SHER: Semantic Databases SHER: Semantic Databases using using - - PowerPoint PPT Presentation

1

SHER: Semantic Databases SHER: Semantic Databases using using

• ntologies
• ntologies

Julian Dolby, Julian Dolby, Achille Fokoue Achille Fokoue, , Aditya Kalyanpur Aditya Kalyanpur, Aaron , Aaron Kershenbaum Kershenbaum, Li Ma, Edith , Li Ma, Edith Schonberg, Schonberg, Kavitha Srinivas Kavitha Srinivas

Watson/ Hawthorne, China Research Lab Watson/ Hawthorne, China Research Lab

2

SHER SHER – – Semantic DB Semantic DB using using ontologies

• ntologies

Breakthrough technology that: Breakthrough technology that:

Is highly scalable

Is highly scalable --

• - reasons on 7.7M triples

reasons on 7.7M triples in 7.9 s on benchmarks, scales to 60M in 7.9 s on benchmarks, scales to 60M triples. triples.

Can cleanse inconsistencies in noisy data.

Can cleanse inconsistencies in noisy data. Identifies thousands of logically inconsistent Identifies thousands of logically inconsistent patterns in minutes. patterns in minutes.

Provides explanations of the chain of

Provides explanations of the chain of semantic reasoning for a result set. semantic reasoning for a result set.

Who is interested in SHER? Who is interested in SHER?

• Government:
• Ordnance survey, the British national mapping agency
• NSA
• Healthcare and informatics:
• Mayo Clinic, Chris Chute, Chair of Biomedical Informatics
• Vanderbilt Medical Center (Dan Masys, Director of Biomedical

Informatics),

• Ohio University Medical Center (Philip Payne)
• Columbia University Medical Center (Clinical and Translational

Award Center)

• Pharmaceutical industry:
• Pfizer
• Telecom:
• DoCoMo, a mobile services client.
• Software/ Services Vendors:
• Clark-Parsia, a semantic consulting services company, interest

in licensing/ subcontracting to SHER.

• RacerPro (Franz Inc.) interest in licensing SHER.

When are semantic When are semantic DBs DBs useful? useful?

Example from healthcare domain -- matching patient records to clinical trials, clinical decision support: Patient data: Queries: Patient on methotrexate Patients on immunosuppresants Patient tested positive for Patients with tuberculosis mycobacterium tuberculosis meningitis. Example from pharmaceutical domain (semantic querying of the metadata on microarray gene expression data): Gene Expression data Queries GSE1402 is data about arthritis Gene expression data on disorders

Complex knowledge domains, where there is a semantic gap between data and queries

5

Clinical trials matching case study Clinical trials matching case study (with Columbia Med) (with Columbia Med)

Patient JS is on Cerner: WarfarinSodium10mg

Clinical trial queries in SNOMED: Patients on drugs with active ingredient

• f warfarin? JS

SHER

1 yr EMR

SNOMED

Standard clinical Ontology in 7 countries

Radiology data Laboratory data Pharmacy data

6

SHER Scalability Results vs. state of the SHER Scalability Results vs. state of the art art

240 2M 7M 50 100 150 200 250 300 350 400 450

SHER versus KAON2

SHER KAON2

Dataset sizes (# assertions) Time in seconds

Performance of SHER vs state of the art (KAON2) on OWL benchmark – KAON2 fails on 7M (112 queries) – AAAI 2007.