def predict(predicted, number): if not trained: return global code predicted_so_far = predicted.get() if number == '0': code += ' ' predicted_so_far += ' ' elif number == '': predicted_so_far = word_prediction.run(code, unigram_letters, bigram_letters, unigram_words, bigram_words) else: code += number predicted_so_far = word_prediction.run(code, unigram_letters, bigram_letters, unigram_words, bigram_words) predicted.set(predicted_so_far)
def predict(predicted, number):
if not trained:
return
global code
predicted_so_far = predicted.get()
if number == "0":
code += " "
predicted_so_far += " "
elif number == "":
predicted_so_far = word_prediction.run(code, unigram_letters, bigram_letters, unigram_words, bigram_words)
else:
code += number
predicted_so_far = word_prediction.run(code, unigram_letters, bigram_letters, unigram_words, bigram_words)
predicted.set(predicted_so_far)
def test():
unigram_letters_filename = 'training/unigram_letter.json'
bigram_letters_filename = 'training/bigram_letter.json'
unigram_words_filename = 'training/unigram_words.json'
bigram_words_filename = 'training/bigram_words.json'
with open(unigram_letters_filename) as data_file:
unigram_letters = json.load(data_file)
with open(bigram_letters_filename) as data_file:
bigram_letters = json.load(data_file)
with open(unigram_words_filename) as data_file:
unigram_words = json.load(data_file)
with open(bigram_words_filename) as data_file:
bigram_words = json.load(data_file)
filenames = ['texto_coloquial.txt', 'texto_culto.txt', 'texto_noticia.txt']
for filename in filenames:
create_set('corpus/' + filename)
with open('corpus/test.json', 'r') as fp:
sentences = json.load(fp)
words_similarities = []
sentences_similarities = []
for sentence in sentences:
code = functions.sentence_to_numbers(sentence).strip()
prediction = word_prediction.run(code, unigram_letters, bigram_letters, unigram_words, bigram_words)
#Sentence similarity
vector_s_1 = functions.text_to_vector(sentence)
vector_s_2 = functions.text_to_vector(prediction)
sentences_similarities.append(functions.get_cosine(vector_s_1, vector_s_2))
sentence_words = sentence.split()
prediction_words = prediction.split()
#Word similarity
for i in range(0, len(sentence_words)):
vector_w_1 = functions.word_to_vector(sentence_words[i])
vector_w_2 = functions.word_to_vector(prediction_words[i])
words_similarities.append(functions.get_cosine(vector_w_1, vector_w_2))
words_avg = sum(words_similarities) / float(len(words_similarities))
sentences_avg = sum(sentences_similarities) / float(len(sentences_similarities))
print("Similaridad media en el {0}:\n\t-Frases: {1:.2f}\n\t-Palabras: {2:.2f}".format(filename.replace('_', ' ').replace('.txt', ''), sentences_avg, words_avg))
