def test_ontology():
kb = KnowledgeBase(
[
("i", "like", "cookies"),
("i", "like", "milk"),
("i", "type", "person"),
("i", "type", "living_thing"),
("cookies", "quality", "good"),
]
)
assert kb.types("i") == {"person", "living_thing"}
def test_query():
kb = KnowledgeBase(
[
("i", "like", "cookies"),
("i", "like", "milk"),
("i", "type", "person"),
("i", "type", "living_thing"),
("cookies", "quality", "good"),
]
)
assert kb.query("i", "like", "quality") == {"good"}
def test_ontology():
kb = KnowledgeBase()
ontology = {
"ontology": {
"season": [
"fall",
"spring",
"summer",
"winter"
],
"month": [
"january",
"february",
"march",
"april",
"may",
"june",
"july",
"august",
"september",
"october",
"november",
"december"
]
}
}
kb.load_json(ontology)
df = DialogueFlow(States.A, Speaker.USER, kb=kb)
df.add_state(States.A)
df.add_state(States.B)
df.add_state(States.C)
df.add_state(States.D)
df.add_state(States.E)
df.set_error_successor(States.A, States.E)
df.set_error_successor(States.B, States.E)
df.set_error_successor(States.C, States.E)
df.set_error_successor(States.D, States.E)
df.add_user_transition(States.A, States.B, "[#ONT(month)]")
df.add_system_transition(States.B, States.C, "B to C")
df.add_user_transition(States.C, States.D, "[$m=#ONT(month), $s=#ONT(season)]")
df.user_turn("january")
assert df.state() == States.B
assert df.system_turn() == "B to C"
df.user_turn("october is in the fall season")
assert df.state() == States.D
assert df._vars["m"] == "october"
assert df._vars["s"] == "fall"
df.set_state(States.A)
df.set_speaker(Speaker.USER)
df.user_turn("hello there", debugging=False)
assert df.state() == States.E
def test_expressions():
kb = KnowledgeBase([('i', 'like', 'cookies'), ('i', 'like', 'milk'),
('i', 'type', 'person'), ('i', 'type', 'living_thing'),
('cookies', 'quality', 'good')])
assert kb.expressions('i') == {'i'}
assert kb.expressions('living_thing') == set()
kb.add_expression('living_thing', 'life')
kb.add_expression('living_thing', 'alive')
assert kb.expressions('living_thing') == {'life', 'alive'}
def test_expressions():
kb = KnowledgeBase(
[
("i", "like", "cookies"),
("i", "like", "milk"),
("i", "type", "person"),
("i", "type", "living_thing"),
("cookies", "quality", "good"),
]
)
assert kb.expressions("i") == {"i"}
assert kb.expressions("living_thing") == set()
kb.add_expression("living_thing", "life")
kb.add_expression("living_thing", "alive")
assert kb.expressions("living_thing") == {"life", "alive"}
def test_constructor():
kb = KnowledgeBase(
[
("i", "like", "cookies"),
("i", "like", "milk"),
("i", "type", "person"),
("i", "type", "living_thing"),
("cookies", "quality", "good"),
]
)
def test_constructor():
kb = KnowledgeBase([('i', 'like', 'cookies'), ('i', 'like', 'milk'),
('i', 'type', 'person'), ('i', 'type', 'living_thing'),
('cookies', 'quality', 'good')])
def test_ontology():
kb = KnowledgeBase([('i', 'like', 'cookies'), ('i', 'like', 'milk'),
('i', 'type', 'person'), ('i', 'type', 'living_thing'),
('cookies', 'quality', 'good')])
assert kb.types('i') == {'person', 'living_thing'}
def test_query():
kb = KnowledgeBase([('i', 'like', 'cookies'), ('i', 'like', 'milk'),
('i', 'type', 'person'), ('i', 'type', 'living_thing'),
('cookies', 'quality', 'good')])
assert kb.query('i', 'like', 'quality') == {'good'}
from emora_stdm.state_transition_dialogue_manager.knowledge_base import KnowledgeBase
from emora_stdm.state_transition_dialogue_manager.dialogue_flow import DialogueFlow
from emora_stdm.state_transition_dialogue_manager.composite_dialogue_flow import (
CompositeDialogueFlow, )
# from time import time
kb = KnowledgeBase([
("lion", "type", "cat"),
("panther", "type", "cat"),
("cat", "type", "animal"),
("animal", "type", "thing"),
("lion", "sound", "roar"),
("roar", "quality", "loud"),
("panther", "sound", "growl"),
("growl", "quality", "scary"),
("also", "type", "also_syns"),
("too", "type", "also_syns"),
("basketball", "type", "unknown_hobby"),
("basketball", "expr", "bball"),
("john", "type", "male"),
("mary", "type", "female"),
("male", "type", "person"),
("female", "type", "person"),
("friend", "type", "person"),
])
ont = {"ontology": {"person": ["mom", "dad", "sister", "brother"]}}
kb.load_json(ont)
df = DialogueFlow(initial_state="start")
df.add_state("start", "start")
df.add_system_transition("start", "start", "hello")
import sys
from time import time
import nltk
try:
nltk.data.find('wordnet')
except:
nltk.download('wordnet')
from nltk.corpus import wordnet as wn
import os
from emora_stdm.state_transition_dialogue_manager.knowledge_base import KnowledgeBase
wordnet_knowledge_base = KnowledgeBase()
import importlib_resources as impr
from . import data
try:
pass
# t1 = time()
# sys.stderr.write('Loading wordnet.json... ')
# json_str = impr.read_text(data, 'wordnet.json')
# wordnet_knowledge_base.load_json_string(json_str)
# sys.stderr.write(' done in {}s'.format(time() - t1))
except FileNotFoundError:
sys.stderr.write(' failed to load.')
