Linked data representation

A Linked-data Model For Semantic Sensor Streams

Authors: P. Barnaghi et al.

Presenter: Haroon Rashid

113/03/15

Problem

• Describe semantically sensor data streams

– Continuous observations and measurements

• Semantic representation of data streams

– Metadata increases the size of transferred data to a greater extent

• Efficient Semantic Queries for large-scale annotated data

213/03/15

Solution

• Use Linked data concept

– Store static, common attributes at one place

– Provide links to static data wherever needed

313/03/15

Approach

• Each observation is associated with

413/03/15

Approach

• Each observation is represented as

5

GEOHASH

SWEET ONT.

13/03/15

RDF Representation

Normal Representation Linked representation

613/03/15

Data stream Representation

Static Source Mobile Source

713/03/15

Publication, Storage, Access Architecture

813/03/15

Data identification

9

No UNIQUE representation to identify data item within a stream

13/03/15

Data Distribution

• Clustering approach

– Store data on distributed repositories

– Results in fast query and resolution mechanisms

• Using K-means Clustering

– Involves both storing and fetching data in/from different clusters

– Needs extensive training

1013/03/15

Evaluation

11

Size of different data stream representation in three different ways

13/03/15

Questions

1. Can we improve data identification which will enhance data resolution/composition?

2. Size of the stream stream/series?

1. Depends on application requirements, bandwidth, caching, freshness

3. On clustering, query efficiency not shown

4. Can we use sample data for showing efficiency of a technique in paper?

1213/03/15

Cont..

5. Efficiency of clustering for mobile scenarios

13/03/15 13