def setUp(self):
self.exDict = {"carry_from_to" :{"sent" : "Carry the meals from the kitchen to the rooms.",
"tree" : Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Carry']), Tree('NP-A', [Tree('DT', ['the']), Tree('NNS', ['meals'])]), Tree('PP-CLR', [Tree('IN', ['from']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['kitchen'])])]), Tree('PP-CLR', [Tree('TO', ['to']), Tree('NP-A', [Tree('DT', ['the']), Tree('NNS', ['rooms'])])])]), Tree('.', ['.'])]),
"possible_frames": [[('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''), ('NP', 'Theme', '', '')],
[('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''), ('NP', 'Theme', '', ''), ('PREP', 'to towards', '', ''), ('NP', 'Destination', '', '')],
[('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''), ('NP', 'Theme', '', ''), ('PREP', 'PREP', '', ''), ('NP', 'Source', '', ''), ('PREP', 'to towards', '', ''), ('NP', 'Destination', '', '')],
[('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''), ('NP', 'Theme', '', ''), ('PREP', 'to towards', '', ''), ('NP', 'Destination', '', ''), ('PREP', 'PREP', '', ''), ('NP', 'Source', '', '')]],
"correct_frame" : [('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''), ('NP', 'Theme', '', ''), ('PREP', 'PREP', '', ''), ('NP', 'Source', '', ''), ('PREP', 'to towards', '', ''), ('NP', 'Destination', '', '')],
"correct_entry" : {'to towards': Tree('TO', ['to']), 'Destination': Tree('NP-A', [Tree('DT', ['the']), Tree('NNS', ['rooms'])]), 'Agent': Tree('NP-SBJ-A', [Tree('-NPNONE-', ['*'])]), 'Source': Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['kitchen'])]), 'Theme': Tree('NP-A', [Tree('DT', ['the']), Tree('NNS', ['meals'])]), 'VERB': Tree('VB', ['Carry'])}
},
"carry_from_to_deep_pp":{"sent" : "Carry the meals from the kitchen to the cafeteria.",
"tree" : Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Carry']), Tree('NP-A', [Tree('DT', ['the']), Tree('NNS', ['meals'])]), Tree('PP-CLR', [Tree('IN', ['from']), Tree('NP-A', [Tree('NP', [Tree('DT', ['the']), Tree('NN', ['kitchen'])]), Tree('PP', [Tree('TO', ['to']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['cafeteria'])])])])])]), Tree('.', ['.'])]),
"possible_frames": [[('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''), ('NP', 'Theme', '', '')],
[('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''), ('NP', 'Theme', '', ''), ('PREP', 'to towards', '', ''), ('NP', 'Destination', '', '')],
[('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''), ('NP', 'Theme', '', ''), ('PREP', 'PREP', '', ''), ('NP', 'Source', '', ''), ('PREP', 'to towards', '', ''), ('NP', 'Destination', '', '')],
[('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''), ('NP', 'Theme', '', ''), ('PREP', 'to towards', '', ''), ('NP', 'Destination', '', ''), ('PREP', 'PREP', '', ''), ('NP', 'Source', '', '')]],
"correct_frame" : [('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''), ('NP', 'Theme', '', ''), ('PREP', 'PREP', '', ''), ('NP', 'Source', '', ''), ('PREP', 'to towards', '', ''), ('NP', 'Destination', '', '')]
},
"defuse_in" : {"sent" : "Defuse in the hallway.",
"tree" : Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Defuse']), Tree('PP-CLR', [Tree('IN', ['in']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['hallway'])])])]), Tree('.', ['.'])]),
"wrong_frame" : [('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''), ('NP', 'Theme', '', '')]
}
}
self.crazyframes = {"SUBPHRASE_PASS":
[('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''), ('DT','DT','',''),('NP', 'Theme', '', ''), ('PREP', 'to towards', '', ''), ('NP', 'Destination', '', '')],
"SUBPHRASE_FAIL":
[('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''), ('NP', 'Theme', '', ''),('DT','DT','',''), ('PREP', 'to towards', '', ''), ('NP', 'Destination', '', '')]
}
self.th = TreeHandler()
self.matcher = ParseMatcher(0,2)
def verb_child_frame_match(child, parent, subject, vp, negated):
result_tree = (parent if not subject else Tree('S', [subject, vp]))
verb = child[0].lower()
match = best_matching_frame(verb, result_tree)
if match != (None, None):
match = FrameMatch(verb, match[1], match[0], negated, result_tree)
return match
def process(parse,verbose=True):
"""Show the steps of transformation for a parse."""
