Title: “Rulelog: Rule-based Knowledge Representation and Reasoning”
The tutorial slides are available here.
- See below for bios. See also more Grosof papers etc. at http://www.mit.edu/~bgrosof/.
Duration: half-day (3.5 hours, excluding 30-minute break).
Two-sentence description of tutorial:
Rulelog is a major basic research advance in fully semantic knowledge representation and reasoning, that has a wide variety of applications and has capable efficient implementations leveraging methods from logic programming and relational/graph databases. It features: highly expressive higher-order formulas, defeasibility, probabilistic uncertainty, bounded rationality, and other strong meta; polynomial-time computational tractability; close combination with natural language processing; and strong complementarity with machine learning as well as other semantic technologies.
Two-paragraph description of tutorial:
In this half-day tutorial, we cover the fundamental concepts, key technologies, recent progress, and outstanding research issues in the area of Rulelog, a leading approach to rule-based knowledge representation and reasoning (KRR). Developed mainly since 2005, Rulelog is much more representationally powerful than the previous state-of-the-art practical approaches, yet is computationally affordable. It is fully semantic and has capable efficient implementations that leverage methods from logic programming and databases, including dependency-aware smart cacheing and a dynamic compilation stack architecture. Rulelog extends Datalog (database logic) with general classical-logic-like formulas – including existentials and disjunctions – and strong capabilities for meta knowledge and reasoning, including higher-order syntax, flexible defeasibility and probabilistic uncertainty, and restraint bounded rationality that ensures worst-case polynomial time for query answering. A large subset of Rulelog is in draft as an industry standard. Rulelog interoperates with graph databases, relational databases and spreadsheets, and expressively simpler rule/ontology systems – and can orchestrate overall hybrid KRR. An exciting research frontier is that Rulelog can combine closely both with machine learning and with natural language processing to interpret and generate English.
The most complete system today for Rulelog is Ergo from Coherent Knowledge, which is supported for both academic and commercial users. A subset of Rulelog is implemented in an open-source Ergo Lite (a.k.a. Flora-2) system. A subset of Rulelog was also implemented in the earlier SILK system from Vulcan. Using Ergo, we will illustrate Rulelog’s applications in deep reasoning and representing complex knowledge – such as policies, regulations/contracts, science, and terminology mappings – across a wide range of tasks and domains in business, government, and academe. Examples include: legal/policy compliance, e.g., in financial services; education/tutoring; and e-commerce marketing. Background assumed of participants is only the basics of first-order-logic and relational databases. This is the first conference tutorial ever offered on Rulelog for a primarily-AI research audience.
Goal of the tutorial:
The target audience is those who are interested in knowledge representation and reasoning (KRR) as a core area in AI. Together with machine learning (NL) and natural language interaction (NLI), KRR forms the tripod basis for the core of AI and cognitive computing. The audience will walk away with an understanding of Rulelog’s key innovative logical and inferencing concepts, its broad applicability, its overall advantages and limitations, a sample of some specific application areas, and its open research topics.
Prerequisite knowledge – more details:
The knowledge will cater to those first learning about declarative logic programs and Rulelog, as well as those who already have some background in them. It will assume only background knowledge of the basics of logical knowledge representation and reasoning: familiarity with the concepts of first order logic and relational database management systems, including querying and theorem proving. Some degree of familiarity with the concepts in one or more of the following will be helpful but is not required: graph database querying (RDF/OWL and SPARQL), ontologies, machine learning, Prolog, production rules, Bayesian probabilistic reasoning, and natural language processing.
Novelty and importance of this tutorial topic:
Rulelog combines several fundamental advances in KRR for complex knowledge (including semantics, theory, and algorithms). It has practical efficient implementations and emerging industrial and academic tools and applications. Greatly extending Datalog, Rulelog includes many of the expressive features that until its advent long eluded practical logical KRR: including defeasibility and probabilistic uncertainty; higher-order general formulas with existentials and disjunctions; and fully semantic bounded rationality enabling polynomial-time scaling. It is perhaps the most flexible and widely-applicable logical KRR available today for deep reasoning and representing complex knowledge – such as regulations/policies, science, and information integration mappings. It has increasingly close relationships to natural language processing and machine learning. There are many exciting open research opportunities both in extending and optimizing the Rulelog KRR itself and in exploring its applications. Both (1) machine learning, including knowledge extraction from natural language, and (2) graph databases/knowledge, including RDF and SPARQL, have each become hot recently in industry as well as research. But both need to be combined with more complex human-authored knowledge and deeper reasoning in order to deliver business/social value more effectively. Techniques for developing such human-authored complex knowledge, including starting from text, have progressed substantially. A large subset of Rulelog is in draft as an industry standard to be submitted to RuleML and W3C as a dialect of Rule Interchange Format (RIF).
Content – Detailed Outline (Preliminary Version):
Note: Examples, and discussion with the audience, are sprinkled throughout
- Introduction and Overview
- Pointers to more info
- Concept of logical Knowledge Representation
- Practical Logic
- Contrast with classical logic
- Main kinds: databases (graph and relational), production rules, FOL, others; RDF, SPARQL, OWL
- Concepts and advantages of “smart data”, “smart rules”, “semantic”, hot areas in industry
- Rulelog features and software
- Strong Meta.
