情感分析於投資溫度評分之應用

Sentiment Analysis for Investment Atmosphere Scoring

網路財經文本和金融商品投資收益的相關性在近年的研究議題上十分熱門。在此類型研究中，以使用自然語言處理領域中的情感分析技術量化投資大眾心理最受研究者青睞。在情感分析中，有類別型及維度型的分析方法，而維度型的分析方法又可分為單維度型與多維度型。本研究使用類別型、單維度型及多維度型三種情感分析方法，將蒐集的網路財經文本量化為投資溫度評分。實驗結果顯示，以多維度型的投資溫度評分與市場漲跌趨勢最具相關性。

The correlation between Internet financial texts and financial commodity investment income has been an emerging research topic in recent years. Sentiment analysis is a useful technique to accomplish the investment psychology. In sentiment analysis, there are two main approaches to emotion state representation: categorical and dimensional. The dimensional approach can be further divided into single dimension and multiple dimensions. This study investigates the use of categorical, single-dimensional and multi-dimensional approaches to quantifying the Internet financial texts as investment atmosphere scoring. Experimental results show that the multi-dimensional method achieves the best correlation between the investment atmosphere scoring and stock price trends.