Myriad: A Platform for Contextualised Information Retrieval and Delivery

About Myriad

Knowledge workers are increasingly dependent upon the availability of the right information in the right form for their daily work. Alarmingly, they spend 15% to 30% of their time looking for information – and are unsuccessful up to 50% of the time.

Despite what you read in the media, Google is not the answer to everything! Search only provides part of the solution for knowledge workers – and there’s often a big gap between providing a list of search results, and actually helping users to efficiently complete tasks. Limitations of the results delivered by current search engines include:

• Providing the user with too much information (not all of it relevant);
• Requiring the user to manually combine and aggregate the information;
• Requiring the user to manually determine how information from multiple sources relates to other information; and
• Requiring the user to organise the data to suit their task.

The information needs of knowledge workers are largely driven by the context in which they make their decisions. This context is dynamic. This context is also evolving. When one information need is satisfied, another is likely to emerge.

In the future, information access systems must be able to:

• Track this dynamic and evolving context;
• Exploit this context to retrieve information from appropriate sources; and
• Deliver that information in a form that is appropriately tailored to the user’s context.

The challenges for such an information access system come in both retrieval and delivery. First, the system must be able to obtain information from heterogeneous databases and document collections – so a unified interface or integrated platform for accessing these knowledge sources is essential. Second, the system must be able to aggregate and deliver the information retrieved from these various sources in a manner that is useful to the user, suiting his or her needs at that time.

Myriad Overview

Myriad is a flexible and configurable software platform, designed and developed by the Information Engagement Stream, in collaboration with Professor Keith Vander Linden from Calvin College, USA. The Myriad platform supports the construction of information access systems that perform context-guided information retrieval and delivery. Myriad attempts to make the best use of the advantages of the two main approaches to information delivery: the delivery-driven approach of adaptive hypermedia systems and the retrieval-driven approach of information retrieval systems (such as Panoptic).

The Myriad platform supports the construction of applications that address current search engine limitations by:

• Providing only information that is relevant and appropriate to the user and their current task;
• Automatically integrating the results of a number of queries across multiple data sources;
• Organising information to facilitate easy comprehension; and
• Customising the delivery of information to suit the user, task and device to help the user better perform their tasks.

Myriad exploits a variety of concepts and techniques from the field of Language Technology to answer complex requests for information, i.e., requests that are satisfied through aggregating the result of a number of single queries. Myriad applications are able to find and extract appropriate information from different data sources, such as databases, web pages or other textual documents, carefully constructing the required queries such that they relate to each other in specific, structured ways, and the information they return contribute to the user’s information need. This allows the information retrieved to be integrated and organised, and ultimately, to be presented to the user in a structured presentation that is tailored to suit the information needs of their task, and optimised to suit the capabilities or constraints of their device and working environment.

Myriad Architecture

The core of the platform is the Virtual Document Planner (VDP), which is a goal-decomposition-based planning engine. The resources which control VDP are declarative (XML) plan libraries, which constrain and control which information is retrieved and how that information is organised and presented.

Supporting the VDP is a set of context models which provide an explicit and dynamic representation of the user’s current context. These context models contain information about the user, the domain, the user’s tasks, the environment, and current and previous interactions with the system.

The Myriad platform is also extensible, providing applications with a:

• Retrieval API to plug in various information retrieval modules – e.g., a Google retrieval module, a Panoptic retrieval module, an IMAP mail server retrieval module, an SQL database retrieval module etc.
• Delivery API for performing specific types of delivery – e.g. ,graphing data, tabulating data, synthesising speech, canned text output, template fillers etc.

Through these APIs, applications can define and use any desired retrieval and delivery modules.

In addition to providing an application platform for constructing applications to perform contextualised information retrieval and delivery, the Myriad platform is shaping up as a very flexible platform for evaluating novel information retrieval and delivery mechanisms. Some preliminary evaluation of ideas and prototypes has been performed. The next step is to further investigate what factors significantly help users and what factors don’t, so that – in the not-too-distant future – when you electronically search for an answer to a complex question you get timely, accurate, useful results every time.

Printed version

last updated 23/09/05
Andrew.Lampert@csiro.au

Contact:
ict@csiro.au