iKnow. Q&A with Benjamin DeBoe
Q1. What is iKnow?
InterSystems iKnow is the InterSystems IRIS™ data platform natural language processing (NLP) technology. It is a unique solution for text exploration that enables customers to gain insight from unstructured data and use it to enrich applications through a “bottom up” approach that discovers concepts and relations within the text itself.
Q2. You just announced that iKnow is now available via open source. What are the benefits of such a choice?
Open source availability of InterSystems iKnow makes it easier for the public to embed the technology deeper into their applications, business processes and ML initiatives, regardless of whether those are running inside or outside of InterSystems IRIS. It also gives customers the ability to customize the technology at a foundational level, and tune and broaden our language models beyond what is achievable within the fixed release cycle of a proprietary product.
Q3. Insights can be derived from data using machine learning, natural language processing, and artificial intelligence. How does iKnow integrate with the rest of InterSystems IRIS data platform ?
When the iKnow technology was acquired by InterSystems in 2010, we put significant effort into embedding it deep within the InterSystems data platform. We ensured the highest possible throughput for iKnow processing and fastest possible response times when querying iKnow output for data stored in SQL tables, coming in through a business process in our integration engine or otherwise hosted on our platform. This remains fully available and supported for users of the InterSystems IRIS data platform. But now, the core engine doing the actual NLP work, is also available in a standalone flavor.
Q4. Who has been successfully using iKnow and for what?
One of our earliest adopters in the UK is still one of the most telling examples. They are in the publishing industry and deploy a powerful knowledge management system on InterSystems technology. They leverage iKnow in a variety of places as an accelerator for research and editorial work, such as automatically suggesting related past articles while an editor is typing, or summarizing articles for reuse in other channels.
Other users include a U.S.- based customer that uses our technology to expedite patient registry input, a major U.S. government agency that uses it for navigating patients’ medical history, and a major university in Belgium, which uses it for patient cohort selection.
Q5. Will you continue to support InterSystems iKnow?
Yes, InterSystems will continue to support InterSystems iKnow as part of the InterSystems IRIS platform for exploration and search use cases, and providing long-term support and maintenance for the open sourced projects.
Q6. What is InterSystems doing to engage the open source community to contribute to iKnow?
We’ll shortly be releasing a Python wrapper around the core engine that’ll make it easier to leverage iKnow for feature engineering as part of machine learning pipelines. ML and feature engineering is also the area most of our team’s efforts are focused on right now, which we hope to publish more about soon.
To solicit external contributions, we’re actively reaching out to some of the academic and other partners we’ve talked to in the past. Also, InterSystems launched its first programming contest for our developer community earlier this month, and iKnow may well figure into one of the next ones.
Q7. Where can you access the open source version of InterSystems iKnow?
You can access the open source version of InterSystems iKnow here.
Benjamin DeBoe
Benjamin is a product manager in the data platforms group at InterSystems, focused on the data management and analytics capabilities of the IRIS data platform. He joined InterSystems in 2010 as part of its acquisition of iKnow, a Belgian startup in Natural Language Processing. Before joining iKnow, he worked at Oracle for 5 years, helping customers take advantage of its data warehousing and analytics products in consulting and sales engineering roles.
Sponsored by InterSystems