Projects‎ > ‎

10 credits projects

Student taking this projects will take full credits from the AUI Technology course and from the AUI Interaction course.

  • Projects can be taken individually or in groups of no more than three students. 
  • The number of available projects is limited.
  • Projects have deadlines: failing to complete the project before the deadline means failing the project


C.1 Design Patterns for Recommender Systems

Preamble. A design pattern is a recurring reusable solution to a commonly-occurring problem in the design for a specific domain. An interface (or interaction) design (ID) pattern is a recurring reusable solution to a commonly occurring problem in interface design or interaction design.  An ID pattern usually consists of the following elements:

  • Problem: Problems are related to the usage of the system and are relevant to the user or any other stakeholder. 
  • Usage: situation(s) in which the problems occur and the pattern applies. 
  • Solution: a proven design solution to the problem, described in terms of design characteristics of the interface and the interaction 
  • Why: why the pattern actually works - The rationale for the solution (principles of UX quality can be used as arguments)
  • Examples: how the pattern has been successfully applied in real life systems. This is often accompanied by a screenshot and a short description. 
  • Related Patterns: Other patterns may be needed to solve sub problems
Examples:
Goals of the project.
  • Identify occurrences of existing design patterns in RS interfaces
  • Discover RS specific new patterns
What to do.
  1. Web site identification and classification.
    Identify a set of web sites that offer recommendations; List them in a shared online document
    Classify them by application domain (refer to, and extend, the classification in gndec.ac.in/~librarian/Articles/ontology/192.pdf)
    e.g., social network, search, news, forum, blog, e-commerce/type of product (e.g., TV, Movies, Music, Books, Tourism Services, Food, Restaurants, …), Learning resources, …
    Outcome: a shared online table 
    <site name, url, domain, proposer (group name)>
  2. Defining scenarios to guide inspection and discovery
    Reflect on stakeholder goals: what the user wants to achieve by using the web site (e.g., exploration, search, decision making, …) or what the service provider want to achieve on web site  users (e.g., persuasion, promotion, advertising, …)
    For each of the sites of the cluster(s) assigned to you/your group define one or more user scenarios, i.e., a story of use of the web site, including the possibly iterated tasks that users (with a given profile) perform on the web site  to achieve their goals.
    Example of scenarios: 
    - a young couple (user profile) is looking for an hotel in town X where they will go for their anniversary (goal=exploration); they will specify the city, the hotel characteristics, and check for room availability (tasks)
    - a young couple (user profile) is deciding which is the «best» hotel in most appropriate among those they have identified in town X where they will spend their anniversary to have the main characteristics they need (goal=decision making)
  3. Inspection: discover occurrences of patterns & identify new patterns
    Consider the patterns in the UI-patterns library ui-patterns.com
    These patterns address «general» design problems , i.e., design problems that are related to the usage of an interface, are relevant to the user or any other stakeholder , but are non RS specific.
    For each web site perform the scenario(s) and
    - Identify  all «general» design problems  addressed by the patterns UI-patterns
    For each general design problem, is the solution provided by the patterns in the library used?  
      + If yes, take note of where the patterns occurrence
      + If not, take note of the different solution used and judge its  effectiveness (is it a good solution or not?)
    - Identify all RS specific design problems
    - For each general design problem, judge the design solution adopted by the web site and take note of it
  4. Analysis and reporting of results.
    A. Existing patterns
    Pattern occurrences reporting: collect all patterns occurrences and describe them in a structured way (template will be provided)
    Pattern violation reporting: collect all patterns violations and describe them in a structured way (template will be provided)
    Compare results among groups
    B. New patterns
    - Provide a good compact formulation of RS specific problems; compare result among groups
    Collect all examples of solutions in all inspected web sites; compare results among groups
    Judge solutions and identify the most appropriate one(s)
    Formulate the «best solution» 
    Complete the definition of the pattern according to the general pattern structure

C.2 On-line evaluation of diversity-based and novelty-based algorithms

Preamble. Recommender systems are often tuned to provide accurate recommendations (i.e., relevant to the user). However, diversity and novelty of recommendations are important attributes to provide useful recommendations. In the literature, different algorithms have been proposed that should be able to provide novel and divers recommendations. The goal of this project is to evaluate with an on-line study which of these algorithms provide the best recommendations in terms of diversity and novelty.

The study requires to follow the following steps:

  1. State-of-art.
    Perform a study of the literature and identify all the algorithms optimized for novelty and/or diversity.
  2. Web app.
    Create a web application for the on-line evaluation of novelty and diversity as perceived by the users.
    The web application will be based on a movie catalog of almost 100'000 movies already crawled from IMDB
    The web application will implement some of the algorithms discovered in step 1.
  3. User study.
    Perform a user study with at least 30 participants for each algorithms
  4. Statistical analysis.
    Perform statistical tests in order to identify which algorithm is significantly better in terms of novelty and diversity as perceived by then users.

C.3 On-line evaluation of different off-line diversity and novelty metrics

Preamble. Recommender systems are often evaluated by means of accuracy metrics (i.e., relevance of recommendations). However, diversity and novelty of recommendations are important attributes to provide useful recommendations. In the literature, different off-line novelty and diversity metrics have been proposed. The goal of this project is to evaluate with an on-line study which of these off-line metrics better capture the perceived diversity and novelty of recommendations.

The study requires to follow the following steps:

  1. State-of-art.
    Perform a study of the literature and identify all the different off-line definitions of diversity and novelty.
  2. Web app.
    Create a web application for the on-line evaluation of novelty and diversity as perceived by the users.
    The web application will be based on a movie catalog of almost 100'000 movies already crawled from IMDB
    The web application will implement at least six recommender algorithms from the literature with different levels of expected novelty and diversity.
  3. User study.
    Perform a user study with at least 150 participants 
  4. Statistical analysis.
    Perform statistical tests in order to identify which off-line metric (if any) is significantly correlated with the perceived opinion of the users on novelty and diversity.