Context & Semantic Similarity Measurement Carsten Keler May - - PowerPoint PPT Presentation

Context & Semantic Similarity Measurement

Carsten Keßler May 15, 2008 • GEOG 288MR

ToDo

• Discuss the papers:
• Dey & Abowd (2000) Towards a Better

Understanding of Context and Context-Awareness

• Rodríguez and Egenhofer (2004) Comparing

geospatial entity classes: an asymmetric and context-dependent similarity measure

• Keßler (2007) Similarity Measurement in Context
• Discuss my ideas

Motivation

More Motivation

More Motivation

Even

So: What is Context?

• Dey & Abowd (2000) Towards a Better

Understanding of Context and Context- Awareness

• Motivation: make the vague notion of

context more concrete

UbiComp Definition

„Context is any information that can be used to characterize the situation of an entity. An entity is a person, place, or object that is considered relevant to the interaction between a user and an application, including the user and applications themselves.“

Context Categories

• Location

e.g. Europe, NY, Del Playa, Room 53, Shelf 12b

• Identity

e.g. users & people/devices (!) they interact with

• Activity

e.g. meeting, shopping, sleeping

• Time

e.g. spring, May 15, playoff season, 1210874400

Context Awareness

„A system is context-aware if it uses context to provide relevant information and/or services to the user, where relevancy depends on the user’s task.“

Categorization

• f Context Aware Features
• Information presentation
• Service execution
• Information tagging

Similarity and Context in GIScience

• Rodríguez and Egenhofer (2004)

Comparing geospatial entity classes: an asymmetric and context-dependent similarity measure

Remember MDSM?

Knowledge Representation

• MDSM comes with its own specific

knowledge representation

• Using Backus Naur Form for entity

class definitions

• Ontology for the paper created from

WordNet and SDTS

Context in MDSM

• Domain of application – the set of entity

classes that are subjects of the user‘s interest

• Selection based on user activities, i.e.,
• perations associated with a set of

entity classes

Examples

Diagnosticity / Feature Relevance

• How significant is a specific feature to

distinguish a specific entity class from the other classes in the set under consideration?

• Variability vs Commonality

Variability

• Stress distinguishing features based on
• number of occurrences oi
• number of entity classes n
• number of features l

Commonality

• Counterpart to the variability measure
• Stress common features
• How much does a specific feature

contribute to the characterization of a domain of application?

MSDM & Context in Practice

Intensional context

MSDM & Context in Practice

Extensional context

HST Results

• Commonality approach works better

for intensional context specifications

• Variability approach works better for

extensional context specifications

A More Generic View

• Me (2007) Similarity Measurement in

Context

• Motivation: Find out about the influence
• f context on similarity measurements

– independent of similarity theory

? „How similar is a to b?“

b a

knowledge base

a b

CONTEXT

explicit context input

automatic context capturing „injection“ of additional context (extend the knowledge base)

b a

selection of domain of application

a b

weighting a b

similarity measurement

• utput of

similarity value

My Context Definition

„A similarity measurement‘s context is any information that helps to specify the similarity

• f two entities more precisely concerning the

current situation. This information must be represented in the same way as the knowledge base under consideration, and it must be capturable at maintainable cost.“

„The more similar two contexts are, the less a similarity measurement should change between those two contexts“

False Assumptions…

Some Recent Ideas

• Context rules: concepts may change in

a specific context

Formally

e.g.,

Concept- vs. Instance Affecting Context

knowledge base CONTEXT QUERY

Instance # Type Name ... 1 A ... ... 2 B ... ... 3 C ... ... 4 A ... ... ... ... ... ...

A B C providing concepts for instance annotation instance database STEP A: Calculate similarity ranking for concepts STEP B: Identify matching instances similarity measurement instance selection

cu cd cr ci cu cr

Context Impact Measure

Canal Stream

Channel Ocean Sea

Aqueduct IrrigationCanal Reservoir

Canal River Stream Aqueduct

IrrigationCanal

Reservoir Spring

CIM: a method to quantify the shifts in the results caused by a change in context

Depends on Result Visualization

• 1. River
• 2. Canal
• 3. Aqueduct
• 4. Lake, Sea, Spring
• 5. IrrigationCanal
• 1. River
• 2. Canal, Lake
• 3. Aqueduct, IrrigationCanal
• 4. Spring
• 5. Reservoir

CIM Characteristics

• Works solely on result sets
• Independent of knowledge

representation

• Returns values from 0 to 1
• Focus on the top results via weights
• Should be cognitively plausible
• HPT is in the pipeline

Result Visualization

Coffee, anyone?