Interfaces for Digital Libraries: what will the future bring? Prof. Fabio Crestani Faculty of Informatics University of Lugano (USI)
Talk outline Me and my background (really quickly) Search interfaces (or UI) for DL: state of the art and current research Present (with limitations) Emerging trends Future (?) 2
A brief introduction Background: Professor in Informatics at USI since 2007 Earlier, Professor at University of Strathclyde, Glasgow, UK, in Dept of Computer and Information Sciences (one of the best library schools in UK) Research interests: Information Retrieval (main area of research) Text Mining Digital Libraries 3
My research interests More specifically, I am interested in: Information retrieval models for classical (e.g. indexing and retrieval of textual documents) and novel (e.g. speech, mobile, and blog search) applications " Text mining to establish links between information items (e.g. in e-books and in police databases)" Personalised, distributed and non-cooperative digital libraries" Opinion coding and sentiment detection in blogs and tweets" Detecting topic shift in topic detection and tracking" " Most of this work is funded by research projects (e.g. SNSF, EU) and is carried out in collaboration with the PhD students that I am advising " 4
Search interfaces for DL today Primarily designed for text search Tailored to keyword search Iterating on the information need Navigation and organization support for fast faceted search In some cases Support for related queries and documents Blended and context-sensitive results 5
Limitations Access needs a large, desktop-style terminal Limited access to mobile users accessing libraries from small terminals (e.g. ipads, mobile phones) Only text-based interface and results presentation No allowance for a more natural style of interaction No access to multimedia information No use of personal and social context 6
Example 7
Example 8
Beyond limitations User studies (also in DL) show that users prefer Longer, more natural queries Mobile search Natural interfaces (hybrid of natural language and command-based queries) Social search More interactive user interface for suggestions and exploration Current UIs for DL are only partially satisfying these needs 9
Emerging trends in UI for DL In recent times some emerging trends are identifiable in DL research that might soon get implemented in real DL New R&D on interfaces enabling: 1. Mobile search 2. Social search 3. Multimedia search 4. More natural search I will briefly review these trends 10
Mobile search Move from small-interface mobile phones to current high-resolution screens had impact on style of interaction People in developing world skipped land-lines phones in favor of mobile phones! 11
Mobile search Current mobile phone are able to capture the user context through their many sensors Context strongly influences information seeking Context can be used to refine results presentation Search queries got shorter than desktop-based search queries, but Less variation in queries More information is provided (e.g. personal info, context, location ) 12
Mobile search Query input is more complex due to the limitations of the device, but Context enables to anticipate common queries Use dynamic term suggestions from previous and popular queries Mobile phone-based voice-entered interaction is becoming possible thanks to advances in speech recognition (thanks to big data) 13
Mobile search Results visualization is more difficult Think of the screen size and information presented But Fewer page visits per query More specific information needs and no exploration tasks 14
Mobile search Presentation of results is slowly changing from desktop style Better use of the limited interface of mobile phones (e.g. windows, vertical results listing, user-based results selection, context-based results selection) Spoken presentation of results is becoming possible, e.g. Development of specialized interfaces for particular question types 15
Social search Move from mediator-style search to individual-style search to social search, via interactive Web (a.k.a. Web 2.0 ) Advances in social ranking using wisdom of crowds and user click-through (e.g. Yahoo MyWeb) Use of social tagging, folksonomy, social bookmarks and results lists saved by users (e.g. Delicious, digg) Collaborative search (pm the realization that people often search together, even in DL) and other people results made visible to users 16
Social search Move towards massive-scale question answering enables a more social use of experience From QA to human QA (e.g. Yahoo Answers and ChaCha search) Users voting the quality of answers provided 17
Multimedia search People in the developing world may skip textual literacy completely in favor of video literacy! UI need to be able to present multimedia information (images, audio, video) in response to search, but still many problems are unsolved SR still has problems with long spoken documents (e.g. recording quality, background noise, variations) Poor interactivity with video and image retrieval systems Realization than universal search (i.e. blended results) is better than vertical search 18
Natural interfaces Move from form-filling UI towards an hybrid command and natural language search But natural language search requires Natural style of queries (e.g. spoken dialogue, proper natural language, teleporting queries) More user interaction with query and results (constant feedback), and user anticipation (e.g. query completion or query suggestion) Full relevance feedback technique (preference to explicit relevance feedback over implicit) 19
The future? In the future we are going to see The continuous decline of text The rise of audio and video Dialogue style UI, with natural voice interaction But what about Personalization? People are social; computers are lonely. Don t personalize search, socialize it! (Marti Hearst) Large screen visualization? Where? Some breakthrough are needed! Who can really predict the future? 20
Conclusions As algorithms get more sophisticated, we can build UIs that allow people to interact more effectively and naturally: More natural language-like queries Hearing/speaking rather than reading/typing Interacting with other people while doing search tasks Leveraging the knowledge in other peoples heads This will require consistent investments in education and services for librarians! 21
Thank you for your attention! 22