def __init__(self):
super(EpisodicMemoryNode, self).__init__("episodic_memory")
rospy.Subscriber('/mhri/person_presence_state', PersonPresenceState,
self.pps_callback)
rospy.Subscriber('/mhri/person_identity_state', PersonIdentity,
self.pi_callback)
rospy.Subscriber('/mhri/mset', MSetInfo, self.mset_callback)
self.graph = KnowledgeGraph()
#self.pub = rospy.Publisher("/mhri/events", PerceptionEvent, queue_size=1)
self.persons = {}
self.episodes = {}
self.load_episodic_memory()
rospack = rospkg.RosPack()
perception_face_pkg_path = rospack.get_path('perception_face')
self.face_registry_path = perception_face_pkg_path + '/data/face_registry'
self.names_file_path = perception_face_pkg_path + '/data/name_info.json'
self.load_people_names(self.names_file_path)
self.face_memorized_pub = rospy.Publisher('/mhri/face_memorized',
String,
queue_size=1)
rospy.loginfo("episodic_memory_node initialized.")
def main():
parser = argparse.ArgumentParser()
args = parser.parse_args()
args.dataset = config["dataset"]
# Create AmazonDataset instance for dataset.
# ========== BEGIN ========== #
print("Load", args.dataset, "dataset from file...")
if not os.path.isdir(TMP_DIR[args.dataset]):
os.makedirs(TMP_DIR[args.dataset])
dataset = AmazonDataset(DATASET_DIR[args.dataset])
save_dataset(args.dataset, dataset)
# Generate knowledge graph instance.
# ========== BEGIN ========== #
print("Create", args.dataset, "knowledge graph from dataset...")
dataset = load_dataset(args.dataset)
kg = KnowledgeGraph(dataset)
kg.compute_degrees()
save_kg(args.dataset, kg)
# =========== END =========== #
# Genereate train/test labels.
# ========== BEGIN ========== #
print("Generate", args.dataset, "train/test labels.")
generate_labels(args.dataset, "train")
generate_labels(args.dataset, "test")
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--dataset',
type=str,
default=BEAUTY,
help='One of {BEAUTY, CELL, CD, CLOTH}.')
args = parser.parse_args()
# Create AmazonDataset instance for dataset.
# ========== BEGIN ========== #
print('Load', args.dataset, 'dataset from file...')
if not os.path.isdir(
TMP_DIR[args.dataset]
): #if the required temp file doesn't exist, create it
os.makedirs(TMP_DIR[args.dataset]) #TMP_DIR is a dict from utils.py
dataset = AmazonDataset(DATASET_DIR[args.dataset])
save_dataset(args.dataset,
dataset) #pickle.dump dataset to file, no return value
# Generate knowledge graph instance.
# ========== BEGIN ========== #
print('Create', args.dataset, 'knowledge graph from dataset...')
dataset = load_dataset(args.dataset)
kg = KnowledgeGraph(dataset)
kg.compute_degrees()
save_kg(args.dataset, kg) #uses pickle.dump
# =========== END =========== #
# Genereate train/test labels.
# ========== BEGIN ========== #
print('Generate', args.dataset, 'train/test labels.')
generate_labels(args.dataset, 'train')
generate_labels(args.dataset, 'test')
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--dataset',
type=str,
default=BEAUTY,
help='One of {BEAUTY, CELL, CD, CLOTH}.')
args = parser.parse_args()
# Create AmazonDataset instance for dataset.
# ========== BEGIN ========== #
print('Load', args.dataset, 'dataset from file...')
if not os.path.isdir(TMP_DIR[args.dataset]):
os.makedirs(TMP_DIR[args.dataset])
dataset = AmazonDataset(DATASET_DIR[args.dataset])
save_dataset(args.dataset, dataset)
# Generate knowledge graph instance.
# ========== BEGIN ========== #
print('Create', args.dataset, 'knowledge graph from dataset...')
dataset = load_dataset(args.dataset)
kg = KnowledgeGraph(dataset)
kg.compute_degrees()
save_kg(args.dataset, kg)
# =========== END =========== #
# Generate train/test labels.
# ========== BEGIN ========== #
print('Generate', args.dataset, 'train/test labels.')