def related_synsets(word):
"""
returns lemmas
"""
def __init__(self, initial_state: Union[Enum, str, tuple], initial_speaker = Speaker.SYSTEM,
macros: Dict[str, Macro] =None, kb: Union[KnowledgeBase, str, List[str]] =None,
default_system_state=None, end_state='__end__', all_multi_hop=True, wordnet=False):
self._graph = GraphDatabase()
self._initial_state = State(initial_state)
self._potential_transition = None
self._initial_speaker = initial_speaker
self._speaker = self._initial_speaker
self._vars = HashableDict()
self._transitions = []
self._update_transitions = []
self.vars()['__state__'] = self._initial_state
self.set_state(self._initial_state)
self._gates = defaultdict(list)
self._prepends = {}
self._var_dependencies = defaultdict(set)
self._error_transitioned = False
self._default_state = default_system_state
self._end_state = end_state
self._goals = {}
self._all_multi_hop = all_multi_hop
self._composite_dialogue_flow = None
self._namespace = None
self.vars()['__stack__'] = []
self.vars()['__system_state__'] = 'None' if initial_speaker == Speaker.USER else self._initial_state
if kb is None:
self._kb = KnowledgeBase()
elif isinstance(kb, str):
self._kb = KnowledgeBase()
self._kb.load_json_file(kb)
elif isinstance(kb, list):
self._kb = KnowledgeBase()
for filename in kb:
self._kb.load_json_file(filename)
else:
self._kb = kb
onte = ONTE(self._kb)
kbe = KBE(self._kb)
goal_exit_macro = GoalExit(self)
self._macros = {
'WN': WN(wordnet),
'ONT': onte, 'ONTE': onte,
'ONTUL': ONTUL(self._kb),
'KBQ': kbe, 'KBE': kbe,
'ONTN': ONTN(self._kb),
'EXP': EXP(self._kb),
'ONT_NEG': ONT_NEG(self._kb),
'FPP': FirstPersonPronoun(self._kb),
'TPP': ThirdPersonPronoun(self._kb),
'PSP': PossessivePronoun(self._kb),
'GATE': Gate(self),
'TRANSITION': Transition(self),
'GOAL': GoalPursuit(goal_exit_macro, self),
'GCOM': GoalCompletion(self),
'GEXT': goal_exit_macro,
'GSRET': SetGoalReturnPoint(),
'GRET': GoalReturn(self),
'GCLR': ClearGoalStack(),
'VT': VirtualTransitions(self),
'CE': CanEnter(self),
'EXTR': ExtractList(self._kb)
}
self._macros.update(macros_common_dict)
self._macros.update(natex_macros_common)
if macros:
self._macros.update(macros)
self._rules = UpdateRules(vars=self._vars, macros=self._macros)
self.add_state(end_state)
self._vars['__user_utterance__'] = None
class DialogueFlow:
Speaker = Speaker
@classmethod
def autostate(cls):
global _autostate
_autostate = str(int(_autostate) + 1)
return _autostate
def __init__(self, initial_state: Union[Enum, str, tuple], initial_speaker = Speaker.SYSTEM,
macros: Dict[str, Macro] =None, kb: Union[KnowledgeBase, str, List[str]] =None,
default_system_state=None, end_state='__end__', all_multi_hop=True, wordnet=False):
self._graph = GraphDatabase()
self._initial_state = State(initial_state)
self._potential_transition = None
self._initial_speaker = initial_speaker
self._speaker = self._initial_speaker
self._vars = HashableDict()
self._transitions = []
self._update_transitions = []
self.vars()['__state__'] = self._initial_state
self.set_state(self._initial_state)
self._gates = defaultdict(list)
self._prepends = {}
self._var_dependencies = defaultdict(set)
self._error_transitioned = False
self._default_state = default_system_state
self._end_state = end_state
self._goals = {}
self._all_multi_hop = all_multi_hop
self._composite_dialogue_flow = None
self._namespace = None
self.vars()['__stack__'] = []
self.vars()['__system_state__'] = 'None' if initial_speaker == Speaker.USER else self._initial_state
if kb is None:
self._kb = KnowledgeBase()
elif isinstance(kb, str):
self._kb = KnowledgeBase()
self._kb.load_json_file(kb)
elif isinstance(kb, list):
self._kb = KnowledgeBase()
for filename in kb:
self._kb.load_json_file(filename)
else:
self._kb = kb
onte = ONTE(self._kb)
kbe = KBE(self._kb)
goal_exit_macro = GoalExit(self)
self._macros = {
'WN': WN(wordnet),
'ONT': onte, 'ONTE': onte,
'ONTUL': ONTUL(self._kb),
'KBQ': kbe, 'KBE': kbe,
'ONTN': ONTN(self._kb),
'EXP': EXP(self._kb),
'ONT_NEG': ONT_NEG(self._kb),
'FPP': FirstPersonPronoun(self._kb),
'TPP': ThirdPersonPronoun(self._kb),
'PSP': PossessivePronoun(self._kb),
'GATE': Gate(self),
'TRANSITION': Transition(self),
'GOAL': GoalPursuit(goal_exit_macro, self),
'GCOM': GoalCompletion(self),
'GEXT': goal_exit_macro,
'GSRET': SetGoalReturnPoint(),
'GRET': GoalReturn(self),
'GCLR': ClearGoalStack(),
'VT': VirtualTransitions(self),
'CE': CanEnter(self),
'EXTR': ExtractList(self._kb)
}
self._macros.update(macros_common_dict)
self._macros.update(natex_macros_common)
if macros:
self._macros.update(macros)
self._rules = UpdateRules(vars=self._vars, macros=self._macros)
self.add_state(end_state)
self._vars['__user_utterance__'] = None
# TOP LEVEL: SYSTEM-LEVEL USE CASES
def run(self, debugging=False):
"""
test in interactive mode
:return: None
"""
t1 = time()
while self.state() != self.end_state():
if self.speaker() == Speaker.SYSTEM:
system_response = self.system_turn(debugging=debugging)
if debugging:
print('Time delta: {:.5f}'.format(time() - t1))
print("S:", system_response)
else:
user_input = input("U: ")
t1 = time()
self.user_turn(user_input, debugging=debugging)
def system_turn(self, debugging=False):
"""
an entire system turn comprising a single system utterance and
one or more system transitions
:return: the natural language system response
"""
t1 = time()
self.vars()['__goal_return_state__'] = 'None'
visited = {self.state()}
responses = []
while self.speaker() is Speaker.SYSTEM:
response, next_state = self.system_transition(self.state(), debugging=debugging)
self.set_state(next_state)
responses.append(response)
if next_state in visited or (not self.state_settings(next_state).system_multi_hop):
self.set_speaker(Speaker.USER)
visited.add(next_state)
t2 = time()
if debugging:
print('System turn in {:.5f}'.format(t2-t1))
full_response = ' '.join(responses)
self.vars()['__selected_response__'] = full_response
return full_response
def user_turn(self, natural_language, debugging=False):
"""
an entire user turn comprising one user utterance and
one or more user transitions
:param natural_language:
:param debugging:
:return: None
"""
t1 = time()
natural_language = ''.join([c.lower() for c in natural_language if c.isalpha() or c == ' '])
self.vars()['__user_utterance__'] = natural_language
self._transitions.clear()
self.apply_update_rules(natural_language, debugging)
visited = {self.state()}
while self.speaker() is Speaker.USER:
next_state = self.user_transition(natural_language, self.state(), debugging=debugging)
if self._error_transitioned and next_state != self.state():
try:
nns = self.user_transition(natural_language, next_state, debugging=debugging)
if nns not in visited:
next_state = nns
except RuntimeError:
if debugging:
print("Couldn't error hop")
self.set_state(next_state)
if next_state in visited or (not self.state_settings(next_state).user_multi_hop):
self.set_speaker(Speaker.SYSTEM)
visited.add(next_state)
self.set_speaker(Speaker.SYSTEM)
t2 = time()
if debugging:
print('User turn in {:.5f}'.format(t2 - t1))
def load_transitions(self, json_dict, speaker=None):
"""
wheeeeeeee!
