News Summary

Understand how News Summary works

Overview

News Summary allows you to instantly search through thousands of news sources to find information about a selected security within a specified timeframe. Boosted Insights then sorts this information into topics, giving you an immediate and holistic idea of current events that may impact the security’s price movement. 

Though the News Summary process is completely automated and takes seconds to generate results, you can read below to understand everything that happens on the backend during this process. 

Visualizing News Summary topic creation

The flow-chart below shows a visual representation of how News Summary topics are created and sustained over time:


How it works

News Summary starts by identifying news articles that mention a selected company within a specified timeframe. It then verifies that each of these articles is relevant to the company and their security’s price movement. Once it’s sure an article is relevant, it aims to sort it into a news topic. 

Sorting an article into an existing topic

First, Boosted Insights attempts to sort the new article into an existing news topic. We use a classifier to identify possible topic matches for the article. It then generates either a successful match or an unsuccessful match.

Successful match
If an article relates to an existing topic, it gets added to the topic. Boosted Insights then adjusts the topic’s information and associated impact where necessary. 

Unsuccessful match
If a new article doesn’t fit into a preexisting news topic, you can expect the following:

  1. The article will enter a pool of other articles with no assigned topic
  2. Boosted Insights may create a new topic

Articles with no assigned topic

Both new and old articles without an assigned topic live in a pool of unassigned articles. Within this pool, we use the same clustering algorithm in an attempt to cluster articles into groups of two or more. 

Articles that are not clustered with at least one other remain in this pool of unassigned articles. If an article never gets clustered with another, we consider the idea of the article as incomplete and not show it in the News Summary.

Creating a new news topic

If at least two articles within the unassigned pool show enough parallels, Boosted Insights clusters them into a new news topic. This topic then appears in the security’s News Summary based on its ranking compared to other topics for the selected security.  

Similarly, if an unassigned article and an assigned article show parallels that don’t match with an existing topic, we will create a new topic. 


Topic rankings

We display topics in a ranked order based on the number of Related Articles within your selected timeframe. For example, a topic with 50 related articles would rank significantly higher than a topic with 10 related articles.

Top Sources and their impact on topic rankings

Articles marked as a Top Source come from one of the sources we’ve marked as trustworthy finance sources. The Top Sources percentage you see for each topic represents the percentage of articles related to the topic that come from top sources. 

A single article coming from a Top Source will count twice for topic ranking purposes. This is because sources identified as Top Sources are more likely to document things that impact a security’s price movement than those that are not considered top sources.

Topic metadata

After the creation of a topic, Boosted Insights generates metadata using news surrounding the topic. This includes labels, descriptions, and an associated impact (positive, negative or neutral) within the specified timeframe. We refresh this metadata once every day to include any new news that may have come in.

Positive/Negative Impact
Boosted Insights determines the topic’s positive or negative impact to the security’s price by analyzing the topic summary. From there, it determines whether the sentiment is positive, negative, or neutral.

We store all historical topic information so that we’re able to generate an accurate representation of topic popularity on the Stock News Interest Over Time graph.