Difference between revisions of "Vulcan"

Revision as of 06:20, 18 September 2013

This page has pointers to documents, resources, and demos.

Project Log

1. Meeting Notes
2. System Development

• Status
  
• Plan
  
• Prototype
  
• Prototype to a Working System
  

3. Experiments

Documents

Examples

• Inference steps for 10 propositions.
• Iron nail example in detail.
• Stephen's examples for quantification inference, and rule templates
• Examples of Open IE coverage issues
• Extended WordNet Overview and example from Stephen.

Design Documents

• Inference System Design

• Rule representation

• Transforming Rules into Tuffy's MLN syntax

• Textual Evidence Finder

• Patterns for extracting definitions from Stephen.

• Representing N-ary and Nested Tuples -- Email thread.

• Handling gaps in inference.

Resources

Questions

NY Regents Sample of 4th-12th grade questions.

Current Training Set.

Definitions

Concept and Definitions extracted from 4th grade specific resources by Harcourt and McMillan.

A subset of definitions covering 4th grade science syllabus. The definitions are extracted from about seven different dictionaries.

Wikipedia sentences covering the target concepts.

Study Guide materials

Barrons Grade 4 Science Topics

Barrons Grade 4 Science Sentences

Demos and Services

UW

Proposition Verifier -- This interface will take in a natural language sentence as input, convert it into a proposition, and find evidence for it using UW's inference system.

Textual Evidence Finder-- Find tuples from study guide and definition sentences.

<b>Solr instance behind Textual Evidence Finder

Open IE 4.0 -- Process input sentence using Open IE 4.0 (similar to Michael's nlpweb) [To Do]

Compound Noun Categorizer -- Process text using fanse parser. [To Do]

Vulcan

Vulcan Evaluation Server -- Select the **SECOND** entry (Direct Question-Answering (XML returned)) and type in a question. If the UW question-answerer can return answers in the same structure, then the tester can interface with it and run the exams through.

Literature

Quantification

Underspecified Quantification

Aurelie Herbelot, Thesis.

BLP and MLNs

1. Learning to Read Between the Lines using Bayesian Logic Programs Raghavan and Mooney, ACL 2012.

2. Tuffy: Scaling up Statistical Inference in Markov Logic Networks using an RDBMS VLDB 2011. Slides

3. RockIt: Exploiting Parallelism and Symmetry for MAP Inference in Statistical Relational Models AAAI 2013.