The Knowledge Graph represents the central Semantic Knowledge Base (SKB) of the webLyzard platform. It integrates and aligns factual and linguistic knowledge from different domains and semantics into a singular, linked data graph. Layers for provenance management and access control facilitate its integration with business applications and R&D use cases. A query abstraction and caching layer based on ElasticSearch mitigates the performance bottlenecks inherent to RDF triplestores. The distributed caching mechanism leads to radically reduced response times and enables a highly scalable Application Programming Interface (API). By accessing this feature-rich REST API, our customers can build powerful and responsive business solutions with minimal effort.
Knowledge Graph Applications and Evolution
The webLyzard Knowledge Graph drives most Natural Language Processing (NLP) tasks including entity annotation, sentiment analysis and emotion classification. It also is a crucial part of the query engine behind the webLyzard dashboard.
Named Entity List – Query for “Tesla”
Factual knowledge is derived and enriched with metadata from external and internal sources. The former category includes Linked Open Data (LOD) sources such as Wikidata for organizations or persons and Geonames for locations. The in-house information extraction pipeline yields additional entity types such as events and anniversaries. Linguistic knowledge such as spelling and syntactic variations, multilingual translations, and trending concepts are provided by state-of-the-art Artificial Intelligence (AI) methods. Such methods require large corpora for training purposes. The creation of these corpora benefits from our team’s extensive experience in automated content analysis and media listening. The Knowledge Graph currently stores over 2.5 million entities and is steadily growing at a rate of over 500 entities per day.
Semantic Knowledge Base Editor
To interactively access and modify the entities contained in the Knowledge Graph, the SKB Editor provides an intuitive visual user interface for webLyzard’s customers and research partners. The embedded search mechanisms support a rich set of filter options. Selecting an entity tile in the search results displays the entity’s individual metadata properties. This includes language-specific labels, descriptions, alternate names, locations and image thumbnails. The built-in access control with customizable user roles provides a safe and efficient way to specify and update all aspects of the Knowledge Graph. It is also straightforward to restrict confidential information to a particular use case.
SKB Editor – Query for UN Organizations