1	Digital Libraries Patterns of Use Week 10 Min-Yen Kan
2	Two parts: Integrating information seeking and HCI in the context of: Digital Libraries The Web
3	Digital Libraries What uses do we commonly use the library for? Accounting for Different age groups? Different professions? Public or private access points?
4	One dimension: expertise “What better contribution could a scholar make than an article which could … provide a clear, but vivid argument to the [secondary school student] but which, if unraveled, could provide the rigor demanded by the most crusty specialist?” Crane (of the Perseus DL) Question: How do DL designers support this in terms of HCI? Answer: Creating different document layers. Allow users to “fold” the document to see the only the relevant portions.
5	"Overview + Details shown as..." Overview + Details shown as best (Hornbaek & Frokjaer 01) Fisheye distortion unsatisfactory Shown better for QA but not for whole document understanding
6	The scientific article How do we use articles? Answer these in groups: Do we use scientific articles as a whole? Or specific components? How do you (personally) determine the relevance of an article? When do you decide to read an article? (Harder) What parts of an article do you use, and for what purpose / task? How do you categorize or label the articles that you read? Typical critical reading patterns: Read the title and the abstract If you still don’t know what this paper is about, then this is a poorly-written paper. Read the conclusion Are you now sure you know what this paper is about? If not, throw it away. Read the introduction Read the section headings Read tables and graphs and captions
7	Usage lifecycle of an article Being found as relevant Assessing relevance Document surrogate “Information finding” Browsing for exploration Searching for specific bits Conveying knowledge not easily rendered in words
8	Being found as relevant Advanced features of search not often used “Just to be safe”, use full text Common and well-understood UI (legacy effect) When features failed, users often don’t try them again Features thus need: To be properly introduced / understood (scaffolding) To have well-understood error messages
9	Searching for specific bits One-shot queries rare: Tip of the larger iceberg of an information seeking pattern I look for specific surface tensions, experimental measurements Looking for best efficiency of electric motors. Ended up reading tons of documents for electric motor I sometimes want to look specifically at other’s methods and theories I often need multiple copies of a specific piece, like a table, for class I need to keep up to date on my research area
10	Browsing Why do people browse? Semi-directed / Undirected learning Initial Exploration Collection Evaluation What’s in this collection? Is it relevant to my objectives? Subject Exploration How well does this collection cover my area of interest? Query Exploration What kind of queries will succeed in this area? How can I access this collection?
11	Using the article Reading has different purposes too: General Learning Identification Skimming Answer questions Defend position Cross-Reference Editing or critical review
12	Using the article (2) Biased to particular user and task Current researcher’s work as “lens” to view the work Different workflow for different users Beginning researchers Seasoned veterans E.g., when to do annotation? Read references? Writing goes hand in hand with reading: Three levels: Creating, note-taking and annotation Annotation serves not so much to add to an article: But to extract / filter important nuggets from an article (e.g., highlighting) Adding a “document layer” to be used to view the document Also inter-document annotation (e.g., labeling)
13	Using Multiple DLs Question: What’s the most common failure when using multiple DLs? A: different layout of UI B: different query operators C: authorization problems D: different materials in collection Same problem in heterogeneous data integration What’s a possible solution?
14	Public or Private? Question: Easier to do information seeking in a public or private place? Need good support of note taking, annotation Access to customization Hardware support Information professions support
15	Teh tarik break time Yay! See you later…
16	Patterns of use on the web How do people use query the web? How do they use the web browser? How can we build a better web browser?
17	Web query types (revisited) What features best for web searches? Discriminate using Mutual Information for 2+ word queries P(x,y) / P(x) P(y) – collocation corrected for chance High MI corresponds to navigational task Navigational (Known item, Home page finding) Relevant pages are mostly entry (root) pages Anchor text and URL information Informational (Topic relevance) Relevant pages are mostly nested pages Content information (e.g., TF ´ IDF)
18	User behavior Users tend not to use monitoring steps Sign up for email alerts, create a channel Even in a formal search mode Users use simple keyword search, not advanced Don’t revise their queries often (75% of all searches) Don’t access help Users don’t seem to have strongly repetitive patterns within a cluster of pages No consistent paths Longest repeated sequence analysis fails Larger volume of queries Higher percentage of repetition Caching is a good strategy
19	Page navigation types ~40% by following hyperlinks ~20-50% by back button navigation 11% new window 10% other (pop-ups count here) Should be counted in hyperlink following 2.5% by bookmarks 0.8% by history
20	URL Vocabulary Observed linear growth, not power law Why?
21	Modes of web browsing Tauscher and Greenberg (1997): First time visit: new URLs observed Revisits: reading in depth (e.g., course notes), flicking to previous page(s) Authoring of pages: reload heavily used Using web-based applications: form submissions Hub-and-spoke: central page Þ specific page and back Guided Tour: Viewing a many-page article
22	Scenario You went to a website this afternoon to do some fact-finding for a project that you’re working on. After going through many sites, some reading you’re currently doing reminds of a link that would be useful to visit on a page that you visited sometime in the last hour or two. How would you go about finding it? Your answers:
23	The Back Button Takes you to the previous page With a reverse-order of chronological pages; i.e. a stack Extremely simple and easy to use How would you improve upon this? A UI feature of web browsers that have made it into operating systems
24	Temporal model of revisiting Promote a previously visited page to the top of the stack if: I go back to visit it and I take a different hyperlink from there
25	The navigation hub Hub: a page that was promoted in the previous algorithm Study shows hubs revisited 1.8 times Ideally, predict which pages would be revisited
26	Algorithm for finding hubs Safari Browser: Search Engine and typed URLs as hubs Previous revisit of a page indicates hub Even across sessions (“new window” commands) Points to per-user customization SmartBack
27	To think about Traditional use studies are very comprehensive But with new IT, new conclusions yet to be drawn What DL use patterns have correlations in the Web? What patterns are unique to the web? To the DL? How do you think web browsers and DL interfaces can be improved in the near future?