Peter Clark is the Senior Research Manager for AI2. His work focuses upon natural language processing, machine reasoning, and large knowledge bases, and the interplay between these three areas. He has received several awards including a AAAI Best Paper (1997), Boeing Associate Technical Fellowship (2004), and AAAI Senior Member (2014). He received his Ph.D. in Computer Science in 1991, and has researched these topics for 30 years with more than 80 refereed publications.
These working notes are unpublished documents, giving informal summaries and discussions of various research issues. They are pre-publication material and of varying quality, and some of the ideas have since been superceded. A brief description of their contents is given below.
- 2014A discussion on how generic sentences can be interpreted and reasoned with.
- 2011Invited presentation at the DARPA Machine Reading Year Three Kickoff Meeting.
42. Creating BLUE Formulations of the Refinement Phase Test Suite (RTS) Questions: Experience and Reflections2010We spent three months adding biology and lexical knowledge so that AURA could answer a test suite of fluently worded English questions. This document summarizes this experience, and its implications for future QA work with AURA.
- 2010An intermediate reflection on the possible roles of query relaxation in AURA's QA module.
- 2009A short literature survey of the main themes in this area in the database literature.
- 2010A fairly informal discussion of what it would take to infer a hypothesis H from a text T for some selected RTE5 examples, using lexical and world knowledge (as opposed to shallow, statistical matching techniques).
- 2010A technical discussion of how to convert KM's prototypes, as build by knowledge engineers using the AURA system, into the F-Logic-like language SILK.Examples Used in Working Note | PowerPoint Overview | Lisp Implementation
- 2010A description of recent developments in AURA's question-answering capabilities.
- 2010A discussion of the formal semantics of some of the questions posed to AURA, with particular focus on the relationship between universals and hypotheticals.
- 2009A short and sweet PowerPoint summaryAlso see the technical paper Boeing's NLP System and the Challenges of Semantic Representation, for longer discussion and illustrations
34.On the Relation between "Semantically Tractable" Queries and AURA's Question Formulation Facility2009A discussion of Popescu et al's notion of "semantically tractable" questions in natural language interfaces to databases, and the extent to which it can be applied to asking questions in the AURA system.
- 2009An analysis of 22 biology questions that the users had trouble expressing in our controlled language CPL in the AURA system.
- 2009The AURA knowledge-based system currently requires users to reformulate questions in a simplified version of English called CPL. This document discusses some of the obstacles to having AURA understand the original English questions directly.
- 2009How question-answering in a textbook-derived knowledge-based system and the Recognizing Textual Entailment (RTE) challenge are closely related.
- 2009A worked example of how general knowledge of transportation can be used to guide interpretation of biomedical text.
- 2008A short discussion of the issues in, and solutions to, the problem of "missing" (unstated) elements in text.
- 2008A discussion on the use of scripts to understand the MUC-3 kidnapping articles.
- 2007A discussion on how to represent the notion of "improve" and "better", set in the context of language processing (textual entailment).
- 2001ome examples of trying to align variations of an input sentence with background knowledge.
- 2001How background knowledge might be used to resolve indirect references in text.
- 2001Some extremely preliminary thoughts.The toy demo code is also available here.
- 2001This little working note presents some brief thoughts on handling variation among members of a concept, and generating examples of that concept.
- 2001Some further, assorted thoughts on the notion of `views' in a KB. Co-authored with John Thompson and the UT KB group.
- 2001Co-authored with John Thompson. Presents some key concepts and examples for representing processes and their executions in KM.
- 2000This note gives a new discussion of role concepts, superceding the earlier (and in places erroneous) Working Note number 11. This working note is currently being revised and will appear shortly.
- 2000Discusses the notion of "views", namely explicit representations of how general concepts can be applied to more domain-specific concepts. Co-authored with John Thompson, Ken Barker, James Fan, Bruce Porter, Dan Tecuci, Peter Yeh.
- 2000This note follows on from Working Note 17, to provide another example (with less discussion), this one from the molecular biology domain about how a virus works.KM KBs: Classes, Prototypes | Graphlet Database
- 2000 A worked KM example of a familiar AI problem. This paper also compares class-based and prototype-based styles in KM.KM KBs: Classes, Prototypes | Graphlet Database
- 1999Comparison of two representation languages, CCALC and KM, for a simple problem. Co-authored with Joohyung Lee, Vladimir Lifschitz, and Bruce Porter.
- 1999A discussion of the gap between NLP-generated semantic graphs, and a fuller representation of meaning, for a short paragraph of text.
- 1999Discusses the representation and generation of aviation incident 'stories'.Prolog Code
- 1998A discussion of components as theory 'patterns' or 'templates', which get instantiated ('morphed') in various ways for reasoning.
- 1998A more detailed discussion of component data-structures and composition.
- 1998A brief discussion of how to place `role concepts' (eg. Teacher, Producer) in a taxonomy.
- 1996An attempt to enumerate general requirements for a KR system, with some example representational challenges.
- 1996A discussion of conceptual graphs, and how they might be represented in the KR language Algernon.
- 1996A pictorial overview of the DCE Help-Desk Project. (0.8MB compressed, uncompresses to 26MB due to color images).
- 1995A view of components as a self-contained system of relationships, with a clearer delineation. Also see the AAAI'97 paper for the evolution of this theme.
- 1995Describes representing stereotypical action sequences (`scripts') compositionally, employing ideas from Batory's GenVoca approach to software composition.
- 1995An overview and discussion of Ken Murray's KI system for Knowledge Integration.
- 1995Describes representing STRIPS-style actions compositionally.
- 1995Thoughts on dealing with contradictions between components during composition.LIFE Code
- 1995An exporation of connections between the idea of components, contexts, composition, and problem-solving.
- 1994Initial thoughts on representations as compositions, visualized as graph superposition.