Dr. David Portabella

picture of David Portabella Clotet
PhD in Computer Science at the Artificial Intelligence Laboratory at EPFL completed in 2007.
In 2008 we have been awarded the Innogrant price to launch a start-up, DB4ALL, together with the EPFL Database laboratory.

Postal Address:
    EPFL - IC - IIF - LBD
    Station 14
    CH - 1015 Lausanne
    Switzerland

Phone and Fax:
    Tel: +41216936707
    Fax: +41216935195

Email:
    david.portabella@gmail.com
    david@portabella.name

My interest in computer science focus on decision support systems (DSS), web information extraction and databases.
I am also interested in natural language processing and understanding, expert systems, multimodal interaction and the semantic web.

Get a copy of my resume (curriculum).

Go to: PhD, Publications, Projects, Courses, Student Projects, Personal (photos and adventures) :)
Check also my Blog and my Linkedin profile.

 

PhD

Title: Improving User Confidence in Decision Support Systems for Electronic Catalogs.

Abstract:
Decision support systems for electronic catalogs assist users in making the right decision from a set of possible choices. Common examples of decision making include shopping, deciding where to go for holidays, or deciding your vote in an election. Current research in the field is mainly focused on improving such systems in terms of decision accuracy, i.e. the ratio of correct decisions out of the total number of decisions taken. However, it has been widely recognized recently that another important dimension to consider is how to improve decision confidence, i.e . the certainty of the decision maker that she has made the best decision.

We first review multi-attribute decision theory --the underlying framework for electronic catalogs-- and present the state-of-the-art research in e-catalogs. We then describe objective and subjective measures to evaluate such systems, and propose a system baseline for achieving more accurate and meaningful comparative evaluations.

We propose a framework to study the building of decision confidence within the query-feedback search interaction model, and use it to compare different types of system feedback proposed in the literature. We argue that different types of system feedback based on constraints ( e.g. conflict and corrective feedback), even if not novel as such, can be combined in order to improve decision confidence. This claim is further validated by simulations and experimental evaluation comparing constraint-based feedback to ranked list feedback.

Take-Away message:
Conflict and corrective feedback are the most promising types of feedback for decision confidence building. This claim is supported by experimental evaluation in the case of ranked list feedback compared to conflict and corrective feedback. In typical situations where there is a mismatch between what the decision maker wants and what is it in the catalog, Confidence Building Tradeoff Explanations (CBTE) provide a clear trade-off question to help the user make a good decision through the global relaxation set, and provides evidence for building decision confidence through the minimal conflict set and the local relaxation set.
In the same experimental evaluation, we saw that CBTE system feedback incited more participants to make more challenging queries, and thus to state more accurate objectives rather than mean objectives. This is a very beneficial behavior because DSSECs can better help them if users state precise objectives.

Download:

 

Publications


Projects


Courses

I have been assistant in the following courses:

Student Projects

I have supervised the following student projects:


Keywords: decision support systems (DSS), web information extraction, databases, query filtering, cooperative answering in databases, recommender systems, case-based reasoning (CBR), collaborative filtering (CF), digital libraries (DL), information visualization seeking (InfoVis), conversational recommenders, decision theory, natural language processing and understanding, expert systems, multimodal interaction, semantic web.

Locations of visitors to this page

Main website: http://david.portabella.name/
Last modified: January 9, 2008