def extract_models_actions(input_text, models_fullInfo):
object_coref = models_char.get_coref(input_text)
get_objectCoref_map(input_text)
lemmatizer = WordNetLemmatizer()
tokinized_sentences = []
doc = nlp(input_text)
for sent in doc.sents:
tokinized_sentences.append(str(sent.text))
models_actions = []
sentence_index = -1
for sent in doc.sents:
sentence_index += 1
tokens_in_currentSentence = []
for word in sent:
tokens_in_currentSentence.append(str(word))
verb = subj = obj = ""
if word.pos_ == "VERB" and (
(lemmatizer.lemmatize(str(word), 'v') in verb_noObject) or
(lemmatizer.lemmatize(str(word), 'v') in verb_prep) or
(lemmatizer.lemmatize(str(word), 'v') in verb_oneObject) or
(lemmatizer.lemmatize(str(word), 'v') in verb_twoObject)):
verb = lemmatizer.lemmatize(str(word), 'v')
obj_word = word
subj_word = word
subj_id = -1
obj1_id = -1
obj2_id = -1
###################### find objects ##################
if (verb in verb_oneObject) or (verb in verb_twoObject):
for child in word.children:
if child.dep_ == "dobj":
if child.pos_ == "PRON":
obj_pronoun = str(child.text).lower()
coref_obj = get_refrencedObject(obj_pronoun)
objInfo = models_char.extract_models_char(
coref_obj)
coref_obj_word = ""
coref_obj_chars = []
for i in range(0, len(objInfo)):
if objInfo[i][1] == 'boy' or objInfo[i][
1] == 'man':
if obj_pronoun in models_char.male_pronoun:
coref_obj_word = objInfo[i][1]
coref_obj_chars = objInfo[i][2]
break
elif objInfo[i][1] == 'girl' or objInfo[i][
1] == 'woman':
if obj_pronoun in models_char.female_pronoun:
coref_obj_word = objInfo[i][1]
coref_obj_chars = objInfo[i][2]
break
elif obj_pronoun in models_char.rigid_pronoun:
coref_obj_word = objInfo[i][1]
coref_obj_chars = objInfo[i][2]
break
obj1_id = get_model_id(models_fullInfo,
coref_obj_word,
coref_obj_chars)
else:
obj_word = child
################################## detect object type ####################################
model_type, object_coref = detect_object_type(
str(obj_word), object_coref)
if model_type == 'none':
continue
################################## detect obj_chars ###########################################
object_chars = []
object_chars = detect_object_char(
model_type, obj_word)
obj1_id = get_model_id(models_fullInfo,
model_type,
object_chars)
if (verb in verb_prep) or (
verb
in verb_twoObject): #----------------------> add coref
for child in word.children:
if child.dep_ == "prep":
current_word = child
for child_current_word in current_word.children:
if child_current_word.dep_ == "pobj":
if child_current_word.pos_ == "PRON":
obj_pronoun = str(
child_current_word.text).lower()
coref_obj = get_refrencedObject(
obj_pronoun)
objInfo = models_char.extract_models_char(
coref_obj)
coref_obj_word = ""
coref_obj_chars = []
for i in range(0, len(objInfo)):
if objInfo[i][
1] == 'boy' or objInfo[i][
1] == 'man':
if obj_pronoun in models_char.male_pronoun:
coref_obj_word = objInfo[
i][1]
coref_obj_chars = objInfo[
i][2]
break
elif objInfo[i][
1] == 'girl' or objInfo[i][
1] == 'woman':
if obj_pronoun in models_char.female_pronoun:
coref_obj_word = objInfo[
i][1]
coref_obj_chars = objInfo[
i][2]
break
elif obj_pronoun in models_char.rigid_pronoun:
coref_obj_word = objInfo[i][1]
coref_obj_chars = objInfo[i][2]
break
obj_id = get_model_id(
models_fullInfo, coref_obj_word,
coref_obj_chars)
if verb in verb_twoObject:
obj2_id = obj_id
else:
obj1_id = obj_id
else:
obj_word = child_current_word
################################## detect object type ####################################
model_type, object_coref = detect_object_type(
str(obj_word), object_coref)
if model_type == 'none':
continue
################################## detect obj_chars ###########################################
object_chars = []
object_chars = detect_object_char(
model_type, obj_word)
obj_id = get_model_id(
models_fullInfo, model_type,
object_chars)
if verb in verb_twoObject:
obj2_id = obj_id
else:
obj1_id = obj_id
############################################## find subjects ###############################################
