Jurgis Skilters - Frame-based meaning activation model of polysemy processing

In this paper I propose a conception of semantic processing according to which a meaning constellation is foregrounded / activated and several other constellations are backgrounded, i.e. constitute a frame or highly complex and at least partly perceptually determined structure of knowledge crucially necessary for the activation / interpretation of the foregrounded constellation. The activated meaning constellation is also called Figure of semantic articulation and backgrounded components – Ground (Skilters 2004, 2006).
   Figure-Ground-Structure is especially unavoidable in the processing of polysemy. Polysemous meanings are frame-dependent on two sorts of additional (i.e. Frame- or Ground-belonging) semantic resources:

(a) Situational, contextual;
(b) Cognitive-experiential (including perceptual experience).

According to these two mutually interacting resources meaning is defined as a dynamic relational cognitive structure with a situation and an experience foundation and is thus perceptually determined.    The current model of semantic processing is compatible with Barsalou’s (Barsalou 1992, 1993, 1999, Barsalou, Hale 1993) and highlights certain aspects of the frame: (a) recursive open-endedness, (b) constraining, (c) flexibility and (d) relative constancy, and is also compatible with KR models (Minsky 1975) in assuming (but also critically discussing) the view that there exist (a) a frame and a sub-frame relation and (b) relations between different frames. The sub-frames are dynamically, processually inherited from larger super-frames according to cognitive and situational determinants.   The current model is also consistent with a knowledge approach emphasizing the role of structured background knowledge (Murphy 2002) and a Frame Semantic model in the lexical processing (Fillmore 1976, 1982, FrameNet II). Terminologically the frame as used in this study is compatible with notions of a domain, script, theory.   My presentation will also contain an implementation-oriented part regarding the natural language processing with a digital content structuring. (A bottom-up frame-based system of semantic structuring is currently being theoretically worked out at the National Library of Latvia.)   The crucial aspects here are as follows: (1) the cognitive agent, i.e., the user of huge amount data bases (in the sense of Digital Libraries (DL)) partly represented in the natural language, always exploits knowledge from a particular frame, (2) the subject as a frame-based user depends on the framed-aspects “before” and “after” the situation in which he uses the databases. (3) The user’s frame is dynamic, natural (intrinsically based on the user’s cognitive background and on the situation), but also incomplete and imprecise and is confronted with a well defined and complete data base structure (an “artificial static and complete frame”, manually constructed for particular purposes). This is a frame confrontation with several non-trivial consequences for both psychological and computational processing.   In conclusion it will be emphasized that the user-centered HCI DL semantic structuring systems that capture the above mentioned frame-dependant, user oriented aspects are more efficient (e.g. more searchable in respect to ambiguity and vagueness, more appropriate for indexing purposes, ontologically more feasible from the user’s perspective) than non-framed (see evidence for that in another terminology: Blanford et al. 2007, Song et. al. 2007, Xie 2006, Weng et al. 2006, Guo et al. 2003, Mueller 2004).