def evaluate_sentence(sentence, vocabulary): """ Translates a string to its equivalent in the integer vocabulary and feeds it to the network. Outputs result to stdout. """ #print(current_milli_time()) x_to_eval = data_helpers.string_to_int(sentence, vocabulary, maxLengthInX) #print(current_milli_time()) #result = sess.run(tf.argmax(network_out,1), feed_dict={data_in: x_to_eval, dropout_keep_prob: 1.0}) #print(current_milli_time()) unnorm_result = sess.run(network_out, feed_dict={ data_in: x_to_eval, dropout_keep_prob: 1.0 }) #print(current_milli_time()) network_sentiment = "NEU" if unnorm_result[0].item(1) > 0.6: network_sentiment = "POS" elif unnorm_result[0].item(1) < 0.4: network_sentiment = "NEG" log("Custom input evaluation:", network_sentiment) log("Actual output:", str(unnorm_result[0])) return network_sentiment
def evaluate_sentence(sentence, vocabulary): x_to_eval = string_to_int(clean_str(sentence), vocabulary, max(len(_) for _ in x)) result = sess.run(tf.argmax(network_out, 1), feed_dict={data_in: x_to_eval, dropout_keep_prob: 1.0}) unorm_result = sess.run(network_out, feed_dict={data_in: x_to_eval, dropout_keep_prob: 1.0}) return result[0]
def evaluate_sentence(sentence, vocabulary): """ Translates a string to its equivalent in the integer vocabulary and feeds it to the network. Outputs result to stdout. """ x_to_eval = data_helpers.string_to_int(sentence, vocabulary, max(len(i) for i in x)) result = sess.run(tf.argmax(network_out,1), feed_dict={data_in: x_to_eval, dropout_keep_prob: 1.0}) unnorm_result = sess.run(network_out, feed_dict={data_in: x_to_eval, dropout_keep_prob: 1.0}) network_sentiment = "POS" if result == 1 else "NEG" log("Custom input evaluation:", network_sentiment) log("Actual output:", str(unnorm_result[0]))
def evaluate_sentence(sentence, vocabulary): """ Translates a string to its equivalent in the integer vocabulary and feeds it to the network. Outputs result to stdout. """ x_to_eval = string_to_int(sentence, vocabulary, max(len(_) for _ in x)) result = sess.run(tf.argmax(network_out, 1), feed_dict={data_in: x_to_eval, dropout_keep_prob: 1.0}) unnorm_result = sess.run(network_out, feed_dict={data_in: x_to_eval, dropout_keep_prob: 1.0}) network_sentiment = 'POS' if result == 1 else 'NEG' log('Custom input evaluation:', network_sentiment) log('Actual output:', str(unnorm_result[0]))
def evaluate_sentence(sentence): x_to_eval = string_to_int(sentence["text_cleaned"], vocabulary, max(len(_) for _ in x)) result = sess.run(tf.argmax(network_out, 1), feed_dict={ data_in: x_to_eval, dropout_keep_prob: 1.0 }) # unnorm_result = sess.run(network_out, feed_dict={data_in: x_to_eval, # dropout_keep_prob: 1.0}) network_sentiment = 'POS' if result == 1 else 'NEG' # log('Custom input evaluation:', network_sentiment) # log('Actual output:', str(unnorm_result[0])) if result is not None: return result[0] else: return 'ERROR'
def evaluate_sentence(id, sentence, vocabulary): """ Translates a string to its equivalent in the integer vocabulary and feeds it to the network. Outputs result to stdout. """ x_to_eval = string_to_int(sentence, vocabulary, max(len(_) for _ in x)) result = sess.run(tf.argmax(network_out, 1), feed_dict={ data_in: x_to_eval, dropout_keep_prob: 1.0 }) unnorm_result = sess.run(network_out, feed_dict={ data_in: x_to_eval, dropout_keep_prob: 1.0 }) network_sentiment = '1' if result == 1 else '-1' log(id + "," + network_sentiment)
def evaluate_sentence(sentence, vocabulary): """ Translates a string to its equivalent in the integer vocabulary and feeds it to the network. Outputs result to stdout. """ #print(current_milli_time()) x_to_eval = data_helpers.string_to_int(sentence, vocabulary, maxLengthInX) #print(current_milli_time()) #result = sess.run(tf.argmax(network_out,1), feed_dict={data_in: x_to_eval, dropout_keep_prob: 1.0}) #print(current_milli_time()) unnorm_result = sess.run(network_out, feed_dict={data_in: x_to_eval, dropout_keep_prob: 1.0}) #print(current_milli_time()) network_sentiment = "NEU" if unnorm_result[0].item(1) > 0.6: network_sentiment = "POS" elif unnorm_result[0].item(1) < 0.4: network_sentiment = "NEG" log("Custom input evaluation:", network_sentiment) log("Actual output:", str(unnorm_result[0])) return network_sentiment
RUN_DIR, 'minimal_graph.txt', as_text=True) if FLAGS.submit: log('Loading submit data') submit_examples = list( open("../twitter-datasets/test_data.txt", "r").readlines()) submit_examples = [s.strip() for s in submit_examples] splitter = [s.split(',', 1) for s in submit_examples] sentences = [s[1] for s in splitter] ids = [s[0] for s in splitter] #evaluate_sentence(tweet_id, tweet_data, vocabulary) max_len = max(len(_) for _ in x) x_to_eval = [ string_to_int(sentence, vocabulary, max_len)[0] for sentence in sentences ] log('generating submissions data') result = sess.run(tf.argmax(network_out, 1), feed_dict={ data_in: x_to_eval, dropout_keep_prob: 1.0 }) result = ['1' if r == 1 else '-1' for r in result] log('saving submissions') with open("../twitter-datasets/tscnnsubmission.csv", "w") as f: f.write("Id,Prediction\n") for i, p in enumerate(result): f.write(str(ids[i]) + "," + str(p) + "\n")