generate_labels(args.dataset, 'train')
generate_labels(args.dataset, 'test')
def load_data(dataset_name, word_emb_size, node_embed_type):
graph_path = "data/" + dataset_name + "/" + dataset_name + "_kg.txt"
train_path = "data/" + dataset_name + "/" + dataset_name + "_train.txt"
test_path = "data/" + dataset_name + "/" + dataset_name + "_test.txt"
vocab_path = "data/" + dataset_name + "/vocab/"
node_embed_path = "data/" + dataset_name + "/" + dataset_name + "_embed_" + node_embed_type + ".npy"
graph = KnowledgeGraph(graph_path, vocab_path)
test = load_questions(train_path, parser)
train = load_questions(test_path, parser)
rel_embedding = init_rel_embedding("glove.840B.300d.txt",
camel_case_spliter, word_emb_size,
graph)
node_embedding = np.load(node_embed_path)
return node_embedding, rel_embedding, graph, test, train
def __init__(self, path_KG, path_QA, split_ratio=0.8, using_cache=True):
self.KG = KnowledgeGraph(path_KG)
self.embedder = Embedder()
self.training = True # 指定是否是训练阶段
self._iter_i = 0
self._split_ratio = split_ratio
# try to load from cache
if using_cache and Utility.Binary.exists('dataset'):
self.questions = Utility.Binary.load('dataset')
print('{} questions loaded'.format(len(self.questions)))
return
# read the original questions
questions = pd.read_csv(
path_QA,
sep='\t',
header=None,
names=['question_sentence', 'answer_set', 'answer_path'])
questions['answer'] = questions['answer_set'].apply(
lambda x: x.split('(')[0])
questions['q_split'] = questions['question_sentence'].apply(
lambda x: x.lower().split(' '))
questions['answer'] = questions['answer_set'].apply(
lambda x: x.split('(')[0])
questions['e_s'] = questions['answer_path'].apply(
lambda x: x.split('#')[0])
# find head entity e_s, answer, and question_list by parsing the question_sentence
questions['q_str'] = [
self.parse_question(row['question_sentence'].split('?')[0],
row['e_s'])
for idx, row in questions.iterrows()
]
# 对问题编码
# NOTE: 这里是正对小数据集采取的空间换时间的方式，避免每一次都重新embed问题，对于大数据集需要单独处理数据
questions['q'] = questions['q_str'].apply(
lambda q: self.embed_question(q))
question_list = questions[['q_str', 'q', 'e_s',
'answer']].values.tolist()
question_list = [tuple(x) for x in question_list]
self.questions = question_list
print('{} questions loaded'.format(len(question_list)))
if using_cache:
Utility.Binary.save('dataset', question_list)
def test_add_relationship():
kg = KnowledgeGraph()
kg.add_relationship('DMG mori CMX', 'Heidenhain TNC 640', 'related_to')
def test_add_node_by_name():
kg = KnowledgeGraph()
kg.add_node_by_name('DMG mori CMX')
def test_search_node_by_name():
kg = KnowledgeGraph()
kg.search_node_by_name('DMG mori CMX')
class EpisodicMemoryNode(PerceptionBase):
'''
Episodic Memory Node
'''
def __init__(self):
super(EpisodicMemoryNode, self).__init__("episodic_memory")
rospy.Subscriber('/mhri/person_presence_state', PersonPresenceState,
self.pps_callback)
rospy.Subscriber('/mhri/person_identity_state', PersonIdentity,
self.pi_callback)
rospy.Subscriber('/mhri/mset', MSetInfo, self.mset_callback)
self.graph = KnowledgeGraph()
#self.pub = rospy.Publisher("/mhri/events", PerceptionEvent, queue_size=1)
self.persons = {}
self.episodes = {}
self.load_episodic_memory()
rospack = rospkg.RosPack()
perception_face_pkg_path = rospack.get_path('perception_face')
self.face_registry_path = perception_face_pkg_path + '/data/face_registry'
self.names_file_path = perception_face_pkg_path + '/data/name_info.json'
self.load_people_names(self.names_file_path)
self.face_memorized_pub = rospy.Publisher('/mhri/face_memorized',
String,
queue_size=1)
rospy.loginfo("episodic_memory_node initialized.")
def load_people_names(self, names_file_path):
rospy.loginfo('episodic_memory - Names File: %s', names_file_path)
self.names = json.load(open(names_file_path))
def save_person_name(self, person_id, name):
if person_id not in self.names:
self.names[person_id] = name
with open(self.names_file_path, 'w') as names_file:
json.dump(self.names, names_file)
def load_episodic_memory(self):
'''
Load all the memory in the episodic memory.
