MVP: Finding the Most Valuable Posts in Financial Social Networks
Feng Liu, Dingding Wang · 2019
With the availability of the Internet, financial social networks such as StockTwits and SeekingAlpha are emerging to provide opportunities for investors around the world to gather and share their experiences and opinions on the stock market. When these social networks become more and more popular, millions of users join and post huge amount of posts every single day to discuss all kinds of topics related to the market, e.g., the news and events about companies, when to buy and sell shares, which stocks to buy. If a user has a question related his/her investment, he/she needs to browse all the posts returned by searching the stock symbols, however due to the high volume of the posts he/she will drown in the information and hardly to find the answers he/she wants to know. In this paper, we propose to apply a language model with dynamic pseudo relevance feedback to obtain relevant posts in StockTwits, so that people can quickly and easily grasp the ongoing events of their interested financial news. Experiments and a case study on the StockTwits dataset demostrate the effectiveness of the dynamic pseudo relevance feedback.