Entity Extraction for Strategic Investment Decisions

December 20, 2019 | Entity Extraction, Geotagging, Intelligence Analysis, Risk Management

Entity Extraction for the financial markets

Strategic Investors Need to Vet Their Investment Targets

Institutions and companies making world-wide investment decisions need to have deep and accurate knowledge of the state of their investment targets. Has a company under consideration for investment experienced any adverse events such as lawsuits? Has one of its principals ever been associated with shady business partners or had legal problems? Are any unusual events affecting the area where the company is located, such as natural disasters or political demonstrations?

Information about these and many other factors will naturally have an impact on an institution or a company that is being considered. It is imperative that there be a good way to gather and analyze such information.

Entity Extraction is an advanced technology that can help. A great deal of relevant information is contained in the world’s media, but unfortunately the world’s media is very large, so having a team of human analysts combing through all that information is not really practical. Entity Extraction automates the identification and extraction of relevant information from unstructured data. It can be used to inform decision makers of critical events and other information regarding an investment target.

How Entity Extraction Supports Informed Investment Decisions

Entity Extraction, aka named entity recognition, starts with extracting basic building blocks of entities from unstructured text data such as people, companies, organizations, countries, monetary amounts, etc. Entity Extraction finds not only names that might be known to the analyst seeking relevant information, but it finds all names, including those previously unknown to the analyst.

Once these names have been identified, they support, among other applications, an advanced semantic search capability: now an analyst can pose queries such as “Show me all documents with company names ranked by frequency.” The statistical frequency of extracted items can be analyzed over a very large document set. The co-occurrence of individuals with organizations and of organizations with other organizations can be analyzed looking for patterns that might affect an investment decision.

In addition to identifying named entities, more advanced Entity Extraction offers additional capabilities that produce a fine-grained understanding of information relevant to an investment decision. For a more detailed discussion of these capabilities, see the following:

In the following sections, we focus on how each of these contributes to improving strategic investment decisions.

Relationship Extraction Identifies Associations among Entities

Relationship Extraction goes beyond the extraction of proper names. It is a powerful means for finding relationships between entities, such as ABC Corporation being a partner or subsidiary of XYZ Corporation or a CEO X is associated with another person Y, which might affect an investment decision. It is superior to simple statistical co-occurrence as it indicates the precise nature of the semantic relationship between two entities. The relationship extraction ontology could include an extensive set of pre-defined semantic relationships that allow the manifold linguistic ways of expressing the same relationship to be normalized into a single structured output format.

This structured output can then be used for link analysis of the relationships among entities and provide analysts with a clear graphical representation of the information.

Event Extraction Identifies Events in which Entities Have Participated

Perhaps the most relevant type of extraction to investment decisions is Event Extraction. This extracts what happened (e.g., an indictment), along with the Who, What, Where, and When of the event (e.g., besides identifying an indictment, it provides fine-grained information such as the identity of the party handing down the indictment, the party who was indicted , the time of the indictment, and the place where it was issued).

Event Extraction is essential for monitoring adverse news such as lawsuits, bankruptcies, and arrests. Typical events that Event Extraction handles include:

  • Business developments relevant to investment decisions:
    • Mergers and acquisitions
    • Plant closings,
    • Appointments or resignations of C-level employees
    • etc.
  • Geopolitical stability indicators for where a company is located:
    • Ethnic conflicts
    • Political violence
    • Protests
    • Elections
    • Coups
    • etc.

Geotagging Also a Key Technology

Geotagging is a technology integrated with Entity Extraction that identifies and disambiguates place names extracted from unstructured data. For example, Geotagging figures out that a name Alexandria mentioned in the article is located in Egypt rather than Louisiana. It then assigns them accurate latitude/longitude values in order to support geospatial analysis. It refines and localizes events extracted by Event Extraction and allows them to be used by GIS software such as Esri. This is critical for good geopolitical analysis, and it provides a detailed geographic context for an investment decision. It indicates where in the world the events of interest are happening.


Entity Extraction is a valuable tool for companies who need to make strategic investment decisions. It empowers them to discover key data found in enormous quantities of unstructured data coming from many different sources in real time. It transforms this key data into structured data that can be used by other analytical tools.