Reasoning capability is one of the greatest hallmarks of human intelligence and is also one of the long standing challenges for Artificial Intelligence. Neuro-scientists and psychologists postulate that human brains are capable of storing infinite number of Concepts. Complex reasoning phenomenon arises from manipulating these concepts and inferring new relations among them. Incorporating reasoning capability into an AI system, therefore, depends on the ability to understand how the concepts in the real world interact with each other. To this end, Relational Reasoning, i.e. learning and inference with relational data, is a promising research direction.
Relational reasoning mostly focuses on learning representations for concepts and relations in existing relational data such as knowledge graphs, that helps to infer new relations among concepts. However, reasoning capability is not limited to simply inferring new relations (e.g. knowledge base completion). It often requires composition and abstraction of several concepts and their interactions (relations). Additionally, in real world, the concept space is not limited to factual knowledge, but also commonsense knowledge plays an important role in reasoning.
Our goal is to design a deep reasoning model, that can combine both factual and common sense knowledge, and has the ability to perform selective composition and abstraction required for complex reasoning.
Broad-coverage knowledge graphs such as Freebase, Wikidata, and NELL are increasingly being used in many NLP and AI tasks. For instance, DB- pedia and YAGO were vital for IBM’s Watson! Jeopardy system. Google’s Knowledge Graph is tightly integrated into its search engine, yielding improved responses for entity queries as well as for question answering. In a similar effort, Apple Inc. is building an in-house knowledge graph to power Siri and its next generation of intelligent products and services.
Despite being rich sources of factual knowledge, cross-domain knowledge graphs often lack a succinct textual description for many of the existing entities. Descriptions of this sort can be beneficial both to humans and in downstream AI and natural language processing tasks, including question answering (e.g., Who is Roger Federer?), named entity disambiguation (e.g., Philadelphia as a city vs. the film or even the brand of cream cheese), and information retrieval, to name but a few. Additionally, descriptions of this sort can also be useful to determine the ontological type of an entity – another challenging task that often needs to be addressed in cross-domain knowledge graphs.