Project: generate multiplie choice answers for given question and answer.
An updated pipeline description was added to the documentation.
- Create configuration system for specifying paths and user accounts (e.g. for the database)
- Create question sets
- Implementation
- Answer candidate UIMA type with tag cloud feature (see #1)
- Open Trivia QA parser [Frank]
- Keyword extraction for parsed questions [Tobias K.]
- Category and hypernym detection [Tobias R.]
- Category search using JWPL
-
Hypernym search using JWI
- Category ranking [Tobias K.]
- Candidate extraction [Tobias K.]
- Synonym resolution [Frank]
- Similarity detection [Tobias R.]
- Candidate selection [Tobias R.]
- The pipeline
- Devlop evaluation formula for tests results
- Analyze and summarize test results
- Write final report
- Statement of the research problem [Tobias K.]
- Methodology and used third-party software
- High-level description of the software (including class/component diagram) [Tobias K.]
- Data sources (how obtained and processed?)
-
Preliminary testing/training (if applicable) - What we did to improve the system's performance, and how these ideas worked out
- Final results of the system (i.e. evaluation results)
- Discussion of results
- Future work (e.g. extend system with WordNet, apply it to other types of questions, maybe other use cases)
- Statement indicating what group members did what
- Create final presentation
- Cover the same content as in the written report
- Highlight particlularly interesing results
- About ten minutes long