"""
if speaker is None:
speaker = self._initial_speaker
if 'state' in json_dict:
source = json_dict['state']
else:
source = DialogueFlow.autostate()
hop = None
switch = False
enter = None
# read settings and transitions for state
transitions = []
for key, value in json_dict.items():
if key == 'transitions':
assert isinstance(value, list)
transitions = value
elif key == 'root':
root = json_dict['root']
elif key == 'hop':
hop = json_dict['hop']
elif key == 'prepend':
prepend = json_dict['prepend']
self.set_state_prepend(source, prepend)
elif key == 'switch':
switch = json_dict['switch']
elif key == 'enter':
enter = json_dict['enter']
elif key not in {'state', 'hop', 'score', 'switch', 'enter'}:
transitions.append((key, value))
# set up state settings
if not self.has_state(source):
self.add_state(source)
if hop:
if speaker == Speaker.USER:
speaker = Speaker.SYSTEM
self.state_settings(source).update(system_multi_hop=True)
elif speaker == Speaker.SYSTEM:
speaker = Speaker.USER
self.state_settings(source).update(user_multi_hop=True)
if switch:
self.update_state_settings(source, switch=True)
if enter:
self.update_state_settings(source, enter=enter)
# set up transitions
expanded_transitions = []
for natex, target in transitions:
natex_with_leading_digits_stripped = ''
i = 0
c = natex[i] if natex else ''
while c and c.isnumeric():
natex_with_leading_digits_stripped += c
i += 1
c = natex[i] if i < len(natex) else ''
if natex == 'error':
if isinstance(target, dict):
if 'state' not in target:
target['state'] = DialogueFlow.autostate()
expanded_transitions.append(target)
target = target['state']
if not self.has_state(target):
self.add_state(target)
self.set_error_successor(source, target)
else:
score = 1.0
if isinstance(target, dict):
if 'state' not in target:
target['state'] = DialogueFlow.autostate()
if 'score' in target:
score = target['score']
expanded_transitions.append(target)
target = target['state']
if not self.has_state(target):
self.add_state(target)
if speaker == Speaker.USER:
if self.has_transition(source, target, Speaker.USER):
intermediate = '_' + self.autostate()
self.add_state(intermediate, target)
self.add_user_transition(source, intermediate, natex + ' #TARGET(%s)' % target, score=score)
else:
self.add_user_transition(source, target, natex, score=score)
elif speaker == Speaker.SYSTEM:
if self.has_transition(source, target, Speaker.SYSTEM):
intermediate = '_' + self.autostate()
self.add_state(intermediate)
self.add_system_transition(source, intermediate, natex + ' #TARGET(%s)' % target, score=score)
else:
self.add_system_transition(source, target, natex, score=score)
# switch turn (will be switched back if multi hop detected on next recursive call)
if speaker == Speaker.USER:
speaker = Speaker.SYSTEM
elif speaker == Speaker.SYSTEM:
speaker = Speaker.USER
# recurse to load nested turns
for transition in expanded_transitions:
self.load_transitions(transition, speaker)
# HIGH LEVEL
def system_transition(self, state: Union[Enum, str, tuple], debugging=False):
"""
:param state:
:param debugging:
:return: a <state, response> tuple representing the successor state and response
"""
if '__gate__' in self._vars:
del self._vars['__gate__']
state = module_state(state)
ti = time()
if state is None:
state = self.state()
else:
state = State(state)
transition_options = []
transitions = list(self.transitions(state, Speaker.SYSTEM))
transition_items = []
for transition in transitions:
natex = self.transition_natex(*transition)
score = self.transition_settings(*transition).score
transition_items.append((natex, transition, score))
while self._transitions:
natex, transition, score = self._transitions.pop()
transition_items.append((natex, transition, score))
while self._update_transitions:
natex, transition, score = self._update_transitions.pop()
transition_items.append((natex, transition, score))
for natex, transition, score in transition_items:
t1 = time()
transition_transition_enter = None
vars = HashableDict(self._vars)
self._potential_transition = transition # MOVED, todo
try:
generation = natex.generate(vars=vars, macros=self._macros, debugging=debugging)
except Exception as e:
print()
print('Transition {}: {} failed'.format(str(transition), natex))
traceback.print_exc(file=sys.stdout)
print()
generation = None
source, target, speaker = transition
if '__source__' in vars:
source = State(module_state(vars['__source__']))
del vars['__source__']
if '__target__' in vars:
target = State(module_state(vars['__target__']))
del vars['__target__']
transition = source, target, speaker
# if not self.is_module() and isinstance(target, tuple):
# continue
if '->' in transition[1]:
_src, _tar = target.split('->')[0], target.split('->')[1]
_tar = State(module_state(_tar))
transition = (_src, _tar, speaker)
try:
appended_generation = self.transition_natex(*transition).generate(vars=vars, macros=self._macros, debugging=debugging)
if appended_generation is None:
generation = None
else:
generation = generation + ' ' + appended_generation
except Exception as e:
print()
print('Transition {}: {} failed'.format(str(transition), natex))
traceback.print_exc(file=sys.stdout)
print()
generation = None
elif isinstance(transition[1], tuple) and '->' in transition[1][1]:
namespace = transition[1][0]
source, target = (namespace, target[1].split('->')[0]), target[1].split('->')[1]
target = State(module_state(target))
transition_transition_enter = source
transition = (source, target, speaker)
try:
appended_generation = self.composite_dialogue_flow().transition_natex(
namespace, *transition).generate(vars=vars, macros=self._macros, debugging=debugging)
if generation is None or appended_generation is None:
generation = None
else:
generation = generation + ' ' + appended_generation
except Exception as e:
print()
print('Transition {}: {} failed'.format(str(transition), natex))
traceback.print_exc(file=sys.stdout)
print()
generation = None
source, target, speaker = transition
if '__source__' in vars:
source = State(module_state(vars['__source__']))
del vars['__source__']
if '__target__' in vars:
target = State(module_state(vars['__target__']))
del vars['__target__']
transition = source, target, speaker
enter_natex_pass = True
transition_transition_enter_vars = vars
if transition_transition_enter is not None:
if self.is_module() and isinstance(transition_transition_enter, tuple):
enter_natex = self.composite_dialogue_flow().state_settings(*transition_transition_enter).enter
else:
enter_natex = self.state_settings(transition_transition_enter).enter
if enter_natex is not None:
try:
enter_natex_pass = enter_natex.generate(vars=transition_transition_enter_vars, macros=self._macros, debugging=debugging)
except Exception as e:
print()
print(e)
print('Enter Natex {}: {} failed'.format(str(transition_transition_enter), enter_natex))
print()
enter_natex_pass = None
if enter_natex_pass:
if self.is_module() and isinstance(target, tuple):
enter_natex = self.composite_dialogue_flow().state_settings(*target).enter
else:
enter_natex = self.state_settings(target).enter
if enter_natex is not None:
try:
enter_natex_pass = enter_natex.generate(vars=vars, macros=self._macros, debugging=debugging)
except Exception as e:
print()
print(e)
print('Enter Natex {}: {} failed'.format(str(target), enter_natex))
print()
enter_natex_pass = None
if generation is not None and enter_natex_pass is not None:
if '__score__' in vars:
score = vars['__score__']
del vars['__score__']
gate_closed = False
gate_var_config = None
gate_target_id = None