found_subj = False
for child in word.children:
######### first subject ###############
if child.dep_ == "nsubj":
found_subj = True
found_conjSubject = True
current_subject = child
while found_conjSubject:
###### check if it is a PRON , then get its coreference , otherwise put it as it is ######
### get the num of pronouns preceding it from the start
if current_subject.pos_ == "PRON":
subj_pronoun = str(
current_subject.text).lower()
coref_subj = get_refrencedObject(subj_pronoun)
subjInfo = models_char.extract_models_char(
coref_subj)
coref_subj_word = ""
coref_subj_chars = []
for i in range(0, len(subjInfo)):
if subjInfo[i][1] == 'boy' or subjInfo[i][
1] == 'man':
if subj_pronoun in models_char.male_pronoun:
coref_subj_word = subjInfo[i][1]
coref_subj_chars = subjInfo[i][2]
break
elif subjInfo[i][1] == 'girl' or subjInfo[
i][1] == 'woman':
if subj_pronoun in models_char.female_pronoun:
coref_subj_word = subjInfo[i][1]
coref_subj_chars = subjInfo[i][2]
break
elif subj_pronoun in models_char.rigid_pronoun:
coref_subj_word = subjInfo[i][1]
coref_subj_chars = subjInfo[i][2]
break
coref_subj_id = get_model_id(
models_fullInfo, coref_subj_word,
coref_subj_chars)
################################# Add a new action to action list #######################
# action --> cat_num,verb_name,subj_id,obj1_id,obj2_id,action_pos
if verb in verb_noObject:
action = (1, verb, coref_subj_id, obj1_id,
obj2_id, word.i)
elif verb in verb_prep:
action = (2, verb, coref_subj_id, obj1_id,
obj2_id, word.i)
elif verb in verb_oneObject:
action = (3, verb, coref_subj_id, obj1_id,
obj2_id, word.i)
else:
action = (4, verb, coref_subj_id, obj1_id,
obj2_id, word.i)
models_actions.append(action)
###### check if there is a conj subject ########
found_conjSubject = False
for child_subj in current_subject.children:
if child_subj.dep_ == "conj":
current_subject = child_subj
found_conjSubject = True
break
else:
subj_word = current_subject
################################## detect subj type ####################################
model_type, object_coref = detect_object_type(
str(subj_word), object_coref)
if model_type == 'none':
continue
################################# find subj chars #######################################
object_chars = []
object_chars = detect_object_char(
model_type, subj_word)
subj_id = get_model_id(models_fullInfo,
model_type,
object_chars)
################################# Add a new action to action list #######################
# action --> cat_num,verb_name,subj_id,obj1_id,obj2_id,action_pos
if verb in verb_noObject:
action = (1, verb, subj_id, obj1_id,
obj2_id, word.i)
elif verb in verb_prep:
action = (2, verb, subj_id, obj1_id,
obj2_id, word.i)
elif verb in verb_oneObject:
action = (3, verb, subj_id, obj1_id,
obj2_id, word.i)
else:
action = (4, verb, subj_id, obj1_id,
obj2_id, word.i)
models_actions.append(action)
###### check if there is a conj subject ########
found_conjSubject = False
for child_subj in current_subject.children:
if child_subj.dep_ == "conj":
current_subject = child_subj
found_conjSubject = True
break
######################################## if there are conj verbs of the same subject ######################
if found_subj == False:
current_verb = word
found_conj_verb = False
while ((found_subj)
== False) and (current_verb.head.pos_
== "VERB") and (current_verb.head.i <
current_verb.i):
found_conj_verb = True
current_verb = current_verb.head
for child_verb in current_verb.children:
if child_verb.dep_ == "nsubj":
found_subj = True
break
if found_conj_verb == True: ### conj_verb --> copy only the subject
for verb_info in models_actions:
if verb_info[1] == lemmatizer.lemmatize(
str(current_verb), 'v'):
subj_id = verb_info[2]
else: # while playing --> copy all the information (subject and object(s) )
for verb_info in models_actions:
if verb_info[1] == lemmatizer.lemmatize(
str(current_verb), 'v'):
subj_id = verb_info[2]
obj1_id = verb_info[3]
obj2_id = verb_info[4]
