K-CAP Tutorials

Tutorial 1:

Natural Language Generation and Natural Language Interfaces to Knowledge Bases.
  • Claire Gardent, CNRS/LORIA Nancy (France)
  • Eva Banik, Computational Linguistics Ltd
  • Laura Perez-Beltrachini, Nancy University (France)

The languages used for interacting with knowledge bases (OWL, Description Logic, RDF) have a syntax that is difficult for knowledge experts and users to comprehend. To facilitate this interaction, graphical ontology editors such as TopBraid Composer, Protégé and SWOOP have long been developed. More recently however, a new strand of research has emerged which aims to produce Natural Language Interfaces (NLI) to ontologies.

In this tutorial, we will focus on natural language interfaces which draw on Natural Language Generation (NLG) techniques to help the user view, query and edit ontologies. We will start with a brief introduction to NLG. We will then survey the various ways in which NLG have been used to promote knowledge management. Finally, we will outline the main issues involved in developing NLG based NL interfaces to Knowledge Bases.

The tutorial will include live demonstrations of existing NLG-based interfaces to Knowledge Bases and will be structured as follows.

  1. Introduction to NLG. The pipeline NLG architecture: Content planning, Document structuring, Microplanning, Surface realisation.
  2. Applications of NLG to KB management
    • Verbalising the content of an ontology (generating
    • individual descriptions and verbalising axioms)
    • Providing NL feedback on user queries
    • Providing NL feedback on user edits
  3. Natural Language Generation and NL interfaces to Knowledge bases
    • Document structuring and Microplanning: how to promote text coherence and fluency when verbalising an ontology, describing an A-Box individual or providing feedback on a user query ? (Aggregation, linearisation and anaphoric reference in NLI to ontologies).
    • Surface realisation: How to verbalise axioms? (practical upper bound complexity; template based vs grammar based approaches)
    • Portability and adaptivity: Automated generation of lexicons and templates for verbalising ontologies; Constraining generation to take into account the user profile.

Claire Gardent is a senior researcher at the French National Center for Scientific Research (CNRS). Her research focuses on the computational treatment of natural language meaning. She has worked on the automatic acquisition of lexical resources for French as well as on grammar development, syntactic parsing, semantic role labelling and natural language generation. Currently, she is working on extending serious games with sophisticated natural language processing tools (e.g., NL generators and dialog systems). Claire Gardent has been nominated Chair of the European Chapter for the Association of Computational Linguistics (EACL), editor in chief of the french journal "Traitement Automatique des Langues" and member of the editorial board of the journals "Computational Linguistics", "Journal of Semantics".

Laura Perez-Beltrachini is a PhD student at Nancy University (France). Her thesis topic is Natural Language Generation in Computer Aided Language Learning and she is currently investigating how knowledge bases can be used to automatically generate teaching exercises and solutions. Laura has done an Argentinian Computer Science Masters with a research thesis on ontology alignment and completed the Erasmus Mundus Masters "Language and Communication Technology" with a Masters thesis on surface realisation optimization and a Masters project on using natural language to provide NL feedback on user queries supervised by Enrico Franconi.

Until recently Eva was a member of the Natural Language Generation group at the Open University in the UK, headed by Donia Scott and Richard Power, where she received her phd from the Department of Computing in 2010.

After graduating Eva has started up her own company, Computational Linguistics Ltd, taking on various projects related to computational linguistics. She has worked on information extraction and data modelling for Lexis Nexis and currently focuses on natural language generation from knowledge bases for a project at SRI International. Eva also holds an MA in linguistics from the University of Pennsylvania, where she was a member of the XTAG group at the Institute of Research in Cognitive Science and worked on the semantics of Tree Adjoining Grammars.

Tutorial 2:

What Does Usability Mean for Knowledge Capture Projects?
Approaches and Principles to Integrate Usefulness and Usability into Knowledge Capture Software Design.

Most people who design tools want them to be used. A key component to effective tools - whether software or hardware - is design: does the presentation of the tool inform either its usefulness or its usableness? How do we know if we're right or wrong or how wrong about that usefulness/usableness?

Challenge for the Half Day Tutorial: Experts in user interaction design (UI), user experience design (UX), human computer interaction research (HCI) spend years learning and developing methods to get at these questions. So what will a half day tutorial achieve?

This tutorial will provide a core suite of principles and approaches in usability design and evaluation. These approaches will help knowledge capture researchers/experts tune their process knowledge to anticipate both where in their design cycle usability/evaluation will play an effective role and how usability methods might be deployed. By mapping out such leverage points in a project cycle, the tutorial will also help participants discern where conversations may need to take place with UI/UX experts, and offer a vocabulary to help those discussions take place sensibly and effectively.

The tutorial has four main components.

  1. Understanding concepts of Interaction design vs interface design: interaction comes first
    • problem being solved: where humans enter the equation
    • understanding context of use: design approaches for elicitation, capture and validation
    • understand who the people are involved in any stage of the process: models of personas for validation
    • methods for communicating this knowledge within a design team
  2. Considerations in Interaction Representations
    • design as consideration of
      • context, - culture of practice, of use
      • material, - cognitive affordances/constraints
      • artefact - form / function
    • attributes and methods of consideration for each component
    • understanding abstractions and information hiding
  3. Mapping interaction requirements with the interface - towards transparent interaction
    • Prototyping methods
    • engaging with stakeholders
    • where heuristics can be applied in design and evaluation
    • assessing success:
      • efficiency, effectiveness and satisfaction or process, outcome. affect
  4. Where HCI research differs from application of usability/user centered design
    • collaboration with HCI researchers; working with Usability specialists

Take Homes:
Participants completing the tutorial will be able to have effective conversations with usability experts/researchers towards design collaborations, and reasonable understanding of how to cost these into their projects' life cycle. Participants will know and be able to use the differences between interaction and interface design effectively; they will have exposure to cognitive perceptural and performance models to inform choices in UI design; they will have learned methods for determining where in their project cycle interaction design consideration effectively comes into play; they will have a set of methods they can begin to use to develop interaction requirements for their projects/tools; they will have practiced with a set of tools they can use to begin to develop and assess those interaction requirements.

Tutorial is lead by
dr. m.c. schraefel
Reader, Electronics and Computer Science
University of Southampton, UK
Senior Resaerch Fellow, Royal Academy of Engineering
Fellow, British Computer Society
Research Fellow, Web Science Trust