CHI 2007 Discussion
From SWUIWiki
ACM CHI Conference 2007
May 2, 2007
Duane Degler, Scott Henninger, Lisa Battle
The following are raw notes from the interactive session "Semantic Web HCI: Discussing Research Implications" at the CHI conference. For background, the slides used to frame the discussion are also available (Semantic Web HCI session, as pdf).
Contents |
Discussion on Ontology Update, Maintenance, and Creation
- These visualization and ontology editing tools are difficult for people generally to use. In terms of scaling to a sizeable population of users, do we think they will be capable of it? Barriers to the initial adoption. (We have tools now that allow people to create HTML without knowing how to code it.)
- Lots of people couldn't deal with raw HTML, even though some of them could.
- Collaborative editing is very important. We have to look at the early stages of ontology development, and how things transform more gracefully into more formal ontology. Getting consensus on an ontology is not an easy thing. There are conflicting views. Later on they may be merged.
- point came up in discussion afterward about how to model conflicting valid perspectives, so as not to lose the reality by over-refining the models...
- Ontology editors are ugly, people who work with them know it.
- What kind of tools do we need to support the social process?
- What are the particular attributes of the semantic web that raise new challenges for user interaction?
- What is the relationship between the semantic web and other recent developments in industry? Social tagging, micro formats, etc.
- Search engine based on categorization was developed in the past--now that does not seem to be done anymore. What lessons learned? Why did it fail? (There is some use of it still, maybe it has moved into different types of situations or into more background enablers of the search/results/navigation process.)
Discussion on Information Seeking
Examples: mSpace, museum site (eCulture), ginseng
- preview feature, flexible categorization, ordering is not fixed which enables changing the perspective
- timeline, filtering, presentation of the detailed metadata when looking at a detailed topic (how do we make the data presentation human readable?)
- natural language query using structured information, allowing searchers to ask more targeted questions than before
Discussion with audience:
- Does this translate across language and culture? What might become possible? (People are working to use multiple languages. The questions of translating across culture are more challenging.)
- A lot of examples are domain specific (e.g. biology), where people have to enter a lot of data to create ontology. Is there something more generic that handle a wide range of data? (There are tools for developers but not necessarily for end users.)
- Tension between domain-specific (increased usability) and generic (widely applicable but not as usable) browsers. A design challenge for the semantic web.
- If you have different sources, do users need to know where the information is coming from? What if some of the sources are more reliable than others?
- Lessons learned from eCulture: There is no way to differentiate info only for machines and some for humans to understand. In user evaluation it became really critical seeing info that has no meaning to them.
- Credibility: We have info from different museums and institutes. The credibility of different sources is an issue.
- cs.aktivespace.org: example of provenance: you select it and see the sources of information, letting you investigate where it is coming from, allowing the end user to look under the hood.
- It should always be possible to expose the source to allow people to make decisions.
- How do you keep the person in the context of what they were doing, while still letting them know how reliable the information it is?
- Might there be standard patterns for how you interact with the information behind the information? Consistent ways to view sources and provenance?
- With natural language, you can't always rely on it to be able to understand every type of question. What to do when it does not understand--how can it respond gracefully?
Note:
- You do not need ontology to get the benefit of the semantic web. All you need is RDF.
- Example: Exhibit (MIT SIMILE group). It only requires a text file and a few lines of code on your HTML page.
Discussion on Information Synthesis
Example: mSpace mobile
- contextual awareness, location, user-entered ratings, increasingly social and mobile technologies
Example: SADIe
- Accessibility portal that uses ontology to translate the HTML and re-display it in a more accessible format
- Benefit of using a shareable format to solve HCI problems
Discussion:
- Annotation by people who don't have a lot of knowledge of the subject
- Services--How can the service oriented architectures help generate user interfaces
