A Taskonomy for the Semantic Web

Paper Title: 
A Taskonomy for the Semantic Web
Authors: 
Tom Heath
Abstract: 
The modalities of search and browse dominate current thinking about interaction with the Web. Given the Web’s origins as a global hypertext system, it is understandable that these document-centric interaction patterns prevail. However, these modalities alone are inadequate as a conceptual model of interaction with the global Linked Data space that is the SemanticWeb. Realising the full potential of the Semantic Web requires a fundamental reconsideration of Web interaction patterns in the light of Linked Data, and this renewed conceptualisation must drive the research agenda related to user interaction and the Semantic Web. This paper argues that a fundamental understanding of user goals and tasks is the appropriate perspective from which to approach the research and development of SemanticWeb applications. However, the Web in Semantic Web should not detract from the potential for cross-platform data interoperability enabled by the Semantic Web technology stack. In this context we propose a taskonomy of data- and object-centric user tasks derived from an analytical abstraction of existing research, not simply in the fields of Web search and browse but also email and Instant Messaging, that can help shape the direction of research and application development in the Semantic Web field.
Full PDF Version: 
Submission type: 
Other
Responsible editor: 
Krzysztof Janowicz
Decision/Status: 
Accept
Reviews: 

Review 1 by Manfred Hauswirth:
The paper argues that the current interaction metaphors on the Web and related systems are (too) document centric and too focused on search / retrieve and thus may be insufficient / inadequate / inefficient for the types of interactions and goals people are using them. To make this point the paper provides a nice and informative survey of recent work on how people interact with / use the Web and email. Unsurprisingly, these systems are used in ways never imagined by their creators (as most successful system which are around long enough) and this creates technical and usages problems. I like the survey but - being not from the field of user interaction analysis - I would have liked also a more direct comparison at the end of the discussion, e.g., a table listing classification criteria, pros & cons, new uses, etc. of the Web, email, etc. to make it easier to understand more obvious where the gaps / problems actually are.

On the basis of this analysis the paper argues that the current interaction models are suboptimal. I agree to this. To support this, I would have liked a more systematic disussion of these types of interaction in Section 3. URIs and Linked Data are means to an end. What do they buy me more concretely? I think this is quite straight-forward to demontrate but necessary to really make the argument strong.

I like the taxonomy in Section 4. It makes a lot of sense to me. However, I have 2 comments: I do not really see a lot of difference between Exploring and Evaluation, probably they can be unified. The same applies in my opinion to Asserting and Discussing. Discussing as described may be seen as a superset of Asserting.

Minor comments:

Reference to Broder, line 3 of 1.2 is missing.

[1] "Spring" is not the volume :-)

[3] misses the year

[6], [7], [16], [17] "," before page numbers

Review 2 by Mark Gahegan:
This paper presents a very approachable summary to the problems we face with legacy metaphors of the document as the means to both describe (search for) and be supplied with web content. It is an out-moded notion and the authors present clear arguments from the perspective of Web pages, email and IM to show that this is the case.

The paper is very well written, easy to follow, clear and concise. It presents an alternative taxonomy of web related tasks, as its title suggests, which is an augmentation and reworking of several existing approaches, but with a clearer focus on the intention of the user, rather than the technology, media or action. it is thus a deeper semantic, and probably carries more useful information--if we can figure out how to capture this information!

This is the only aspect I would like to see added...how do we recognise the intention, so that we can support it better? Are you imagining that these intentions form the interface. How is this idea to be operationalised. Adding some ideas about this would be very helpful to this reader.

From the abstract, agree that it is inescapably important to have an understand of goals & tasks. a friend pointed me back to Wittgenstein's Zettel recently, the first note is on the primacy of INTENTION.

Section 1.2 The work of Broder needs a real reference.

Another way to think about classifying Web search might be via the kind of outcome desired: often, this may be a 'document' of some kind, but increasingly a user may wish to locate, say, a social network (ideally a graph), a scientific experiment or workflow (another graph) a route plan (ideally map), a date (ideally a calendar). All I'm saying is that we can try to break this unhelpful 'document' mould at both ends of the query.

Section 4. Monitoring: to me, implies some regularity, or repeating pattern., otherwise it is locating or grazing. It might help differentiate these to think of it that way.

