進階搜尋


   電子論文尚未授權公開,紙本請查館藏目錄
(※如查詢不到或館藏狀況顯示「閉架不公開」,表示該本論文不在書庫,無法取用。)
系統識別號 U0026-2906202016465600
論文名稱(中文) 基於財經新聞知識圖譜的公司關係聊天機器人
論文名稱(英文) A Company Relation Chatbot Based on Financial News Knowledge Graph
校院名稱 成功大學
系所名稱(中) 資訊工程學系
系所名稱(英) Institute of Computer Science and Information Engineering
學年度 108
學期 2
出版年 109
研究生(中文) 張允揚
研究生(英文) Yun-Yang Chang
學號 P76071080
學位類別 碩士
語文別 英文
論文頁數 53頁
口試委員 指導教授-盧文祥
口試委員-梁勝富
口試委員-楊中平
口試委員-張景新
口試委員-張嘉惠
中文關鍵字 公司關係  知識圖譜  聊天機器人 
英文關鍵字 company relation  knowledge graph  Chatbot 
學科別分類
中文摘要 隨著近年AI的火熱,能夠理解語意的聊天機器人也變成熱門的話題,並廣泛應用在各個領域,幾乎所有場景都可以看見聊天機器人的蹤影。知識圖譜也是這個時代相當熱門的應用,不管是搜尋引擎還是聊天機器人,或多或少都有使用到它的技術。
但是,我們發現,目前市面上沒有一個聊天機器人可以回答有關公司關係的問題,舉例來說: “鴻海收購哪家公司?”。就連擁有海量資料的知識圖譜也無法回答上面的問題。
因此,為了解決上述問題,本研究提出了一種基於公司關係提取的方法,分析財經新聞的句法與常見的公司關係後,透過自然語言處理工具從中提取公司關係,並建構財經新聞知識圖譜。基於此知識圖譜,我們建立了一個公司關係聊天機器人,透過分析使用的意圖與辨識到的實體,回答使用者常見的公司關係的問題。
英文摘要 In recent years, with the popularity of AI, chatbot who can understand semantic sense also a popular topic. Chatbot is used in various fields, it can be seen everywhere. Knowledge graph also a hot application, both search engine and chatbot are used it in backend.
However, we found that there is no a chatbot on relation between companies. For example, “Which company does Google work with?”. Even knowledge graph with huge amounts of data source cannot answer the question.
Therefore, to solve above problem, we propose a method to extract relation between companies in this work. After analyze syntax of finance news articles and common company relations, using nature language process tool to extract company relation from news articles and construct a knowledge graph with company relation. We build a chatbot with intent classification and entity recognition to response the questions with company relation.
論文目次 摘要 I
Abstract II
致謝 III
List of Tables VI
List of Figures VII
Chapter 1 Introduction 1
1.1 Background 1
1.2 Motivation 2
1.3 Goal 3
1.4 Method 3
1.5 Contribution 4
Chapter 2 Related Work 5
2.1 Relation Extraction 5
2.2 Studies on Nature Language Understanding 6
2.3 Studies on Nature Language Generation 6
Chapter 3 Method 8
3.1 System Framework 8
3.2 Data Resource 9
3.2.1 Finance News 9
3.2.2 Company Lexicons 9
3.3 Company Relation Sentence Extraction 10
3.3.1 Observation of Relation between companies 10
3.3.2 Text pre-processing using CkipTagger 12
3.3.3 Relation Lexicons 14
3.3.4 Company Relation Sentence Extraction 18
3.4 Company Relation Triple Extraction 19
3.4.1 Company Relation Triple Extraction 19
3.4.2 Company Event Term Triple Extraction 23
3.5 Company Relation Triple Filtration 28
3.5.1 Observation of Company Relation Triple 28
3.5.2 Uncertain and Negative words Filter 28
3.5.3 Company Relation Confliction 30
3.6 Finance news Knowledge graph 31
3.6.1 Entity Alignment 31
3.6.2 Company Relation triple Preservation 32
3.6.3 Knowledge graph Construction 33
3.7 Intent Classification 35
3.7.1 Entity Recognition 35
3.7.2 Intent Recognition 36
3.8 Service Matching 37
3.9 Response Generation 40
Chapter 4 Experiments 42
4.1 Dataset 42
4.1.1 Finance news articles 42
4.1.2 Questions from CQA platform 42
4.2 Evaluation Metrics 43
4.3 Experiment on CkipTagger NER Enhancement 44
4.3.1 Experiment Setup 44
4.3.2 Experiment Result 44
4.4 Experiment on Event-term Extraction 45
4.4.1 Experiment Setup 45
4.4.2 Experiment Result 45
4.5 Experiment on Company Relation triple Extraction 47
4.5.1 Experiment Setup 47
4.5.2 Experiment Result 47
4.6 Experiment on Intent Classification 49
4.6.1 Experiment Setup 49
4.6.2 Experiment Result 50
Chapter 5 Conclusions 51
5.1 Conclusions 51
5.2 Future Work 51
Reference 52
參考文獻 Reference
[1] Michael Färber and Achim Rettinger, Which Knowledge Graph Is Best for Me?, 2018.
