I. Introduction
News is a major factor that drives changes in a stock price. Financial investors make decisions based on the available data considering the probability that a new piece of information will have an impact on the market. They analyze recently published news when judging equitable market prices. The information about a company's fundamentals, the activities in which a company is involved and the expectations of other market participants are incorporated into the news articles. The release of major news items often produces speculation among traders which results in price movements [1], [2]. With the development of the internet, real-world trading applications provide a huge amount of textual data with a tremendously increased broadcasting speed [3]. A number of efforts were made to create an automated framework that analyses a large amount of textual data relevant to a particular stock, extracts relevant information and uses it for financial forecasting. A strong relationship between the fluctuation of a stock price and the publication of a related news article has been shown in previous research works [4]. Many researchers have employed or expanded the existing data mining approaches to investigate how news can impact traders' actions and hence affect the stock price [5]–[7]. In the reviewed literature, news articles are considered to be relevant to an analyzed stock based on a predefined criterion, then retrieved from a large collection of documents and used for further analyses. Researchers tend to define rules for selecting relevant news items from a whole data set and then analyze every article as having the same potential impact on price changes. So far, no previous work has been done in combining stock-related and industry-related news items with learning the degree of influence for these two categories of news.