論 文Papers

JOURNAL (INTERNATIONAL)

GINN: Gradient Interpretable Neural Networks for Visualizing Financial Texts

Tomoki Ito (UTokyo), Hiroki Sakaji (UTokyo), Kiyoshi Izumi (UTokyo), Kota Tsubouchi, Tatsuo Yamashita

International Journal of Data Science and Analytics 2018, 2018/12

Category:

機械学習 (Machine Learning) データサイエンス (Data Science)

Abstract:
We are concerned with how to visualize financial documents in the way that even nonexperts can swiftly understand their sentiments. To achieve this, we propose a novel text visualizing method using interpretable neural network (NN) architecture called gradient interpretable NN (GINN). GINN can visualize the market sentiment score from a whole financial document and the sentiment gradient scores in both word and concept units. Moreover, GINN can visualize which concept is important considering sentence contexts. This type of visualization helps nonexperts easily understand financial documents. We theoretically analyze the validity of GINN and experimentally demonstrate the validity of text visualization produced by GINN using real financial texts.
Download:

GINN: Gradient Interpretable Neural Networks for Visualizing Financial Texts(外部サイト/External Site Link)