HYOWON LEE B.Eng., M.Sc., Ph.D. |
|
| |||||||||
|
|
MY WORK & PROGRESS in 2003 |
| Here is the list of the major events & work (in reverse-chronological order) I have been
doing in 2003. |
|
|
TRECVID 2003 Workshop, Gaithersburg, Maryland
(17-18 November 2003) | |
|
Físchlár-TREC2003: Interface design
(August - September 2003) | |
|
Físchlár-News-Stories: Calendar with
Recommendation
(July 2003) | |
|
Físchlár TV schedule visualisation with SVG
(June 2003) | |
|
| |
|
Físchlár-News desktop: story-based presentation
(April 2003) | |
|
Físchlár-Nursing Deployment
(10 February - mid March 2003) | |
|
Questionnaire design & testing for RTE library
(March 2003) | |
|
|
|
| my reading in 2003 |
One of the most enjoyable one I read this year was the recent book by
Ben Schneidermand from Maryland HCI Lab (it was my Easter Holiday reading):
This book had been summarised in two previous papers which I had had a taste before: The book is full of interesting future applications of information & communication technology, but the constant emphasis is to think about what the users can do rather than what the software can do . Chapter 6 on e-Learning is especially inspirational and I have some modifications to make for next semester's lectures for my students. Leonardo da Vinci's observation, mixing art and science, sketching and notemaking were interesting to know. Regarding the framework of Activities (collect - relate - create - donate) and Relationship (self - family & friends - colleagues & neighbors - citizens & markets), I think it is a very good idea to have something like this so that we can come up with new possibilities, as well as being able to slot in any existing technology. We all like to develop some kind of framework (a theoretical basis or orientation within which related phenomena can be understood), even based on a lot of assumptions, because they do help understanding better and predict better a phenomenon. In fact I have a kind of framework myself, which categorises keyframe-based video browsers, to help better design new browsers as well as to understand existing browsers in a set of pre-specified attributes (see my Ph.D thesis 2001). Even if it tends to distort the real picture by implying some hard segmentation or crude categorisation, ignoring middle grounds or gradiently-changing nature, it is at least worth trying because at the end it does add insights and understanding. In video browsing study, I have even divided all users into 8 crude groupings, and one of the dimensions was whether a user has seen the video before or not - surely it will not be merely 'have seen' or 'have not seen' distinction... as my PhD external examiner asked me during my viva, there will be users who know the structure or characters in the video but haven't seen that particular instance (e.g. people who watch soap opera regularly... they know characters and backgrounds of events, but now browsing a new episode) - my categorisation cannot distinguish this, it fails to address this type of users. Traditional Korean medicine categorises human beings into 8 broad types, and for each type there are different kinds of prescriptions even with a same symptom. If a patient visits a doctor complaining about a skin problem or a chronic headache, the doctor doesn't care much about the symptom itself but firstly checks (by pulse and tapping here and there, asking a few questions like 'do you easily get angry?', or 'do you like hot bath?') what type of the 8 types this patient belongs to, and prescribes appropriate food and life style suitable for that type. Once the person gets the right food and life style for certain period of time, symptoms will disappear - the symptoms appeared because the person lived with wrong food and wrong style. The basic philosophy for prescription is based on the people's inherent characteristics rather than any particular symptom itself. While this makes much more sense than the Western medicine - in which each symptom or condition has a matching set of prescriptions thus ignoring people's individual body differences (individual differences are covered by the idea of 'allergy' as exception rather than the main part of the cases) - still an important distortion or cover-up happens: those 8 different types are only artificial rough chunking for convenience of understanding and convenience of prescription, and in fact there are people who are at the boundaries of these types. People's characteristics are gradiently different, not discretely different. Whenever I tell my Web Design students "about 5 percent of all male users are red-green colour blind...", I hasten to add that "... to some degree or other". In those 5 percent of population, some will surely be worse than others. If doctors really care, they would have to divide people into more fine-grained groupings, maybe 50 different types, and each type having its own prescriptions. Moving on to other papers I read this year, all related to digital video & interface design: I really enjoy and benefit by reading highly qualitative studies with full details of usage addressed. Without much availability of guidance in usage of new technology products, we benefit more by investigating the very details of individual cases rather than trying to quantify with statistical methods. Such studies on home/mobile appliances have appeared in recent years, such as: The Petersen paper above is on the same vein, with family visits and interviewing. Very rich and interesting data are collected and interpreted. As mentioned at the editorial paper: The new line of IT products will require good understanding at the beginning, with different usability criteria and different methods for measuring them. Many "distractions" occuring while interacting with these appliances should not be editted out in measuring usability as they are now, because those distractive elements in the user context are the main concerns of the study and the centre of understanding. And the evaluations for these are inevitably more qualitative than quantitative, as the Shneiderman's book refers to as 'new computing': "advocates of the new computing often seek subjective measures and apply ethnographic approaches to conduct evaluations. They observe and interview users while they are doing their work or enjoying their entertainment. The results are not numbers but understanding, not percentages but insights" (Leonardo's Laptop, 2002, p238). Papers I read this year dealing with news video story segmentation - related to our group's recent work on Físchlár-News - are: Ide's & Merlino's & Pickering's papers have front-end user interfaces for searching and browsing news stories, and Informedia (Hauptmann) has fancy timeline & map visualisation of news interfaces (Christel 2002 paper on collage, see below). Other useful papers I read this year follows:
Lecturing has been a beneficial aspect of my research - 'mutually reinforcing' is how it should be between my teaching and research. Three books I read this year in regard to Web design and visual design that were helpful for both my lecturing and (partly) for my research were: With what I learned from the e-learning chapter in the Shneiderman's book, experience from this year's lectureing and above web design books, I have some improvements I will make for the next semester: good use of emails with students (email facility is low cost and high payoff - well worth making good use of it), ask students to critique each other's projects, assign the groups from my direction rather than allowing them to form themselvse, richier references in web design guidelines, and more number of lab tutorials at the beginning of the module. I'm sure these will help the students better. |
Followed by TREC Video Track in 2001, 2002, we again participated in
TRECVID 2003.