if '__gate__' in vars:
gate_var_config = vars['__gate__']
gate_target_id = (self.namespace(), target) if (not isinstance(target, tuple) and self.is_module()) else target
for vc in self.gates()[gate_target_id]:
if gate_var_config == vc:
gate_closed = True
del vars['__gate__']
tt_gate_var_config = None
tt_gate_target_id = None
if transition_transition_enter is not None and '__gate__' in transition_transition_enter_vars:
tt_gate_var_config = transition_transition_enter_vars['__gate__']
tt_gate_target_id = (self.namespace(), transition_transition_enter) if \
(not isinstance(transition_transition_enter, tuple) and self.is_module()) else transition_transition_enter
for vc in self.gates()[tt_gate_target_id]:
if tt_gate_var_config == vc:
gate_closed = True
del transition_transition_enter_vars['__gate__']
transition_transition_enter_vars.update(vars)
vars = transition_transition_enter_vars
if not gate_closed:
transition_options.append((score, natex, generation, transition, vars, gate_var_config, gate_target_id, tt_gate_var_config, tt_gate_target_id))
t2 = time()
if debugging:
print('Transition {} evaluated in {:.5f}'.format(transition, t2-t1))
while self._transitions:
natex, transition, score = self._transitions.pop()
transition_items.append((natex, transition, score))
self._transitions.clear()
if transition_options:
if debugging:
print('Transition options: ------------')
for option in transition_options:
print('{} {}: {}'.format(option[0], option[3][1], option[1]))
print('--------------------------------')
score, natex, response, transition, vars, gate_var_config, gate_target_id, tt_gate_var_config, tt_gate_target_id =\
random_max(transition_options, key=lambda x: x[0])
if gate_var_config is not None:
self.gates()[gate_target_id].append(gate_var_config)
if tt_gate_var_config is not None:
self.gates()[tt_gate_target_id].append(tt_gate_var_config)
if debugging:
updates = {}
for k, v in vars.items():
if k not in self._vars or v != self._vars[k]:
updates[k] = v
if updates:
print('Updating vars:')
for k, v in updates.items():
if k in self._vars:
print(' {} = {} -> {}'.format(k, self._vars[k], v))
else:
print(' {} = None -> {}'.format(k, v))
self.update_vars(vars)
next_state = transition[1]
if debugging:
tf = time()
print('System transition in {:.5f}'.format(tf-ti))
print('Transitioning {} -> {}'.format(self.state(), next_state))
if '__response_prefix__' in self.vars() and self.vars()['__response_prefix__'] != 'None':
response = self.vars()['__response_prefix__'] + ' ' + response
self.vars()['__response_prefix__'] = 'None'
return response, next_state
else:
if self._default_state is not None:
self.set_state(self._default_state)
if debugging:
print('No valid system transitions found, going to default state...')
return self.system_transition(self.state(), debugging=debugging)
raise AssertionError('dialogue flow system transition found no valid options from state {}'.format(state))
def user_transition(self, natural_language: str, state: Union[Enum, str, tuple], debugging=False):
"""
:param state:
:param natural_language:
:param debugging:
:return: the successor state representing the highest score user transition
that matches natural_language, or None if none match
"""
if '__gate__' in self._vars:
del self._vars['__gate__']
if '__user_utterance__' in self.vars() and self.vars()['__user_utterance__'] is not None:
natural_language = self.vars()['__user_utterance__']
else:
natural_language = ''.join([c.lower() for c in natural_language if c.isalpha() or c == ' '])
state = module_state(state)
self._error_transitioned = False
ti = time()
if state is None:
state = self.state()
else:
state = State(state)
transition_options = []
transition_items = []
for transition in self.transitions(state, Speaker.USER):
natex = self.transition_natex(*transition)
score = self.transition_settings(*transition).score
transition_items.append((natex, transition, score))
while self._transitions:
natex, transition, score = self._transitions.pop()
transition_items.append((natex, transition, score))
ngrams = Ngrams(natural_language, n=10)
for natex, transition, score in transition_items:
self._potential_transition = transition
if not self.is_module() and isinstance(transition[1], tuple):
continue
t1 = time()
if debugging:
print('Evaluating transition {}'.format(transition[:2]))
vars = HashableDict(self._vars)
try:
match = natex.match(natural_language, vars, self._macros, ngrams, debugging)
except Exception as e:
print()
print('Transition {}: {} failed'.format(str(transition), natex))
traceback.print_exc(file=sys.stdout)
print()
match = None
source, target, speaker = transition
if '__source__' in vars:
source = State(module_state(vars['__source__']))
del vars['__source__']
if '__target__' in vars:
target = State(module_state(vars['__target__']))
del vars['__target__']
transition = source, target, speaker
if self.is_module() and isinstance(target, tuple):
enter_natex = self.composite_dialogue_flow().state_settings(*target).enter
else:
enter_natex = self.state_settings(target).enter
enter_natex_pass = True
if enter_natex is not None:
try:
enter_natex_pass = enter_natex.generate(vars=vars, macros=self._macros, debugging=debugging)
except Exception as e:
print()
print(e)
print('Enter Natex {}: {} failed'.format(str(target), enter_natex))
print()
enter_natex_pass = None
if match and enter_natex_pass is not None:
if debugging:
print('Transition {} matched "{}"'.format(transition[:2], natural_language))
if '__score__' in vars:
score = vars['__score__']
del vars['__score__']
gate_closed = False
gate_var_config = None
gate_target_id = None
if '__gate__' in vars:
gate_var_config = vars['__gate__']
gate_target_id = (self.namespace(), target) if (
not isinstance(target, tuple) and self.is_module()) else target
for vc in self.gates()[gate_target_id]:
if gate_var_config == vc:
gate_closed = True
del vars['__gate__']
if not gate_closed:
transition_options.append((score, natex, transition, vars, gate_var_config, gate_target_id))
t2 = time()
if debugging:
print('Transition {} evaluated in {:.5f}'.format(transition, t2-t1))
while self._transitions:
natex, transition, score = self._transitions.pop()
transition_items.append((natex, transition, score))
self._transitions.clear()
if transition_options:
if debugging:
print('Transition options: ------------')
for option in transition_options:
print('{} {}: {}'.format(option[0], option[2][1], option[1]))
print('--------------------------------')
score, natex, transition, vars, gate_var_config, gate_target_id = random_max(transition_options, key=lambda x: x[0])
if gate_var_config is not None:
self.gates()[gate_target_id].append(gate_var_config)
if debugging:
updates = {}
for k, v in vars.items():
if k not in self._vars or v != self._vars[k]:
updates[k] = v
if updates:
print('Updating vars:')
for k, v in updates.items():
if k in self._vars:
print(' {} = {} -> {}'.format(k, self._vars[k], v))
else:
print(' {} = None -> {}'.format(k, v))
self.update_vars(vars)
next_state = transition[1]
if debugging:
print('User transition in {:.5f}'.format(time() - ti))
print('Transitioning {} -> {}'.format(self.state(), next_state))
return next_state
else:
self._error_transitioned = True
next_state = self.error_successor(self.state())
if debugging:
print('User transition in {:.5f}'.format(time() - ti))
print('Error transition {} -> {}'.format(self.state(), next_state))
return next_state
def precache_transitions(self, process_num=1):
"""
Make DialogueFlow fast from the start with the power of precache!
