 Hello and welcome to the presentation for Thinking Fast and Slow, Combining Vector Spaces and Knowledge Class. First, let me introduce myself. My name is Sudeep Mithur and I am a PhD candidate in Computer Science and part of the Accelerated Coordinated Cybersecurity Lab at UNPC. Knowledge representation is one of the central challenges in the field of artificial intelligence. While representing knowledge through different methods, there is an inherent information plus. For example, by representing knowledge just as vector embeddings, we lose the declarative character of information. While knowledge graphs are adept at asserting declarative information, there is important contextual information around the entity while restricted by the expressibility of the baseline ontology schema that is used to represent the knowledge. To overcome limitations of both and take advantage of their complementary strengths, we have developed the VKT structure that is part knowledge graph and part vector embeddings. Together, they can be used to develop powerful inference methods. Some agents which can be developed over the VKT structure are the query processing system, an agent which adds domain knowledge to embeddings using the knowledge graphs, an agent which uses the vector representation to do link prediction in a knowledge graph and some others. Here we have an example for a VKT structure. The knowledge graph part asserted using Unified Cybersecurity Ontology includes the information that a product Microsoft Internet Explorer has available to execute arbitrary code and analysis. That can be exploited by remote attackers using the means crafted website. The knowledge graph entities are linked to their vector embeddings using the relation as vector. Here is a query processing agent built on the VKT structure. An input query QVKT can be decomposed to multiple components that can run on different parts, namely the knowledge graph part and the vector part. For a cybersecurity example, let the query input be inferred and alert if a vulnerability similar to denial of service is listed in MySQL. The idea here is a part of the query about searching should run on the vector space model and the list and the info part should run on the knowledge graph part. Thank you for listening to this presentation. You can reach me at my email id. You sit here. Have a good day.