'''
query = 'match (x:Episode) return x.person_id, x.name, x.confidence, x.eyeglasses,\
x.cloth_color, x.hair_length, x.start_time, x.end_time, x.gender'
categories = [
'name', 'confidence', 'eyeglasses', 'cloth_color', 'hair_length',
'start_time', 'end_time', 'gender'
]
result = self.graph.Run(query)
records = []
for record in result:
records.append(record)
rospy.loginfo("EPISODIC_MEMORY: Total No of episodes = %d",
len(records))
persons = {}
for record in records:
person_id = record['x.person_id']
if person_id not in persons:
persons[person_id] = {}
for category in categories:
if category not in persons[person_id]:
persons[person_id][category] = {}
value = record['x.' + category]
if value not in persons[person_id][category]:
persons[person_id][category][value] = 0
persons[person_id][category][value] += 1
for person_id in persons:
for category in categories:
max_num = 0
max_val = ''
values = persons[person_id][category]
for key in values:
rospy.loginfo('EPISODES(%s): %s is %s = %d', person_id,
category, key, values[key])
if key == '':
continue
if max_val == '' or values[key] >= max_num:
max_num = values[key]
max_val = key
persons[person_id][category] = max_val
rospy.loginfo('EPISODES(%s): Decision for %s is %s', person_id,
category, max_val)
if category == 'name':
rospy.set_param('/{0}/{1}'.format(str(person_id),
str(category)),
max_val) # 170118
rospy.loginfo('/%s/%s = %s', str(person_id), str(category),
max_val)
self.episodes = persons
rospy.loginfo("EPISODIC_MEMORY: Memory of %d people recalled.",
len(self.episodes.keys()))
def person_appeared(self, trk_id):
'''
새로운 사람의 등장!
'''
if trk_id not in self.persons:
self.persons[trk_id] = {}
self.persons[trk_id]['start_time'] = time.time()
def person_disappeared(self, trk_id):
'''
사람 사라짐!
'''
if trk_id in self.persons:
self.persons[trk_id]['end_time'] = time.time()
self.memorize_episode(trk_id)
else:
rospy.logerr("I DO NOT KNOW A PERSON WITH AN ID=%s", trk_id)
def memorize_episode(self, trk_id):
'''
Store an episode with a person(trk_id).
'''
if trk_id not in self.persons:
rospy.logerr("I DO NOT KNOW A PERSON WITH AN ID=%s", trk_id)
return
episode = self.persons[trk_id]
# if the person has not been identified, no memory lives on.
# if an episode ended in 10 seconds, no memory lives on.
duration = episode['end_time'] - episode['start_time']
if not episode.has_key('person_id') or duration < 10:
rospy.loginfo('EPISODIC_MEMORY: No Memory with %s', trk_id)
return
rospy.loginfo("MEMORIZING EPISODE (%s):======\n%s\n=======", trk_id,
episode)
self.graph.Create('Episode', 'human_id', trk_id)
self.graph.Set('Episode', 'human_id', trk_id, 'person_id',
episode['person_id'])
self.graph.Set('Episode', 'human_id', trk_id, 'confidence',
episode.get('confidence', 0))
self.graph.Set('Episode', 'human_id', trk_id, 'name',
episode.get('name', ''))
self.graph.Set('Episode', 'human_id', trk_id, 'gender',
episode.get('gender', 0))
self.graph.Set('Episode', 'human_id', trk_id, 'eyeglasses',
episode.get('eyeglasses', 0))
self.graph.Set('Episode', 'human_id', trk_id, 'cloth_color',
episode.get('cloth_color', ''))
self.graph.Set('Episode', 'human_id', trk_id, 'hair_length',
episode.get('hair_length', ''))
self.graph.Set('Episode', 'human_id', trk_id, 'start_time',
episode.get('start_time', 0))
self.graph.Set('Episode', 'human_id', trk_id, 'end_time',
episode.get('end_time', 0))
# 초면인 경우 얼굴을 기억한다.
if trk_id == episode['person_id']:
self.memorize_face(trk_id)
self.face_memorized_pub.publish("A")
rospy.set_param('/{0}/{1}'.format(trk_id, 'name'),
episode.get('name', '')) # 170214
def memorize_face(self, trk_id):
'''
얼굴을 기억한다.
