LLilog(ID:5414/lli001)

KR Lanugage 

for Language for LInguistic and LOGic methods

written LLILOG

Knowledge representation language from IBM Stuttgart


Related languages
KL-ONE => LLilog   Incorporated features of
Prolog III => LLilog   Built on
LLilog => FORKADS   Based on
References:
  • Studer, Rudi "A Conceptual Model for Time" LILOG-Report 2 December 1986 view details Abstract: For capturing static and dynamic aspects of an application domain on a conceptual level, THM-Nets based on a semantic data model and Petri net concepts have been proposed. In this paper THM-Nets are generalized to timed THM-Nets, thus providing modeling concepts for capturing physical and logical time aspects of a slice of reality. These modeling concepts are based on an appropriate notion of physical and logical time within the semantic: data model THM.
  • Studer, Rudi, and Walter, Bernd "Implementation Aspects of a Natural Language Understanding System in a PROLOG/DB Environment" LILOG-REPORT 3 September 1986 view details Abstract: LILOG is a project for exploring linguistic and logic methods for an automatic understanding of German texts and for an adequate representation of the acquired knowledge. In order to maintain knowledge bases of realistic size, database technology will be used. This paper discusses some of the problems that occur when an existing database system (SQL/DS) is used for representing the various types of knowledge. Additionally, the design of a rapid prototype PROLOG/SQL system will be presented, which supports the exploration of various mapping and access schemes and considers the fact that the used knowledge representation methods will most certainly evolve during the course of the project.
  • Bosch, Peter "Representation and Accessibility of Discourse Referents" LILOG-Report 24 September 1987 view details Abstract: Linguists and philosophers long looked upon definite reference as if it were entirely a matter of the descriptive content of referential expressions. In ordinary discourse, however, such information is usually insufficient to bring about unambiguous reference, certainly for pronouns with their extremely attenuate semantic content, but also for full NPs.

    As a framework for the questions I want to discuss I shall adopt a proposal for a division in short-term working memory which was put forward by Sanford and Garrod (1981) (Section 1.1). Subsequently I shall amend this proposal to account for the role of the descriptive content of referential expressions in accessing the different memory registers. The remainder of Section 1 provides some linguistic evidence for this amended model. Section 2 discusses some problems which this model faces with regard to antecedentless pronouns (2.1) and contrastive reference (2.2) and leads up to the final version of our model for the relation between the representation and accessibility of focused referents, which I shall present in Section 2.2.

  • Pletat, Udo, Beierle, Christoph "The Semantics of Asserting and Retracting Clauses to Logic Programs" July 1987 LILOG-Report 7 view details Abstract: We discuss two approaches for defining the operational semantics of modifying logic programs by means of asserting and retracting clauses. The first approach defines a "logically clean" behavior for assert and retract. Logically clean means that this operational semantics for logic programs including asserts and retracts is equivalent to the model theoretic semantics of the logical skeleton of a program, i.e., where the asserts and retracts are removed. This is achieved by delaying the modification of the program due to the asserts and retracts passed during a proof after the successful evaluation of a goal. We contrast this clean semantics with the PROLOG style of modifying logic programs and discuss the reasons for losing the logical cleanness.
  • Rehkamper, Kalus "Mental Images and Route Descriptions" LILOG-Report 20 August 1987 view details Abstract: Mental images are of great importance in the text comprehension of human beings. Text understanding computer systems which intend to meet the demands of cognitive adequacy must take this fact into account. Human beings use these images to represent knowledge. Thus images are-in addition to propositions-another way of gaining and representing knowledge. In this paper I want to show some of the new possibilities opened up by this second form of representation as well as the restrictions connected with it. Route descriptions form a class of texts which obviously require mental images for their generation and comprehension. During the generation of a route description the informant uses a cognitive map of a quasipictorial format, on which he locates his position, the destination and the route between them. To understand the following description the hearer must re-transform the verbal information into an appropriate format-presumably a combination of propositional and quasi-pictorial representation.
  • Rollinger, Clause, Studer, Rudi, Uszkoreit, Hans and Wachsmuth, Ipke "Text Understanding in LILOG-Sorts and Reference Objects" LILOG-Report 25 August 1987 view details Abstract: The main objective of the project LILOG (linguistic and logic methods) is to develop concepts and methods for understanding German texts and dialogs. "Understanding", in this context, refers to the construction of a semantic representation of a piece of text or of a dialog statement, that is a (partial) model of the situation described in the text. This representation is held in a computer memory and is used by the knowledge processing component, e.g., for extracting information to augment a knowledge base, or for answering questions about the text, etc. As a prerequisite, appropriate means for constructing such a model must be available in a permanent knowledge base. These means must be retrieved and applied to the actual situation by appropriate processes.
  • Schmitt, P. H. "Mathematical Logic and Artificial Intelligence" LILOG-Report 6 January 1987 view details Abstract: This paper discusses some research topics of mutual interest in mathematical logic and artificial intelligence. Among the topics treated are mathematical theorem proving, modal logic, manyvalued logic, reasoning under uncertainty and monotonic logic. While some issues are treated in detail for others only a selected guide to the literature is given.
  • Schmitt, P. H. Computational Aspects of Three-valued Logic LILOG-Report 26 August 1987 view details Abstract: This paper investigates a three-valued logic L , that has been introduced in the study of natural language semantics. A complete proof system based on a three-valued analogon of negative resolution is presented. A subclass of L.sub.3 corresponding to Horn clauses in two-valued logic is defined. Its model theoretic properties are studied and it is shown to admit a PROLOG-style procedure.
  • Seiffert, Roland "Chart Parsing of Unification-Based Grammars with ID/LP Rules" September 1987 LILOG-Report 22 view details Abstract: Earley-style chart parsers for unification-based grammars are now commonly used in implementations of formalisms like PATR-II and others. Some of the most essential extensions to the standard Earley algorithm are shown: the subsumption check for the insertion of new edges into the chart and the restriction of feature structures within the predictor step.

