Entity Extraction Is a Critical Technology for Developing Investment Intelligence

Gaining Investment Intelligence Is a Critical Part of a Business Strategy

Investment intelligence is an essential aspect of modern financial decision-making. Its goal is to discover valuable insights to guide investment strategies and improve financial forecasting of specific investments and to do so as quickly as possible to stay ahead of the market.

Investment intelligence entails deriving valuable information from vast amounts of data and diverse sources, including:

• • News articles
  • Press releases
  • Earnings reports
  • Financial statements
  • Social media
  • etc.

The kinds of information any company needs to be aware of in its business sector include:

• • A critical partnership among companies in a hot area
  • The announcement of an antitrust lawsuit
  • The replacement of an embattled CEO
  • The backlash of the public against a company’s actions
  • etc.

In a specific industry, a company may have other requirements for investment intelligence. For a pharmaceutical company, for example, it also needs to know about:

• • Promising clinical trial results for a new drug being developed by a competitor
  • Changes in FDA regulations
  • Moves by governments to control pharmaceutical prices
  • etc.

The main challenges, however, are that the data in question is unstructured, the data volumes are huge, and insights are needed in a timely manner to beat a fast-moving financial market.

Extracting Relevant Data for Investment Intelligence with Entity Extraction

One powerful tool for gathering investment intelligence is Entity Extraction. It’s an AI technology that, in its basic form, helps identify concepts such as names of people, organizations, dates, and locations in unstructured text data. In the more advanced form, it can uncover relationships and events that entities are involved in, which is critical to investment intelligence.

By applying Entity Extraction to unstructured data, market analysts can more efficiently identify relevant information, gain deeper insights into market trends, and forecast investment performance.

Entity Extraction processes large volumes of text at scale to identify critical information. For instance, Entity Extraction can process press releases, news, and social media to identify important developments that can affect the performance of a company:

• • Adverse information on a company or its executives, such as:
    • Lawsuits filed against the company
    • Bankruptcy filings
    • SEC fines
    • etc.

Or any other event that could make an investment in it inadvisable:

• • Geopolitical events such as political unrest, coups, demonstrations, natural disasters, tariffs, etc.
  • Gain or loss of major customers
  • Changes regarding suppliers and partners
  • A delayed timeline for a future product release
  • Regulatory changes
  • M&A activities such as a planned acquisition or merger
  • Leadership changes
  • Changes in main investors

Entity Extraction can process various sources to identify the organizations and individuals involved, as well as relevant dates that may impact investment strategies.

Entity Extraction enables multiple analytical tools and types of analysis. For instance:

• • Real-time alerts for when certain companies or individuals of interest are mentioned in the context of adverse events in news stories, providing immediate access to critical updates
  • Trend monitoring such as for public sentiment towards a company’s products or brand
  • Due diligence to research a company’s links to other companies (partners, subsidiaries, suppliers) and to identify any associations that the company or its executives may have with questionable organizations or persons, and any crime or allegations they may be implicated in.

How Entity Extraction Works

Entity Extraction is an AI-based technology that recognizes key concepts in unstructured text. At its most basic level, it identifies:

• • Names of people, organizations, and places
  • Numerical amounts (e.g., money, etc.)
  • Dates
  • Times

Advanced Entity Extraction identifies more complex concepts such as relationships and events. It can analyze, for example, news reports on overall M&A activity in an industry, which would provide investment companies who track that industry with valuable insights. For example, consider the following unstructured data from a media report:

“New York Energy and Southern States Power have agreed to a $10 billion stock merger which will create the biggest investor-owned utility in the country. The merger is expected to be finalized by December 31, 2025.”

Advanced Entity Extraction will process this and produce structured output like the following:

MERGER_EVENT

CORPORATION1: New York Energy

CORPORATION2: Southern States Power

DATE_TO_BE_FINALIZED: December 31, 2025

In order to produce this output, Entity Extraction first recognizes the names of the two corporations as well as the date. For the names, Entity Extraction does not rely on long lists of known names. It uses the context to identify names and what kind of entities they represent. This is usually termed dynamic recognition, and it is one of the great strengths of Entity Extraction. Entity Extraction then recognizes the relationship between the companies through analyzing the surrounding sentences. In this instance the two companies are merging by the end of the year.

Here’s another example of Advanced Entity Extraction finding and structuring an event. Suppose a company contemplating a partnership with another wants to monitor the business press. Advanced Entity Extraction discovers the following negative event about the potential partner:

“XYZ Corporation was hit with a $5M fine Thursday imposed by the SEC.”

And here is the resulting structured event:

FINING_EVENT

REGULATORY_AGENCY: SEC

CORPORATION: XYZ Corporation

FINE_AMOUNT: $5M

TIME: Thursday

One of the strengths of Entity Extraction is that it is very fast, so it can analyze very large amounts of structured data very quickly. Since the output is structured, it can easily be stored in a database to be utilized by various analytical tools. It can also be aggregated – event ontology labels like MERGER_EVENT or FINING_EVENT are standardized and will not vary like natural language sentences, so it’s easy to develop an aggregated view of such events over time and across an industry, thus supporting trend analysis and alerts.

Sentiment Analysis Enables a Company to Gauge Market Sentiment

In addition to name, relationship, and event extraction, some Entity Extraction tools also offer an additional capability, Entity-Based Sentiment Analysis, that gauges market sentiment about many aspects of a company:

• • Brand
  • Product
  • Pricing
  • Executives
  • Customer Support
  • etc.

It builds upon Entity Extraction, but in addition it focuses on the sentiments expressed about the entities being extracted. Entity-Based Sentiment Analysis can process all mentions of a company in social media, company reviews, and other forums such as Reddit and Quora, and can calculate whether sentiments expressed towards it in a post are positive, negative, or neutral.

Simplistic versions of Sentiment Analysis can identify positive or negative sentiments in a text or a sentence by looking for sentiment-bearing words such as “excellent” or “awful.” They do not, however, capture what the sentiment is about — the so-called target of the sentiment — because they do not understand the context in which those words occur.

By contrast, Entity-Based Sentiment Analysis doesn’t just pick out sentiment-bearing words and apply them to a whole sentence. It explicitly identifies which entity is being characterized by the sentiment. It goes even further. Entity-Based Sentiment Analysis can identify when multiple sentiments are expressed in, say, the same post about different aspects of a company:

“ABC Corp.’s products are very expensive, but its customer support is excellent.”

Entity-Based Sentiment Analysis would produce the following two outputs:

SENTIMENT_TYPE: Negative

SENTIMENT_PREDICATE: very expensive

TARGET: ABC Corporation

ASPECT: products

and

SENTIMENT_TYPE: Positive

SENTIMENT_PREDICATE: excellent

TARGET: ABC Corporation

ASPECT: customer support

All output is in this structured, uniform format, which can be stored in a database. Analysts can search for all positive and negative mentions of a company and the specific aspects that sentiment is directed to (e.g., customer support, a specific product feature). These mentions can be aggregated and displayed on a timeline to see if a company is trending positive or negative.

Summary

Entity Extraction has become a powerful AI tool in the arsenal of market analysts and investment companies. By automating the extraction of key entities, relationships, and events from unstructured data sources, Entity Extraction enables investment professionals to more efficiently track important developments and identify emerging trends. Sentiment Analysis allows analysts to gauge a company’s reputation with the public.

As financial markets continue to evolve and become even more data-driven, leveraging Advanced Entity Extraction and Entity-Based Sentiment Analysis for investment intelligence is increasingly crucial in making informed, data-backed investment decisions.