def run_word_embeddings_model():
print('Loading Test Data...')
#question, answer = get_test_data('dataset/MSR_Sentence_Completion_Challenge_V1/Data/')
question, answer = get_test_data('dataset/SAT_Questions/')
print('Loading Model Data...')
m = Model('./vectors.txt')
m.load()
right = 0.0
wrong = 0.0
print('Predicting Answers... ')
for i in range(1, len(question) + 1):
#print('Predicting Q' + str(i))
q_num = str(i)
best = find_best_answer(question[q_num], m, wmd)
if answer[q_num] in best:
right += 1
else:
wrong += 1
print('The accuracy is ' + str(right / (right + wrong)))
def fill_in_choices(datafolder):
question, answer = get_test_data(datafolder)
sentences = [question[x]['statement'] for x in question]
n = len(sentences)
new_sentences = []
for i in range(1, n):
for choice in "abcde":
word_choice = question[str(i)][choice]
sentence = question[str(i)]['statement']
sentence = sentence.replace('_____', word_choice)
# replace punctuation marks using logic above
new_sentence = replace_punctuation_marks(sentence)
new_sentences.extend(new_sentence)
return new_sentences
def _raw_data(data_path=None, backwards=False):
"""Load training/test raw data from data directory "data_path".
Reads text files, converts strings to integer ids,
and performs mini-batching of the inputs.
Args:
data_path: string path to the directory where simple-examples.tgz has
been extracted.
Returns:
tuple (train_data, test_data, vocabulary)
where each of the data objects can be passed to Iterator.
"""
train_path = "dataset/treebank2/raw/wsj/"
test_path = "dataset/SAT_Questions/"
question, answer = get_test_data(test_path)
word_to_id = _build_vocab(train_path)
train_data, train_sentences, train_data_in_list_of_lists = _file_to_word_ids(
train_path, word_to_id, True, backwards)
test_data, test_sentences, test_data_in_list_of_lists = _file_to_word_ids(
test_path, word_to_id, False, backwards)
vocabulary = len(word_to_id)
return word_to_id, train_data, test_sentences, test_data_in_list_of_lists, question, answer
def _read_test_stop_at_blank(datafolder, backwards=False):
question, answer = get_test_data(datafolder)
sentences = [question[x]['statement'] for x in question]
n = len(sentences)
new_sentences = []
for i in range(1, len(question) + 1):
sentence = question[str(i)]['statement']
forward_sentence, backward_sentence = sentence.split('_____')
word_choices = ""
for choice in "abcde":
word_choice = question[str(i)][choice] + " "
word_choices += word_choice
if backwards:
reversed_backward_sentence = reverse_words_in_string(
replace_punctuation_marks(backward_sentence))
new_reversed_backward_sentence = word_choices + reversed_backward_sentence
new_sentences.append(new_reversed_backward_sentence)
else:
new_forward_sentence = word_choices + replace_punctuation_marks(
forward_sentence)
new_sentences.append(new_forward_sentence)
return new_sentences
def _read_test(datafolder):
question, answer = get_test_data(datafolder)
sentences = [question[x]['statement'] for x in question]
return sentences
import sys
from NGram import get_test_data
from operator import add
test_data_location = 'dataset/MSR_Sentence_Completion_Challenge_V1/Data'
forward_file = 'forward_out_MSR.txt'
backward_file = 'forward_out_MSR.txt'
output_file = 'bidirectional_out_MSR.txt'
_, test_answer = get_test_data(test_data_location)
probs = {}
question_nums = []
with open(forward_file, 'r') as f:
lines = f.readlines()
for line in lines:
values = line.split()
# all data lines will have length 6
if len(values) == 6:
question_num = str(int(values[0]) + 1)
question_nums.append(question_num)
probs[question_num] = [float(x) for x in values[1:]]
with open(backward_file, 'r') as f:
lines = f.readlines()
for line in lines:
values = line.split()
if len(values) == 6:
question_num = str(int(values[0]) + 1)
for i in range(len(probs[question_num])):