MultirDevelopmentTimetable
From Knowitall
Development Timetable
Green means DONE
Blue means IN PROGRESS
Red means TODO
10/28 - 11/1
- Develop extensible Database Interface for Preprocessed Corpus Abstraction
- Load Preprocessed Corpus Information into Database
11/4 - 11/8
- Run Distant Supervision (Argument Identification and Relation Matching)
- Run Feature Generation
- Convert to Multir Input Format
- Train Multir Model
- Test Multir Model and Validate Performance Results
11/11 - 11/15
- Convert Batch-mode preprocessing code to in-memory preprocessing
- Resolve discrepancies in NER, POS tags, and dependency parses
- Develop sentential extractor
11/18 - 11/22
- Process KBP Corpus with Newer Models
11/25 - 11/27
- Process KBP Corpus with Newer Models
12/2 - 12/6
- Implement New MultirExtractor with New Models
12/9 - 12/14
- Evaluate Sentential Extraction
- Debug ADEPT implementation of Multir
12/16 - 12/20
- Debug ADEPT implementation of Multir
- Annotate Distant Supervision Errors for NER Argument Identification
12/30 - 1/3
- Generate new Wikification Data from UIUC 2013 Wikifier in batch
- Run UIUC 2013 Wikifier as Server for faster test time Argument Identification
1/6 - 1/10
- Train a New Multir Model that uses old Wikification Data for Argument Identification
1/13 - 1/17
- Implement Coref Data Into Multir
- Implemented new manual evaluation scheme
1/21 - 1/24
- Implemented Various Generalized Feature Generators
- Chunked the Corpus with Hadoop
1/27 - 2/1
- Evaluated Feature Generation
- Generate new Wikification Data from UIUC 2013 Wikifier in batch
2/3 - 2/7
- Generate new Wikification Data from UIUC 2013 Wikifier in batch
Longer Term
- Move Corpus Representation to a SOLR Database as an infrastructure for Information Omnivore
- Develop Web App