A Text Extraction-Based Smart Knowledge Graph Composition for Integrating Lessons Learned During the Microchip Design

Abstract

The production of microchips is a complex and thus well documented process. Therefore, available textual data about the production can be overwhelming in terms of quantity. This affects the visibility and retrieval of a certain piece of information when it is most needed. In this paper, we propose a dynamic approach to interlink the information extracted from multisource production-relevant documents through the creation of a knowledge graph. This graph is constructed in order to support searchability and enhance user's access to large-scale production information. Text mining methods are firstly utilized to extract data from multiple documentation sources. Document relations are then mined and extracted for the composition of the knowledge graph. Graph search functionality is then supported with a recommendation use-case to enhance users' access to information that is related to the initial documents. The proposed approach is tailored to and tested on microchip design-relevant documents. It enhances the visibility and findability of previous design-failure-cases during the process of a new chip design.

Transferrable Framework Based on Knowledge Graphs for Generating Explainable Results in Domain-Specific, Intelligent Information Retrieval

“…Those subgraphs are then processed to produce feature vectors that represent the relations between nodes. A similar approach is used in [21], where vectors of the textual documents are considered as graph nodes, while the relations amongst them are calculated using the cosine similarity scores between document pairs. The importance of model-agnostic solutions comes from their ability to use the same explanation concept and apply it for different intelligent models [11,17,22].…”

“…The knowledge graph is capable of representing multiple dimensions of a process. This is accomplished by: (1) defining node-types in the KG [21], (2) expanding the relation definition to connect different nodes from the same type or different types, while defining different levels of relevancies between the nodes [26]. In the proposed approach, graph nodes are defined based on two pillars: information sources and expert-defined rules.…”

“…We implemented a customized text mining pipeline to extract the textual content of the three document types. Extraction steps included cleaning the text from less meaningful terms, which can affect text weighing and similarity algorithms, such as cosine similarity scores calculated from Term Frequency Inverse Document Frequency (TF-IDF) weights [21]. After cleaning the textual content of each document, document vectors were calculated to enable their representation in a machine-readable format.…”

Knowledge Integration in Smart Factories

“…After the raw text is pre-processed, several approaches have been addressed in the literature for the extraction of knowledge from the text, including Named Entity Recognition (NER) [53] and Relation Extraction (RE) [54]. Both methods are directly relevant for a semantic representation of data, since entities and relations can be identified in a meaningful way to define and populate the nodes and edges, in graph based representations of texts, as, e.g., knowledge graphs, while utilizing additional pre-existing domain knowledge [55].…”