# Original parse
parse_tree = Tree.parse(parse)
print_parse(parse_tree, "Parse")
frames = extract_frames_from_parse(parse, verbose=verbose)
if verbose:
print
for frame in frames:
print frame.pprint()
if frame.condition:
print "Condition:"
print frame.condition.pprint()
print
# Bail if no frames matched
if not frames:
if verbose: print "No frames matched."
return
# Extract semantic structures
semantic_structures = create_semantic_structures(frames)
if semantic_structures:
if verbose: print semantic_structures
return semantic_structures
else:
if verbose: print "No semantic structures returned."
def pos_split(self,tree,pos,possibleParents=["S","VP","NP"],desiredCC="and",ccpos="CC"):
"""The strategy for this is to find the CC in the VP (like the VB in matching),
find the nearest left or right siblings and if they are heads, pop
@input tree input tree to split on
@input pos is the part of speech we are splitting on
@input possibleParents are the possible phrase parents currently supported
@input desiredCC is the lemma of the CC we are looking for
@input ccpos is the cc part of speech we are looking for
"""
res = []
cursor = [-1]
cccount = self.num_ccs(tree,desiredCC)
for i in range(cccount):
ccpath = list(self.th.get_main_pos_path(tree, ccpos, -1, cursor=cursor))
parentPhrase = self.th.which_parent(tree, ccpath, possibleParents, -1)
parentpos = parentPhrase.split(self.th.depthdelim)[0].split(self.th.posdelim)[0]
if parentpos == pos:
left = self.sibling_cc_path(ccpath)
right = self.sibling_cc_path(ccpath,tree=tree);
#ADDED
#if siblings of CC are NN (e.g. "Go to office1 and office2")
#make them NP instead
#this is needed otherwise NNs lead to bad creation of commands
for child_pos, child in enumerate(tree[ccpath[:-2]]):
if child.node.split(".")[0] == "NN":
temp_path = ccpath[:-2]
temp_path.append(child_pos)
child_path = temp_path
new_child = Tree("NP-A", [child])
tree[child_path] = new_child
if len(left) != len(ccpath) != len(right):
raise UnlevelCCSiblings
lefttree = copy.deepcopy(tree)
#Instead of popping, need to replace parent with correct branch
temp = self.th.pop_path_cc(lefttree,right,ccpath)
if temp:
#If splitting on S, temp will be the parent replacement
lefttree = temp
#Recurse on "," list
lefttrees = self.pos_split(lefttree,pos,possibleParents=possibleParents,desiredCC=self.listCC,ccpos=self.listCC)
res.extend(lefttrees) #Copy and put left branch in results and keep going
self.th.pop_left(tree,ccpath)#Pop everything to the left of ccpath and keep looking
#self.th.pop_path_cc(tree, left,ccpath)#Pop for real, keep looking
cursor = [-1]
else:
if DEBUG: print pos,' CC: ',ccpath
cursor = ccpath
if len(tree) == 1:
#Only one branch-> split on S, consume
tree = tree[0]
res.append(tree)
return res
def process(parse):
"""Show the steps of transformation for a parse."""
# Original parse
parse = Tree.parse(parse)
print_parse(parse, "Parse")
split_trees = split_conjunctions(parse)
for split_parse, conjunction in split_trees.values():
for subtree in split_parse:
print_if_diff(subtree, parse, "Subtree ({})".format(conjunction))
def process(parse):
"""Show the steps of transformation for a parse."""
# Original parse
parse = Tree.parse(parse)
print_parse(parse, "Parse")
split_trees = split_conjunctions(parse)
for split_parse, conjunction in split_trees.values():
for subtree in split_parse:
print_if_diff(subtree, parse, "Subtree ({})".format(conjunction))
def match_parse(self, parse_tree):
"""Takes a Treebank parse tree compiled into NLTK's tree structure.
