本研究的主要目的為,透過多重編碼器的模型架構,結合語言模型的訓練,達成可以在保留輸出格式的限制下,完成能以無監督式學習的方式對原歌詞下一句的作改寫預測。本研究主要分成二個部分,一部分探討改寫歌詞的品質。透過自動評價以及人工評價的標註,對改寫結果在中心主旨、押韻、可唱度上可以比擬原歌詞甚至好於原歌詞,在自動評價方面亦顯示出押韻的高準確度。另一個部分基於前面部分所建構的模型,觀察模型在學習正確韻律、詞性、情感的改寫上。本研究觀察了改寫條件在情緒、押韻的改動(transfer)結果,以及隨著訓練回數的注意力分布,和每次預測規一化結果的(softmax)分布上前10名在韻律、詞性上的準確率。
Creative writing has become a standard task to showcase the power of artificial intelligence. This work tackles a challenging task in this area, the lyrics rewriting. This task possesses several unique challenges. First, we require the outputs to be not only semantically correlated with the original lyrics, but also coherent in segmentation structure, rhyme as the rewritten lyrics must be performed by the artist with the same music. Second, there is no parallel rewriting lyrics corpus available for supervised training. We propose a deep neural network based model for this task and exploit both general evaluation metrics such as ROUGE and human study to evaluate the effectiveness of the model.