論文提要內容：
    對政府機關來說，如何快速並正確的了解輿論走向一直是各行政機關努力的目標，政府部門希望可以在第一時間根據輿論去修正政策的執行方向，但是以傳統的問卷、電話訪查方式無法第一時間得知輿論的走向，除了耗費大量時間外，也浪費許多寶貴的人力成本，因此希望可以透過人工智慧的技術來處理此議題。
    相較於傳統人力的方式，透過人工智慧可以更加快速、廣泛地蒐集資料，並且透過自然語言處理的相關技術使電腦認知並理解問題後，再對巨量的資料進行分析與學習，進而應用在情感分析，得知句子背後所含有的情緒與意圖，而更加廣泛地應用在分析大量網路上的評論即是輿情分析。
    本論文提出一套基於人工智慧分析輿情的方法，以人工智慧分析出輿論的走向，希望在政府機關發布政策的第一時間即可透過網路上民眾的留言即時取得民眾對政策的滿意度。本論文的重要工作主要分成以下幾大部分，首先，利用爬蟲技術自動化蒐集網路上的大量輿論，由於數位化的趨勢現今民眾會在各大網路平台上發表對議題的看法，像是臉書專頁、PTT論壇…等，其次，對蒐集到的資料進行處理，像是對句子斷詞斷句後再給予標籤，最後，透過人工智慧訓練大量具有標籤的資料後，得到各種分類的模型，例如，針對發布的新政策判斷出是屬於哪個政府單位所發布，以及民眾對此政策的施政滿意度等都可以透過分類模型進行判斷，進而去預測出相關議題的未來輿論走向。本論文提出一套自動化蒐集資料，並且自動對資料進行標籤的演算法，透過以上方式解決人工智慧訓練模型時資料量不足的問題。

Abstract:
    For government agencies, how to quickly and correctly understand the trend of public opinion has always been the goal of administrative organs. Government departments hope to revise the implementation direction of policies according to public opinion in the first time. However, it is impossible to know the direction of public opinion in the first time through traditional questionnaires and telephone interviews. In addition to wasting a lot of time, it also wastes a lot of valuable human resources, Therefore, it is hoped that this issue can be dealt with through the technology of artificial intelligence.
    Compared with the traditional human way, artificial intelligence can collect data more quickly and widely, and through the related technology of natural language processing, the computer can recognize and understand the problem, then analyze and learn the huge amount of data, and then apply it to emotional analysis to get the emotion and intention behind the sentence, and it is more widely used in analyzing large number of public opinion which is the public opinion analysis.
    This paper proposes a set of methods based on artificial intelligence to analyze public opinion. It can analyze the trend of public opinion with artificial intelligence. It is hoped that the government can obtain the satisfaction of the public on the policy through the public message on the Internet as soon as the government releases the policy. The important work of this paper is mainly divided into the following parts. Firstly, the use of crawler technology to automatically collect a large number of online public opinion, due to the trend of digitalization, nowadays people will express their opinions on topics on major network platforms, such as Facebook page, PTT forum, etc. Secondly, the collected data will be processed, such as sentence segmentation and sentence tagging. Finally, after a large number of tagged data are trained through artificial intelligence, various classification models are obtained, such as for publishing the classification model can be used to predict the future public opinion trend of the relevant issues. This paper proposes an algorithm for automatically collecting data and automatically labeling the data to solve the problem of artificial intelligence lack of training data.

目錄
目錄	VI
圖目錄	VIII
表目錄	IX
第一章、	簡介	1
第二章、相關研究	5
第三章、背景知識	8
3.1、爬蟲技術	8
3.2、Google Trends	8
3.3、CKIP	9
3.4、TF-IDF	9
3.5、Topic Modeling	10
3.6、情緒詞庫	11
3.7、LSTM	11
3.8、BERT	11
第四章、系統架構	13
4.1、環境與問題描述	13
4.2、系統架構	13
第五章、	實驗分析	21
5.1、實驗環境	21
5.2、數據集	21
5.3、實驗比較	21
第六章、結論	24
參考文獻	25
附錄-英文論文	26
圖目錄
圖 1、TF-IDF	10
圖 2、機構單位辨識與分類 (訓練期)	15
圖 3、機構單位辨識與分類 (使用期)	15
圖 4、提取文章討論議題	16
圖 5、議題權重演算法	17
圖 6、計算輿情滿意度(訓練期)	18
圖 7、輿情分數演算法	19
圖 8、計算輿情滿意度(使用期)	19
圖 9、建構影響力評估模型	20
圖 10、實驗分析比較圖	23
表目錄
表 1、相關文獻比較表	7
表 2、混淆矩陣表	22

[1]H. Liu, "Internet Public Opinion Hotspot Detection and Analysis Based on Kmeans and SVM Algorithm," 2010 International Conference of Information Science and Management Engineering, Xi'an, 2010, pp. 257-261, doi: 10.1109/ISME.2010.207.
[2]Cuixin Yuan, Hao Lin, Xu Zhang, Chunyang Liu and Lihong Wang, "OPO: Online public opinion analysis system over text streams," 2017 International Conference on Service Systems and Service Management, Dalian, 2017, pp. 1-6, doi: 10.1109/ICSSSM.2017.7996299.
[3]V. Sathya, A. Venkataramanan, A. Tiwari and D. D. P.S., "Ascertaining Public Opinion Through Sentiment Analysis," 2019 3rd International Conference on Computing Methodologies and Communication (ICCMC), Erode, India, 2019, pp. 1139-1143, doi: 10.1109/ICCMC.2019.8819738.
[4]V. S. Pagolu, K. N. Reddy, G. Panda and B. Majhi, "Sentiment analysis of Twitter data for predicting stock market movements," 2016 International Conference on Signal Processing, Communication, Power and Embedded System (SCOPES), Paralakhemundi, 2016, pp. 1345-1350, doi: 10.1109/SCOPES.2016.7955659.
[5]L. Li, Y. Wu, Y. Zhang and T. Zhao, "Time+User Dual Attention Based Sentiment Prediction for Multiple Social Network Texts With Time Series," in IEEE Access, vol. 7, pp. 17644-17653, 2019, doi: 10.1109/ACCESS.2019.2895897.
[6]G. Xu, Y. Meng, X. Qiu, Z. Yu and X. Wu, "Sentiment Analysis of Comment Texts Based on BiLSTM," in IEEE Access, vol. 7, pp. 51522-51532, 2019, doi: 10.1109/ACCESS.2019.2909919.
[7]Devlin, Jacob; Chang, Ming-Wei; Lee, Kenton; Toutanova, Kristina (11 October 2018). "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding". arXiv:1810.04805v2 [cs.CL].