Outputs a result dictionary mapping predicates to arguments"""
result_dict = {}
subtree_list = []
subtrees = parse_tree.subtrees()
# We need a list of subtrees, not a generator
for subtree in subtrees:
subtree_list.append(subtree)
current_subtree = 0
matches = 0
# For each frame element, find the next subtree that matches
for frame_tag in self.frame_list:
# Go through the subtrees until you run out of subtrees
while current_subtree < len(subtree_list):
subtree = subtree_list[current_subtree]
# If there is no explicit NP for the Agent role, insert one
# NOTE: Assumes Agent role is the first in the frame, may not be true
if (current_subtree == 0 and frame_tag[1] == 'Agent'
and not self.__match_subtree(subtree, frame_tag)):
result_dict[frame_tag[1]] = Tree("(NP-SBJ-A (-NONE- *))")
matches += 1
break
if self.__match_subtree(subtree, frame_tag):
#print 'Match -> ' + str(frame_tag)
# If the subtree matches, add role->phrase to the dictionary
#result_dict[frame_tag[1]] = ' '.join(subtree.leaves())
result_dict[frame_tag[1]] = subtree
matches += 1
break
current_subtree += 1
if current_subtree == len(subtree_list):
return None
# Only return something if every frame was matched
if matches == len(self.frame_list):
return result_dict
else:
return None
def interactive_mode(window, first_input):
"""Interactively get input from the user and parse it."""
input_frame, input_win, parse_win, semantic_win = setup_windows(window)
# Initialize pipeline and knowledge base
pipeline = PipelineClient()
kb = KnowledgeBase()
# Send some data through the pipeline
result = pipeline.parse("This is a test.")
input_frame.addstr(1, 1, 'Enter your input, then press Ctrl+G. '
'Enter "quit" or press Ctrl+C to exit.')
input_frame.refresh()
# Until the input is q/quit, process data
last_input = first_input
while True:
# Display the first input if needed
input_win.erase()
input_win.refresh()
if last_input:
input_win.addstr(0, 0, last_input)
# Get text from the input box, removing any embedded newlines
if first_input:
text = first_input
first_input = None
else:
text = get_input(input_win).replace("\n", "").strip()
last_input = text
# Quit if needed
if text == "q" or text == "quit":
return
# Get input again if it was empty
if not text:
continue
# Echo input and display status, clearing both windows
parse_win.clear()
parse_win.addstr(text)
parse_win.addstr('\nParsing and restoring null elements...')
parse_win.refresh()
semantic_win.clear()
semantic_win.refresh()
# Run the parse pipeline
result = pipeline.parse(text)
result_tree = Tree(result)
# Output the longest parse that will fit. We try to draw the
# possible output in order of decreasing length.
parse_max_width = parse_win.getmaxyx()[1]
possible_formats = (result_tree.pprint(margin=parse_max_width, force_multiline=True),
result_tree.pprint(margin=parse_max_width),
result)
for formatted_result in possible_formats:
parse_win.clear()
try:
parse_win.addstr(text + '\n')
parse_win.addstr(formatted_result)
except _curses.error:
continue
else:
# We've successfully printed, stop trying formats
break
else:
parse_win.clear()
parse_win.addstr("Parse too large to show.\n")
parse_win.refresh()
# Do the same for semantics
# Echo input and display status, after clearing the window
semantic_win.clear()
semantic_win.addstr(text)
semantic_win.addstr('\nPerforming semantic analysis...')
semantic_win.refresh()
frames, new_commands, kb_response = process_parse_tree(result, text, kb)
semantic_win.clear()
try:
if frames:
semantic_win.addstr("Frames matched:\n")
for frame in frames:
semantic_win.addstr("\t" + str(frame) + "\n")
if new_commands:
semantic_win.addstr("New commands:\n")
for command in new_commands:
semantic_win.addstr(str(command) + "\n")
if kb_response:
semantic_win.addstr("KB response:\n")
semantic_win.addstr(str(kb_response) + "\n")
if not any((frames, new_commands, kb_response)):
semantic_win.addstr("No frames matched.\n")
except _curses.error:
semantic_win.clear()
semantic_win.addstr("Semantic representation too large to show.")
semantic_win.refresh()
return
def setUp(self):
self.th = TreeHandler()
self.exDict = {
"carry_from_to": {
"sent":
"Carry the meals from the kitchen to the rooms.",
"tree":
Tree(
'''(S\n (NP-SBJ-A (-NONE- *))\n (VP\n (VB Carry)\n (NP-A (DT the) (NNS meals))\n (PP-CLR (IN from) (NP-A (DT the) (NN kitchen)))\n (PP-CLR (TO to) (NP-A (DT the) (NNS rooms))))\n (. .))'''
