In my thesis work I developed a software framework that supports contextualisation. Contextualisation is described in this work as a process consisting of three asynchronous activities. Contextualisation involves Identifying contextual information, Collecting contextual information, and Using contextual information. This framework can be reused to develop new components for identifying, collecting and using contextual information. The framework provides us with a reusable and extensible architecture that allows new interfaces that support further scenarios of information-seeking behaviour.

Identifying. Extract contextual information from documents the user views, monitor the user’s interaction and navigation patterns to extract further contextual information.

Collecting. Store the contextual information in a knowledge store that supports reasoning about the contextual information.

Using. Make contextual information available to the user.

This collection of activities has been labeled the ICU paradigm in this work. In order to apply this paradigm to developing tools that help people find information on the World-Wide Web, a software framework that describes these activities in terms of software components has been developed. This framework is referred to as the ICU framework.

Top-level ICU Framework Architecture

In the next few posts I'll be drilling down on the structure of each of those components. 

Completeness

Since an ontology will be developed to represent context in a domain, rather than just for a single application, the set of entities specified by the contextual ontology provided a useful reference for application developers working within that domain.

Consistency

The elements of an ontology allow us to refer to domain elements in a consistent way, reducing ambiguity and allowing for the construction of thesauri that allow the mapping of vocabularies between groups of specialists operating in the domain.

Formality

The rigour and structure imposed on the domain elements being described in a contextual model mean that assumptions about structure are minimised, and allow for the creation of standardised interfaces that work with the formal structure.

 Reusability

Since the contextual models are well defined and can be mapped to different vocabularies, a model created for one application can be used by other applications that are aware of the ontological structure. It is also possible to provide mappings between ontological structures to support this kind of reuse.

Interoperability

Ontologies can be implemented as interfaces rather than as just data elements, so they can be made available as standalone components that can be readily plugged in to other systems or applications. It is also possible to create various tools that utilise ontologies without having to impact on existing tools or systems that are in place.

The word ontology translates as the study of what exists. Ontology can be described as the study of being in so far as this is shared in common by all entities, both material and immaterial. It deals with the most general properties of beings in all their different varieties. As part of metaphysics, ontology is variously concerned with substances, properties, relations, events, times, places and states of affairs. These concepts have readily apparent parallels to the concepts that have been identified in this work as constituting context.

Ontologies are widely used in knowledge-based systems, and are increasingly being applied to web-based systems. In knowledge-based systems, ontologies provide a conceptualisation of a domain of knowledge. This work is interested in being able to express a conceptualisation of context in hypermedia systems.

Ontologies provide a way of modelling the concepts of context in a formal, structured way that is also readily implementable for integration with existing hypermedia systems. Having previously described context as a form of knowledge, I believe it is entirely appropriate to use a knowledge modelling technique to model context. The development of the Ontology Web Language as part of the Semantic Web project provides a way for ontologies to be made available in a form readily consumable by Web applications.

This “something more” is the context of the user. Context itself is a widely ambiguous term. A key researcher into the nature of context is Brenda Dervin. Dervin concludes that “there is no term that is more often used, less often defined, and when defined so variously as context”. She asks the question “What is context?” and provides a broad discussion of context in the social sciences. Some key elements relevant to hypermedia systems are the following.

1)      Knowledge is partial and temporary. When perceiving a message or an event, a receiver will only interpret as much of it as is relevant to his or her current interest. Therefore, what a receiver knows about a message or event depends on the context on which it is being received.

2)      The knower and the known are inextricably bound. Each person has a unique collection of a priori knowledge, goals and experiences that affect how a message is interpreted. Thus, what a user knows because of receiving a message depends on the nature of the user. The result of interpretation of a message is not constant. Because it depends on the knowledge, goals and experiences of the user, the interpretation will vary across users, and will even for the same user as their knowledge or goals or experiences change over time.

3)      Context is not usefully conceptualised as an independent entity It is neither meaningful, nor useful to identify an entity independent of the received message and label it as context. Context is more than just a set of related facts that alter perception of a message. Any such entity is an artificial construct that trivialises the impact of external factors on perception.

4)      Context is necessary as a source of meaning. Communication relies on common experiences and understandings in order that the interpretation of messages by sender and receiver are close enough that the ideas understood by the receiver are those generated by the sender. It is not possible to draw a neat box and put all the things that are context inside it and label the box as context. Context is not easily labelled as set of state variables associated with a message, or as a set of variables describing a user‘s state. These are possible representations of context in a system, but they are by no means the only possible, or sometimes not even the appropriate, representation.

Contextual information can come in any number of forms. For example, visual context, information context, user context are all forms of contextual information that affect how the user interprets the information component. By altering any of these peripheral inputs, an author can influence the user‘s interpretation.

For application developers it is useful to consider what contextual cues can be made available to the user of an information management application to assist them with their information management tasks. For content authors where a specific meaning is intended to be communicated by the content, it is important to ensure the contextual elements surrounding the message help convey the appropriate message. It is not sufficient to just supply pieces of information, the context of these information elements must also be considered.



