Vulcan/MeetingNotes/Sep06 2013

Meeting Minutes

Points of integration

1. Add UW folks to e-mail list [Peter?]

2. Make a single wiki at Vulcan. Move UW's content to the new wiki. [Niranjan will work with Wil Smith]

3. Integrate w/ Vulcan testing framework. [Niranjan: UW's inference system will be a web service that outputs the reqd xml.]

4. Training data [Niranjan will work with Peter to make sure all data is available at UW.]

5. Share URLs for demos and services at Vulcan [Peter]

6. Send information about various knowledge sources inside vulcan [Peter]

7. Send Stephen pointers about dicitionary processing. [Peter]

System Dev. To Do

1. Convert current prototype into a working system. [Niranjan, by next Friday].

2. Textual evidence finder [Greg]

-- minimal web interface [next Friday] -- tuple representation design [next Friday] -- lexical variants (head word, stemming etc.) [next Friday?]

3. Update the system architecture figure to emphasize textual matching using procedural escapes. [Niranjan]

Discussion

1. Oren on integration -- Let's share systems when possible. For now begin with sharing information/resources etc. Aim for common rule language format to enable plug-n-play in different reasoners?

2. Peter on inference -- There are likely to be gaps in the proofs and you would need a way to handle these gaps.

3. Musam -- A framework such as MLN will allow us to handle these gaps and uncertaintities in inference.

4. Oren -- Next hire will be an MLN expert.

5. Stephen on extracting from defintions -- Open IE needs to have higher coverage. Stephen needs someone to implement definition extraction patterns.

6. Peter on definitions -- There have been several PhDs in dictionary processing. We ought to be careful not to repeat the mistakes or reinvent the wheel.

Update

1. Built prototype system using hand-generated rules and axioms that work on the three questions. [Will add two more.]

• Prototype details.
• Sample output.

2. Started a plan for converting the prototype system into a working system. See steps involved.

3. Discussed specs for TEF (Textual Evidence Finder) with Greg. See Greg's plan.

• Query generation. See examples of the current system's output.
• Tuple representation
• Scoring

4. Stephen's notes on Open IE coverage issues.

To Do

0. Get a working system soon.

1. Figure out how to do have externally defined functions as predicates in Tuffy.

2. Spec out requirements based on prototype output and identify mismatch with current capabilities.

...

Agenda

• Toward a working system for inference.

• Textual evidence finder

• Definitions extraction