With each passing year the Internet and its many services, data sources and general capabilities, improves and expands. And in improving, it changes with offerings of different interfaces, access methods and functionalities. The current “state” of the Internet/Web is replete with services upon which real progress in new capabilities are available for research, business optimization, data science, and more. This democratization has sprung new inventions with ideas shared, mixed with other ideas and inventions produced at stunning rates. The pace of innovation has never been greater. The place where we are now, along this maturity and innovation curve, is exciting in one particular aspect that catches my (and ours at DeepSee) attention. It is the use, and creation, of platforms and the ability to produce a platform offering a suite of services and data access to solve real problems and provide a view into previously unseen insights, correlations and actions that can be taken from these insights.
At DeepSee we think deeply about platforms and what a platform might offer for our customers and partners. We have a focus on using machine learning (ML) and deep learning and, in particular, ML models around Natural Language Processing (NLP) that will lead to Natural Language Understanding (e.g. what was meant in that document in addition to what the words were). We see NLP, and its variants, as key components to DeepSee’s ability to ingest documents and provide action and insight from them. The combination of NLP models for unstructured data AND more traditional ML models for structured data offer DeepSee customers a rich and fertile space to innovate from the base of services, insights, and capabilities of the DeepSee platform.
What is a platform?
The DeepSee platform analyzes and gives insights and actions from unstructured documents (read: text) and unstructured data from sources like audio files, email, social media artifacts, and others. These unstructured documents can be legal documents, financial instruments, insurance documents and contracts, and many others. From these documents, DeepSee has a set of ML models that analyze for bespoke solutions or activities (activities pertinent to the document ingested) and/or ML models that provide “insights” from an array of documents of similar types. These insights may include aggregate financial data or metrics, collections of similarities found within a document set or outliers from a prescribed or derived comparative sampling.
DeepSee has additional tools that help subject matter experts (SMEs) “teach” the DeepSee models the particulars of document types and relationships within or between documents that would be difficult or impossible to determine without SME input. The tool within the DeepSee platform that helps capture this “tribal knowledge”, or feature engineering, is called DeepSee Atlas.
In addition to the curated and trained NLP ML models mentioned above, the DeepSee platform also has a set of curated and trained ML models available for structured and/or dense data in many forms.
Lastly, the platform accommodates the inclusion of customer or partner provided ML models to add specific value or functionality, where appropriate. These customer-introduced ML models can leverage the many insights the platform already has obtained from a set of documents which provides additional value to introduced models to solve or explore a specific opportunity. For data scientists, the DeepSee platform simplifies many of the difficult steps in data preparation for ML modeling.
With this general overview, what is the DeepSee platform? For a deeper look and description request access to our White Paper on this topic. For purposes of this document, however, the DeepSee platform is loosely defined as:
- A suite of APIs to access the capabilities of the system,
- A data model, data lake repository, and APIs to gain access to data,
- Methods to add insights to a document’s descriptive schema, which we call the D3O, or DeepSee Document Data Object,
- An ever-growing collection of various machine learning models that DeepSee has built or curated that add value to handling complex and simple unstructured and structured data sources,
- Capabilities to extend the platform with customer, or partner, provided ML models,
- A series of bespoke solutions for industries or customers within industries that are somewhat specific (e.g. a reconciliation set of interfaces for OTC derivative contracts for capital markets),
- User interfaces and conventions for developers or data scientists to use or extend.
RPA and Discrete Solutions
The set of capabilities of the DeepSee platform accommodates a large swath of opportunities in Robotic Process Automation (RTA) of existing processes within companies. As we have demonstrated in the complex task of using the RPA techniques from the capabilities of the platform to add substantial efficiencies in reconciling complex OTC derivatives contracts with “truth” that was booked in our customer’s risk system, we see, and are working on, various other bespoke solutions utilizing the many capabilities of the platform. This is a virtuous cycle of growing platform capabilities combined with more bespoke RPA solutions leading to more desired capabilities from the platform which, when added, then opens more opportunities to use the platform to solve other bespoke customer challenges. We are only at the beginning but already can identify solutions and customers in capital markets, insurance, legal, government and others.
Come back and visit often as we highlight more and more features, capabilities, tools, and solutions that can help solve your difficult challenges. We would invite you to consider the richness of information, trends, corollaries, outliers, and what may be hidden groupings or insights within your dense document repositories. Discuss with us how we can help find efficiencies in your processing of these texts. Further, we invite you to talk with us about how we can open up the many capabilities in our platform to bring insight and efficiencies to your efforts.