"""
if process_num == 1:
for transition in self._graph.arcs():
data = self._graph.arc_data(*transition)
data['natex'].precache()
for rule in self.update_rules().rules:
rule.precondition.precache()
if rule.postcondition:
rule.postcondition.precache()
else:
# transition_data_sets = []
# for i in range(process_num):
# transition_data_sets.append([])
# count = 0
# for transition in self._graph.arcs():
# transition_data_sets[count].append(self._graph.arc_data(*transition))
# count = (count + 1) % process_num
#
# print("multiprocessing...")
# p = Pool(process_num)
# results = p.map(precache, transition_data_sets)
# for i in range(len(results)):
# result_list = results[i]
# t_list = transition_data_sets[i]
# for j in range(len(result_list)):
# parsed_tree = result_list[j]
# t = t_list[j]
# t['natex']._compiler._parsed_tree = parsed_tree
raise NotImplementedError()
def check(self, debugging=False):
all_good = True
for state in self._graph.nodes():
has_system_fallback = False
has_user_fallback = False
for source, target, speaker in self._graph.arcs_out(state):
if speaker == Speaker.SYSTEM:
if self.transition_natex(source, target, speaker).is_complete():
has_system_fallback = True
if self.error_successor(state) is not None:
has_user_fallback = True
in_labels = {x[2] for x in self.incoming_transitions(state)}
if Speaker.SYSTEM in in_labels:
if not has_user_fallback:
if debugging:
print('WARNING: Turn-taking dead end: state {} has no fallback user transition'.format(state))
all_good = False
if Speaker.USER in in_labels:
if not has_system_fallback:
if debugging:
print('WARNING: Turn-taking dead end: state {} may have no fallback system transitions'.format(state))
all_good = False
return all_good
def add_user_transition(self, source: Union[Enum, str, tuple], target: Union[Enum, str, tuple],
natex_nlu: Union[str, NatexNLU, List[str]], **settings):
source, target = module_source_target(source, target)
source = State(source)
target = State(target)
if self.has_transition(source, target, Speaker.USER):
raise ValueError('user transition {} -> {} already exists'.format(source, target))
natex_nlu = NatexNLU(natex_nlu, macros=self._macros)
if not self.has_state(source):
self.add_state(source)
if not self.has_state(target):
self.add_state(target)
self._graph.add_arc(source, target, Speaker.USER)
self.set_transition_natex(source, target, Speaker.USER, natex_nlu)
transition_settings = Settings(score=1.0)
transition_settings.update(**settings)
if self._all_multi_hop:
self.update_state_settings(source, user_multi_hop=True)
self.set_transition_settings(source, target, Speaker.USER, transition_settings)
if target in self._prepends:
prepend = self._prepends[target]
natex = self.transition_natex(source, target, Speaker.USER)
self.set_transition_natex(source, target, Speaker.USER, prepend + natex)
def add_system_transition(self, source: Union[Enum, str, tuple], target: Union[Enum, str, tuple],
natex_nlg: Union[str, NatexNLG, List[str]], **settings):
source, target = module_source_target(source, target)
source = State(source)
target = State(target)
if self.has_transition(source, target, Speaker.SYSTEM):
raise ValueError('system transition {} -> {} already exists'.format(source, target))
natex_nlg = NatexNLG(natex_nlg, macros=self._macros)
if not self.has_state(source):
self.add_state(source)
if not self.has_state(target):
self.add_state(target)
self._graph.add_arc(source, target, Speaker.SYSTEM)
self.set_transition_natex(source, target, Speaker.SYSTEM, natex_nlg)
transition_settings = Settings(score=1.0)
transition_settings.update(**settings)
self.set_transition_settings(source, target, Speaker.SYSTEM, transition_settings)
if self._all_multi_hop:
self.update_state_settings(source, system_multi_hop=True)
if target in self._prepends:
prepend = self._prepends[target]
natex = self.transition_natex(source, target, Speaker.SYSTEM)
self.set_transition_natex(source, target, Speaker.SYSTEM, prepend + natex)
def add_state(self, state: Union[Enum, str, tuple], error_successor: Union[Union[Enum, str, tuple], None] =None, **settings):
state = module_state(state)
state = State(state)
if self.has_state(state):
raise ValueError('state {} already exists'.format(state))
state_settings = Settings(user_multi_hop=False, system_multi_hop=False, switch=False, enter=None)
state_settings.update(**settings)
self._graph.add_node(state)
self.update_state_settings(state, **state_settings)
if error_successor is not None:
error_successor = State(error_successor)
self.set_error_successor(state, error_successor)
# LOW LEVEL: PROPERTIES, GETTERS, SETTERS
def transition_natex(self, source: Union[Enum, str, tuple], target: Union[Enum, str, tuple], speaker: Enum):
source, target = module_source_target(source, target)
source = State(source)
target = State(target)
return self._graph.arc_data(source, target, speaker)['natex']
def set_transition_natex(self, source, target, speaker, natex):
source, target = module_source_target(source, target)
source = State(source)
target = State(target)
if isinstance(natex, str):
if speaker == Speaker.USER:
natex = NatexNLU(natex, macros=self._macros)
else:
natex = NatexNLG(natex, macros=self._macros)
self._graph.arc_data(source, target, speaker)['natex'] = natex
def transition_settings(self, source: Union[Enum, str, tuple], target: Union[Enum, str, tuple], speaker: Enum):
source, target = module_source_target(source, target)
source = State(source)
target = State(target)
return self._graph.arc_data(source, target, speaker)['settings']
def set_transition_settings(self, source, target, speaker, settings):
source, target = module_source_target(source, target)
source = State(source)
target = State(target)
self._graph.arc_data(source, target, speaker)['settings'] = settings
def update_transition_settings(self, source, target, speaker, **settings):
source, target = module_source_target(source, target)
source = State(source)
target = State(target)
self.transition_settings(source, target, speaker).update(**settings)
def state_settings(self, state):
state = module_state(state)
state = State(state)
return self._graph.data(state)['settings']
def add_global_nlu(self, state, nlu, score=0.5, post_nlu=None):
state = module_state(state)
state = State(state)
if not self.has_state(state):
self.add_state(state)
if isinstance(state, tuple):
state = ':'.join(state)
if isinstance(nlu, list) or isinstance(nlu, set):
nlu = '{' + ', '.join(nlu) + '}'
if post_nlu is None:
self._rules.add('{} ({})'.format(nlu, score), '#TRANSITION({}, {})'.format(state, score))
else:
self._rules.add('{} ({})'.format(nlu, score), '#TRANSITION({}, {}, {})'.format(state, score, post_nlu))
