Written by

InterSystems Corporation
InterSystems Official Benjamin De Boe · Sep 21, 2023

Deprecation of InterSystems IRIS NLP, formerly known as iKnow

InterSystems has decided to stop further development of the InterSystems IRIS Natural Language Processing, formerly known as iKnow, technology and label it as deprecated as of the 2023.3 release of InterSystems IRIS. InterSystems will continue to support existing customers using the technology, but does not recommend starting new development projects outside of the core text exploration use cases it was originally designed for. Other use cases involving natural language are increasingly well-served using novel techniques based on Large Language Models, an area InterSystems is also investigating in the context of specific applications.

Customers with questions on their current or planned use of InterSystems IRIS NLP are invited to reach out to their account team, or get in touch with @Benjamin De Boe 

The open-source version of the core iKnow engine, packaged as a Python module, can be used independently of InterSystems IRIS and will continue to be available.

Please note the InterSystems IRIS SQL Search feature, also known as iFind, is only partially affected. Only the Semantic and Analytic index types make use of the iKnow engine and therefore are deprecated. All other functionality and index types are not affected by this decision, and continue to be the recommended choice for applications requiring a flexible and high-performance full text search capability.

Comments

Dmitry Maslennikov · Sep 21, 2023

I think it is worth mentioning, that the alternative is possible

The latest technologies based on Vectors, may help to replace it, with even more capabilities.

0
Tirthankar Bachhar  Oct 31, 2023 to Dmitry Maslennikov

Can you please share some more details about it?

0
Dmitry Maslennikov  Oct 31, 2023 to Tirthankar Bachhar

Well, I did some notes about Vectors in my article, about the project I tried to implement.

Basically, it's possible by using neural network based algorithms calculate vectors for any texts, index them in the database, and search using vector search for any text query. The results in this case will not find texts which are exact to the search query, but with using similarity, the closest to the query. And it can be used with mostly any language, types of the texts, files and so on, even pictures, or videos.

0
Benjamin De Boe  Oct 31, 2023 to Dmitry Maslennikov

FYI - we plan a native datatype for vector content and fast similarity functions in 2024.1, with deeper integration planned for the next few releases. Stay tuned...

0
Enrico Parisi · Nov 10, 2023

Does this deprecation include iFind text search?

Enrico

0
Evgeny Shvarov  Jan 16, 2024 to Enrico Parisi

If I'm not mistaken, iFind is a part of 'Basic' text search which and according to documentation it is not deprecated 

0
Benjamin De Boe  Jan 16, 2024 to Evgeny Shvarov

indeed. please see the bottom paragraph of the post above, and feel free to reach out to me directly if you have any specific questions about your use of the technology.

0