Goals of IRIS

Key requirements in the design of IRIS are:

• Support an ontology-based knowledge store. We require the ability to model rich semantic structures that can capture every aspect of a user’s work environment.
• Support organization of personal knowledge assets. We should provide users the ability to organize their information resources in ways which suit individual needs (“just for me”) while maintaining semantic interoperability with other semantic desktop installations.
• “Real” enough to do daily work. In order to convince people to exchange their current mail program, web browser, or calendar for new, semantically enhanced versions, IRIS should offer a full-featured experience that supports all everyday needs – mail encryption, spam filters, calendar servers, synchronization with PIMs, embedded flash, etc. Rather than implement these features ourselves, we aim to integrate the most mature third-party applications available into IRIS.
• Implemented in, or able to easily integrate with Java. This requirement comes from the fact that many of the machine learning components we will include from the CALO project are implemented in Java. While IRIS is currently only supported on Windows platforms, it aims for cross-platform support.

Approach

The main elements of the system are aptly described by its own acronym:

• Integrate: IRIS harvests and unifies the data from multiple, independently-developed applications such as email (Mozilla), web browser (Mozilla), file manager, calendar (OpenOffice), and Chat (Jabber).
• Relate: IRIS stores this data an ontology-based KB that supports rich representation and connection to the user's worklife. In IRIS, you can express things like: "this file, authored by this person, was presented at this meeting about this project".
• Infer: IRIS comes with a learning framework that makes it possible for online learning algorithms (e.g. clustering, classification, extraction, prioritization, association, summarization, various predictors) to plug-in and reason about the rich data and events presented to them. In addition to learning through observation of user activity, CALO's learning algorithms have access to interface mechanisms in IRIS where they can get feedback from the user.
• Share: The knowledge created in IRIS by the user and by CALO will eventually be made sharable with selected team members. Currently, the ability to share content across IRIS users is a future capability.

IRIS will help move the research community from focusing largely on processing offline data to creating algorithms that work online, in real time, incorporating and adapting to "in-the-wild" user feedback, and that have access not only to core data types (e.g. a mail message), but additionally the rich context in which that data lives (e.g., the author of the email plays this role on this project and has these upcoming meetings about these task).

Status

In spite of an advanced version number, IRIS is a research platform, and very much a work in progress. If you have low expectations, you will be impressed! If you expect a commercial-level product, you will be disappointed. With the help of the open source community, we intend to continue improving IRIS and taking it in new directions.

Currently, IRIS runs only under Windows XP. Most of the code is platform-independent, so other operating systems could be supported in the future. 

Release

License IRIS is released as open source under the LGPL.

Background

By "augmenting human intellect" we mean increasing the capability of a man to approach a complex problem situation, to gain comprehension to suit his particular needs, and to derive solutions to problems. --Douglas Engelbart, "Augmenting Human Intellect, 1962

In the early 1960's, J.C.R. Licklider began the augmentation program at DARPA in concert with the existing artificial intelligence (AI) program. The augmentation program funded much of the work of Douglas Engelbart at SRI. In recent years, the AI program and the augmentation program have joined forces in the PAL project, which funded the Cognitive Assistant that Learns and Organizes (CALO) project at SRI. IRIS provides user interface, inferencing, and knowledge storage facilities to CALO.

Further reading

Adam Cheyer, Jack Park, Richard Giuli, "IRIS: Integrate. Relate. Infer. Share", Workshop on the Semantic Desktop: Next Generation Personal Information Management and Collaboration Infrastructure, ICSC 2005, Galway, Ireland (November 6, 2005).

Contributors

SRI International: Adam Cheyer, Jack Park, Rich Giuli, Colin Evans, Steve Hardt, Leslie Pound, Girish Acharya, Jim Carpenter, David Dunkley, Ken Nitz, Ayse Onalan, Mark Gondek, Talia Shaham, Julie Wong, Chris Brigham, Jason Rickwald, Ken Doran.

Radar Networks: Nova Spivack, Jim Wissner, Rich Schiavi, Chris Jones.

Open Source Contributions from the CALO Research Community: Andrew McCallum (UMass), Alon Halevy (UWash), Michael Jordan (Berkeley), and others.

Contact

Feel free to contact us with comments and questions!