def update_state_settings(self, state, **settings):
state = module_state(state)
state = State(state)
if 'settings' not in self._graph.data(state):
self._graph.data(state)['settings'] = Settings()
if 'global_nlu' in settings:
self.add_global_nlu(state, settings['global_nlu'])
if 'enter' in settings and isinstance(settings['enter'], str):
settings['enter'] = NatexNLG(settings['enter'], macros=self._macros)
self.state_settings(state).update(**settings)
def remove_transition(self, source, target, speaker):
source, target = module_source_target(source, target)
source = State(source)
target = State(target)
MapMultidigraph.remove_arc(self.graph(), source, target, speaker)
def states(self):
return self.graph().nodes()
def state(self):
return self._vars['__state__']
def set_state(self, state: Union[Enum, str, tuple]):
state = module_state(state)
state = State(state)
if self.speaker() == Speaker.SYSTEM:
if '__state__' in self.vars():
st_str = self.vars()['__state__'][1] if isinstance(self.vars()['__state__'],tuple) else self.vars()['__state__']
if not st_str.startswith('_'):
self.vars()['__system_state__'] = self.vars()['__state__']
if '__system_state__' not in self.vars():
self.vars()['__system_state__'] = 'None'
else:
self.vars()['__system_state__'] = 'None'
self._vars['__state__'] = state
def has_state(self, state):
state = module_state(state)
state = State(state)
return self._graph.has_node(state)
def error_successor(self, state):
state = module_state(state)
state = State(state)
data = self._graph.data(state)
if 'error' in data:
return data['error']
else:
return None
def set_error_successor(self, state, error_successor):
state, error_successor = module_source_target(state, error_successor)
state = State(state)
error_successor = State(error_successor)
self._graph.data(state)['error'] = error_successor
def speaker(self):
return self._speaker
def set_speaker(self, speaker: Enum):
self._speaker = speaker
def graph(self):
return self._graph
def vars(self):
return self._vars
def set_vars(self, vars):
self._vars = vars
self.update_rules().set_vars(vars)
def transitions(self, source_state, speaker=None):
"""
get (source, target, speaker) transition tuples for the entire state machine
(default) or that lead out from a given source_state
:param source_state: optionally, filter returned transitions by source state
:param speaker: optionally, filter returned transitions by speaker
:return: a generator over (source, target, speaker) 3-tuples
"""
source_state = module_state(source_state)
source_state = State(source_state)
if speaker is None:
yield from self._graph.arcs_out(source_state)
elif self._graph.has_arc_label(source_state, speaker):
yield from self._graph.arcs_out(source_state, label=speaker)
else:
return
def has_transition(self, source, target, speaker):
source, target = module_source_target(source, target)
source = State(source)
target = State(target)
return self._graph.has_arc(source, target, speaker)
def incoming_transitions(self, target_state):
target_state = module_state(target_state)
target_state = State(target_state)
yield from self._graph.arcs_in(target_state)
def change_speaker(self):
if self.speaker() is Speaker.USER:
self.set_speaker(Speaker.SYSTEM)
elif self.Speaker is Speaker.SYSTEM:
self.set_speaker(Speaker.USER)
def reset(self):
self._transitions.clear()
self._speaker = self._initial_speaker
self._vars = HashableDict()
self.vars()['__state__'] = self._initial_state
self.vars()['__stack__'] = []
self.vars()['__user_utterance__'] = None
self.vars()['__system_state__'] = 'None' if self._initial_speaker == Speaker.USER else self._initial_state
self.set_state(self._initial_state)
self._rules.set_vars(self._vars)
self._gates = defaultdict(list)
def update_vars(self, variables: HashableDict):
if not isinstance(variables, HashableDict):
variables = HashableDict(variables)
for k in variables:
if k in self._var_dependencies:
dependencies = self._var_dependencies[k]
for dependency in dependencies:
if dependency in self._vars:
self._vars[dependency] = None
self._vars.update({k: variables[k] for k in variables if k != '__score__' and k in variables})
def potential_transition(self):
return self._potential_transition
def gates(self):
return self._gates
def var_dependencies(self):
return self._var_dependencies
def set_state_prepend(self, state, prepend):
state = module_state(state)
self._prepends[state] = prepend
if self.has_state(state):
for transition in self._graph.arcs_in(state):
natex = self.transition_natex(*transition)
self.set_transition_natex(*transition, prepend + natex)
def add_update_rule(self, precondition, postcondition=None):
self._rules.add(precondition, postcondition)
def apply_update_rules(self, user_input, debugging=False):
result = self._rules.update(user_input, debugging)
if result is not None:
response, score = result
self._update_transitions.append(
(response, (self.state(), self.state(), Speaker.SYSTEM), score))
#self.set_speaker(Speaker.SYSTEM)
def knowledge_base(self):
return self._kb
def set_is_module(self, composite_dialogue_flow):
self._composite_dialogue_flow = composite_dialogue_flow
def is_switch(self, state):
return self.state_settings(state)['switch']
def end_state(self):
return self._end_state
def update_rules(self):
return self._rules
def goals(self):
return self._goals
def set_goals(self, goals_dict):
self._goals = goals_dict
def dynamic_transitions(self):
return self._transitions
def composite_dialogue_flow(self):
return self._composite_dialogue_flow
def is_module(self):
return self.composite_dialogue_flow() is not None
def namespace(self):
return self._namespace
def add_macros(self, macro_dict):
self._macros.update(macro_dict)
def set_namespace(self, namespace):
self._namespace = namespace
def set_gates(self, gates):
self._gates = gates
def load_global_nlu(self, transitions, default_score=0.5):
orig_score=default_score
for nlu, followup in transitions.items():
default_score=orig_score
if nlu == 'state':
continue
if isinstance(followup, str):
state = followup
else:
if 'state' not in followup:
state = DialogueFlow.autostate()
followup['state'] = state
else:
state = followup['state']
if 'score' in followup:
default_score = followup['score']
self.add_global_nlu(state, nlu, default_score, post_nlu='`/.*/ #GEXT`')
self.load_transitions(transitions, Speaker.USER)
def load_update_rules(self, rules_dict, score=None):
for pre, post in rules_dict.items():
self.update_rules().add(pre, post, score=score)
def macros(self):
return self._macros
def add_goal(self, id_string, return_state=None, return_phrase=None, doom_counter=None):
goal = {
'id': id_string,
'return_state': return_state,
'return_phrase': return_phrase,
