- 學位論文
人工智慧個人助理之設計與實作
Design and Implementation of an AI-Based Personal Assistant
摘要
AI（人工智慧）是一門很有趣的科學，利用NLP（自然語言處理）進行中文自然語言處理的研究，更是另人著迷。藉由NLP能夠讓人類以最自然的方式與電腦進行溝通，使電腦能夠理解、分析對話的句型與語意，擷取句子中重要的概念，進而歸納、學習成為知識。透過Q/A( 問答 )系統的設計，讓電腦與使用者能夠進行互動，理解對話中的語意，藉由句子的詞性(POS)、句型結構、文法分析、資料擷取與知識整理、問答系統將能回應正確的資訊給使用者。 本篇論文藉由網路搜尋，結合類神經網路概念去設計一套Artificially Intelligence-Based Person Assistant(AIPA)系統，並且經由Triple-mode 與 Hash Function 去建構一個 ontology，利用Class Algebra為理論基礎，同時以IS-A hierarchy架構搭配HNN關聯式模組進行設計。本篇論文的研究與實作的目的：首先，並非讓電腦擁有創造智慧的能力，而是藉由實作一套QA系統，實際應用於電腦的使用，讓使用者能利用最單純的口語進行對話，使電腦可以理解對話的語意，進而回應相關的資料給使用者。透過人工智慧的自然語言技術，除了能讓使用者以口語化的方式與AIPA進行中文的溝通外，並且希望藉由電腦指令的辨識與Google搜尋引擎的橋接，幫助使用者能更快速的查詢資訊與執行電腦指令，增進工作效率與節省時間。 藉由AIPA的人機介面輔助與學習模型的設計，就如同兒童學習語言一般，讓AIPA的學習能夠循序漸進，透過使用者的指導，將能降低句型、詞彙誤判的情形並讓AIPA學習的效果加倍。雖然本篇論文無法創造AIPA能夠擁有人類創造智慧的能力，但藉由本篇論文所研究與實作的AIPA系統，能實際便利人們的生活，透過最自然口語化的方式，縮短操作電腦的時間、提高工作效率，並且AIPA利用網路查詢與互動式學習的特性，更能擴增AIPA的知識，豐富NTL的Knowledge Token，讓語句的判斷與分析越趨正確。如此，藉由AIPA的輔助，便能有效的幫助使用者，快速的查詢資訊與口語化執行電腦指令，以達增進工作效率與節省時間之目標。
並列摘要
Artificial Intelligence (AI) is a very interesting science. Using Natural Language Process (NLP) for Chinese natural language processing is also a fascinated research topic. NLP lets people communicate with the computer more naturally. It enables the computer to understand the sentence meaning, to analyze the sentence pattern and semantic of dialogue, to extract the important concept from the article, to conclude its learning and finally to become useful knowledge. Through the design of Question Answering (Q/A) system to comprehend the semantics of the dialogue, part of speech (POS), sentence structures, grammatical parsing, data arrangement and information extraction, the computer will provide accurate responses to users and interact with people more friendly. We describe the design of an Artificially Intelligent-Based Personal Assistant (AIPA) system that uses concepts from the artificial neural networks, a triple-based knowledge representation that is used for both an ontology and its binary relationships, fast access via hash functions, and Class Algebra as the theoretical basis. Instead of providing the computers with the ability to create intelligence, the purpose of the research and implementation herein is to apply a QA system for actual use by users who may converse via the Chinese language with the computer, which then will understand the semantics and respond to the users with corresponding data. Through artificial intelligent natural language technology, not only are the users able to colloquially to communicate with AIPA, but the system can also identify the users’ commands to bridge the Google search engine. Therefore, users can efficiently search for information or execute programs by using AIPA. Like children gradually learn the new language, the user interface of AIPA gains knowledge step by step from users. With the help of users’ guide, AIPA reduces the likelihood of misjudgment in sentence patterns and vocabulary identifications so that it can increase the computer’s learning efficiency. However, the aim of this paper is not to make the computer have the capabilities to create intelligence as human beings. With this thesis, research and implementation of the AIPA system can actually facilitate people’s lives. Through the most natural colloquial way, the system helps users to shorten the computer’s operating time and improve their work efficiency. Also, through Internet searching and interactive learning, AIPA can increase its knowledge and enrich the Knowledge Tokens of NTL to obtain more and more correct sentence analyses. Therefore, users can quickly search for information and easily operate computer instructions by colloquial conversations via AIPA assistance, to improve work efficiency and save time.