Showing papers in "Theoretical Linguistics in 1990"

Connectionism and the computational effectiveness of reasoning

Abstract: It is generally acknowledged that tremendous computational activity underlies some of the most commonplace cognitive behavior. If we view these computations as systematic rule governed operations over symbolic structures (i.e., inferences) we are confronted with the following challenge: Any generalized notion of inference is intractable, yet our ability to perform cognitive tasks such as language understanding in real-time suggests that we are capable of performing a wide range of inferences with extreme efficiency almost as a matter of reflex. One response to the above challenge is that the traditional formulation is simply inappropriate and it is erroneous to view computations underlying cognition as inferences. An alternate response and the one pursued in this paper is that the traditional account is basically sound: The notion of symbolic representation is fundamental to a computational model of cognition and so is the view that computations in a cognitive system correspond to systematic rule governed operations. However, there is much more to a computational account of cognition than what is captured by these assertions. What is missing is an appreciation of the intimate and symbiotic relationship between the nature of representation, the effectiveness of inference, and the computational architecture in which the computations are situated. We argue that the structured connectionist approach offers the appropriate framework for explicating this symbiotic relationship and meeting the challenge of computational effectiveness.

Abstraction, context, and perspectivization – evidentials in discourse semantics

Abstract: ION, CONTEXT, AND PERSPECTIVIZATION EVIDENTIALS IN DISCOURSE SEMANTICS The paper analyzes the role of evidentials such as epistemic muß 'must', offensichtlich/offenbar 'obviously', sicherlich 'surely'/'certainly', schein'seem', and angeblich 'supposedly' within discourse semantics. It is shown that the evidentials under discussion can be distinguished via a small number of parameters. In appropriate contexts these distinctions may be neutralized. Moreover, the interaction with problems of non-monotonicity and perspectivization is investigated. The results are relevant to a \"semantics of pretending\".

Spreading activation and connectionist models for natural language processing

Abstract: High-level cognitive tasks performed by an artificial neural network require both knowledge over a domain and inferencing abilities. To operate in a complex, natural environment neural networks must have robust, reliable and massively parallel inference mechanisms. This paper describes various spreading activation and connectionist mechanisms for inferencing as part of natural language processing systems, including possible techniques to enrich these systems by machine learning. In particular, models which attack one or more important problems such as variable binding, knowledge-intensive learning, avoidance of cross-talk and false classifications are selected for this overview.

Review of Buszkowski/Marciszewski/van Benthem: Categorial Grammar

Abstract: Categorial Grammar is a collection of papers more or less closely related to the topic which the title suggests. It includes a few historical papers which are milestones in the history of categorial grammar. Most of the remaining papers deal with the mathematical aspects of categorial grammar formalisms. Among these are several ones which are worth reading for logicians and computational linguists. In spite of one of the interpretations which the titel of \"categorial grammar\" allows for, the volume does not contain more recent contributions concerning the use of categorial grammar as a formalism to describe natural language syntax.

Meaning metamorphosis in the semiotic topos

Abstract: Denotational semantics is based on the tarskian notion of truth, which grounds in turn on the cartesian introduction of \"clear\" and \"distinct\" ideas as primitive elements of the deductive method that forms the core of rationalistic scientific analysis. To say that a statement expresses a true proposition, if this proposition denotes a fact of reality, presupposes that there is an objective reality consisting of logical atoms, called facts. The propositions denoting these atoms are considered to be context free, i. e. they represent ideas, which are true in all contexts. Prototypes of these ideas are, of course, the mathematical ones. A mathematical proof has an \"essentially\" context free content, since it describes a construction that can be carried out under all circumstances. The quintessence of natural language expressions is, on the contrary, their contextsensitivity, which enables that the same expression can be used in different contexts to mean different things. Therefore denotational semantics cannot by considering meaning as a context free denotational mapping offer an adäquate formalization of natural language meaning. Situation semantics is a contextsensitive extension of model theoretical denotational semantics, since it starts from a notion of proposition, in which a model of the context is already embedded.

