Our data science team has built the RiskGenius Platform to support the integration of our insurance-specific algorithms and machine learning tools into your policy automation solution. The Platform incorporates the latest in Artificial Intelligence to structure, categorize and deliver policy data. Our software development team then delivers up your solution in a simple, elegant user interface depending on the selected RiskGenius Modules.
We realize that there is interest in how we have formulated the RiskGenius algorithms. We are pleased to describe basic information about our algorithms without revealing business trade secrets.
Clausematic is the name of the overall RiskGenius algorithm. Clausematic includes numerous Elements and Sub-Elements used to review an insurance policy, classify the policy language and deliver the most relevant results to users in the appropriate workflow. We also use distributive semantics to evaluate these Elements.
Elements & Sub-Elements
RiskGenius evaluates a number of Elements, and also tracks Sub-Elements within each Element. The Elements include:
o Policy Metadata
o Document Metadata
o Clause Data
o Character Styling
o Text Block
Just like machine learning is only a component of Google’s Hummingbird algorithms, machine learning is also only a component of RiskGenius’ Clausematic algorithms.
Also, please note that the weight of these Elements depends on how we are using the Clausematic algorithms. For example, in one instance, we may use the Clausematic algorithms to parse clauses in a policy. Alternatively, the Clausematic algorithms may be used to categorize or deliver content via our Genius Search. Each of these use cases requires a different algorithm configuration.
The Clausematic algorithms have and will continue to evolve as we learn more about insurance policy language and as policy language and technology changes. The following is a history of our algorithms.
Classification Project 1.
During this project, we started with baseline Elements to identify and categorize policy language. This project focused on one limited Line of Business and previously constructed set of truth data (provided by the customer).
During this project, we layered in additional Elements and Sub-Elements to identify and categorize policy language across all Lines of Business. From this project, we learned to track and rely upon a variety of new Sub-Elements depending on the Line of Business.
Scoring Project 1.
This project involved comparing and scoring policy language from one sub-category. From this project, we configured our algorithms to score similar clauses to identify subtle differences between each.
Based on previous learnings, we then evolved Clausematic to include more Sub-Elements to account for policy language differences.
Scoring Project 2.
Similar to Scoring Project 1, this project involved comparing and scoring policies. This project was used to identify identical or most similar documents. From this project, we created a workflow to allow RiskGenius users to quickly match a document against other documents stored in our library.
We are currently working on this project. This project focuses primarily on deriving classifiers of Sub-Elements being fed to our machine learning application.