'''
files = []
home_dir = expanduser("~") + '/.ros'
img_file_names = list_files(home_dir, '.jpg')
for file_name in img_file_names:
pid = file_name.split('/')[-1].split('_')[-1].split('.')[0]
if pid == trk_id:
files.append(file_name)
rospy.loginfo(">>>>FILE NO: %s = %d", trk_id, len(files))
if len(files) == 0:
return
elif len(files) < 3:
count = len(files)
else:
count = 3
for num in range(count):
i = randint(0, len(files) - 1)
file_name = files[i]
rospy.loginfo('>>>>FILE: %s', file_name)
copy(file_name, self.face_registry_path)
files.remove(file_name)
def identity_recognized(self, trk_id, person_id):
'''
얼굴이 인식되었을 때 호출되는 콜백
'''
if trk_id not in self.persons:
rospy.logerr("I DO NOT KNOW A PERSON WITH AN ID=%s", trk_id)
else:
if person_id == 'unknown':
person_id = trk_id
rospy.loginfo('>>> EPISODIC_MEMORY_NODE: person_id=%s', person_id)
self.persons[trk_id]["person_id"] = person_id
# 기억을 기반으로 소셜 이벤트 생성
if person_id in self.episodes:
self.write_memory_recall_event()
def write_memory_recall_event(self):
'''
Generate a social event by recalling episodic memory.
'''
wr_data = self.get_memory_data()
rospy.loginfo("EPISODIC MEMORY_RECALLED: %s", wr_data)
self.save_to_memory(self.conf_data.keys()[0], data=wr_data)
self.raise_event(self.conf_data.keys()[0], 'memory_recalled')
def get_memory_data(self):
'''
Convert retrieved memory into a JSON document.
'''
data = json.loads('{}')
data['human_id'] = []
data['person_id'] = []
data['name'] = []
data['gender'] = []
data['eyeglasses'] = []
data['cloth_color'] = []
data['hair_length'] = []
data['confidence'] = []
for trk_id in self.persons:
if 'person_id' not in self.persons[trk_id]:
continue
person_id = self.persons[trk_id]['person_id']
if person_id in self.episodes:
memory = self.episodes[person_id]
data['human_id'].append(trk_id)
data['person_id'].append(person_id)
data['name'].append(memory['name'])
data['gender'].append(memory['gender'])
data['eyeglasses'].append(memory['eyeglasses'])
data['cloth_color'].append(memory['cloth_color'])
data['hair_length'].append(memory['hair_length'])
data['confidence'].append(memory['confidence'])
return json.dumps(data)
def pi_callback(self, pi_msg):
'''
pi_msg structure:
string human_id
string category
string value
int i_value
float f_value
'''
if pi_msg.human_id not in self.persons:
rospy.logerr("I DO NOT KNOW A PERSON WITH AN ID=%s",
pi_msg.human_id)
return
rospy.logdebug(
"EPISODIC_MEMORY: PI MSG = human_id=%s category=%s value=%s i_value=%i f_value=%f",
pi_msg.human_id, pi_msg.category, pi_msg.value, pi_msg.i_value,
pi_msg.f_value)
if pi_msg.category == 'person_id':
rospy.loginfo('EPISODIC_MEMORY_NODE: person id notified: %s',
pi_msg.value)
self.identity_recognized(pi_msg.human_id, pi_msg.value)
else:
if pi_msg.i_value != 0:
self.persons[pi_msg.human_id][pi_msg.category] = pi_msg.i_value
elif pi_msg.f_value != 0:
self.persons[pi_msg.human_id][pi_msg.category] = pi_msg.f_value
else:
self.persons[pi_msg.human_id][pi_msg.category] = pi_msg.value
'''
else:
rospy.logerr("EPISODIC_MEMORY: Invalid Value = %s %s %s",
pi_msg.value, pi_msg.i_value, pi_msg.f_value)
'''
def pps_callback(self, pps_msg):
'''
pps_msg 구조
------------
Header header
int16 count
string[] appeared
string[] disappeared
'''
for human_id in pps_msg.appeared:
self.person_appeared(human_id)
for human_id in pps_msg.disappeared:
self.person_disappeared(human_id)
def mset_callback(self, mset_msg):
'''
mset_msg 구조
-------------
string Key
string value
'''
rospy.loginfo("mset_callback called: %s %s", mset_msg.key,
mset_msg.value)
keys = mset_msg.key.split('/')
tag = keys[0]
human_id = keys[1]
# TODO: 형태소 분석기 활용에 적정 장소
if tag == 'name': # 사용자의 이름 정보
if human_id in self.persons and self.persons[human_id].has_key(
'person_id'):
rospy.loginfo("mset_callback: name is going to be updated...")