################################# Add a new action to action list #######################
# action --> cat_num,verb_name,subj_id,obj1_id,obj2_id,action_pos
if verb in verb_noObject:
action = (1, verb, subj_id, obj1_id, obj2_id, word.i)
elif verb in verb_prep:
action = (2, verb, subj_id, obj1_id, obj2_id, word.i)
elif verb in verb_oneObject:
action = (3, verb, subj_id, obj1_id, obj2_id, word.i)
else:
action = (4, verb, subj_id, obj1_id, obj2_id, word.i)
models_actions.append(action)
return models_actions
####################### files for the following modules ###########################
file_models_char = open("models_char.txt", "w")
file_models_actions = open("model_actions.txt", "w")
file_models_relations = open("models_relations.txt", "w")
##################################################################################
file_input_text = open("input_text.txt", "r")
###################### input text ############################################### ---------- > take from GUI
input_text = file_input_text.read()
#print(input_text)
#input_text = "There is an old , tall and smart gentleman in a room. He has a small white cat and he carries it. There is a small black table behind a huge red chair. There is a black laptop to the right of a large brown bed."
################################################################################
models_info = extract_models_char(input_text)
relations_models = Objs_relations(input_text, models_info)
model_actions = extract_models_actions(input_text, models_info)
sequence(input_text)
########################### write model chars ######################
for i in range(0, len(models_info)):
current_model_name = models_info[i][1]
current_model_chars = models_info[i][2]
file_models_char.write(current_model_name + " ")
if len(current_model_chars) == 3: #human
for j in range(0, len(current_model_chars) - 1):
if j == 0 and current_model_chars[
j] == -1: #not mentioned age then set it not old
current_model_chars[j] = 0
def Objs_relations(input_text,models_fullInfo):
object_coref = models_char.get_coref(input_text)
get_objectCoref_map(input_text)
doc = nlp(input_text)
relations = []
for sentence in doc.sents:
for word in sentence:
current_prep = pnoun = pobj = " "
pnoun_id = -1
pobj_id = -1
if word.pos_ == "ADP" and (str(word) in rel_avail )and ( (str(word.head.pos_) != "VERB") or (lemmatizer.lemmatize(str(word.head),'v') == "be" )): #make sure that it isn't a prep with a verb
current_prep = str(word)
pnoun_word = word
pobj_word = word
pnoun_list = []
#################################### find pnoun #########################################
if word.head.pos_ == "NOUN": # There is a chair on the left of a table
pnoun_word = word.head
################################# get conj_pnoun ########################################
current_pnoun = pnoun_word
found_conjPnoun = True
while found_conjPnoun:
found_conjPnoun = False
if current_pnoun.dep_ == "conj":
current_pnoun = pnoun_word.head
found_conjSubject = True
########### detect type of current pnoun ################
model_type_conjPnoun, object_coref = detect_object_type(str(current_pnoun), object_coref)
if model_type_conjPnoun == 'none':
continue
########## detect char of current pnoun #################
object_chars = []
object_chars = detect_object_char(model_type_conjPnoun, current_pnoun)
new_pnoun_id = get_model_id(models_fullInfo, model_type_conjPnoun, object_chars)
pnoun_info = [model_type_conjPnoun, new_pnoun_id]
pnoun_list.append(pnoun_info)
break
################################## detect pnoun type ####################################
model_type_mainPnoun, object_coref = detect_object_type(str(pnoun_word), object_coref)
if model_type_mainPnoun == 'none':
continue
################################## get pnoun char #######################################
object_chars = []
object_chars = detect_object_char(model_type_mainPnoun, pnoun_word)
pnoun_id = get_model_id(models_fullInfo, model_type_mainPnoun, object_chars)
