以多語系自然語言理解與機器學習為基之智慧型專利摘要系統

Intelligent patent summarization system incorporating multiple natural language understanding and machine learning capability

王韋智(Wei-Chih Wang)

指導教授 ： 張瑞芬

國立清華大學/工學院/工業工程與工程管理學系/碩士(2018年)

摘要

隨著近年來第四次工業革命及人工智慧的崛起，巨量的智慧財產資訊伴隨著研發技術的進步而不斷的產生且更新，資訊超載的問題也因此產生，因此如何能讓公司或研究者快速且有效率地掌握IP重要技術以及訊息，已成為重要的研究課題。而以往公司或研究者在面對大量智財文檔上，其需能大量檢索相關前案、系爭專利、先前技術，且需閱讀理解文義、分析相關判例才能獲得有利參考情資；此工作往往耗費大量時間與人力，若檢索精準度與分析結果不佳，將難以做出具時效且合理的智財防禦及權力主張。因此本計畫運用人工智慧下之序列到序列關注機制、深度類神經網路及循環類神經網路、詞嵌入等演算法，針對多語系智財文件 (中、英文專利技術文檔)進行智財文檔智慧技術摘要彙編，以節省原本人工研擬彙整文意、翻譯、撰寫摘要的大量人力與物力，並提供研究者快速掌握IP重要技術資訊之契機。另外本研究鎖定智慧機械領域之多語系技術專利文件為資料集進行自動技術彙編之功能，運用智慧文字探勘與語意辨識之方法，進行智財文件群組之自動研讀、關鍵詞句彙整、技術報告摘要彙編，本研究最後運用ROUGE的Precision與 Recall去衡量摘要產生的品質，以驗證摘要產出涵蓋關鍵重要資訊的一致性。

關鍵字

人工智慧 ； 循環類神經網路 ； 多語系智財文件 ； 智慧機械 ； 詞嵌入

並列摘要

With the growing awareness of applying advanced technologies in hardware, software, and their integration for industrial applications, such as intelligent manufacturing (Industry 4.0), there are increasing demands in fast securing intellectual properties (IPs), to commercially protect competitive products and innovations. In order to allowing machine to learn rapid growing amount of IP documents, such as patent documents, Natural Language Processing (NLP) and Deep Learning (DL) algorithms should be deployed for their context e-discovery. The means to explain the related patent documents in a short summary remains a significant challenge. In this research, we develop an intelligent patent summarization system based on artificial intelligence (AI) approaches that include Recurrent Neural Network (RNN), Word Embedding, and Attention Mechanisms. The aim of this system is to automatically summarize multi-lingual patents in Chinese and English. The AI-based solution for summarization is used to capture the key technical keywords, popular terminologies. The ROUGE- Precision ratio and recall ratio are used to evaluate the accuracy and consistency of the output pf summarization.

並列關鍵字

Artificial intelligence ； Natural language processing ； Recurrent Neural Network ； Intellectual property

參考文獻

[1] Afzal, N., Wang, Y., & Liu, H. (2016). MayoNLP at SemEval-2016 Task 1: Semantic Textual Similarity based on Lexical Semantic Net and Deep Learning Semantic Model. In SemEval@ NAACL-HLT (pp. 674-679).

[2] Allman, T. (2009). Deterring E-Discovery Misconduct with Counsel Sanctions: The Unintended Consequences of Qualcomm v. Broadcom. Yale LJ.

[3] Anjum, N. A., Harding, J. A., Young, R. I., & Case, K. (2012). Mediation of foundation ontology based knowledge sources. Computers in Industry, 63(5), 433-442.

[4] Arel, I., Rose, D. C., & Karnowski, T. P. (2010). Deep machine learning-a new frontier in artificial intelligence research [research frontier]. IEEE computational intelligence magazine, 5(4), 13-18.

[5] Arel, I., Rose, D. C., & Karnowski, T. P. (2010). Deep machine learning-a new frontier in artificial intelligence research [research frontier]. IEEE computational intelligence magazine, 5(4), 13-18.