),
"solution": {
'Destination':
Tree('NP-A', [Tree('DT', ['the']),
Tree('NNS', ['rooms'])]),
'Agent':
Tree('NP-SBJ-A', [Tree('-NPNONE-', ['*'])]),
'to towards':
Tree('TO', ['to']),
'Theme':
Tree('NP-A', [Tree('DT', ['the']),
Tree('NNS', ['meals'])]),
'VERB':
Tree('VB', ['Carry']),
'DT':
Tree('DT', ['the'])
}
}
}
self.crazyframes = {
"SUBPHRASE_PASS": [('NP', 'Agent', '', ''),
('VERB', 'VERB', '', ''), ('DT', 'DT', '', ''),
('NP', 'Theme', '', ''),
('PREP', 'to towards', '', ''),
('NP', 'Destination', '', '')],
"SUBPHRASE_FAIL":
[('NP', 'Agent', '', ''), ('VERB', 'VERB', '', ''),
('NP', 'Theme', '', ''), ('DT', 'DT', '', ''),
('PREP', 'to towards', '', ''), ('NP', 'Destination', '', '')]
}
def extract_frames_from_parse(parse_tree_string, verbose=False):
"""Take a string representing the parse tree as input, and print the
semantic parse. The result list consists of a list of tuples, with each
tuple containing the VerbNet frame and its associated tree."""
result_list = []
# In case we're handed an bad string, bail somewhat gracefully
try:
parse_tree = Tree.parse(parse_tree_string)
except ValueError:
print "Warning: semantics could not parse tree", repr(
parse_tree_string)
return result_list
# Temporarily (and maybe permanently) disabled features:
# 1. Clause splitting: we have not found any example where it does something
# 2. Activizing clauses: for now, passives do not matter.
# Split clauses to handle them separately
#split_clause_dict = split_clauses(parse_tree)
# Activize clauses
#for key, (clause, conjunction) in split_clause_dict.items():
# activized_clause = activize_clause(clause)
# split_clause_dict[key] = (activized_clause, conjunction)
#for (clause, conjunction) in split_clause_dict.values():
for clause, conjunction in ((parse_tree, ''), ):
# Split conjunctions and duplicate arguments if necessary
split_tree_dict = split_conjunctions(clause)
if conjunction != '':
result_list.append(conjunction)
for (split_tree, conjunction) in split_tree_dict.values():
if conjunction != '':
result_list.append(conjunction)
for tree in split_tree:
tag_list = []
# Store whether there was an existential there
if is_existential(str(tree)):
tag_list.append('ex')
# Transformational grammar stuff
tree = existential_there_insertion(tree)
tree = invert_clause(tree)
tree = wh_movement(tree)
if EXTRACT_DEBUG:
print 'Transformed tree:'
print str(tree)
verbs = find_verbs(tree)
# Create VFOs for each verb, then match them to the parse tree
for verb, negation in verbs:
lemmatized_verb = morphy(verb, 'v')
vfo_list = create_VerbFrameObjects(lemmatized_verb)
match_list = []
if EXTRACT_DEBUG:
print 'VFO list for %s:' % verb
print '\n'.join(
str(vfo.frame_list) for vfo in vfo_list)
for vfo in vfo_list:
match = vfo.match_parse(tree)
if match:
if EXTRACT_DEBUG:
print 'Matched:'
print '\t', str(vfo.frame_list)
print 'with'
print '\t', str(tree)
match_list.append((match, vfo.classid))
if EXTRACT_DEBUG:
print 'Match list:'
for m in match_list:
print 'Sense:', m[1]
for a, b in m[0].items():
print a, str(b)
print '\n\n'
(best_match, sense) = pick_best_match(match_list)
if EXTRACT_DEBUG:
print 'Chose: '
if best_match:
print sense
for a, b in best_match.items():
print a, str(b)
else:
print str(None)
print '\n\n'
if not best_match is None:
result_list.append((best_match, tree, tag_list, sense,
verb, negation))
return result_list
def extract_frames_from_parse(parse_tree_string, verbose=False):
"""Take a string representing the parse tree as input, and print the
semantic parse. The result list consists of a list of tuples, with each
tuple containing the VerbNet frame and its associated tree."""
