• 沒有找到結果。

立 政 治 大 學

N a tio na

l C h engchi U ni ve rs it y

第六章 結論與未來規劃

以往對於使用者行為的研究大多侷限於單一文字或圖片,同時做圖文分析的 研究更是少之又少,是因為這些問題很難被定義,本研究使用 Word2vec 找出文 字的向量,使得文字特徵可以跟圖片特徵做結合,進一步做抽象概念的分析,藉 由提出此框架初步驗證在質化議題下只要擁有好的資料集,或做好適度的資料清 洗,仍然可以被投射至量化。另外,以往對文字的分析通常是在資料集中尋找詞 彙出現頻率,以此分析出特定的使用者行為,然而本研究透過圖文的結合所定義 出的特徵向量,當有一筆新的資料進來時,便可直接進行預測,比起單純透過詞 頻去分析,本研究提出的框架更有彈性空間。

本研究在觀光與非觀光類的類別下以階層式分群法進行資料分群,雖然在細 分類上的效果還不如預期,但在廣義上各個群集的資料內容的確符合各自的大類 別,因此本研究也期望在未來能藉由質化的定義,建立更多類別的標準,以便後 續訓練與分析,有助於各種質化研究的需求。

除此之外,還有很多議題可以探討,即使本研究的主軸是分析推文是否為觀 光類型,但在資料清洗與標註的過程中,我們也觀察到原始資料包含形形色色的 推文,例如被本研究歸類至負樣本的偶像類別或政治類別,在 Twitter 上確實常 常造成話題的風向,因此未來也可往流行趨勢、政治等議題進行分析。

[1] Google Clound Vision API Documentation.

https://cloud.google.com/vision/docs/.

[2] Amazon Rekognition. https://aws.amazon.com/rekognition/?nc1=h_ls.

[3] 中華民國交通部觀光局,觀光統計圖表。

https://admin.taiwan.net.tw/public/public.aspx?no=315

[4] GU, Chunhui, et al. AVA: A video dataset of spatio-temporally localized atomic visual actions. arXiv preprint arXiv:1705.08421, 2017, 3.4: 6.

[5] HUBEL, David H.; WIESEL, Torsten N. Receptive fields, binocular interaction and functional architecture in the cat's visual cortex. The Journal of physiology, 1962, 160.1: 106-154.

[6] HINTON, Geoffrey E.; OSINDERO, Simon; TEH, Yee-Whye. A fast learning algorithm for deep belief nets. Neural computation, 2006, 18.7: 1527-1554.

[7] RANJAN, Rajeev; PATEL, Vishal M.; CHELLAPPA, Rama. Hyperface: A deep multi-task learning framework for face detection, landmark localization, pose estimation, and gender recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 41.1: 121-135.

[8] KRIZHEVSKY, Alex; SUTSKEVER, Ilya; HINTON, Geoffrey E. Imagenet classification with deep convolutional neural networks. In: Advances in neural information processing systems. 2012. p. 1097-1105.

[9] HE, Kaiming, et al. Deep residual learning for image recognition. In: Proceedings of the IEEE conference on computer vision and pattern recognition. 2016. p. 770-778.

[10] HU, Jie; SHEN, Li; SUN, Gang. Squeeze-and-excitation networks. arXiv preprint arXiv:1709.01507, 2017, 7.

[11] 林之昫,HubertLin (2017 )。 最 後一屆 ImageNet 大 規 模視覺 識別 大賽

(ILSVRC2017)順利落幕,而 WebVision 圖像大賽會是下一個 ImageNet 大賽嗎?。https://goo.gl/5rHG1y。

[12] LI, Wen, et al. Webvision database: Visual learning and understanding from web data. arXiv preprint arXiv:1708.02862, 2017.

[13] WebVision. https://www.vision.ee.ethz.ch/webvision/2017/index.html.

[14] WebVision Challenge Results.

https://www.vision.ee.ethz.ch/webvision/2017/challenge_results.html.

[15] HU, Yuheng, et al. What We Instagram: A First Analysis of Instagram Photo Content and User Types. In: Icwsm. 2014.

[16] SZEGEDY, Christian, et al. Going deeper with convolutions. In: Proceedings of the IEEE conference on computer vision and pattern recognition. 2015. p. 1-9.

[17] IOFFE, Sergey; SZEGEDY, Christian. Batch normalization: Accelerating deep network training by reducing internal covariate shift. arXiv preprint arXiv:1502.03167, 2015.

[18] SZEGEDY, Christian, et al. Rethinking the inception architecture for computer vision. In: Proceedings of the IEEE conference on computer vision and pattern recognition. 2016. p. 2818-2826.

[19] CHOLLET, François. Xception: Deep learning with depthwise separable convolutions. arXiv preprint, 2017, 1610.02357.

[20] HOWARD, Andrew G., et al. Mobilenets: Efficient convolutional neural

networks for mobile vision applications. arXiv preprint arXiv:1704.04861, 2017.

[21] Keras Documentation. https://keras.io/applications/.

[22] HARRIS, Zellig S. Distributional structure. Word, 1954, 10.2-3: 146-162.

‧ 國

立 政 治 大 學

N a tio na

l C h engchi U ni ve rs it y

[23] Vector Representations of Words. https://www.tensorflow.org/tutorials/word2vec.

[24] MIKOLOV, Tomas, et al. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781, 2013.

[25] 李維平、張加憲(2013)。使用 N 組連結平均法的階層式自動分群。電子商

務學報,第十五卷(第一期),35-56。

[26] MAATEN, Laurens van der; HINTON, Geoffrey. Visualizing data using t-SNE. Journal of machine learning research, 2008, 9.Nov: 2579-2605.

[27] HINTON, Geoffrey E.; ROWEIS, Sam T. Stochastic neighbor embedding.

In: Advances in neural information processing systems. 2003. p. 857-864.

[28] Flood and Fire Twitter Capture and Analysis Toolset, ff-tcat.

https://github.com/Sparklet73/ff-tcat.git

[29] Sara Robinson(2016), Google Cloud Vision – Safe Search Detection API.

https://cloud.google.com/blog/big-data/2016/08/filtering-inappropriate-content-with-the-cloud-vision-api

[30] Caffe. http://caffe.berkeleyvision.org/

[31] Open NSFW Model, yahoo. https://github.com/yahoo/open_nsfw.git

[32] GODIN, Fréderic, et al. Multimedia Lab $@ $ ACL WNUT NER Shared Task:

Named Entity Recognition for Twitter Microposts using Distributed Word

Representations. In: Proceedings of the Workshop on Noisy User-generated Text.

2015. p. 146-153.

相關文件