Originally from Sydney, Australia, Pete Fitzsimmons
has an extensive background in banking and financial services. But in the last 10 years he’s gotten a much larger exposure to technology with a compliance and risk management business, before deciding to start Contexture
with partner Craig Lovell
We caught up with Fitzsimmons for this new member profile.
SIPA: How long has Contexture been around?
Pete Fitzsimmons: We’ve been in business for a little over a year now. But we were working on strategy and developing materials for the last 3-4 years.
Tell us about your partner.
Craig and I have worked together for over a decade, he’s a first-class software architect and computer scientist with over 20 years of experience designing and building mission-critical information systems in the financial, biotech and intelligence industries. He holds a BS and MS in Logic and Computation from Carnegie Mellon University.
What’s the gist of what Contexture does?
We bring together cutting-edge technology in natural language processing, machine learning and artificial intelligence to add value to our customers through greater insight and understanding of the unstructured content that impacts their business. We can work with anything that can be converted to text. Our focus is bringing actionable insight to complex and voluminous content challenges across any industry vertical.
And for SIPA members?
With SIPA members, the information they’re providing is highly specialized like best practices in aging healthcare or wealth management for millennials or plumbing supplies, and they each have unique audiences. That works for us because we’re industry and content agnostic; our platform curates, indexes and maps to any forms of content including text, audio and video extracts—whether they are from open source or paywall media and publishing to proprietary corporate repositories or Contexture’s own legislative and policy content or media feeds.
And then you apply your systems?
Yes. By applying our indexing and concept mapping, along with semantic network, our machine learning recommendation engine and clustering algorithms, we continuously identify changing and new information extracts that are important to our customers across international, national and local markets.
What are some of the markets you serve?
We serve the legislative and policy market with our Bill Watch solution, and we also have implementations starting in media, entertainment, membership and information publishing; education technology; legal services (contract management, legal document analysis, etc.) and the general enterprise content market.
Can you explain machine learning a bit for us?
Artificial intelligence is like an overarching umbrella of this area of technology. There are many definitions of machine learning—supervised and unsupervised, etc. It basically means providing a set of instructions to an engine to execute a process, and using training sets to enable the engine to learn how to perform. In our case, the objective is for the machine to identify “what a particular piece of content is.” We are then able to take an untagged set of text content and apply classifications to it that will convey information with meaning and context. So through machine learning, we’re able to propose and recommend to users/subscribers the specific pieces of content they might have a high degree of interest in.
And then that information only increases the more users interact?
Yes, as users interact, we’re able to learn from those interactions at a deeper level. What are they really into? It’s continuous learning. Our technology will combine with publishing platforms of SIPA members. We can enrich their users’ experience by providing highly relevant content based on their user profile and continued interaction with the publisher’s content. We can then also separately inform the publisher on what kind of content would be relevant to which subscribers and what areas they might not have identified yet.
Your timing is good with everything happening now.
Yes, the technologies are evolving quickly including in-voice recognition and translation, which are getting really powerful. We’re at a space where we can get high-quality transcription of audio, podcasts, voice extracts from video— and we can tag where passages of text are being spoken that are highly relevant to you, so you can just go to the 5-minute segment that you want. Our technology can do that too.
That’s a real time-saver.
We can also classify the content by topic areas because our content index is contextual and not just keyword search. We’re looking at context in and around what the reader is trying to drive towards instead of a runoff of keywords.
Sounds like a good value proposition.
Our customers have been very excited about the possibilities; they value our capability to make NLP and machine learning technology easy to understand and implement—and most importantly affordable.