pnoun_info = [model_type_mainPnoun, pnoun_id]
pnoun_list.append(pnoun_info)
elif word.head.pos_ == "VERB": # a chair is on the left of a table
verb_word = word.head
for child_verb in verb_word.children:
if child_verb.dep_ == "nsubj" or child_verb.dep_ == "attr": # a box is on the table # there is a chair in the room
if child_verb.pos_ == "PRON":
pnoun_pronoun = str(child_verb.text).lower()
coref_pnoun = get_refrencedObject(pnoun_pronoun)
pnounInfo = models_char.extract_models_char(coref_pnoun)
coref_obj_word = ""
coref_obj_chars = []
for i in range(0, len(pnounInfo)):
if pnounInfo[i][1] == 'boy' or pnounInfo[i][1] == 'man':
if pnoun_pronoun in models_char.male_pronoun:
coref_obj_word = pnounInfo[i][1]
coref_pnoun_chars = pnounInfo[i][2]
break
elif pnounInfo[i][1] == 'girl' or pnounInfo[i][1] == 'woman':
if pnoun_pronoun in models_char.female_pronoun:
coref_obj_word = pnounInfo[i][1]
coref_pnoun_chars = pnounInfo[i][2]
break
elif pnoun_pronoun in models_char.rigid_pronoun:
coref_obj_word = pnounInfo[i][1]
coref_pnoun_chars = pnounInfo[i][2]
break
pnoun_id = get_model_id(models_fullInfo, coref_obj_word,coref_pnoun_chars)
pnoun_info = [coref_obj_word, pnoun_id]
pnoun_list.append(pnoun_info)
else:
pnoun_word = child_verb
################################# get conj_pnoun ########################################
current_pnoun = pnoun_word
found_conjPnoun = True
while found_conjPnoun:
found_conjPnoun = False
for child_pnoun in current_pnoun.children:
if child_pnoun.dep_ == "conj":
current_pnoun = child_pnoun
found_conjSubject = True
########### detect type of current pnoun ################
model_type_conjPnoun, object_coref = detect_object_type(str(current_pnoun),
object_coref)
if model_type_conjPnoun == 'none':
continue
########## detect char of current pnoun #################
object_chars = []
object_chars = detect_object_char(model_type_conjPnoun, current_pnoun)
new_pnoun_id = get_model_id(models_fullInfo, model_type_conjPnoun,
object_chars)
pnoun_info = [model_type_conjPnoun, new_pnoun_id]
pnoun_list.append(pnoun_info)
break
################################## detect pnoun type ####################################
model_type_mainPnoun, object_coref = detect_object_type(str(pnoun_word), object_coref)
if model_type_mainPnoun == 'none':
continue
################################## get pnoun char #######################################
object_chars = []
object_chars = detect_object_char(model_type_mainPnoun, pnoun_word)
pnoun_id = get_model_id(models_fullInfo, model_type_mainPnoun, object_chars)
pnoun_info = [model_type_mainPnoun, pnoun_id]
pnoun_list.append(pnoun_info)
################################# find pobj ############################################
for child in word.children:
if child.dep_ == "pobj":
child_refer_to_prep = False
for Child in child.children:
if Child.dep_ == "prep":
child_refer_to_prep = True
break
if not child_refer_to_prep:
pobj = str(child)
pobj_word = child
################################## detect pnoun type ####################################
model_type, object_coref = detect_object_type(pobj, object_coref)
if model_type == 'none':
continue
################################## get pnoun char #######################################
object_chars = []
object_chars = detect_object_char(model_type, pobj_word)
pobj_id = get_model_id(models_fullInfo, model_type, object_chars)
######################################### add new relation ##############################
for i in range(0,len(pnoun_list)):
rel = [current_prep, pnoun_list[i][0], pnoun_list[i][1], model_type, pobj_id]
relations.append(rel)
elif (word.pos_ == "ADP") and (str(word) == "of") and (str(word.head) in rel_avail):
current_prep = str(word.head)
pnoun_word = word
pobj_word = word
pnoun_list = []
#################################### find pnoun #########################################
current_word = word
word_in_on = current_word.head.head
if word_in_on.head.pos_ == "NOUN": # There is a chair on the left of a table
pnoun_word = word_in_on.head
