UMBC ebiquity

A Deep Learning Approach to Understanding Cloud Service Level Agreements

Authors: Srishty Saha, Karuna Pande Joshi, and Aditi Gupta

Book Title: Fifth International IBM Cloud Academy Conference

Date: May 24, 2017

Abstract: Educational organizations, like Universities and School Systems, are rapidly adopting Cloud based services to provide Information Technology (IT) infrastructure to their students. These include course offerings, class materials, ​data storage, emailing and collaboration software, virtual computing environment, etc. Moreover, cloud providers, like Amazon, are also providing free computing credits targeted to students. The legal documents associated with cloud based services, such as Service Level Agreements (SLAs), provide information regarding quality and use of cloud services. These documents are often long text-based documents containing domain specific terminology. In addition, this terminology varies from one document or service provider to another. We propose a framework to extract semantically similar terms and entities across cloud service documents using word embeddings and neural networks. Our work is intended to aid cloud service consumers across a variety of fields by providing the ability to understand the services and requirements offered by large-scale commercial cloud services. In some of our previous papers, we have used semantic web and natural language processing to analyze SLA and privacy policy documents for cloud services [1, 2, 3]. In this work, we extend our approach to propose a deep learning-based technique to analyze these documents and populate cloud service ontologies. The preliminary analysis of cloud SLAs documents performed by us showed that deep learning techniques are useful in context disambiguation and identifying semantically similar terminology across services.

Type: InProceedings

Organization: IBM

Tags: cloud service, deep learning, word embedding, natural language processing

Google Scholar: search

Number of downloads: 388


Available for download as

size: 139744 bytes

Related Projects:

Active Project

 ALDA: Automated Legal Document Analytics.