As always, we did Search task by developing an interactive video retrieval system.
This year, we developed a system that incorporates a feature in which
relevance feedback is supported based on shot-content addition to the query. Georgina and I
designed the interface for the system, which turned out to allow smoother and more efficient
interaction
for our test users than last year. Playback feature as part of the browsing (partly because the playback
screen is embedded into the browsing interface, partly because the playback started immediately
on clicking on a keyframe unlike the usual streamed playback for other Fishclar systems)
considerably helped this enhanced interaction, compared to last year's system. We knew this from
3 years ago... but for LAN-based internet usage, streaming was the only way for playback, whereas
for this year's experiment we used network disk mapping for the test machines.
Físchlár-News-Stories system automatically records and
processes daily 9 o'clock news: thus the collection in this application is daily news programmes.
Kieran suggested having a calendar-style presentation of the collection instead of news title
repeatedly presented chronologically as in Físchlár-TV.
Updated interface uses monthly calendar on the left column, to present archive. This illustrates how
we should find out a suitable presentation format by considering the particular characteristics of that
archive.
Also the recommendation from
ClixSmart Institute kicked in, thus 'Recommendation' icon appeared at the top of the calendar.
This Recommendation button allows alternative access to the archive, reducing the navigation
effort by the user (thus highly suitable for mobile interaction). Clicking on the button
shows a list of recommended news stories to the user.
Físchlár-TV programme schedule listing for recording has been text: you browse today's schedule
for RTE1, then Network2, then TV3...
it takes quite long to browse all 8 channel pages to browse. Making use of
colour-coding and high level of interactivity, I designed & implemented TV schedule visualisation interface in SVG
for Físchlár-TV's recording section. With this interface, you can see all 8 channels of a day at
one glance, also what's available for recording, what programmes have been requested by users,
how long each programme is, what genre each programme is are all colour-coded, but by using
chrominance for genre and luminance for availability, distinguising between different types
of programmes is more or less instant, requiring no interpretation.
If widgits and features
have been designed with clear 'affordances' (something that suggests interaction possibility) users will
be able to more easily guess what something does correctly and use it.
If most elements have been
designed this way, the result is an obvious, easy to understand interface that clearly
indicates what it is about. While requiring no user's guide is a neat idea, it doesn't mean
that the designer should be too proud of his/her design that s/he need not provide any
manual.
Also, "the manuals and instructions accompanying a product, that is, the whole system
image, plays a role in forming the users expectations or model of the artifact [Petersen et al. 2002 - see bibliography in the
next section below]." When nicely designed and supplimented along with the product itself,
a manual can enhance the usability and positive system image (see image beside). With these considerations in mind
(though rather late), I added User's Guides explaning some frequently asked questions from Físchlár-TV users
since its deployment regarding specific features such as what the thumbs-up and -down icons do,
accessibility, why
some programmes are unable to be recording, etc. along with brief instruction for the
first-time users.
The underlying techniques to automatically determine story
boundaries of news programmes has been nearly complete with its initial results.
To get
user feedback for UCD to personalise the story-based presentation (as opposed to programme-based
presentation), desktop version of a system has to be up and running. The original
I modified the interface of Físchlár-News system to accommodate this progress. Overview
of each news programme is no longer a fixed 30 keyframes of evenly-selected shots, but by the
news stories of the programme, which is a big step forward semantic indexing. Detailed
(shot-level) browsing is no longer continuous shot-level space but pages of stories.
Físchlár-News
to libraries, and this time Físchlár-Nursing. As the start of the term,
I did active demonstration and installation
of Físchlár-Nursing to lecturers in School of Nursing
(appointment - installation - demo)
on their office machines.
Lecturers got individual installation and demo, 3 of whom adopted
the system as teaching material for their students. Nursing computer lab's PCs were checked with Oracle plug-in for
the Nursing students to use, and those PCs were attached with a sticker (as shown beside). I did demonstration to those students in
small groups.
To capture what video library users want and what kind of search
they do, we collaborated with RTE video library and worked on questionnaire design and
initial testing. From our group's discussion and consulation with RTE, I designed the
initial questionnaire form. With feedback from our group and RTE (iterative refinement has
been a regular feature in my sector), the form was improved (screen shot on the right)
and piolte survey was conducted.
Dublin City University
Design by Hyowon Lee 2003