TAM has been repeatedly shown to be a reliable instrument for predicting the adoption of a new technology for a particular task. The factors measured by TAM are Perceived Usefulness (PU) and Perceived Ease Of Use (PEOU).

Perceived usefulness is defined as ‘the degree to which a person believes that using a particular system would enhance his or her job.’ It is a measure of the extent to which a user believes a system would help them with the task they are trying to achieve.

Perceived ease-of-use is defined to be ‘the degree to which a person believes that using a particular system would be free of effort.’ It is a measure of how easy a user finds a system to use for the task they are carrying out. These constructs are measured by a survey that uses a set of established questions to collect user’s responses.

The questions used to measure PU are shown in Table 1. The questions used to measure PEOU are shown in Table 2. Responses are measured using a 7 item Likert scale ranging from Extremely Likely through to Extremely Unlikely.

The TAM Evaluation was performed using students from three Masters level subjects at the University of Technology, Sydney. These subjects were chosen because it is expected that students have a reasonable level of proficiency using the web to find information, and are also expected to use the web for study-related activities during the subject. Participation in the evaluation was voluntary and anonymous in accordance with the guidelines of the UTS Human Research Ethics Committee.

Participants were asked to download and install ISeeYou. They then used ISeeYou while they engaged in their regular information-seeking behaviours, whether study related or otherwise. Participants were then prompted to fill out the survey instrument, which was hosted on the Institute for Interactive Media and Learning’s Survey Manager application.

The results were collated and then analysed using a Factor Analysis and a Reliability Analysis. The scale items were found to measure the expected two factors, and the Cronbach’s alpha calculation for the factors indicated that the measurements were reliable. The medians for the factors are shown in Table 3.

The median Perceived Usefulness of ISeeYou is 3. From this it can be concluded that it is likely that ISeeYou is perceived to be useful. The median Perceived Ease of Use of ISeeYou is 3. From this it can be concluded that it is likely that ISeeYou is perceived to be easy to use. These results indicate that a tool built using the ICU framework is perceived to be useful for people performing information-seeking behaviours on the World-Wide Web. This result supports further research into the use of the ICU framework as an approach to developing tools for Personal Information Management and tools for contextualising web-based information.

The recognised starting point for hypermedia research is the Memex system proposed by Vanevar Bush in his famous paper ‘As We May Think’. Memex was proposed as a ‘supplement to memory‘, a device where an individual could store books, records and communications for fast and flexible recall. In addition to storing information, Bush proposed that the Memex be able to associate items. It is this ability to associate information, as well as to store and retrieve it, which is recognised as being a defining characteristic of hypermedia. One of the most remarkable aspects of Memex is that its proposal predated digital information storage. Bush proposed a mechanical system using information stored on microfiche and recording user behaviour using a stylus and phonograph system. Revisiting this is a useful reminder that the problems we face organising information are not purely due to the availability of the Web – they are human problems.

The first implemented hypermedia system was Augment/NLS. This system was part of Douglas Engelbart’s work to develop systems that would augment human intelligence. Augment/NLS emerged from the Bootstrap project which sought to improve the way people worked by supporting collaborative working with online technology. Engelbart demonstrated his system at the famous ‘Mother of All Demos’ in 1968, which was the public debut of technologies such as the mouse and the hyperlink. The video recording of the demonstration is available online. The technology developed in this project lead to the adoption of the mouse, the graphical user-interface and the World-Wide Web. The great significance of this work is the premise that technology be used to augment human activity. Technology is not the end goal, rather it should improve existing or create new modes of human behaviour.

Ted Nelson‘s Xanadu system was another early system that has had a lot of influence on the thinking behind current hypermedia systems (and is itself a current hypermedia system, being an ongoing work in progress). The term hypertext was coined by Ted Nelson to refer to documents that could be distributed online. In particular, it was important to Nelson that documents be reusable. Reusability in this case meaning that documents should be annotatable and quotable without altering the original document or comprising the copyright of the original author.

A number of operational advantages to using hypertext were identified in the early days of hypertext research by Jeff Conklin. It is interesting to consider how current systems perform in each of these areas when thinking about why they’re successful.

• Ease of tracing references.
• Ease of creating new references.
• Information structuring. Links provide a way of structuring pieces of information.
• Global views. The link structures provide an effective mechanism for summarising hypermedia information.
• Customised documents. The link structure can be used to present different views of information.
• Modularity of information. Information can used in different ways as part of different link structures.
• Consistency of information. Using links allows information to be maintained in one place.
• Task stacking. Several trails of information can be explored simultaneously.
• Collaboration. Link structures can be shared and information can be managed by multiple people.

For example, consider Facebook and Twitter in terms of the following list. In which areas do they excel? In which areas do they fall short? In the areas they fall short, what possible innovations could occur to address these shortfalls? Which innovations can be done by a third party to add value to the existing technology? Which shortcomings could be exploited by a competitor?

