World Learning Improvement Utilizing Devices of Precisiated Common Dialect Yu Sun (PAMIR) Managers Prof. F. Karray Prof.

0
0
1897 days ago, 702 views
PowerPoint PPT Presentation
World Knowledge Enhancement. Utilizing Tools of. Precisiated Natural ... Present a novel methodology bestowing human learning into factual methodology; ...

Presentation Transcript

Slide 1

World Knowledge Enhancement Using Tools of Precisiated Natural Language Yu Sun (PAMIR) Supervisors Prof. F. Karray Prof. O. Basir Yu Sun (PAMIR)

Slide 2

Previous Works on Term Related Unified Semantic Tree (TRUST) Previous work on TRUST depends on unadulterated factual approach Disadvantages: Huge measurements prompt to many-sided quality Estimated parameters depend on deficient information and thusly loose Yu Sun (PAMIR)

Slide 3

Previous Works (cont) Questions left: How to give human Knowledge into the factual approach How to utilize such human Knowledge to tune the TRUST Yu Sun (PAMIR)

Slide 4

How to Tackle the Issues The novel approach is actualized through: Epistemic Lexicon (EL) Precisiated Natural Language (PNL) Yu Sun (PAMIR)

Slide 5

A lexine utilizes granule fluffy system to characterize a term and its properties, its relations to other lexine terms. Epistemic Lexicon (EL) r ij lexine j lexine i Yu Sun (PAMIR)

Slide 6

Examples of EL is a static representation of information Buyer: establishment (organization, bank) /extremely conceivable , administration staff /conceivable, government /not exceptionally conceivable Seller: foundation (organization, bank) /extremely conceivable , administration staff /conceivable, government /not extremely conceivable, Yu Sun (PAMIR)

Slide 7

Precisiated Natural Language ( PNL ) EL is a static word reference . There need sum up imperatives to repay the inadequacy. PNL is a sub-dialect of precisiable suggestions in NL which is outfitted with a lexicon (EL) characterized by space specialists and Generalized Constraint (standards of conclusion) GC progressively applies EL on NLU Yu Sun (PAMIR)

Slide 8

Examples of GC Generalized Constraint in PNL is utilized to upgrade the static EL through principles Any lexine in EL (word) is bi-sense-coordinated : has a place with one of two headings: in/out or up/down reliant on its unique situation; One lexine may have a place with multi ideas (classes) [ Self ]; If Positive( + ,, for example, concur , professional ) meets Negative( - ,, for example, against , con ), the outcome is Negative(- ) [ blend ] ; Lexine " s relationship can be acquired by its characteristics. For example, Institute has the accompanying characteristics: administration staff , execution , legitimate , embarrassment (which are characterized by the first EL). On the off chance that activity Against (Institute), then Against ( Institute.management-individuals , Institute.performance, Institute.legal , Institute. embarrassment ) . What's more, the other way around [ Inheritance ]. Yu Sun (PAMIR)

Slide 9

Example of GC Enhancing NLU This illustration demonstrates how GC improves the framework to effectively comprehend NL: "Beech-Nut Corp. harm d it s picture over the offer of squeezed apple that ended up being water " EL: Performance-up(sale) [ default ]; Against(damage); Company(image, deal) GC: Against(damage) Against(company.image) Against(company)Against(company.sale) Against(Performance-up(sale)) Performance-down(sale) [ derived ] Yu Sun (PAMIR)

Slide 10

Experiment Pre-handling (manual explanation) 5 records from WSJ with extraordinary subject: "organization takeover" Epistemic Lexicon: including 35 concepts(classes, for example, activity in, execution up, purchaser, vender, and so on. Summed up Constraints: rules representing combinational operations of ideas. Yu Sun (PAMIR)

Slide 11

Results of Applying PNL Experimental results demonstrated the tuned TRUST much nearer to human sound judgment Enlarged and Adjusted EL in the wake of tuning TRUST Buyer: activity in /exceptionally conceivable , agree /extremely conceivable, administration staff /not exceptionally conceivable legitimate /not exceptionally conceivable Seller: activity out /exceptionally conceivable , lawful /conceivable administration staff /conceivable deal /difficult to tell Yu Sun (PAMIR)

Slide 12

Possible Applications To distinguish the semantic importance of a given term ( word sense disambiguation ): " The juice embarrassment constrained Beech-Nut to pay a $ 2.2 million fine and $ 7.5 million to settle a claim " Problem: Beech-Nut might be purchaser or merchant Analysis: Against(scandal)  nearer to Seller; Against(fine)  nearer to Seller; Legal(lawsuit)  nearer to Seller; Solution: Beech-Nut is a Seller Yu Sun (PAMIR)

Slide 13

Future Possible Applications Work as a colleague to a semantic parser: For example: First Pennsylvania had consented to be obtained by Marine Midland in a few month prior… . Midland chooses to venture out the obtaining and it begins a claim. (claim is nearer to dealer than to purchaser) Work as an associate to a syntactic parser: For example: Ralston Co. consented to purchase Beech-Nut Nutrition Corp . with a most loved offer. anaphor determination pp connection Yu Sun (PAMIR)

Slide 14

Conclusions Present a novel approach conferring human information into measurable approach; The preparatory examination indicates PNL and EL enhance the factual approach; The test utilizes one of a kind subject reports and as a part of future, research work will extend to more mind boggling archives. Yu Sun (PAMIR)

Slide 15

Publications Submitted to Journal of Fuzzy Sets and Systems Yu Sun (PAMIR)

Slide 16

Overview of Architecture Yu Sun (PAMIR)

Slide 17

Comparing Procedure of Untuned TRUST and Tuned TRUST Yu Sun (PAMIR)

Slide 18

Yu Sun (PAMIR)

Slide 19

Yu Sun (PAMIR)

SPONSORS