当前位置:
X-MOL 学术
›
arXiv.cs.DC
›
论文详情
Our official English website, www.x-mol.net, welcomes your
feedback! (Note: you will need to create a separate account there.)
IoT Virtualization with ML-based Information Extraction
arXiv - CS - Distributed, Parallel, and Cluster Computing Pub Date : 2021-06-10 , DOI: arxiv-2106.06022 Martin Bauer
arXiv - CS - Distributed, Parallel, and Cluster Computing Pub Date : 2021-06-10 , DOI: arxiv-2106.06022 Martin Bauer
For IoT to reach its full potential, the sharing and reuse of information in
different applications and across verticals is of paramount importance.
However, there are a plethora of IoT platforms using different representations,
protocols and interaction patterns. To address this issue, the Fed4IoT project
has developed an IoT virtualization platform that, on the one hand, integrates
information from many different source platforms and, on the other hand, makes
the information required by the respective users available in the target
platform of choice. To enable this, information is translated into a common,
neutral exchange format. The format of choice is NGSI-LD, which is being
standardized by the ETSI Industry Specification Group on Context Information
Management (ETSI ISG CIM). Thing Visors are the components that translate the
source information to NGSI-LD, which is then delivered to the target platform
and translated into the target format. ThingVisors can be implemented by hand,
but this requires significant human effort, especially considering the
heterogeneity of low level information produced by a multitude of sensors.
Thus, supporting the human developer and, ideally, fully automating the process
of extracting and enriching data and translating it to NGSI-LD is a crucial
step. Machine learning is a promising approach for this, but it typically
requires large amounts of hand-labelled data for training, an effort that makes
it unrealistic in many IoT scenarios. A programmatic labelling approach called
knowledge infusion that encodes expert knowledge is used for matching a schema
or ontology extracted from the data with a target schema or ontology, providing
the basis for annotating the data and facilitating the translation to NGSI-LD.
更新日期:2021-06-14