result_list = []
# In case we're handed an bad string, bail somewhat gracefully
try:
parse_tree = Tree.parse(parse_tree_string)
except ValueError:
print "Warning: semantics could not parse tree", repr(parse_tree_string)
return result_list
# Temporarily (and maybe permanently) disabled features:
# 1. Clause splitting: we have not found any example where it does something
# 2. Activizing clauses: for now, passives do not matter.
# Split clauses to handle them separately
#split_clause_dict = split_clauses(parse_tree)
# Activize clauses
#for key, (clause, conjunction) in split_clause_dict.items():
# activized_clause = activize_clause(clause)
# split_clause_dict[key] = (activized_clause, conjunction)
#for (clause, conjunction) in split_clause_dict.values():
for clause, conjunction in ((parse_tree, ''),):
# Split conjunctions and duplicate arguments if necessary
split_tree_dict = split_conjunctions(clause)
if conjunction != '':
result_list.append(conjunction)
for (split_tree, conjunction) in split_tree_dict.values():
if conjunction != '':
result_list.append(conjunction)
for tree in split_tree:
tag_list = []
# Store whether there was an existential there
if is_existential(str(tree)):
tag_list.append('ex')
# Transformational grammar stuff
tree = existential_there_insertion(tree)
tree = invert_clause(tree)
tree = wh_movement(tree)
if EXTRACT_DEBUG:
print 'Transformed tree:'
print str(tree)
verbs = find_verbs(tree)
# Create VFOs for each verb, then match them to the parse tree
for verb, negation in verbs:
lemmatized_verb = morphy(verb, 'v')
vfo_list = create_VerbFrameObjects(lemmatized_verb)
match_list = []
if EXTRACT_DEBUG:
print 'VFO list for %s:' % verb
print '\n'.join(str(vfo.frame_list) for vfo in vfo_list)
for vfo in vfo_list:
match = vfo.match_parse(tree)
if match:
if EXTRACT_DEBUG:
print 'Matched:'
print '\t', str(vfo.frame_list)
print 'with'
print '\t', str(tree)
match_list.append((match, vfo.classid))
if EXTRACT_DEBUG:
print 'Match list:'
for m in match_list:
print 'Sense:', m[1]
for a, b in m[0].items():
print a, str(b)
print '\n\n'
(best_match, sense) = pick_best_match(match_list)
if EXTRACT_DEBUG:
print 'Chose: '
if best_match:
print sense
for a, b in best_match.items():
print a, str(b)
else:
print str(None)
print '\n\n'
if not best_match is None:
result_list.append((best_match, tree, tag_list, sense, verb, negation))
return result_list
def setUp(self):
self.splitter = Split()
self.th = TreeHandler()
self.easyDict = {"Defuse_VP_VP" : {"sent": "Defuse the bomb and go to the hallway.",
"tree": Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VP-A', [Tree('VB', ['Defuse']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['bomb'])])]), Tree('CC', ['and']), Tree('VP-A', [Tree('VB', ['go']), Tree('PP-CLR', [Tree('TO', ['to']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['hallway'])])])])]), Tree('.', ['.'])])
},
}
self.exDict = {"Extracts VP, produces S trees with no verb" : {"sent" : "Go and defuse the bomb in the cellar.",
"tree" : Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Go']), Tree('CC', ['and']), Tree('VB', ['defuse']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['bomb'])]), Tree('PP-MNR', [Tree('IN', ['in']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['cellar'])])])]), Tree('.', ['.'])])
},
"Two carrys only one source-dest because of wrongly embedded 'to' pp": {"sent" : "Carry the hostages from the kitchen and the cafeteria to the cellar.",
"tree" : Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Carry']), Tree('NP-A', [Tree('DT', ['the']), Tree('NNS', ['hostages'])]), Tree('PP-CLR', [Tree('IN', ['from']), Tree('NP-A', [Tree('NP', [Tree('DT', ['the']), Tree('NN', ['kitchen'])]), Tree('CC', ['and']), Tree('NP', [Tree('NP', [Tree('DT', ['the']), Tree('NN', ['cafeteria'])]), Tree('PP', [Tree('TO', ['to']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['cellar'])])])])])])]), Tree('.', ['.'])])