self.persons[human_id]['name'] = mset_msg.value
person_id = self.persons[human_id]['person_id']
self.save_person_name(person_id, mset_msg.value)
self.update_memory_on_name(person_id, mset_msg.value)
else:
rospy.loginfo("mset_callback: name is not updated...")
def update_memory_on_name(self, person_id, name):
self.graph.Set('Episode', 'person_id', person_id, 'name', name)
if len(tuples) > MAX_CANDIDATE_TUPLES:
tuples = set(random.sample(tuples, MAX_CANDIDATE_TUPLES))
sources = extract_dimension_from_tuples_as_list(tuples, 0)
relations = extract_dimension_from_tuples_as_list(tuples, 1)
targets = extract_dimension_from_tuples_as_list(tuples, 2)
output_row = {
'question': question,
'qn_entities': get_str_of_seq(qn_entities),
'ans_entities': get_str_of_seq(ans_entities),
'sources': get_str_of_seq(sources),
'relations': get_str_of_seq(relations),
'targets': get_str_of_seq(targets)
}
writer.writerow(output_row)
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='Specify arguments')
parser.add_argument('--input_examples', help='the raw qa pairs', required=True)
parser.add_argument('--input_graph', help='the graph file', required=True)
parser.add_argument('--input_doc', help='the doc file', required=False)
parser.add_argument('--stopwords', help='stopwords file', required=False)
parser.add_argument('--output_examples', help='the processed output file', required=True)
args = parser.parse_args()
#global variables
knowledge_base = KnowledgeGraph(args.input_graph, unidirectional=False)
search_index = SearchIndex(args.input_doc, args.stopwords)
stop_vocab = read_file_as_dict(args.stopwords)
question_parser = QuestionParser(knowledge_base.get_entities(), stop_vocab)
main(args)
textExtractor4.extract()
textExtractorPipe = TextExtractorPipe()
textExtractorPipe.addTextExtractor(textExtractor1)
textExtractorPipe.addTextExtractor(textExtractor2)
textExtractorPipe.addTextExtractor(textExtractor3)
textExtractorPipe.addTextExtractor(textExtractor4)
nlp = spacy.load('en_core_web_sm')
nlp.add_pipe(nlp.create_pipe('sentencizer')) # updated
doc = nlp(textExtractorPipe.extract())
andOtherPatternMatcher = AndOtherPatternMatcher(nlp)
suchAsMatcher = SuchAsPatternMatcher(nlp)
orOtherMatcher = OrOtherPatternMatcher(nlp)
includingPatternMatcher = IncludingPatternMatcher(nlp)
especiallyPatternMatcher = EspeciallyPatternMatcher(nlp)
matcherPipe = MatcherPipe()
matcherPipe.addMatcher(andOtherPatternMatcher)
matcherPipe.addMatcher(suchAsMatcher)
matcherPipe.addMatcher(orOtherMatcher)
matcherPipe.addMatcher(includingPatternMatcher)
matcherPipe.addMatcher(especiallyPatternMatcher)
relations = matcherPipe.extract(doc)
for relation in relations:
print(relation.getHypernym(), relation.getHyponym())
knowledgeGraph = KnowledgeGraph(relations)
knowledgeGraph.build()
knowledgeGraph.show()
def test(nlp,
src,
gen,
bert=False,
print_annotations=False,
print_latex=False,
verbose=False):
if print_annotations:
print("source:", src_line[:50])
print("summary:", gen_line[:50])
src = nlp(src)
gen = nlp(gen)
if verbose:
print("clusters:", src._.coref_clusters, gen._.coref_clusters)
ce = CompoundEquivalency()
spe = SpeakerPronounEquivalency()
spe.register(src)
spe.register(gen)
kg = KnowledgeGraph(nlp,
use_bert=bert,
equivalencies=[ce, spe],
verbose=verbose)
if print_annotations:
annotator = Annotator(src, gen, latex=print_latex)
kg.add_document(src)
contained = 0
contained_bert = 0
missing = 0
missing_verb = 0
missing_actors = 0
missing_acteds = 0
contradiction = 0
contradiction_bert = 0
invalid_simplification = 0
total = 0
for token in gen:
if token.pos_ == "VERB":
total += 1
relation = kg.get_relation(token)
r = kg.query_relation(relation)
if r[0] == KnowledgeGraph.entailment:
if print_annotations:
print(util.format("contained", "blue", latex=print_latex),
"|", relation, "|", r[1])
contained += 1
if r[0] == KnowledgeGraph.entailment_bert:
