def initialize_models(src_corpus, trg_corpus): transition_model = TransitionModel(src_corpus, trg_corpus) translation_model = TranslationModel(src_corpus, trg_corpus) return transition_model, translation_model
def initialize_models(src_corpus, trg_corpus): prior_model = PriorModel(src_corpus, trg_corpus) translation_model = TranslationModel(src_corpus, trg_corpus) return prior_model, translation_model
]) len(word_count_de) word_count_de.most_common()[:100] word_count_de['<u>'] model_de.predict_next_word('<\s>') max_de = 0 max_en = 0 for text_de, text_en in zip(texts_de[:n_text], texts_en[:n_text]): max_de = max(max_de, len(split(text_de))) max_en = max(max_en, len(split(text_en))) print(max_de) print(max_en) de_en = TranslationModel(model_de, model_en, 11) de_en.train(texts_de[:n_text], texts_en[:n_text], epochs=epochs_translation) pd.DataFrame({'loss': de_en.history.history['loss']}).plot() plt.yscale('log') plt.grid() plt.show(block=False) i = 4000 target_text = texts_de[i] target_text = 'ich aß einen apfel.' target_text = 'du aßest einen apfel.' target_text texts_en[i] de_en.predict_word_count(texts_en[i], length=20, ref=texts_de[i])