[2] Wikidata. Available from: https://www.wikidata.org/wiki/Wikidata:Main_Page.
[3] knowledge graph generation for finance databases. Available from: https://blog.bitext.com/knowledge-graph-generation-for-finance-databases.
[4] Yuen-Hsien Tseng, Lung-Hao Lee and Shu-Yen Lin Chinese Open Relation Extraction for Knowledge Acquisition, 2014.
[5] freebase. Available from: https://developers.google.com/freebase.
[6] Mike Mintz, Steven Bills, Rion Snow, Dan Jurafsky, Distant supervision for relation extraction without labeled data, 2009.
[7] Zhiheng Huang, Wei Xu and Kai Yu, Bidirectional LSTM-CRF Models for Sequence Tagging, 2015.
[8] 鉅亨網. Available from: https://www.cnyes.com/.
[9] 經濟日報. Available from: https://money.udn.com/money/index.
[10] Wall Street Journal. Available from: https://www.wsj.com/.
[11] Yahoo Stock. Available from: https://tw.stock.yahoo.com/.
[12] E-HowNet. Available from: http://ehownet.iis.sinica.edu.tw/index.php.
[13] Tomas Mikolov, Ilya Sutskever and Kai Chen, Distributed Representations of Words and Phrases and their Compositionality, 2013.
[14] CkipTagger. Available from: https://github.com/ckiplab/ckiptagger.
[15] Hanlp. Available from: https://hanlp.hankcs.com/.
[16] 莊勝棠, Product Recommendation BOT Based on Related Task Structure Using News Event and Yahoo Answer, 2018.
[17] John Lafferty, Andrew McCallum and Fernando C.N. Pereira, Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data, 2001.
[18] Lee-Feng Chien, Chun-Liang Chen, Wen-Hsiang Lu and Yuan-Lu Chang, Recent Results on Domain-Specific Term Extraction from Online Chinese Text Resources, 1999.
[19] Neo4J. Available from: https://neo4j.com/.
[20] Cypher. Available from: https://neo4j.com/developer/cypher-query-language/.
[21] yahoo knowledge+. Available from: https://tw.answers.yahoo.com/?guccounter=1&guce_referrer=aHR0cHM6Ly93d3cuZ29vZ2xlLmNvbS8&guce_referrer_sig=AQAAAHSZzs7iz1zfofWDQKEE2CUiq-bAeDUHX3uy4ss7HZ6c-afLgNZ123-Jk14QD1S8YD6GRIqOabpOWRn1NcFpZ-3fSzAtpPCg8zeMXt6Grh6IinDnXLbr3pWQJku99W2L6URUFqRDM-kMFV_MQsubH-GdSW-_aH4mOZTiRYxnOfhM.
[22] 百度知道. Available from: https://zhidao.baidu.com/.
論文全文使用權限
  • 同意授權校內瀏覽/列印電子全文服務,於2025-06-23起公開。
  • 同意授權校外瀏覽/列印電子全文服務,於2025-06-23起公開。


  • 如您有疑問,請聯絡圖書館
    聯絡電話:(06)2757575#65773
    聯絡E-mail:etds@email.ncku.edu.tw