################################# get conj_pnoun ########################################
current_pnoun = pnoun_word
found_conjPnoun = True
while found_conjPnoun:
found_conjPnoun = False
if current_pnoun.dep_ == "conj":
current_pnoun = pnoun_word.head
found_conjSubject = True
########### detect type of current pnoun ################
model_type_conjPnoun, object_coref = detect_object_type(str(current_pnoun),object_coref)
if model_type_conjPnoun == 'none':
continue
########## detect char of current pnoun #################
object_chars = []
object_chars = detect_object_char(model_type_conjPnoun, current_pnoun)
new_pnoun_id = get_model_id(models_fullInfo, model_type_conjPnoun, object_chars)
pnoun_info = [model_type_conjPnoun, new_pnoun_id]
pnoun_list.append(pnoun_info)
break
################################## detect pnoun type ####################################
model_type_mainPnoun, object_coref = detect_object_type(str(pnoun_word), object_coref)
if model_type_mainPnoun == 'none':
continue
################################## get pnoun char #######################################
object_chars = []
object_chars = detect_object_char(model_type_mainPnoun, pnoun_word)
pnoun_id = get_model_id(models_fullInfo, model_type_mainPnoun, object_chars)
pnoun_info = [model_type_mainPnoun, pnoun_id]
pnoun_list.append(pnoun_info)
elif word_in_on.head.pos_ == "VERB": # a chair is on the left of a table
verb_word = word_in_on.head
for child_verb in verb_word.children:
if child_verb.dep_ == "nsubj" or child_verb.dep_ == "attr": # a box is on the table # there is a chair in the room
if child_verb.pos_ == "PRON":
pnoun_pronoun = str(child_verb.text).lower()
coref_pnoun = get_refrencedObject(pnoun_pronoun)
pnounInfo = models_char.extract_models_char(coref_pnoun)
coref_obj_word = ""
coref_obj_chars = []
for i in range(0, len(pnounInfo)):
if pnounInfo[i][1] == 'boy' or pnounInfo[i][1] == 'man':
if pnoun_pronoun in models_char.male_pronoun:
coref_obj_word = pnounInfo[i][1]
coref_pnoun_chars = pnounInfo[i][2]
break
elif pnounInfo[i][1] == 'girl' or pnounInfo[i][1] == 'woman':
if pnoun_pronoun in models_char.female_pronoun:
coref_obj_word = pnounInfo[i][1]
coref_pnoun_chars = pnounInfo[i][2]
break
elif pnoun_pronoun in models_char.rigid_pronoun:
coref_obj_word = pnounInfo[i][1]
coref_pnoun_chars = pnounInfo[i][2]
break
pnoun_id = get_model_id(models_fullInfo, coref_obj_word,coref_pnoun_chars)
pnoun_info = [coref_obj_word, pnoun_id]
pnoun_list.append(pnoun_info)
else:
pnoun_word = child_verb
################################# get conj_pnoun ########################################
current_pnoun = pnoun_word
found_conjPnoun = True
while found_conjPnoun:
found_conjPnoun = False
for child_pnoun in current_pnoun.children:
if child_pnoun.dep_ == "conj":
current_pnoun = child_pnoun
found_conjSubject = True
########### detect type of current pnoun ################
model_type_conjPnoun, object_coref = detect_object_type(str(current_pnoun),
object_coref)
if model_type_conjPnoun == 'none':
continue
########## detect char of current pnoun #################
object_chars = []
object_chars = detect_object_char(model_type_conjPnoun, current_pnoun)
new_pnoun_id = get_model_id(models_fullInfo, model_type_conjPnoun,
object_chars)
pnoun_info = [model_type_conjPnoun, new_pnoun_id]
pnoun_list.append(pnoun_info)
break
################################## detect pnoun type ####################################
model_type_mainPnoun, object_coref = detect_object_type(str(pnoun_word), object_coref)
if model_type_mainPnoun == 'none':
continue
################################## get pnoun char #######################################
object_chars = []
object_chars = detect_object_char(model_type_mainPnoun, pnoun_word)
pnoun_id = get_model_id(models_fullInfo, model_type_mainPnoun, object_chars)
pnoun_info = [model_type_mainPnoun, pnoun_id]
pnoun_list.append(pnoun_info)
################################# find pobj ############################################
