The overarching aim of our work is to develop tools and methods that overcome the challenge of efficiently accessing heterogeneous and distributed research data. The focus of this work is to answer questions regarding the action of pharmaceutical drugs. Specifically, we will:
1) Assess the state of the art in question answering and knowledge bases.
- Task 1.1: each presenter prepares a short summary
2) Design an open platform for question answering.
Documentation on the progress : Google Doc
- Task 2.1: Identify main components and define architecture diagram
- (Machine Aided) Question Formulation
- Entity Recognition
- Disambiguation [details]
- Multi-lingual support
- Template Generation
- SPARQL query Generation
- SPARQL query Execution
- Rendering (Answer Presentation, Verbalization) [details]
- User Feedback service
- Evaluation service?
- Controller: coordinates modules to take system from question to answer
- Task 2.2: Define REST API for each component
- Task 2.3: Identify how services will be discovered
3) Create a reference implementation to answer questions posed in natural language.
- Task 3.1 implement core modules
- Task 3.2 create integrated kb, search indices
- Task 3.3 create user interface
4) Evaluate the accuracy and performance of our system.
- Task 4.1 Evaluate and Extend Biomedical QALD
- New Questions using Datasets from Bio2RDF, Biogateway
- Documentation: Google Doc
- Task 4.3 Identify new directions
- Hybrid QA
- Fact Checking
- Entity Ranking
- Task 4.4 Evaluate System with QALD (DBpedia, biomedical)
Task 1.1 Michel, Axel, Erick
Task 2.1 Axel
Task 3.1 Axel
Task 3.2 Michel
Task 4.1 Michel
Task 4.2 Michel
Task 4.3 Erick