- Reasoning methods and scalability
- Integration points with databases, ontologies, natural language processing (NLP), machine learning (ML)
- Example Rulelog system architecture
- Software tools
- Applications (examples)
- Horizontally: Policy-based decisions, Info Integration, Analytics, HCI, Search, business intelligence, risk management
- Vertically: E-commerce & marketing, Financial services, Personalized E-Learning, E-Science, Security & defense, Health treatment, Insurance
- Textual Rulelog: combining closely with natural language processing
- Textual terminology
- Rule-based mapping between logic and NL for text generation and text interpretation
- Templates (in Ergo, a.k.a. ErgoText)
- Authoring process & methodology (for authoring of knowledge)
- Knowledge management challenges addressed
- Value, variety, and veracity – as well as volume and velocity
- Case Study Demo and Features Tour
- Financial Regulatory/Policy Compliance domain
- Public application pilot by Enterprise Data Management Council
- Drilldowns to facts, non-fact rules, exception rules, data/ontology import, ontology/terminology mapping
- Representational use of higher-order, meta knowledge, defeasibility
- Explanations, text generation, interactive navigability
- Editor tools: syntax checking; dependency analysis; term search
- Concepts and Foundations
- Horn LP, with Functions (LP = declarative logic programs)
- Horn FOL (FOL = First Order Logic)
- Horn LP syntax and semantics
- Comparison to Horn FOL, incl. fundamental theorem
- The “spirit” of LP: avoid disjunction, stay grounded
- Logical Functions: general requirements analysis for KRR
- Computational complexity analysis
- Propositional: linear
- Function-free: polynomial if VB (VB = bounded # of variables per rule)
- Functions without restraint: undecidable
- Functions with restraint: polynomial if VB
- Horn Datalog: function-free Horn LP
- Well-Founded Negation
- Logical nonmonotonicity
- Negation-as-Failure (naf; a.k.a. weak or default negation)
- Declarative LP beyond Horn
- Well-Founded Semantics for LP
- Third truth value: “undefined”. Use to represent paradox.
- Comparison to: Stable Semantics for LP; Answer Set Programs
- RDF, SPARQL, OWL: Revisited
- “Tabling” Algorithms for LP & Rulelog
- Smart cacheing of inferencing results (mixed direction not backward)
- Forest of subgoal attempts and results
- Challenge from nonmonotonicity and higher-order syntax
- Dependency-awareness: table incrementally upon updates
- In-memory: but combine with external querying to scaled-out DBMS
- External Querying
- Orchestrating ensembles of databases and knowledge-based systems
- Restraint: semantic bounded rationality using “undefined”
- Radial restraint: generalized term depth/size limit
- Polynomial time complexity with dial-able degree, for function-ful
- Skipping restraint: control per rule instance
- Unsafety in weak negation and external querying
- Restraint for unsafety and failure to return
- Frame syntax (a.k.a. F-Logic), Object Oriented style
- Transformation sketch
- Relationship to RDF and OWL
- Higher-Order Syntax via Hilog
- Transformation sketch
- Rule identifiers (ID’s) within the logical language
- Meta knowledge to represent source provenance of rules
- Defeasibility via Argumentation Rules
- Conflict (inconsistency flavor)
- Prioritization knowledge: its bases, its mechanics
- Representational uses for defeasibility
- Transformation sketch
- Remedying FOL’s Fragility
- General Formulas (Omniformity)
- Existentials and Skolems
- Omni-directional Disjunction
- Transformation sketch
- Representing Text: Revisited
- Textual Terminology
- Text generation, using Rulelog, e.g., for explanations
- Text interpretation, using Rulelog, e.g., for authoring via templates
- Importing structured knowledge automatically
- External querying via SPARQL. Spreadsheets. RDF. Full OWL.
- Axiomatic semantics for description logic variants
- Ontology/Terminology Mapping
- Explanation as a critical tool
- Role in development lifecycle of knowledge bases
- Justification graphs
- Transformations sketch
- Probabilistic uncertainty
- Evidential reasoning
- Combine results from multiple sources: machine learning, search, heuristic algorithms
- Distribution semantics: probabilities on more complex rules
- Represent more than Bayesian Networks
- Contrast with Markov Logic Networks
- Reactiveness: knowledge update Events, side-effectful Actions
- Lesser Features
- Datatypes, Aggregation, Integrity Constraints, Inheritance, Equality, “Constraints”
- Conclusions and Future Work
- Other summary observations
- Essential expressiveness features
- Transformation/compilation stack
- Advantages and limitations for knowledge management
- Complex knowledge. Deep reasoning. The 5 V’s.
- Big picture KRR-wise. The 5 V’s (variety, value, etc.).
- Open Research Topics in the KRR itself
- Distributed reasoning. Optimization, e.g., for probabilistic uncertainty.
- Reasoning-by-cases. Integration and hybridization with classical logic, “constraint solving”, and answer set programs. Hypotheticals, abduction.
