Need for Multilingual Stock Sentiment Analysis.
Multilingual Stock Sentiment Analysis. Today, data is changing people’s lives. In the financial industry, hundreds of millions of pages are produced daily in SEC and global fillings, broker research, conference calls, investor’s relations presentations, social media, real-time news and press releases … across many languages. It is physically impossible to read all this.
So, the news analysis does matter. There is a need to get cutting-edge technology to analyze the humongous flow of information produced by the financial industry. Technology allows you to truly leverage this big data by uncovering low signals, alerting you in real time, cutting the noise and feeding your systems. Sentiment analysis applied to stocks trading for instance helps detecting Market patterns and monitoring market risk and so on to help make smarter decisions.
In the process of news analysis, the first step is collecting sources. Reliable information sources are really significant. And different sources lead to varying widely different analysis results.
In this article, we are analyzing an Indonesian company : Pelindo data from 26/2/2014 to 14/5/2014 as an example.
Most of the content we found is in the local language: Bahasa Indonesian. There is very few content in English, outside of Jakarta Post.
We quite easily detect 2 cases:
- case 1: A wave of protests/strikes, with a continuous (low) flow of articles, even starting from December 2013 (not shown in the graph)
- case 2: The alleged corruption case which found a deep echo in the local press, but almost nothing in the international one.
In fact, most of English news are translated from Indonesian news. Thus, we can easily find following conclusions from this example:
- English contents have limitations in this situation.
- Compared with English, the local press in Bahasa can offer information to us more rapidly.
This Indonesian stocks sentiment analysis is a good example of the added value of a multilingual stock sentiment analysis framework.