'doom_counter': doom_counter
}
self._goals[id_string] = goal
def serialize(self):
"""
Returns json serialized dict of
{'vars': vars, 'gates': gates, 'state': state}
"""
config = {'vars': self.vars(),
'gates': self.gates(),
'state': self.state()}
return json_serialize_flexible(config, speaker_enum_mapping)
def deserialize(self, config_str):
config = json_deserialize_flexible(config_str, speaker_enum_rmapping)
self.reset()
self.set_state(config['state'])
self.set_vars(config['vars'])
self.set_gates(config['gates'])
import pytest
from emora_stdm import NatexNLU, NatexNLG
from emora_stdm.state_transition_dialogue_manager.macros_common import *
from emora_stdm.state_transition_dialogue_manager.natex_common import *
from emora_stdm.state_transition_dialogue_manager.knowledge_base import KnowledgeBase
from emora_stdm.state_transition_dialogue_manager.dialogue_flow import DialogueFlow
from emora_stdm.state_transition_dialogue_manager.composite_dialogue_flow import CompositeDialogueFlow
from time import time
kb = KnowledgeBase([
('lion', 'type', 'cat'), ('panther', 'type', 'cat'),
('cat', 'type', 'animal'), ('animal', 'type', 'thing'),
('lion', 'sound', 'roar'), ('roar', 'quality', 'loud'),
('panther', 'sound', 'growl'), ('growl', 'quality', 'scary'),
('also', 'type', 'also_syns'), ('too', 'type', 'also_syns'),
('basketball', 'type', 'unknown_hobby'), ('basketball', 'expr', 'bball'),
('john', 'type', 'male'), ('mary', 'type', 'female'),
('male', 'type', 'person'), ('female', 'type', 'person'),
('friend', 'type', 'person')
])
ont = {"ontology": {"person": ["mom", "dad", "sister", "brother"]}}
kb.load_json(ont)
df = DialogueFlow(initial_state='start')
df.add_state('start', 'start')
df.add_system_transition('start', 'start', 'hello')
macros = {
'ONT': ONTE(kb),
'KBQ': KBE(kb),
for lemma in lemmas:
expr_arcs.add((hyponym.name(), lemma))
else:
type_arcs.add((hyponym.name(), synset.name()))
for lemma in hyponym.lemmas():
expr_arcs.add(
(hyponym.name(), lemma.name().replace('_', ' ')))
i += 1
if i % 1000 == 0:
print(i, 'processed')
print('done')
print(len(type_arcs))
kb = KnowledgeBase(arcs={(x, 'type', y)
for x, y in type_arcs}
| {(x, 'expr', y)
for x, y in expr_arcs})
output = open('data/wordnet.json', 'w')
output.write(kb.to_json())
'''
while True:
i = input('>> ')
try:
synsets = wn.synsets(i)
result = set()
for synset in synsets:
result.update(kb.expressions(kb.subtypes(synset.name())))
print(result)
except:
print('nope')
def test_integration_opening_component():
kb = KnowledgeBase()
kb.load_json_file("opening_database.json")
receive_how_are_you = ("{"
"[how are you],"
"[how you doing],"
"[what about you],"
"[whats up with you],"
"[how you are],"
"[how about you]"
"}")
feelings_pos_and_not_received_how_are_you = (
"{"
"[!#ONT_NEG(ont_negation), -%s, [#ONT(ont_feelings_positive)]],"
"[! -%s, [#ONT(ont_negation)], [#ONT(ont_feelings_negative)]],"
"#IsPositiveSentiment"
"}" % (receive_how_are_you, receive_how_are_you))
nlu = NatexNLU(
feelings_pos_and_not_received_how_are_you,
macros={
"ONT": ONTE(kb),
"ONT_NEG": ONT_NEG(kb)
},
)
m = nlu.match("im not too bad", debugging=False)
assert m
m = nlu.match("great i guess", debugging=False)
assert m
m = nlu.match("great", debugging=False)
assert m
m = nlu.match("i seem to be pretty great thanks", debugging=False)
assert m
m = nlu.match("not bad", debugging=False)
assert m
m = nlu.match("thanks im not bad dude", debugging=False)
assert m
m = nlu.match("well how are you", debugging=False)
assert not m
m = nlu.match("pretty well how are you", debugging=False)
assert not m
m = nlu.match("not well how are you", debugging=False)
assert not m
m = nlu.match("how are you im not well", debugging=False)
assert not m
m = nlu.match("im bad", debugging=False)
assert not m
m = nlu.match("bad", debugging=False)
assert not m
m = nlu.match("im doing ok", debugging=False)
assert not m
feelings_neg_and_not_received_how_are_you = (
"{"
"[!#ONT_NEG(ont_negation), -%s, [#ONT(ont_feelings_negative)]],"
"[! -%s, [#ONT(ont_negation)], [{#ONT(ont_feelings_positive),#ONT(ont_feelings_neutral)}]],"
"#IsNegativeSentiment"
"}" % (receive_how_are_you, receive_how_are_you))
nlu = NatexNLU(
feelings_neg_and_not_received_how_are_you,
macros={
"ONT": ONTE(kb),
"ONT_NEG": ONT_NEG(kb)
},
)
m = nlu.match("bad i guess", debugging=False)
assert m
m = nlu.match("bad", debugging=False)
assert m
m = nlu.match("i seem to be pretty bad thanks", debugging=False)
assert m
m = nlu.match("not good", debugging=False)
assert m
m = nlu.match("thanks im not ok dude", debugging=False)
assert m
m = nlu.match("great i guess", debugging=False)
assert not m
m = nlu.match("great", debugging=False)
assert not m
m = nlu.match("i seem to be pretty great thanks", debugging=False)
assert not m
m = nlu.match("not bad", debugging=False)
assert not m
m = nlu.match("thanks im not bad dude", debugging=False)
assert not m
m = nlu.match("ok how are you", debugging=False)
assert not m
m = nlu.match("pretty well how are you", debugging=False)
assert not m
m = nlu.match("not well how are you", debugging=False)
assert not m
m = nlu.match("how are you im not well", debugging=False)
assert not m
feelings_neutral_and_not_received_how_are_you = (
"[!#ONT_NEG(ont_negation), -%s, [#ONT(ont_feelings_neutral)]]" %
(receive_how_are_you))
nlu = NatexNLU(
feelings_neutral_and_not_received_how_are_you,
macros={
"ONT": ONTE(kb),
"ONT_NEG": ONT_NEG(kb)
},
)
m = nlu.match("ok i guess", debugging=False)
assert m
m = nlu.match("ok", debugging=False)
assert m
m = nlu.match("i seem to be ok thanks", debugging=False)
assert m
m = nlu.match("great i guess", debugging=False)
assert not m
m = nlu.match("great", debugging=False)
assert not m
m = nlu.match("i seem to be pretty great thanks", debugging=False)
assert not m
m = nlu.match("not bad", debugging=False)
assert not m
m = nlu.match("thanks im not bad dude", debugging=False)
assert not m
m = nlu.match("ok how are you", debugging=False)
assert not m
m = nlu.match("pretty well how are you", debugging=False)
assert not m
m = nlu.match("not well how are you", debugging=False)
assert not m
m = nlu.match("how are you im not well", debugging=False)
assert not m
m = nlu.match("bad i guess", debugging=False)
assert not m
m = nlu.match("bad", debugging=False)
assert not m
m = nlu.match("i seem to be pretty bad thanks", debugging=False)
assert not m
m = nlu.match("not good", debugging=False)
assert not m
m = nlu.match("thanks im not ok dude", debugging=False)
assert not m
feelings_pos_and_received_how_are_you = (
"{"
"[!#ONT_NEG(ont_negation), [#ONT(ont_feelings_positive)], [%s]],"
"[#ONT(ont_negation), #ONT(ont_feelings_negative), %s],"
