Applying BiLSTM and CNN to assist actionable ontology building from FMEA documents

Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

Perustieteiden korkeakoulu | Master's thesis

Authors

Date

2019-10-21

Department

Major/Subject

Data Science

Mcode

SCI3095

Degree programme

Master's Programme in ICT Innovation

Language

en

Pages

55

Series

Abstract

Heterogeneous data sources can lead to high efforts for data integration, data access and data analytics. A possible solution to this problem can be the creation of a Semantic Information model also popularly known as a knowledge graph acting as an intermediate data source on which all data access applications are built. Knowledge graph allows huge organization to tap into the connection from various data sources and efficient data storage and data access techniques. A Failure mode and effect analysis(FMEA) document are used as a starting point for creating semantic information models. FMEA is a critical document for providing predictive failure analysis for product and processes in an automotive supplier company. It consists of simple statements which can be used for information extraction tasks such as named entity recognition and relation extraction. The goal of this thesis is to extract information in the form of entity-relation-entity triples that serves as the basis for knowledge graph creation. As the data is highly company and domain-specific, the entities are labelled manually. Later, Neural network models like BiLSTM and CNN are used for relation classification task from sentences consisting of the entities. Information extraction task typically utilizes an ontology to identify the type of entity and therefore establishing a relation between the entities. However, due to lack of ontology, we make use of an end to end model for entity recognition and relation extraction. Therefore, the results can also serve as the basis for a domain ontology creation for the automotive supplier industry.

Description

Supervisor

Gionis, Aristidis

Thesis advisor

Neale, Carl

Keywords

information extraction, recurrent neural network, convolutional neural network, knowledge discovery, semantic Information model, natural language processing

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-201910275891

Collections

[dipl] Perustieteiden korkeakoulu / SCI