for child in word.children:
if child.dep_ == "pobj":
child_refer_to_prep = False
for Child in child.children:
if Child.dep_ == "prep":
child_refer_to_prep = True
break
if not child_refer_to_prep:
pobj = str(child)
pobj_word = child
################################## detect pnoun type ####################################
model_type, object_coref = detect_object_type(pobj, object_coref)
if model_type == 'none':
continue
################################## get pnoun char #######################################
object_chars = []
object_chars = detect_object_char(model_type, pobj_word)
pobj_id = get_model_id(models_fullInfo, pobj, object_chars)
######################################### add new relation ##############################
for i in range(0,len(pnoun_list)):
rel = [current_prep, pnoun_list[i][0], pnoun_list[i][1], model_type, pobj_id]
relations.append(rel)
###########################################################################################
#displacy.serve(parsed_sentence, style='dep')
return relations
def nlp_module(input_text):
####################### files for the following modules ###########################
file_models_char = open("models_char.txt", "w")
file_models_actions = open("model_actions.txt", "w")
file_models_relations = open("models_relations.txt", "w")
##################################################################################
models_info = extract_models_char(input_text)
relations_models = Objs_relations(input_text, models_info)
infered_models, infered_relations = extract_totalInference(
models_info, relations_models)
print("infered_models : ", infered_models)
print("infered_relations : ", infered_relations)
model_actions = extract_models_actions(input_text, models_info)
############## add infered models and relations to models info and models relations #############
for i in range(0, len(infered_models)):
models_info.append(infered_models[i])
for i in range(0, len(infered_relations)):
relations_models.append(infered_relations[i])
########################### write model chars #################################################
for i in range(0, len(models_info)):
current_model_name = models_info[i][1]
current_model_chars = models_info[i][2]
file_models_char.write(current_model_name + " ")
if len(current_model_chars) == 4: # human
for j in range(0, len(current_model_chars) - 1):
if j == 0 and current_model_chars[
j] == -1: # not mentioned age then set it not old
current_model_chars[j] = 0
file_models_char.write(str(current_model_chars[j]) + " ")
elif j == 2 and current_model_chars[
j] == -1: # not mentioned height then set it meduim height
current_model_chars[j] = 1
file_models_char.write(str(current_model_chars[j]) + " ")
else:
file_models_char.write(str(current_model_chars[j]) + " ")
else:
for j in range(0, len(current_model_chars) - 1):
if j == 1 and current_model_chars[
j] == -1: # not mentioned size then set it meduim size
current_model_chars[j] = 1
file_models_char.write(str(current_model_chars[j]) + " ")
else:
file_models_char.write(str(current_model_chars[j]) + " ")
file_models_char.write(str(models_info[i][3]) + "\n")
file_models_char.close()
########################## write model actions #######################
for i in range(0, len(model_actions)):
file_models_actions.write(model_actions[i][1] + " " +
str(model_actions[i][2]) + " " +
str(model_actions[i][3]) + " " +
str(model_actions[i][4]) + " " +
str(model_actions[i][5]) + "\n")
file_models_actions.close()
########################### write models relations ####################
for i in range(0, len(relations_models)):
file_models_relations.write(relations_models[i][0] + " " +
relations_models[i][1] + " " +
str(relations_models[i][2]) + " " +
relations_models[i][3] + " " +
str(relations_models[i][4]) + "\n")
file_models_relations.close()
############################# Extract action sequence ###############
sequence(input_text)
######################################################################
return