Principles of orthography

Abstract: Principles are necessary guides to knowledge acquisition and use. We define and exemplify 7 principles for alphabetic orthographies which, when learned and observed, render orthographic representations biunique. When violated, various types of ambiquity are formed and resolved in ways that are quite regular. The principles of ambiguity formation and resolution are discussed and exemplified in some detail, with the presentation following the cycle: (1) Non-Ambiguity, (2) Ambiguity Formation, (3) Ambiguity Resolution, (4) Non-Ambiguity. Stress falls on the mental movements needed to effect each transition and the similarities among them. Finally, an attempt is made to define the processes of Ambiguity Formation (2) and Ambiguity Resolution (3) in terms of the Movement Features which compose them ambiguity types are treated as dynamical systems.

The organization of orthography

Abstract: Words are displayed on unified trees which may be spoken, heard, read, or written, depending upon the intention of the language knower-user. Despite the widespread view that English orthography is grossly ambiguous, we show that ambiguity bears an inverse relation to context the more the context, the less the ambiguity. In particular, we permit the context of lexical semantics to bind the relations between phonemes and graphemes.

Approaches to parallel processing in linguistics – remarks on a panel at coling 1988 –

Abstract: form always was and still is to define it via equivalence relations. In our case, we should be concerned with equivalences of representations. It is in this context that we should be interested in representations which are not formal definitions of patterns of stored symbols and relations between such patterns but rather constraints of interactivity which determine patterns of activity in a field of simple activity units — similar to the physical wave equations determining the distribution of measurable quantities in a physical representation space. Neurons are like interactive points in space; it must be possible that the internal constraints of interactivity in the brain areas controlling speech are potentially equivalent to constraints over patterns of stored symbols. If such an equivalence representation could be defined, it would provide the best idea about the abstractness of linguistic content, precisely because the representations shown to be equivalent are so different in nature. My own attempts to develop net-linguistics have been guided by this aim. This aim has not yet been attained. We still struggle with the technicalities of representing by means of activity patterns emerging from an underlying substance, appropriately constrained by its interactivity structure. The precise character of this aim is unfortunately blurred by the fact that many researchers are not interested in determining representations which are equivalent to symbolic representations but rather want to show that the interactivity paradigm does allow to represent and to explain empirical facts from different disciplines (linguistics, psychology, neurobiology) indiscriminately, i. e. in one and the same structure. This direction of research characterizes most approaches in Connectionism. They aim at an indiscriminate explanation of regular facts of language together with the explanation of the typical errors of speech, the flexibility of contextual use, the graceful degradation which occurs under psychological or physiological disturbance of speech, and the naturalness of learning. Though there cannot be any doubt that these features must all find their explanation, it is rather doubtful to try to come to an understanding by approaches which, due to their being indiscriminate, will not allow to check their models against the empirical facts established in the different scientific disciplines involved.

A multiprocessing model of natural language processing

Abstract: Computational linguistics has two major goals: (1) develop a technology of natural language processing that can be used in machine translation, userinterfaces, and other applications, (2) model human's language behavior in computers. The multiprocessing model of computation is examined for both these goals and found to be adequate for the former. On the basis of limited information on human language processing phenomena available today the multiprocessing model is found to be inappropriate for characteri2ing human's linguistic ability.

Communicative characteristics of the sentence paradigm: a quantitative approach. article two: sentence type and its immediate context

Abstract: This is the second of a series of articles describing a quantitative approach to the dialogue study. The approach proposed aims at giving a communicative characteristics to the categories of the sentence paradigm proceeding from their functioning in the dialogic text. The point of departure is the \"scheme\" of a real dialogue where every sentence is replaced by a set of categories defining it. In the second article the tendencies of sentences to be \"attracted\" and \"repushed\" by their immediate context are studied from the viewpoint of this scheme and some hypotheses formulated in Article 1 find a confirmation this way.