利用Attentive來改善端對端中文語篇剖析遞迴類神經網路系統

Using Attentive to improve Recursive LSTM End-to-End Chinese Discourse Parsing

篇章剖析，可以幫助我們以不同角度來理解文句之間的關係與連結，但篇章剖析資料結構目前仰賴人工標記，使這項技術無法直接利用在任意篇章中。因此至目前為止，有許多研究著手於讓電腦能夠自動對篇章進行篇章剖析，並建構出一個完整的剖析樹。以中文語料庫CDTB來說，欲建立完整的篇章剖析程式，其問題主要可以被分成四項，分別是子句分割、剖析樹建立、子句關係辨識與中心關係辨識。由於深度學習近幾年發展快速，因此針對篇章剖析的建構方法也從傳統的SVM，CRF 等方法，進展到目前以遞迴類神經的方式來建構剖析篇章程式。在本篇論文中，我們使用了許多目前最新的深度學習技術，例如Attentive RvNN、self-attentive、BERT等方法，來提高模型的準確度。最後，我們成功將每一項任務的F1都提高了近10％左右，達到目前我們所知研究中最好的效能。

Discourse parser helps us understand the relationship and connection between sentences and sentences from different angles, but the tree structure data still need to rely on manual marking, which makes this technology unable to be directly used in daily life. So far, there have been many research and studies on how to automatically construct the complete tree structure on the computer. Since deep learning has progressed rapidly in recent years, the construction method for discourse parser has also changed from the traditional SVM, CRF method to the current recursive neural network. In the Chinese corpus tree library CDTB, the parsing analysis problem can be divided into four main problems, including elementary discourse unit (EDU) segmentation, tree structure construction, center labeling, and sense labeling. In this paper, we use many state-of-the-art deep learning techniques, such as attentive recursive neural networks, self-attentive, and BERT to improve the performance. In the end, we succeeded in increasing the accuracy by more than 10% of F1 in each task, reaching the best performance we know so far.