Asserting to me also sounds like Contributing--which I think in the case of your example is probably closer to what the author may think they are doing

Tags: 

Comments

The paper argues that the current view of users doing tasks using the Web is based on a document-centric view of the web, and that in a Linked Data web of things, people etc. a fresh new way of classifying user tasks would be useful for developing new technologies that make use of the Semantic Web infrastructure. The author presents such a "taskonomy", based on an analysis of previous literature, and claims that it can be used to classify the ends and purposes of people using Internet platforms.

The paper is well written and clear, and the organization is logical. It places quite much emphasis on analysis of literature, with relatively little space devoted to original ideas. The main contribution of the paper, the "taskonomy" presented in section 4, is unfortunately not explained in very much detail, in particular its intended use.

The paper includes a thorough and well summarized review of the literature on classifications about web search and other ways of using the web. Most of the referenced studies and the paper itself seem to take a big-picture view of web/internet usage, trying to classify all the possible "activities and goals" (section 4) people may have when using the Internet. However, I felt that the presented "taskonomy" is still a fairly low level view of internet usage, and does not include overarching goals such as arranging a holiday (author's example from end of section 1) or researching digital cameras for the purpose of buying one. Fulfilling either of these goals may involve a possibly long sequence of individual tasks such as exploring available products and services, looking up information about stores or hotels, evaluating different options by e.g. reading reviews, discussing options with friends, arranging accommodation, and paying for purchases. The paper doesn't mention these overarching goals except in passing (as in the holiday example).

Understanding the underlying goals and not just the task components which together enable the fulfilling of a goal is crucial for designing usable and useful systems (see, e.g., [1]). The paper should state that the presented taxonomy is limited to the task level and not claim that it represents a classification of user goals, although some of the task classes such as writing a review (asserting) or socializing (discussing) may sometimes be end goals by themselves.

In Section 3, the author criticizes the current document-centric way of doing tasks, e.g., that it is necessary to create an e-mail message (document) and attach a photo to that, rather than just sending the photo itself to the right person. While I agree that the document is unnecessary here, it can be avoided -- at least by the recipient -- by using the "Send to.." mechanism found in many file and photo organizers. The communication channel example is a bit unconvincing, as the sender's choice of channel (e.g., e-mail, IM, SMS or phone call) involves important social considerations about, e.g., urgency and expectation of reply. Of course, some of the fragmentation in current communication channels (such as the wide choice of social networking sites with different user identities) does not serve the needs of either sender or recipient, but it is questionable whether a single URI-based identity could realistically be implemented solve this issue due to the commercial walled garden nature of most current social internet services.

The term "taskonomy" (i.e., taxonomy of tasks) appears to be a neologism that has been used at least since 1982. Perhaps a reference to the American Ethnologist paper [2] (which may or may not be first source to coin the term) or a paper that applies the idea to the Web [3] would be appropriate here?

The taxonomy is (perhaps by necessity, given the current state of the Semantic Web) rather speculative and is not based on any first hand empirical study. Has the author thought about how it could be empirically verified or its usefulness for constructing future applications evaluated? Finally, the paper concludes that the taskonomy "can form the basis for discussion and innovation" but gives little concrete ideas about how that could happen. Does the taskonomy help in designing SW applications? In publishing Linked Data? In designing ontologies? In integrating services using SW technologies?

The topic of the paper is quite original. However, there was a poster paper at SWUI 2006 which had a similar scope [4] and which could perhaps be referenced.

In all the paper provides a fresh look at the LD/SW ecosystem from an end-user and activity-oriented perspective, which in my view is a very important but often neglected point of view. Whether the presented taskonomy will become a useful tool for researchers and practitioners or not, just providing a different viewpoint that may help illustrate the potential of the Semantic Web is a valuable contribution.

References

[1] User and Task Analysis for Interface Design. Hackos, J. and Redish, J. Redish & Associates, 1998.

[2] Taskonomy: A Practical Approach to Knowledge Structures. Janet W. D. Dougherty and Charles M. Keller. American Ethnologist, Vol. 9, No. 4, Symbolism and Cognition II (Nov., 1982), pp. 763-774

[3] The Tangled Web We Wove: A Taskonomy of WWW Use. Byrne, M., John, B., Wehrle, N., Crow, D. Proceedings of CHI 99, 1999.

[4] Preliminary Inventory of Users and Tasks for the Semantic Web. Battle, L. Proceedings of SWUI 2006.

After reading this paper, I felt both stimulated and unsatisfied. Stimulated, because it contains a wealth of ideas and pointers. Unsatisfied, because it end rather abruptly and vague. After section 4, the natural question to ask is: "What can I do now with the insights I just gained?" "What are possible next steps?" It would be good if the author could address these, even if briefly.