},
"VP_NPNP_VP" : { "sent": "Defuse the bomb and the bomb and go to the hallway.",
"tree" : Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VP-A', [Tree('VB', ['Defuse']), Tree('NP-A', [Tree('NP', [Tree('DT', ['the']), Tree('NN', ['bomb'])]), Tree('CC', ['and']), Tree('NP', [Tree('DT', ['the']), Tree('NN', ['bomb'])])])]), Tree('CC', ['and']), Tree('VP-A', [Tree('VB', ['go']), Tree('PP-CLR', [Tree('TO', ['to']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['hallway'])])])])]), Tree('.', ['.'])])
},
"go_to_np_comma_np_comma" : {
"sent" : "Go to the cellar, kitchen, and bedroom.",
"tree" : Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Go']), Tree('PP-CLR', [Tree('TO', ['to']), Tree('NP-A', [Tree('NP', [Tree('DT', ['the']), Tree('NN', ['cellar'])]), Tree(',', [',']), Tree('NP', [Tree('NNP', ['kitchen'])]), Tree(',', [',']), Tree('CC', ['and']), Tree('NP', [Tree('NN', ['bedroom'])])])])]), Tree('.', ['.'])]),
"correct_list" : [Tree('S', [Tree('NP-SBJ-A', [Tree('-NPNONE-', ['*'])]), Tree('VP', [Tree('VB', ['Go']), Tree('PP-CLR', [Tree('TO', ['to']), Tree('NP', [Tree('DT', ['the']), Tree('NN', ['cellar'])])])]), Tree('.', ['.'])]), Tree('S', [Tree('NP-SBJ-A', [Tree('-NPNONE-', ['*'])]), Tree('VP', [Tree('VB', ['Go']), Tree('PP-CLR', [Tree('TO', ['to']), Tree('NP', [Tree('NNP', ['kitchen'])])])]), Tree('.', ['.'])]), Tree('S', [Tree('NP-SBJ-A', [Tree('-NPNONE-', ['*'])]), Tree('VP', [Tree('VB', ['Go']), Tree('PP-CLR', [Tree('TO', ['to']), Tree('NP', [Tree('NN', ['bedroom'])])])]), Tree('.', ['.'])])]
}
}
self.ccConditionalDict = {"CC and adv conditional" : {"sent" : "Go to the cellar and if you see a bomb, activate your camera.",
"tree" : Tree('S', [Tree('S-A', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Go']), Tree('PP-CLR', [Tree('TO', ['to']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['cellar'])])])])]), Tree('CC', ['and']), Tree('S-A', [Tree('SBAR-ADV', [Tree('IN', ['if']), Tree('S-A', [Tree('NP-SBJ-A', [Tree('PRP', ['you'])]), Tree('VP', [Tree('VBP', ['see']), Tree('NP-A', [Tree('DT', ['a']), Tree('NN', ['bomb'])])])])]), Tree(',', [',']), Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['activate']), Tree('NP-A', [Tree('PRP$', ['your']), Tree('NN', ['camera'])])])]), Tree('.', ['.'])]),
"frames" : []
},
"CC and tmp conditional" : {"sent" : "Go to the cellar and when you see a bomb, activate your camera.",
"tree" : Tree('S', [Tree('S-A', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Go']), Tree('PP-CLR', [Tree('TO', ['to']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['cellar'])])])])]), Tree('CC', ['and']), Tree('S-A', [Tree('SBAR-TMP', [Tree('WHADVP-0', [Tree('WRB', ['when'])]), Tree('S-A', [Tree('NP-SBJ-A', [Tree('PRP', ['you'])]), Tree('VP', [Tree('VBP', ['see']), Tree('NP-A', [Tree('DT', ['a']), Tree('NN', ['bomb'])]), Tree('ADVP-0', [Tree('-NONE-', ['*T*'])])])])]), Tree(',', [',']), Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['activate']), Tree('NP-A', [Tree('PRP$', ['your']), Tree('NN', ['camera'])])])]), Tree('.', ['.'])]),
"frames": [],
}
}
#Examples the current methods do not correctly parse
self.hardDict = { "Carry_NP_NP" : {"sent" : "Carry the hostages from the kitchen and bedroom to the cellar.",
"tree" : Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Carry']), Tree('NP-A', [Tree('DT', ['the']), Tree('NNS', ['hostages'])]), Tree('PP-CLR', [Tree('IN', ['from']), Tree('NP-A', [Tree('NP', [Tree('DT', ['the']), Tree('NN', ['kitchen'])]), Tree('CC', ['and']), Tree('NP', [Tree('NP', [Tree('NN', ['bedroom'])]), Tree('PP', [Tree('TO', ['to']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['cellar'])])])])])])]), Tree('.', ['.'])])
