1 / 6

Task 2: Ex trinsic Evaluation

Task 2: Ex trinsic Evaluation. Vasile Rus , Wei Chen, Pascal Kuyten , Ron Artstein, Elnaz Nouri. New modality. Generate questions from pictures Provide text/metadata description as a seed for QG systems (to be dispensed with in the future) Users/evaluators only see the picture

jubal
Télécharger la présentation

Task 2: Ex trinsic Evaluation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Task 2: Extrinsic Evaluation VasileRus, Wei Chen, Pascal Kuyten, Ron Artstein, ElnazNouri

  2. New modality • Generate questions from pictures • Provide text/metadata description as a seed for QG systems (to be dispensed with in the future) • Users/evaluators only see the picture • Answer = region in image; users interact graphically

  3. Users evaluate each other • Users provide answers to generated questions • Users evaluate answers from other users • http://anawiki.essex.ac.uk/phrasedetectives/

  4. Evaluate Q-A pairs • Motivate children to answer questions • Correct answer = plant seed in virtual garden • http://pbskids.org/arthur/games/groovygarden/groovygarden.html • Systems provide questions and answers • Perhaps questions, answers and distractors? • Rate of correct responses = quality of q-a pair

  5. Guess who • User chooses a picture element • System asks yes/no questions • User can answer yes, no, don’t understand • Evaluates question answerability • Goal: best strategy to get to user’s choice • Metric: number of steps • Question formulation might be trivial • Hard part is deciding what to ask • Build model of user inentions • Create taxonomy on the fly? In a short time?

  6. What systems should win • QG has two essential components • What to ask • How to ask it • (Optional: answer your question) • Either what or how should be non-trivial • Winner = the most templates: is that good?

More Related