Natural Language Query Answering Using Deep Logical Reasoning and Machine Learning

We show a flexible and powerful system for natural language query answering over diverse sources of information, both structured and unstructured, that combines deep logical reasoning with the results of machine learning. Business end users can directly query and understand the answers, and add new knowledge. For each answer, the system provides a fully detailed, interactively navigable explanation. Benefits include strong extensibility to new information sources, very strong expressiveness of knowledge, and good scalability qualitatively similar to databases. These lead to increased scope of automation in analytics and decision making, at lower cost, with greater agility and higher accuracy, as compared to previous methods.

We employ the Ergo implementation of Textual Rulelog, an expressively powerful yet computationally scalable form of logical knowledge representation and reasoning (KRR) that closely integrates natural language processing (NLP). This combines:

the results of multiple systems for text extraction and other text analysis such as NL parsing, coreference, and entity recognition – based on machine learning (ML);
complex text sentences encoded logically with deep semantics;
graph databases and ontologies;
relational databases;
CSV, XML, and other structured data formats.

We illustrate a subset of these capabilities with a case study from national intelligence analysis.

Dr. Janine Bloomfield is director of operations and marketing, co-founder, and lead knowledge engineer at Coherent Knowledge. Her experience and passion lies especially in data science, education, and communicating technical material to less-technical audiences. Previously she was a senior scientist, on global climate change, at Environmental Defense Fund, an influential USA environmental non-profit. There she analyzed impacts and did science communications, including national media; state and federal policy advising; and science curriculum development. Her background includes a Yale PhD in ecosystem ecology, a Stanford MS in biology, and a post-doc at US Forest Service researching acid rain.

Benjamin Grosof is CTO, CEO, and co-founder of Coherent Knowledge, a software-centric startup that is commercializing a major research breakthrough in logic-based artificial intelligence combined with natural language processing. He is an industry leader in the theory and practice of how to reason with and acquire logical knowledge. He has pioneered semantic technology and industry standards for rules combined with ontologies, their acquisition from natural language, and their applications in finance, e-commerce, policies, e-learning, life science, defense intelligence analysis, legal regulations, and security/privacy. He co-founded RuleML and had driving roles in W3C RIF and OWL-RL. Previously he was senior research program manager in advanced AI for Paul G. Allen, MIT Sloan professor and IBM Research scientist. His background includes a part-time expert consulting practice, Harvard BA, Stanford PhD, 60+ refereed publications, three patents, and five major industry software products.