},
"Go_NP_NP" : {"sent": "Go to the cellar and kitchen.",
"tree": Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Go']), Tree('PP-CLR', [Tree('TO', ['to']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['cellar']), Tree('CC', ['and']), Tree('NN', ['kitchen'])])])]), Tree('.', ['.'])])
},
"Carry_NP_NP_2" : {"sent": "Carry meals to the cafeteria and lounge.",
"tree": Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Carry']), Tree('NP-A', [Tree('NNS', ['meals'])]), Tree('PP-CLR', [Tree('TO', ['to']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['cafeteria']), Tree('CC', ['and']), Tree('NN', ['lounge'])])])]), Tree('.', ['.'])])
},
"Carry_NPNP_NPNP" : {"sent" : "Carry meals from the kitchen and cafeteria to the office and lounge.",
#Wrong pp attachment for this tree ("to the office" modifies cafeteria)
"tree" : Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Carry']), Tree('NP-A', [Tree('NNS', ['meals'])]), Tree('PP-CLR', [Tree('IN', ['from']), Tree('NP-A', [Tree('NP', [Tree('DT', ['the']), Tree('NN', ['kitchen'])]), Tree('CC', ['and']), Tree('NP', [Tree('NP', [Tree('NN', ['cafeteria'])]), Tree('PP', [Tree('TO', ['to']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['office'])])])]), Tree('CC', ['and']), Tree('NP', [Tree('NN', ['lounge'])])])])]), Tree('.', ['.'])])
},
"Two carrys only one source-dest because of wrongly embedded 'to' pp": {"sent" : "Carry the hostages from the kitchen and the cafeteria to the cellar.",
"tree" : Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Carry']), Tree('NP-A', [Tree('DT', ['the']), Tree('NNS', ['hostages'])]), Tree('PP-CLR', [Tree('IN', ['from']), Tree('NP-A', [Tree('NP', [Tree('DT', ['the']), Tree('NN', ['kitchen'])]), Tree('CC', ['and']), Tree('NP', [Tree('NP', [Tree('DT', ['the']), Tree('NN', ['cafeteria'])]), Tree('PP', [Tree('TO', ['to']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['cellar'])])])])])])]), Tree('.', ['.'])])
},
"Extracts VP, produces S trees with no verb" : {"sent" : "Go and defuse the bomb in the cellar.",
"tree" : Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Go']), Tree('CC', ['and']), Tree('VB', ['defuse']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['bomb'])]), Tree('PP-MNR', [Tree('IN', ['in']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['cellar'])])])]), Tree('.', ['.'])])
},
"Carry_NPNP_NP" : {"sent" : "Carry meals to the library and cellar from the kitchen.",
"tree": Tree('S', [Tree('NP-SBJ-A', [Tree('-NONE-', ['*'])]), Tree('VP', [Tree('VB', ['Carry']), Tree('NP-A', [Tree('NP', [Tree('NNS', ['meals'])]), Tree('PP', [Tree('TO', ['to']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['library']), Tree('CC', ['and']), Tree('NN', ['cellar'])])])]), Tree('PP-CLR', [Tree('IN', ['from']), Tree('NP-A', [Tree('DT', ['the']), Tree('NN', ['kitchen'])])])]), Tree('.', ['.'])]),
}
}
def append_period(self,tree):
if len(tree) > 0 and tree[-1].node != '.':
node = '.'+self.depthdelim+str(1)
children = ['.'+'__'+str(self.get_ulid())+'__']
tree[1].append(Tree(node,children))
def split_conjunctions_sparse(parse_tree):
'''Find conjunctions, split them based on syntax tree. Return the result dictionary
'''
splitter = Split()
#trees = splitter.split_on_cc(parse_tree)
#DEBUG
g1 = Digraph(format='png')
utils.print_tree_to_png(g1, [parse_tree])
g1.render("/home/gian/hri_project/figures/parse_trees/command.gv")
#ADDED
#check whether it's 'if-and' or 'if-or' phrase
#example: 'If you see a hostage and|or a bomb...'