- Transactionality and process flow
- Research Directions – Other Aspects
- Combination with Machine Learning
- Combination with Natural Language Processing
- Industry standards design
- Appendix: References and Resources
- Other summary observations
- Evidential reasoning
Content – List of Additional Materials:
Note: We intend to post the PDF version of the tutorial slideset publicly on our website(s) as well as to grant (non-exclusive) copyright to IJCAI to post it and/or otherwise redistribute it.
Presenters’ Bios – including background in the tutorial area:
Benjamin Grosof (lead presenter) is CTO and CEO (2013-present) of Coherent Knowledge. He is an industry leader in AI knowledge representation, reasoning, and acquisition. He has pioneered semantic technology and industry standards for rules combined with ontologies, their acquisition from natural language (NL), and their applications in finance, e-commerce, policies (including contracts, regulations, and security), and e-learning. He co-founded Coherent Knowledge, a software-centric startup that is commercializing a major research breakthrough in AI logical/probabilistic knowledge representation and reasoning (Rulelog) combined with natural language processing. He also is president of the expert consulting firm Benjamin Grosof & Associates, founded while he was at MIT. Previously he was a senior research program manager at Vulcan Inc. (2007-2013), the asset management company of Paul G. Allen (co-founder of Microsoft). There he conceived and led a large research program in the area of rule-based semantic technologies and artificial intelligence (AI). Before Vulcan, he was an IT professor at MIT Sloan (2000-2007) and a senior software scientist at IBM Research (1988-2000). He co-founded the influential RuleML industry standards design effort and prototyped it in SweetRules, the main bases for the W3C Rule Interchange Format (RIF) standard. He co-founded the International Conference on Rules and Rule Markup Languages for the semantic web (which since became the RR and RuleML conferences). He led the invention of several fundamental technical advances in knowledge representation, including courteous defeasibility (exception-case rules), restraint bounded rationality (scalability in complex reasoning), and rule-based description logic ontologies. He also has extensive experience in user interaction design, and in combining logical methods with machine learning and probabilistic reasoning uncertainty. His background includes 5 major industry software releases, 2 years in previous software startups, a Stanford PhD in AI, a Harvard BA in applied math, 3 patents, and over 60 refereed publications.
Michael Kifer is a Professor with the Department of Computer Science, Stony Brook University, USA. He received his Ph.D. in Computer Science in 1984 from the Hebrew University of Jerusalem, Israel, and the M.S. degree in Mathematics in 1976 from Moscow State University, Russia.
Kifer is a co-founder of Coherent Knowledge, a new startup on semantic technology, and since 2012 he has been serving as the President of the Rules and Reasoning Association (RRA). His interests include Web information systems, knowledge representation, and database systems. He has published four text books and numerous articles in these areas. In particular, he co-invented F-logic, HiLog, and Transaction Logic, which are among the most widely cited works in Computer Science and, especially, in Semantic Web research. Kifer serves on the editorial boards of several computer science journals and chaired a number of conferences. Twice, in 1999 and 2002, he was a recipient of the prestigious ACM-SIGMOD “Test of Time” awards for his works on F-logic and object-oriented database languages. In 2013, Kifer’s paper on Transaction Logic Programming was awarded the Association of Logic Programming’s “Test of Time” award as the most influential paper of 20 years ago. In 2006, Kifer was Plumer Fellow at Oxford University’s St. Anne’s College and, in 2008, he received SUNY Chancellor’s Award for Excellence in Scholarship. He has taught numerous courses at Stony Brook University since 1984.
Paul Fodor is a Research Assistant Professor with the Department of Computer Science, Stony Brook University, USA. He received his Ph.D. in Computer Science in 2011 from the Stony Brook University, New York, preceded by his M.S. degree in 2006 from Stony Brook University, and B.Sc. in Computer Science in 2002 from the Technical University of Cluj-Napoca, Romania.
Dr. Fodor is a co-founder of Coherent Knowledge with over 10 years’ experience in databases research, natural language processing, artificial intelligence and stream processing systems. His work on declarative rule languages and logic used as a specification language and implementation framework for knowledge bases was applied in areas ranging from natural language processing to complex event processing and semantic Web technologies. Through his research, Dr. Fodor has contributed to several large software projects: the IBM Watson natural language processing system for the Jeopardy! Challenge with human champions, the OpenRuleBench suite of benchmarks for analyzing the performance and scalability of rule systems for the semantic Web, the ETALIS declarative complex event processing and stream reasoning system, and the SILK Semantic Inferencing on Large Knowledge. Dr. Fodor was Principal Investigator (PI), Co-PI and contractor for projects funded by both public governmental sources and private companies, such as, PI for the SILK project funded by Vulcan Inc. to develop intelligent textbooks, contractor for the IBM Watson project, contractor for XSB Inc. for the DARPA Component, Context, and Manufacturing Model Library (C2M2L-1) using XSB Prolog, and PI for the Stony Brook University Hospital’s Lung Cancer Evaluation Center management program. He has taught numerous courses at Stony Brook University since 2011.