"<#IsPositiveSentiment, %s>"
"}" % (receive_how_are_you, receive_how_are_you, receive_how_are_you))
nlu = NatexNLU(
feelings_pos_and_received_how_are_you,
macros={
"ONT": ONTE(kb),
"ONT_NEG": ONT_NEG(kb)
},
)
m = nlu.match("pretty well how are you", debugging=False)
assert m
m = nlu.match("great how are you", debugging=False)
assert m
m = nlu.match("not too bad how are you", debugging=False)
assert m
m = nlu.match("not bad how are you", debugging=False)
assert m
m = nlu.match("ok i guess", debugging=False)
assert not m
m = nlu.match("ok", debugging=False)
assert not m
m = nlu.match("i seem to be ok thanks", debugging=False)
assert not m
m = nlu.match("great i guess", debugging=False)
assert not m
m = nlu.match("great", debugging=False)
assert not m
m = nlu.match("i seem to be pretty great thanks", debugging=False)
assert not m
m = nlu.match("not bad", debugging=False)
assert not m
m = nlu.match("thanks im not bad dude", debugging=False)
assert not m
m = nlu.match("ok how are you", debugging=False)
assert not m
m = nlu.match("not well how are you", debugging=False)
assert not m
m = nlu.match("how are you im not well", debugging=False)
assert not m
m = nlu.match("bad i guess", debugging=False)
assert not m
m = nlu.match("bad", debugging=False)
assert not m
m = nlu.match("i seem to be pretty bad thanks", debugging=False)
assert not m
m = nlu.match("not good", debugging=False)
assert not m
m = nlu.match("thanks im not ok dude", debugging=False)
assert not m
feelings_neg_and_received_how_are_you = (
"{"
"[!#ONT_NEG(ont_negation), [#ONT(ont_feelings_negative)], [%s]],"
"[#ONT(ont_negation), {#ONT(ont_feelings_positive),#ONT(ont_feelings_neutral)}, %s],"
"<#IsNegativeSentiment, %s>"
"}" % (receive_how_are_you, receive_how_are_you, receive_how_are_you))
nlu = NatexNLU(
feelings_neg_and_received_how_are_you,
macros={
"ONT": ONTE(kb),
"ONT_NEG": ONT_NEG(kb)
},
)
m = nlu.match("not well how are you", debugging=False)
assert m
m = nlu.match("bad how are you", debugging=False)
assert m
m = nlu.match("im bad how are you", debugging=False)
assert m
m = nlu.match("ok i guess", debugging=False)
assert not m
m = nlu.match("ok", debugging=False)
assert not m
m = nlu.match("i seem to be ok thanks", debugging=False)
assert not m
m = nlu.match("great i guess", debugging=False)
assert not m
m = nlu.match("great", debugging=False)
assert not m
m = nlu.match("i seem to be pretty great thanks", debugging=False)
assert not m
m = nlu.match("not bad", debugging=False)
assert not m
m = nlu.match("thanks im not bad dude", debugging=False)
assert not m
m = nlu.match("ok how are you", debugging=False)
assert not m
m = nlu.match("bad i guess", debugging=False)
assert not m
m = nlu.match("bad", debugging=False)
assert not m
m = nlu.match("i seem to be pretty bad thanks", debugging=False)
assert not m
m = nlu.match("not good", debugging=False)
assert not m
m = nlu.match("thanks im not ok dude", debugging=False)
assert not m
m = nlu.match("pretty well how are you", debugging=False)
assert not m
m = nlu.match("great how are you", debugging=False)
assert not m
feelings_neutral_and_received_how_are_you = (
"[!#ONT_NEG(ont_negation), [#ONT(ont_feelings_neutral)], [%s]]" %
(receive_how_are_you))
nlu = NatexNLU(
feelings_neutral_and_received_how_are_you,
macros={
"ONT": ONTE(kb),
"ONT_NEG": ONT_NEG(kb)
},
)
m = nlu.match("ok but how are you", debugging=False)
assert m
m = nlu.match("not ok how are you", debugging=False)
assert not m
m = nlu.match("im not ok how are you", debugging=False)
assert not m
m = nlu.match("not well how are you", debugging=False)
assert not m
m = nlu.match("bad how are you", debugging=False)
assert not m
m = nlu.match("im bad how are you", debugging=False)
assert not m
m = nlu.match("ok i guess", debugging=False)
assert not m
m = nlu.match("ok", debugging=False)
assert not m
m = nlu.match("i seem to be ok thanks", debugging=False)
assert not m
m = nlu.match("great i guess", debugging=False)
assert not m
m = nlu.match("great", debugging=False)
assert not m
m = nlu.match("i seem to be pretty great thanks", debugging=False)
assert not m
m = nlu.match("not bad", debugging=False)
assert not m
m = nlu.match("thanks im not bad dude", debugging=False)
assert not m
m = nlu.match("bad i guess", debugging=False)
assert not m
m = nlu.match("bad", debugging=False)
assert not m
m = nlu.match("i seem to be pretty bad thanks", debugging=False)
assert not m
m = nlu.match("not good", debugging=False)
assert not m
m = nlu.match("thanks im not ok dude", debugging=False)
assert not m
m = nlu.match("pretty well how are you", debugging=False)
assert not m
m = nlu.match("great how are you", debugging=False)
assert not m
decline_share = (
"{"
"[#ONT(ont_negation), {talk, talking, discuss, discussing, share, sharing, tell, telling, say, saying}],"
"[#ONT(ont_fillers), #ONT(ont_negative)],"
"[#ONT(ont_negative)]"
"<{dont,do not}, know>,"
"<not, sure>"
"}")
nlu = NatexNLU(decline_share,
macros={
"ONT": ONTE(kb),
"ONT_NEG": ONT_NEG(kb)
})
m = nlu.match("i dont know", debugging=False)
assert m
m = nlu.match("im not sure", debugging=False)
assert m
m = nlu.match("i dont want to tell you", debugging=False)
assert m
m = nlu.match("no", debugging=False)
assert m
m = nlu.match("i dont want to talk about it", debugging=False)
assert m
m = nlu.match("ok but how are you", debugging=False)
assert not m
m = nlu.match("not ok how are you", debugging=False)
assert not m
m = nlu.match("im not ok how are you", debugging=False)
assert not m
m = nlu.match("not well how are you", debugging=False)
assert not m
m = nlu.match("bad how are you", debugging=False)
assert not m
m = nlu.match("im bad how are you", debugging=False)
assert not m
m = nlu.match("ok i guess", debugging=False)
assert not m
m = nlu.match("ok", debugging=False)
assert not m
m = nlu.match("i seem to be ok thanks", debugging=False)
assert not m
m = nlu.match("great i guess", debugging=False)
assert not m
m = nlu.match("great", debugging=False)
assert not m
m = nlu.match("i seem to be pretty great thanks", debugging=False)
assert not m
m = nlu.match("not bad", debugging=False)
assert not m
m = nlu.match("thanks im not bad dude", debugging=False)
assert not m
m = nlu.match("bad i guess", debugging=False)
assert not m
m = nlu.match("bad", debugging=False)
assert not m
m = nlu.match("i seem to be pretty bad thanks", debugging=False)
assert not m
m = nlu.match("not good", debugging=False)
assert not m
m = nlu.match("thanks im not ok dude", debugging=False)
assert not m
m = nlu.match("pretty well how are you", debugging=False)
assert not m
m = nlu.match("great how are you", debugging=False)
assert not m