if splitter.num_ccs(parse_tree, "if") > 0:
for tree_pos in parse_tree.treepositions():
#we don't want the leaves
if not isinstance(parse_tree[tree_pos], Tree):
continue
#we look for the 'if' tree and we divide it from
#the command tree
if "SBAR-ADV" in parse_tree[tree_pos].node:
if_tree = deepcopy(parse_tree[tree_pos])
temp_tree = deepcopy(parse_tree)
del temp_tree[tree_pos]
cmd_tree = temp_tree
num_ands = splitter.num_ccs(if_tree, "and")
num_ors = splitter.num_ccs(if_tree, "or")
#check whether if subtree has 'and|or' conjunctions inside
if num_ands > 0:
#it's an 'if-and' phrase
#e.g. 'If you see a hostage and a bomb...'
trees = []
#check whether command tree contains CCs, in which case split it
for cmd in splitter.split_on_multiple_ccs(cmd_tree, 'and'):
children_of_tree = [if_tree]
for child in cmd:
children_of_tree.append(child)
trees.append(Tree('S', children_of_tree))
elif num_ors > 0:
#it's an 'if-or' phrase
#e.g. 'If you see a hostage or a bomb...'
trees = []
for temp_tree in splitter.split_on_multiple_ccs(
parse_tree, "or"):
for tree in splitter.split_on_multiple_ccs(
temp_tree, "and"):
trees.append(tree)
else:
#it's a simple 'if' phrase
#e.g. 'If you see a hostage activate the radio and the camera'
trees = splitter.split_on_multiple_ccs(parse_tree, "and")
break
else:
#it's a simple command
#e.g. 'Activate the radio and the camera'
trees = splitter.split_on_multiple_ccs(parse_tree, "and")
#DEBUG
for i, item in enumerate(trees):
g2 = Digraph(format='png')
utils.print_tree_to_png(g2, [item])
g2.render(
"/home/gian/hri_project/figures/parse_trees/splitted_command{}.gv".
format(str(i)))
#Abstracting away path to conjunction because it isn't used anyway
#Only support 'and' right now
res = {}
res[0] = (trees, "and")
return res
def extract_frames_from_parse(parse_tree_string, verbose=False):
"""Take a string representing the parse tree as input, and print the
semantic parse. The result list consists of a list of tuples, with each
tuple containing the VerbNet frame and its associated tree."""
# TODO: Use verbose argument
result_list = []
# In case we're handed an bad string, bail somewhat gracefully
try:
parse_tree = Tree.parse(parse_tree_string)
except ValueError:
print "Error: semantics could not parse tree", repr(parse_tree_string)
return result_list
# Temporarily (and maybe permanently) disabled features:
# 1. Clause splitting: we have not found any example where it does something
# 2. Activizing clauses: for now, passives do not matter.
# Split clauses to handle them separately
# split_clause_dict = split_clauses(parse_tree)
# Activize clauses
# for key, (clause, conjunction) in split_clause_dict.items():
# activized_clause = activize_clause(clause)
# split_clause_dict[key] = (activized_clause, conjunction)
# TODO: This strange loop is because split_clauses may not work
# for (clause, conjunction) in split_clause_dict.values():
for clause, conjunction in ((parse_tree, ''), ):
# Split conjunctions and duplicate arguments if necessary
#split_tree_dict = split_conjunctions(clause)
split_tree_dict = split_conjunctions_sparse(clause)
#DEBUG
#print "SPLIT TREE DICT: ---------------"
#print split_tree_dict
if conjunction != '':
result_list.append(conjunction)
for (split_tree, conjunction) in split_tree_dict.values():
for tree in split_tree:
if conjunction and verbose:
print "Subtree ({}):".format(conjunction)
print tree.pprint(force_multiline=True)
# TODO: Deactivated for now
# Store whether there was an existential there
# if is_existential(str(tree)):
# tag_list.append('ex')
# Transformational grammar stuff
orig_tree = tree
tree = assertion_filter(tree)
tree = existential_there_insertion(tree)
tree = invert_clause(tree)
tree = wh_movement(tree)
if verbose and tree != orig_tree:
print 'Transformed tree:'
print tree.pprint(force_multiline=True)
match = match_verb(tree, verbose=verbose)
if match:
result_list.append(match)
return result_list