Difference between revisions of "Vulcan/MeetingNotes/Aug16 2013"
From Knowitall
(→Update) |
(→Update) |
||
Line 63: | Line 63: | ||
</blockquote> | </blockquote> | ||
: 3. Method: Input sentences that correspond to each assertion. Score assertions using our system and submit to Vulcan's web interface.<br/> | : 3. Method: Input sentences that correspond to each assertion. Score assertions using our system and submit to Vulcan's web interface.<br/> | ||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
; Analysis | ; Analysis |
Revision as of 18:33, 16 August 2013
Update
- System development ( Details on architecture and status)
- 1. Online inference components implemented.
- Proposition generator -- Extract tuples from input sentence and convert into a proposition.
- Evidence finder -- Tuple matching over Open IE Clueweb data.
- MLN Inference -- A wrapper around Tuffy's MLN inferencer.
- 2. Offline components -- axioms and rule generation -- NOT implemented.
- 3. Planning to use Tuffy MLN Inference system directly.
Why Tuffy and not Jena or another inference engine? Why not Alchemy?
- Inference engines such as Jena/OWLim don't directly support multiple inference paths. Community's response is to suggest Datalog/prolog implementations.
- Tuffy supports MLN capabilities in Alchemy but is orders of magnitude faster (what takes 6 hours in Alchemy takes 2 minutes in Tuffy).
- Experiments and Evaluation
Not ready to do evaluation yet but here are some useful details.
- 1. Framework: Vulcan has a good evaluation interface setup. We will use this for starters. (Example output from the evaluation framework.)
- 2. Data: Training/Test splits set up by Vulcan. The questions cover 4-12th and AP exams.
Training = 474 questions.
Test = 290 questions.
Training data distribution and Vulcan's current performance:
Grade All Questions #Mult.Choice and
Non-diag. (MC-ND)Vulcan Performance
on MC-ND4th grade 249 108 55.09% 8th grade 476 125 55.07% 12th grade 446 160 25.83% AP 116 81 45.68% All 1287 474
- 3. Method: Input sentences that correspond to each assertion. Score assertions using our system and submit to Vulcan's web interface.
- Analysis
- 1. Selected 10 propositions that are single Open IE tuples as starting targets.
- 2. Wrote down steps involved in verifying these propositions.
Agenda
To Do (Copied over from previous week)
- System building
- 1. Implement "template matching" using the ClueWeb corpus.Pending
- URL for Open IE backend is available.
- For an assertion A, find sentences that have high overlap. Generate regex patterns for the proposition. Score sentences by how well they match the regex patterns.
- 2. Continue system building.
- Create a derivation scorer stub. This will be replaced with a MLN or a BLP scorer. Done.
- Test with iron nail example.
- 3. Jena API doesn't readily support multiple derivations.
- Ask Jena community to find out if this is possible. Done. Not possible.
- OWLIM as replacement. Done. Doesn't look promising. No response from community.
- 4. Try out Tuffy MLN implemenatation. Done.
- Use output of iron nail example
- If easy to use write wrappers around Tuffy to hook into our system.
- 5 Write evaluation code. Vulcan has a good interface set up.
- Check with Peter.
- 6. Create a system architecture page with a figure and overview of the main components.
Created a System status page instead.
- Created a figure. Added it to system design document.
- Need to create a wiki page for system architecture and overview.
- Experiments Pending
- 1. Run template matching approach as a baseline.
- 2. Run inference system as a baseline.