A Microfeature Based Approach Towards Metaphor Interpretation
Department of Computer Science
The University of Alabama
Tuscaloosa, AL 35487, USA
This paper advocates a microfeature based approach towards developing computational models for metaphor interpretation. It is argued that the existing models based on semantic networks and mappings of complex symbolic structures are insufficient and inappropriate for modeling metaphors. A connectionist model of metaphor interpretation based on microfeatures is presented, which tries to take into account some important issues, such as accurate capturing of similarity, automatic formation of features, contextual effects, elimination of long paths in conceptual hierarchies, salience imbalance, and feature enhancement. Some of these issues have broad implications in cognitive modeling.
Metaphor is an important cognitive phenomenon, and it is of great interest to AI, philosophy, psychology, linguistics, and literary studies. In fact there has been a surge of interests in the past few decades in the philosophy and linguistics communities in the nature and the process of metaphor interpretation; more recently, there are also increasingly more interests in the AI community in non-literal language and analogical reasoning, in both of which metaphor occupies a prominent place. Among the voluminous studies of metaphor, computational models are relatively few. Most existing computational models of metaphor are based on semantic networks and mappings of complex symbolic structures, or in other words, based entirely on the traditional symbolic AI methods. I would like to argue that such symbolic models are insufficient and inappropriate for modeling metaphor in general, on the basis of a number of important considerations. I will instead propose a microfeature based (connectionist) approach towards developing computational models for metaphor interpretation.
2 Semantic Network Based Approaches
Most existing models of metaphor interpretation take semantic network based approaches: they represent the requisite linguistic (and world) knowledge in some
kinds of semantic networks with hierarchically structured concepts, and metaphor interpretation is accomplished through traversing and mapping the hierarchical conceptual structures in some ways. Similarities of words and other linguistic entities are computed from structural relations among entities within a hierarchy. Such models (especially, for example, Martin 1988 and Fass 1991) achieved certain successes within limited domains and/or strictly controlled environments. However, there are many valid questions and objections to these models. Let us look into some representative existing models. Martin (1988) presents a system for dealing with conventional metaphors, i.e., metaphors that reflect a core set of correspondences that can manifest in various ways (Lakoff and Johnson 1980). For example, How can I enter Emacs?" or I am in Lisp", where a computer program is uniformly viewed as an enclosure. His system consists of large semantic networks with conceptual hierarchies and a set of procedures that operate on them. When a verb (e.g. enter) is found to be incompatible with (violating some constraints of) the object (e.g. Emacs), a procedure is called to find a coherent mapping between the domain of the verb and the domain of the object, from a core set of conventional metaphors stored in the system; once such a mapping is found, the verb in the source domain is mapped into a corresponding concept in the target domain (i.e., a concept suitable for the object, e.g. invoke Emacs). New variations of known conventional metaphors can also be handled, through detecting their similarity to existing ones by finding a path through the conceptual hierarchy between the new instance and the known instances. For example, once the system knows I am in Lisp", it can also make sense of I am in Emacs" by finding the relation between Lisp and Emacs in the hierarchy (the sibling relation in this case, since they are both computer programs).
Veale and Keane (1992) deal also with conventional metaphors. They divide the interpretation process into two steps: first a scaffold of core (semantic network) structures is constructed in accordance with some conventional metaphors which is already known and stored in the system; then the structure is fleshed out with details, from general world knowledge and/or the context. What attributes to transfer from the source domain (e.g. physical enclosures) to the target domain (e.g. computer programs) and their respective pre-conditions for valid