應用深度學習與注意力機制分析網路社群文本之情緒：以日本動畫社群Twitter資料為例

Applying Deep Learning and Attention Mechanisms to Analysis the Sentiment of Cyber Community Texts: A Study of Animation Community Twitter Data in Japan

李承翰(Li Cheng-Han)

指導教授 ： 陳灯能

國立屏東科技大學/管理學院/資訊管理系所/碩士(2020年)

https://doi.org/10.6346/NPUST202000318

摘要

近年隨著社群媒體的興起，人們習慣於在社群媒體上發表帶有其情緒的推文，除了會利用文字來抒發自己的情緒之外，使用文字配合聲音訊號、影像和影片來表達自我情緒的人也有逐漸增長的趨勢。因此，許多研究者也紛紛開始在研究中嘗試加入分析聲音訊號、影像和影片的資料，希望能分析這類的多媒體資料的情緒。本研究提出結合文字與影像來進行情緒分析的深度學習多模態模型，利用該模型分析與日本動畫評論相關之Twitter推文情緒。研究結果顯示，在文字加上影像的原推文上，純文字分析的模型即可達到良好的辨識準確率，而結合文字與影像特徵的多模態模型並沒有達到較好的辨識準確率。

關鍵字

情緒分析 ； 自然語言處理 ； 日語 ； 詞嵌入 ； 卷積神經網路 ； 長短期記憶 ； 注意力機制

並列摘要

With the rise of social media in recent years, people are accustomed to publishing tweets with their emotions on social media. In addition to expressing their emotions with text only, they also use text with voice, images, and videos to express themselves. Therefore, many researchers have begun to add the analysis of sound signals, images, and videos in their sentiment analysis research, hoping to improve the accuracy of the overall sentiment analysis research. This study proposes a deep learning hybrid model that combines text and images for sentiment analysis and uses this model to analyze the sentiment of Twitter tweets related to Japanese animation reviews. The research results show that, in the original tweet of text and image, the pure text analysis model can achieve good classification accuracy, but the combination of text and image analysis does not achieve better accuracy.

並列關鍵字

sentiment analysis ； natural language processing ； Japanese ； word embedding ； convolutional neural network ； long short-term memory ； attention mechanism

參考文獻

1. Bahdanau, D., Cho, K., and Bengio, Y. 2014. "Neural Machine Translation by Jointly Learning to Align and Translate," arXiv preprint arXiv:1409.0473).

2. Bates, M. 1995. "Models of Natural Language Understanding," Proceedings of the National Academy of Sciences (92:22), pp. 9977-9982.

3. Bengio, Y., Ducharme, R., Vincent, P., and Jauvin, C. 2003. "A Neural Probabilistic Language Model," Journal of machine learning research (3:Feb), pp. 1137-1155.

4. Borth, D., Ji, R., Chen, T., Breuel, T. M., and Chang, S.-F. 2013. "Large-Scale Visual Sentiment Ontology and Detectors Using Adjective Noun Pairs," MM '13.

5. Cao, D., Ji, R., Lin, D., and Li, S. 2014. "A Cross-Media Public Sentiment Analysis System for Microblog," Multimedia Systems (22), pp. 479-486.