chart
):
96 ''' Give the inner probability of having the node LHS cover whatever's
97 between s and t in sentence sent, using grammar g.
99 Returns a pair of the inner probability and the chart
101 For DMV, LHS is a pair (bar, h), but this function ought to be
104 e() is an internal function, so the variable chart (a dictionary)
105 is available to all calls of e().
107 Since terminal probabilities are just simple lookups, they are not
108 put in the chart (although we could put them in there later to
116 '''Chart has lists of probability and whether or not we've attached
117 yet to L and R, each entry is a list [p, Rattach, Lattach], where if
118 Rattach==True then the rule has a right-attachment or there is one
119 lower in the tree (meaning we're no longer adjacent).'''
120 if (i
, j
, LHS
) in chart
:
121 return chart
[i
, j
, LHS
]
123 debug( "trying from %d to %d with %s" % (i
,j
,LHS
) , "IO")
125 if (LHS
, O(i
,j
)) in g
.p_terminals
:
126 prob
= g
.p_terminals
[LHS
, O(i
,j
)] # b[LHS, O(s)] in L&Y
129 print "\t LACKING TERMINAL:%s -> %s : %.1f" % (LHS
, O(i
,j
), prob
)
130 debug( "\t terminal: %s -> %s : %.1f" % (LHS
, O(i
,j
), prob
) ,"IO")
131 # terminals have no attachment
134 if (i
,j
,LHS
) not in chart
:
135 # by default, not attachment yet
137 for rule
in g
.rules(LHS
): # summing over rules headed by LHS, "a[i,j,k]"
138 debug( "\tsumming rule %s" % rule
, "IO")
141 for k
in range(i
+1, j
): # i<k<j
144 chart
[i
, j
, LHS
] += rule
.p() * p_L
* p_R
145 debug( "\tchart[%d,%d,%s] = %.2f" % (i
,j
,LHS
, chart
[i
,j
,LHS
]) ,"IO")
146 return chart
[i
, j
, LHS
]
149 inner_prob
= e(i
,j
,LHS
)
152 for k
,v
in chart
.iteritems():
153 print "\t%s -> %s_%d ... %s_%d : %.1f" % (k
[2], O(k
[0]), k
[0], O(k
[1]), k
[1], v
)
154 print "---CHART:end---"
155 return [inner_prob
, chart
]
164 if __name__
== "__main__":
165 print "IO-module tests:"
167 s
= CNF_Rule(0,1,2, 1.0) # s->np vp
168 np
= CNF_Rule(1,3,4, 0.3) # np->n p
169 b
[1, 'n'] = 0.7 # np->'n'
170 b
[3, 'n'] = 1.0 # n->'n'
171 b
[4, 'p'] = 1.0 # p->'p'
172 vp
= CNF_Rule(2,5,1, 0.1) # vp->v np (two parses use this rule)
173 vp2
= CNF_Rule(2,2,4, 0.9) # vp->vp p
174 b
[5, 'v'] = 1.0 # v->'v'
176 g
= Grammar({0:'s',1:'np',2:'vp',3:'n',4:'p',5:'v'},
177 {'s':0,'np':1,'vp':2,'n':3,'p':4,'v':5},
181 # for i in range(0,5):
182 # for r in g.rules(i):
186 test1
= inner(0,1, 1, g
, ['n'], {})
188 print "should be 0.70 : %.3f" % test1
[0]
191 test2
= inner(0,3, 2, g
, ['v','n','p'], test1
[1])
192 if test2
[0] != 0.0930:
193 print "should be 0.0930 : %.4f" % test2
[0]
194 test2
= inner(0,3, 2, g
, ['v','n','p'], test2
[1])
195 if test2
[0] != 0.0930:
196 print "should be 0.0930 : %.4f" % test2
[0]