The World-Wide Web emerged from a significant body of work into how information could be organised to support human activity. As most information is now available online, the focus has naturally moved to online technologies. The problem space is by no means closed though. Despite (or maybe because of) Google, Twitter and Facebook, people still face problems managing information in routine tasks. Remembering these are people problems, and seeking to find innovative ways to develop new and apply existing technologies to solving these people problems should be our focus as web developers and authours.

The first component, identifyApp, is an implementation of an Identify component. It is implemented as a Browser Helper Object using C++ and ATL. It is loaded by Internet Explorer and monitors the user behaviour using the DWebBrowserEvents2 dispatch interface. Contextual information recognised by the identifyApp is encoded in a SOAP message and sent to the collectApp.

The second component is an implementation of a Collect component, collectApp. The collectApp component is implemented in Java. It publishes a set of Web Services using SOAP. These interfaces are used by distributed components to update and query the collection of contextual information. The contextual information is persisted using Prevayler. This results in a fast implementation as all the information is available in memory, but as the dataset grows the memory footprint becomes significantly large. The next iteration of the collectApp will use a lightweight embedded database solution.

The third component, useApp, is an implementation of a Use component. It is implemented as an Explorer Band for Internet Explorer using C++ and ATL. The user interface is provided using HTML. Contextual information retrieved from the collectApp is encoded using XML. The XML is run through an XSLT script to generate HTML that is presented to the user. This allows the user interface to be easily modified or customised as data elements are changed or added, and as new user interface (UI) techniques are identified as being useful to the user. The contextual information is retrieved by sending SOAP-based queries to the collectApp when changes in the content being displayed are detected, or when the user selects a new option in the Explorer Band.

In the first implementation it was assumed that the collectApp is running on the same machine as the identifyApp and storeApp. It is envisaged that collectApps can be shared by communities of users and that contextual information can be shared as a form of collective intelligence. This will be a focus of future implementations.

The usability evaluation showed that ISeeYou was perceived by users to be useful and easy-to-use when carrying out information-seeking behaviours. This result encourages further development of the tool. The implementation developed for the usability study was developed as a proof of concept and had a minimal feature set and was not designed to be robust or scalable. Additionally, a decision was made to only implement support for Internet Explorer. This constraint was imposed to make the implementation schedule manageable. There is no reason identifyApp and useApp components for FireFox cannot be developed, and this is another focus of future implementations. A final issue for future implementation is to decouple useApp’s from the web browser. This decoupling is readily supported by the use of SOAP for communication between components.

Future work on a production system will focus on providing a richer experience, integrating community contextual information, a broader set of identifyApps and useApps, and using a robust scalable implementation that can be rolled out across the Web. This will move ISeeYou from being a research tool to being a widely used tool that augments peoples information-seeking behaviours by providing contextual information that helps people select, search and filter information on the Web.

Information-as-Thing

An entity that contains information is considered information. Applying this concept, a web page, a music file, and a video clip are information because they contain information. The meaning extracted by different users will depend on their context and their information-need. In a similar vein it has been suggested that there is no difference between data and information and that in fact the two are just different names for the same thing. Data means literally (from the Latin) ‘the givens’. Data is nothing more than the things a user is given; in the case of hypermedia, text, audio clips, video clips and images. This concept appears to be what is most commonly meant when people discuss information.

Information-as-a-Property-of-Matter

The main point of this concept is that information can be viewed as a property of matter. In this case, matter is ‘things’ – things being physical entities, such as a video clip, or conceptual entities, such as a philosophy. This concept describes two types of information, subjective and objective. Objective information is a property of matter, while subjective information is a reflection of an individual‘s consciousness of objective matter. For example, objective information about a video clip would be its running time and its encoding format, while subjective information about the video clip would be its entertainment value. Information is said to be a fundamental property of matter and of consciousness. The information about things can fall into categories, and what someone sees as the information of an object depends on the categories of information they recognise or are interested in.

Information-as-Structure

In this concept, information is a set of data that has been organised a specific way. Structure is applied when an author organises a set of data, and structure is also extracted by the reader when they view the information. The reader’s structure may differ from the author’s structure. This is especially relevant for database driven systems where sets of information can be organised in to different tables depending on the user’s needs. For example in an accounting system the set of financial data for a company can be organised in to various reports for different functions within a company.  In a travel system, data about flights available can be organised based on criteria such as departure point, destination, date of travel and/or fare price. The information is only generated when the system is told to structure the information according to specific criteria. The growing field of information visualisation is another way of restructuring data using graphical techniques.

Information Concepts for the World-Wide Web

Based on the concepts discussed above, we can view information on the web in three ways.

1. Information is an entity on the web.
2. Information is the result of a reader interpreting an entity on the web.
3. Information is the result of structuring entities on the web.

There are many more concepts of information and there is certainly no definitive definition of information that captures all its nuances. As an author or developer, the reality is that all of these are true. Information researchers often speak of operationalising a concept where they apply a working definition for the current conversation. There are numerous information concepts, and there is no ‘correct’ concept. Rather, information concepts should be used to encourage and focus discussion when working in information-related fields. When working with information, looking at what it is you’re working with from various differing perspectives can provide deeper insight into the site being created or the application being developed.