Deployment and Usage Scenarios

The EKG has 200 million entities, 1 billion attribute value pairs, and 200 million relations. It takes few hours to extract entities, attribute value pairs, and relations from data sources. The EKG is rebuilt once a month to incorporate newly added enterprise data. There are multiple user cases of the EKG:

  1. Finding an enterprise’s real controllers

  2. Innovative enterprise analysis

  3. Enterprise path discovery

  4. Multidimensional relationships discovery

For the first use case, we want to know who the real decision makers are. Generally the person who owns the biggest equity share is the real decision maker. However, a person can control a company indirectly as shown in the figure below. We can develop an algorithm that traverse the KG to find the real controllers. We do this by calculating the equity shares of all the shareholders of the target company recursively until the shareholders are natural persons (real controllers). We then multiply all the equity shares on the recursion path to determine the total equity share of the person. The person with the largest equity shares is the real controllers of the target company. Not only does this graph algorithm allows us to identify the real controllers but it also allows us to identify the control paths as shown by the different investment structures in the figure below.

The second use case is innovative enterprise analysis. Ideally we want to find new innovative companies to invest in. To do this, we can use our EKG to analyse how many patents does a company hold. The users can provide the field / area they are interested in and they EKG can return a list of companies that owns the most patents in the chosen field.

Enterprise path discovery allows companies to discover paths to reach new customers as well as analyse how their potential customers can discover their competitors. This use case of EKG allows us to find the path between any two companies or persons. This is shown in the figure below.

The last use case is the multidimensional relationships discovery. Given two companies, there are many relationships between them. It could be competitive relationships, patent transferring, acquisition relationship or investment relationship. Our EKG provides visualised graphs highlighting the relationships between companies.

Related Work

DBPedia and YAGO are general-purposed cross-domain KGs. Other related work covers areas such as:

  1. D2R

  2. Information Extraction

  3. Graph Database

Different methodologies have been researched to improve extraction of knowledge from lists, tables, and webpages. Another area of work is improving entity linking, the task of linking named entity mentions with referent entities in specific KB.

There are many graph databases such as Jena, Blazegraph, sesame, AllegroGraph, and Hexastore.

Ryan

Ryan

Data Scientist

Leave a Reply