def make_x_train():
# создаём словарь со всеми текстами ручной разметки {индекс:текст}
dict_texts = {}
all_train_texts = db.select_text('train_corpus')
for inst in all_train_texts:
dict_texts[inst.index_text] = inst.text
# создаём словарь {идекс1.индекс2: категория}
dict_rels = {}
all_train_relations = db.select_relations('train_relations')
for inst in all_train_relations:
dict_rels[inst.first_text + '.' + inst.second_text] = inst.relation
# создаём вектор каждой пары из предыдущего словаря
vectors = []
y_train = []
for pair, rel in dict_rels.items():
text1 = dict_texts[pair.split('.')[0]]
text2 = dict_texts[pair.split('.')[1]]
vectors.append(mv.compare(text1, text2))
y_train.append(rel)
return vectors, y_train
def make_x_test():
collection = {}
ids = []
for inst in db.select_text('texts', **{'publication_date': " BETWEEN '2015-01-05' AND '2015-01-05'"}):
collection[inst.id] = inst.text
ids.append(inst.id)
vectors = []
pairs = []
for index, first_id in enumerate(ids[:-1]):
for second_id in ids[index+1:]:
pairs.append(str(first_id) + '.' + str(second_id))
vectors.append(mv.compare(collection[first_id], collection[second_id]))
return vectors, pairs
