Logic for Dynamic Real-World Information
June 25, 2018

Workshop at UNILOG'2018 organized by

Erik Thomsen
CTO Blender Logic, Cambridge, Mass, USA

In constructing symbolic logic, Frege, Peano and Russell always had their eye on its application to mathematics alone, and they never gave any thought to the representation of real states of affairs.
(Wittgenstein’s conversation with Waismann, 1929)

Over the past fifty years, the fabric of our society has been radically transformed by successful logic-based applications. In today’s world, logic chips (i.e., CPUs) and the logic-grounded software that controls them support nearly all socio-economic infrastructure, from banking to defense.

Semantic Information Systems ‘SIS’ are software that process information based on some formal (i.e., logical) understanding of the meaning of that information. SIS typically provide interpretation, representation and/or reasoning capabilities. They include:
  .    1. Data-/knowledge-bases (or languages) that represent and support querying and calculations over an internal canonical form. For example: Oracle (2016), Teradata (2016), DB2 (2016), CYC (2016), OWL (2016) and RDF (2014); and
  .    2. Natural Language Processing ‘NLP’ that converts text into an internal canonical form that expresses facts, rules and definitions. For example: OpenNLP (2016), Stanford parser (2016), Berkley parser (2016)

Overwhelmingly, SIS are grounded in some (possibly restricted variant of) First Order Logic ‘FOL’. For example, Relational (or, more properly, SQL) databases, which are still the predominant form of data management for organizations around the world, are varyingly faithful implementations of Codd’s Relational Model (Codd 1970; Mallede et. al. 2013; Libkin 2014), which itself was explicitly grounded in FOL. NLP also uses FOL as its predominant target representation (Collins 1999; Kwiatkowski et.al. 2013). In this sense, logic provides the abstract material technology from which SIS are engineered. SIS where the semantics are pre-defined and the domain knowledge is static within any particular transaction (e.g., banking and e-commerce applications), form the information backbone of our global economy. They are incredibly reliable. FOL has proven to be an extremely successful abstract material technology for these sorts of software applications.

However, for other SIS where the semantics are hard to predefine, or where domain knowledge may need to change within a transaction, what we call “Dynamic Real-World Information Domains”, FOL-based SIS’s exhibit real world problems. Sometimes the problem is that the software generates incorrect inferences based on the information entered into the system. This happens, for example, when the assertions made represent only some of the domain that needs to be reasoned over and bivalency, tout court, is insufficient (McGoveran 1994). Sometimes the problem is that the software cannot compose a formal representation to interpret what a normal adult can speak or understand (Zarri 2017). Sometimes the problem is that the software assumes certain attributes of the objects it is representing, and performs analysis based on those assumptions even though there was information ingested about the objects that a human would have understood as signaling that background assumptions about the domain were false (Al-Fedaghi 2017).

References, Links and Suggested Readings

  • Al-Fedaghi, A. Context-aware software systems: Toward a diagrammatic modeling foundation. Journal of Theoretical and Applied Information Technology. February 2017
  • Barwise J., Etchemendy J. (1999). Language, Proof and Logic. New York: Seven Bridges Press
  • Codd, T. (1970). A relational model of data for large shared data banks. Communications of the ACM, 13.
  • Codd, T. (1990). The Relational Model for Database Management. Version 2. Addison-Wesley Publishing Company
  • Collins, M. (1999). Head-driven Statistical Models for Natural Language Parsing. PhD thesis, University of Pennsylvania, 1999
  • Croft, W. (1991). Syntactic Categories and Grammatical Relations: The Cognitive Organization of Information. The University of Chicago Press
  • Date, C. J. (2000). An Introduction to Database Systems. 8th edition Boston: Addison-Wesley
  • Date, C. J., Darwen, H. (2000). Foundations for Future Database Systems. Boston: Addison Wesley
  • Domingos, Kok, Poon et al., (2006). Unifying Logical and Statistical AI. In AAAI, 6, 2-7
  • Kwiatkowski, T., Choi, E., Artzi, Y., and Zettlemoyer, L. (2013). Scaling semantic parsers with on-the-fly ontology matching. Proceeding of EMNLP. Seattle, Washington
  • Libkin, L (2014). Incomplete data: what went wrong, and how to fix it. in Proceedings of the 33rd ACM SIGMOD-SIGACT-SIGART Symposium on Principles of Database Systems, PODS'14, Snowbird, UT, USA, June 22-27, 2014. ACM, pp. 1-13
  • Lukasiewicz, J. (1970). Philosophical remarks on many-valued systems of propositional logic (1920/1930). in Selected Works. L. Borkowski, ed.; O. Wojtasiewicz, trans. Amsterdam: North-Holland Pub.
  • Mallede, W., Marir, F., Vassiles, V., Algorithms for mapping RDB schema to RDF for facilitating access to deep web. Proceedings of WEB 2013. Seville, Spain
  • McGoveran, D. (Dec.1993 - Mar. 1994). Nothing from Nothing. In Database Programming & Design.
  • Russell S., Norvig P. (2003). Artificial Intelligence. A modern Approach. Pearson Education New Jersey
  • Shavel S., Thomsen E. (1990). A Tractarian Approach to Information Modeling. Wittgenstein-Towards a Re-evaluation. Wien: Verlag Holder-Pichler-Tempsky
  • Steedman M. (2000). The Syntactic Process. Cambridge MA: MIT Press