    Graham et al. (1980) propose a method for the efficient encoding of all parse trees for context-free grammars. The representation of the parse forest provides efficient access to every single parse tree for a given sentence. This method is slightly extended for unification-based grammars. It is shown that the parse forest can be built at very little extra cost while the Earley chart is being constructed.

    To combine the well-known advantages of the ID/LP formalism with the full power of unification-based grammars, Unification-ID/LP (UID/LP for short) grammars are defined. ULP acceptability of parse trees for UID/LP grammars can be decided for every complete analysis of an input sentence by checking every local tree in the analysis, just as it is the case for simple ID/LP grammars. The problems with UID/LP parsing that arise from allowing unrestricted unification in combination with constraining LP rules are discussed.

    A parsing algorithm consisting of two steps is proposed. In the first step, an extension of the Earley/Shieber algorithm for unification grammars is used to build a chart with a representation of the parse forest for the input sentence. In the second step, every single parse tree is extracted from the forest and is checked for ULP acceptability.
  • Studner, Rudi, Borner, Stefan "An Approach to Manage Large Inheritance Networks" March 1987 LILOG-Report 8 view details Abstract: When developing large-scale knowledge-based systems, concepts are required for handling knowledge bases on external storage. In this paper we present an approach for managing structured inheritance networks in a database. Actually, we develop a representation of a KL-ONE-Iike formalism in an extended relational database system supporting non-first-normal-form relations. Of special importance is the handling of the general hierarchical structure provided by KL-ONE within the tree structure offered by non-first-normal-form relations.
  • Wachsmuth, Ipke "On Structuring Domain-Specific Knowledge" March 1987 LILOG-Report 12 view details Abstract: This paper presents a proposal on how domain-specific knowledge of both conceptual and assertional nature can be structured. The aim is to devise a way that allows large amounts of domain-dependent knowledge to be used by a knowledge-based system while keeping the system manageable. The proposal grounds on findings from empirical research on the acquisition of domain-specific knowledge. It is presented abstractly in the form of principles that are to be understood as a specification rather than a symbol-level description for a representation scheme. The model comprised by these principles suggests domain-specific knowledge be organized in nested packets of knowledge elements. The central notions of "visible" and "reachable" knowledge are used to characterize static and dynamic access conditions.
  • Wesche, Birgit "At Ease with "at"" August 1987 LILOG-Report 23 view details Abstract: In representing the meaning of a preposition, the matn question that arises is whether to assume a polysemous lexem or a number of homonyms according to the various senses the respective preposition can adopt. In this paper it will be shown that by taking a prototypical approach each preposition can be represented as one coherent concept. Exemplified by the concept of the English preposition "at"---distinguished by its diverse range of meanings such as spatial, temporal, causal, etc.--the systematic relation between the different senses will be illustrated. The argumentation is centered around the assumption of a prototypical concept, which in the case of "at" is a spatial one. Departing from this prototype new senses develop successively, motivated by the fact that a certain domain is conceptualized similarly or analogously to an already existing concept of the respective preposition.
  • Wesche, Birgit, Renz, Ingrid "Word Order and Focus Projection" LILOG-Report 20 April 1987 LILOG-Report 13 view details Abstract: One of the major problems that has to be confronted in a natural language system for German is its relatively free word order. If one takes into account, however, that word order is to a large extent motivated by pragmatic considerations the choices for placing the single constituents within a sentence narrow down considerably. A factor by which pragmatic aspects are reflected quite explicitly is stress. In this paper we will outline the regularities that result from the strong interaction of word order and stress in German, and show how these can be exploited even within an NL system which is based on written input only. Taking such an approach will have an impact on the encoding of lexical entries, on the formulation of syntactic rules, on parsing strategies, and it will, furthermore, support the component of semantic representation in that we will have a clearer insight as to which elements constitute the core message of a sentence.
  • Godehard Link, Hinrich Schütze "The Treatment of Plurality in L-LILOG" Text Understanding in LILOG, Integrating Computational Linguistics and Artificial Intelligence, Final Report on the IBM Germany LILOG-Project January 1991 view details Abstract: The logic-based knowledge representation language L_LILOG, which is used to represent both the semantic background knowledge as well as the information extracted from German texts within the LILOG project, is discussed. L_LILOG integrates frame-like features-value descriptions used in computational linguistics into an order-sorted predicated logic framework. The basis design principles of L_LILOG and examples of how L_LILOG is used to model real world knowledge are presented. The implementation of the first LILOG prototype is described. A formal semantics definition is provided.
  • Klose, G. "Ontologie Und Axiomatik Der Wissensbasis Von Lilog Wissenmodellinerung Im IBM Deutschland Lilog-Project" Springer Verlag 1992 view details
  • Beierle, C.; Pletat, U.; Studer, R. "Knowledge Representation for Natural Language Understanding: The L_LILOG Approach" IEEE Transactions on Knowledge and Data Engineering 5(3) June 1993 view details Abstract: The logic-based knowledge representation language L_LILOG, which is used to represent both the semantic background knowledge as well as the information extracted from German texts within the LILOG project, is discussed. L_LILOG integrates frame-like features-value descriptions used in computational linguistics into an order-sorted predicated logic framework. The basis design principles of L_LILOG and examples of how L_LILOG is used to model real world knowledge are presented. The implementation of the first LILOG prototype is described. A formal semantics definition is provided.