Presentation is loading. Please wait.

Presentation is loading. Please wait.

User Characterization in Search Personalization

Similar presentations


Presentation on theme: "User Characterization in Search Personalization"— Presentation transcript:

1 User Characterization in Search Personalization
RDA Data Discovery IG SJS Khalsa User Characterization in Search Personalization

2 Search Personalization
Keyword-based search is imperfect Ambiguity in meaning Varies with context What meaning does the user assign to the term? Personalization in information retrieval Long history (“ask a Librarian”) Area of intense research since advent of search engines Commercial interest in tailoring information presented to individuals

3 User Characterization
Also called User Profiling Through actions and inputs during search session Search terms Click throughs, dwell time Response to survey questions or registration forms Through information gleaned from other sources Browsing history Actions in previous sessions Activities or profiles on social media Affiliations (institutional, personal, foaf) Publications, presentations

4 Creating User Profiles
Explicit (direct user intervention) information Implicit information (monitoring user activity, inferring interests) Static vs. dynamic Short term vs. long term

5 User Profiles from Explicit Information
Explicit (direct user intervention) information Information gathered at time of user registration Response to questions/selections presented during session Tagging, annotations, feedback forms OrchID, ResearcherID, etc.

6 User Profiles from Implicit Information
Client-side Browser history, bookmarks, other activity logs Local software agents to track activity Mining calendar entries, s, local documents (<= Privacy Issues) General personalized environment Server-side Search history Cookies (persistent) Session IDs (ephemeral) User behavior while interacting with site Doesn’t capture negative feedback

7 User Profiles in Discovery
Discovery services involves both providing response to user queries as well as supporting browse (“you might also be interested in…”) Algorithms for combining preferences from profiles with query parameters (keywords, facets) Re-ranking What a user identifies as relevant in the initial query result set provides feedback that an algorithm can use to re-rank the result set and possibly show new results Relevance determined e.g. by seeing what the user selects for further inspection or by having the user tag results they feel are relevant

8 Breakouts What should be the nature of the Data Discovery Paradigms Interest Group (DDPIG)? Discussion group, sharing challenges, solutions and ideas Focus on identifying bite-sized problems and spinning off WGs to find solutions (18 month timeframe) Fostering coherence and coordination of data discovery issues in other interest and working groups, advising, contributing to use cases, etc. Breadth of interests among membership Strategies for entraining SMEs for focus areas

9 Possible Breakout Topics
Common API’s for data discovery: inventory and comparison Benchmarking search results for data discovery Overlap and interactions with other IG/WGs Methods for content enrichment and faceted search Algorithms for relevance ranking Issues and guidelines for user profiling in search personalization

10 Possible focus topics Deduplication and cross-repository issues
Collections and granules: build tool that enables guidance for data submitters on how data is organized Identifiers and how they help in search Data citation: how do we access/use? Guidelines for making your data findable! Best practices based on experiences. Relevancy ranking for structured data? Enrichment tools for faceting and ranking Identify collections of use cases for users: e.g. browsing vs search Domain-specific vs. generic issues: interfaces and enrichment Measures of data quality: and impact of findability Different discovery platforms for Open Search, science-focused OS profile? Define series of reference datasets – can be used to do these metrics Metadata standards to enhance data discovery, e.g. schema.org and such Identify list of prototyping tools, use by WG! Models and methods of personalisatoin Cross over between domains: how to enable cross-walk between domains Identify core elements of Findability “Return to the semantic”: schema has been populsated by crowdsourcing rather than 1 researcher. Automated integration of records; granularity and findability Common APIs (e.g. OpenSearch) Implementing schema.oirg as it exists! How does it apply to science? Upper-level ontologies for search Creating test collections for search evaluation and methods of evaluation


Download ppt "User Characterization in Search Personalization"

Similar presentations


Ads by Google