The goal of this workshop is to explore:

  • Where and how real-world SIS problems (e.g., those that occur in the analytical information systems of large corporations and governments) can be traced to specific characteristics of the logic (e.g., FOL) upon which they are constructed;
  • Modifications to FOL (e.g., the alternative views espoused by Wittgenstein in the Tractatus or by Peirce) that would enable the construction of SIS that can operate more successfully in dynamic information domains.

    Call for papers

    Topics of interest to the workshop include but are not limited to:

    Real world information system problems that can be traced to the logic upon which the information system is built; e.g.;

    • Missing data
    • Non-applicable data
    • Nulls
    • Inaccessible data
    • Word sense disambiguation
    • Semi-autonomous representational layers
    • Autonomous systems
    • Application contexts
    • Data-driven schema updates
    • Multi-agent planning systems with imperfect information

    The justification for specific non-classical logic (features) to solve specific classes of real world information problems such as the representation of missing and meaningless data:

    • Temporal logics
    • Spatial logics
    • Relevance logics
    • Dialetheist logics
    • Mereological/Mereo-topological logics
    • Modal logics
    • Higher order logics
    • Sub-structural logics
    • Tractarian logics
    • Peircean logics

    Principled approaches to specializing/extending logics (constraints, operators) for new domains:

    • Merging specialized features into a new/enhanced multi featured logic
    • Specialized logic blades (akin to specialized RDB blades)
    • Upper ontology (e.g., BFO) logics
    • Cascading constraint-based processing

    Principled approaches to determining the computational properties of semantic representations:

    • Data-driven approaches
    • Definitional and reasoning-based approaches

    Principled approaches to defining computational properties (e.g., atomic operators) that can be used to support semantic information systems:

    • Well formed types/domains
    • Lambda calculus
    • Types as propositions
    • Martin-Löf (or intuitionistic) – inspired type approaches

    ll proposed papers/talks should keep an eye on both theoretical and real world practical aspects. Abstracts (one page) should be sent by December 1st, 2017 via e-mail to: ethomsen@blenderlogic.com  


    Semantic softwares

    		•   .        .  

     SCHEDULE 

    Keynote Speaker


    David McGoveran
    Alternative Technologies, Deerfield Beach, USA
    “Foundational Issues: Still Meaningful”

    Contributing Speakers

    Jean Krivine, Institute of fundamental research in Informatics, CNRS, Paris, France, “Introducing iota: a logic for biological modeling”

    John Samuel and Christophe Rey, CPE, University of Lyon, France and LIMOS, University Clermont-Auvergne, France, “Datalog access to real-world web services”

    Erik Marcade, VP of Advanced Analytics at SAP, “Impacts of Statistical Learning Theory for Enterprise Software”

    Carlos Mario Márquez Sosa, Pontifical Catholic University of Rio de Janeiro, Brazil, “Singular reference, dynamic thoughts and spatial representation”

    David Stodder, The Data Warehousing Institute, Research for Business Intelligence, “Smart, Sentient and Connected: Trends and Directions in Information-Driven Applications”

    Erik Thomsen, Blender Logic, Cambridge, USA, “Logic-Grounded Ontological Fusion of Sensor Data\\ and Natural Language”

    Uwe Wolter and Cyril Pshenichny, Department of Informatics, University of Bergen, Norway and Saint-Petersburg National Research University of Information Technologies, Russia, “A Universal (?) Framework for Representing Knowledge\\ about Real World Phenomena”



    Back to the 6th Universal Logic Congress