if print_annotations:
print(
util.format("contained (BERT)",
"blue",
latex=print_latex), "|", relation, "|",
r[1])
contained_bert += 1
if r[0] == KnowledgeGraph.contradiction_bert:
if print_annotations:
print(
util.format("contradiction (BERT)",
"red",
latex=print_latex), "|", relation, "|",
r[1])
contradiction_bert += 1
elif r[0] == KnowledgeGraph.missing_dependencies:
missing += 1
if print_annotations:
print(
util.format("generic missing dependency",
"yellow",
latex=print_latex), "|", relation, "|",
r[1])
elif r[0] == KnowledgeGraph.missing_actors:
missing_actors += 1
if print_annotations:
print(
util.format("missing actors",
"magenta",
latex=print_latex), "|", relation, "|",
r[1])
elif r[0] == KnowledgeGraph.missing_acteds:
missing_acteds += 1
if print_annotations:
print(
util.format("missing acteds",
"magenta",
latex=print_latex), "|", relation, "|",
r[1])
elif r[0] == KnowledgeGraph.missing_verb:
missing_verb += 1
if print_annotations:
print(
util.format("missing verb",
"magenta",
latex=print_latex), "|", relation, "|",
r[1])
elif r[0] == KnowledgeGraph.invalid_simplification:
invalid_simplification += 1
if print_annotations:
print(
util.format("invalid simplification",
"magenta",
latex=print_latex), "|", relation, "|",
r[1])
elif r[0] == KnowledgeGraph.contradiction:
contradiction += 1
if print_annotations:
print(
util.format("contradiction", "red", latex=print_latex),
"|", relation, "|", r[1])
if print_annotations:
annotator.annotate(relation, r)
if print_annotations:
annotated_document, annotated_summary = annotator.annotated()
print("Document:", " ".join(annotated_document))
print("Summary:", " ".join(annotated_summary))
if total == 0:
return 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0
return 100.0 * contained / total, \
100.0 * contained_bert / total, \
100.0 * missing / total, \
100.0 * missing_verb / total, \
100.0 * missing_actors / total, \
100.0 * missing_acteds / total, \
100.0 * contradiction / total, \
100.0 * contradiction_bert / total, \
100.0 * invalid_simplification / total
from flask import Flask, render_template, request, jsonify
from knowledge_graph import KnowledgeGraph
from google_sheets import GoogleSheets
app = Flask(__name__)
kg = KnowledgeGraph()
gsheets = GoogleSheets()
@app.route("/", methods=['GET', 'POST'])
def welcome():
data = request.get_json()
#print(data)
intent = data['queryResult']['intent']['displayName']
#message_text = data['queryResult']['queryText']
actual_value = data['queryResult']['parameters']['float_measure']
#print(f'Intent: {intent}')
#print(f'Message text: {message_text}')
if intent == 'write_measurement':
#print(type(actual_value))
#print(f'Actual value: {actual_value}')
#gdocs.insert_text(message_text)
gsheets.write_actual_size(actual_value)
response = {
"fulfillmentMessages": [{
"text": {
"text":
["Der Wert " + actual_value + " wurde eingetragen."]
}
def retrieve(seeds):
crawled_one_hop = [crawl_one_hop(seed) for seed in seeds]
crawled_two_hop = [crawl_two_hop(seed) for seed in seeds]
kg = KnowledgeGraph()
for crawled_from_seed, seed in zip(crawled_one_hop, seeds):
kg.add_node(seed)
kg.nodes[seed].add_candidates([seed])
kg.nodes[seed].set_type(get_type(seed))
for path in crawled_from_seed:
x = kg.add_node()
kg.add_edge(seed, x, path[0])
kg.nodes[x].set_type(path[1])
cands = get_candidates(seed, path[0:1])
kg.nodes[x].add_candidates(cands)
for crawled_from_seed, seed in zip(crawled_two_hop, seeds):
for path in crawled_from_seed:
predge = kg.find_edge(seed, None, path[0])[0]
if predge:
med = predge.split('--')[2]
x = kg.add_node()
kg.add_edge(med, x, path[1])
kg.nodes[x].set_type(path[3])
cands = get_candidates(seed, path[0:2])
kg.nodes[x].add_candidates(cands)
else:
x1 = kg.add_node()
x2 = kg.add_node()
kg.add_edge(seed, x1, path[0])
kg.add_edge(x1, x2, path[1])
kg.nodes[x1].set_type(path[2])
kg.nodes[x2].set_type(path[3])
return kg