IR and AI: traditions of representation and anti-representation in information processing
IEE Two-day Seminar. Searching for Information: Artificial Intelligence and Information Retrieval Approaches
1980's brought some clarification of the issue within AI, partly because it came in both representationalist (localist) and non-representationalist (distributed) forms, which divided on precisely this issue. Matters were sometimes settled not by argument or experiment but by declarations of faith, as when Charniak said that whatever the successes of Connectionism he didn't like it because it didn't give him any perspicuous representations with which to understand the phenomena of which AI
... . Within psychology, or rather computational psychology, there have been a number of recent assaults on the symbolic reasoning paradigm of AI-influenced Cognitive Science, including areas such as rule-driven expertise which was an area where AI, in the form of Expert Systems, was thought to have had some practical success. In an interesting revival of classic associationist methods, Schvaneveldt developed an associative network methodology for the representation of expertise (Pathfinder, 1990)--producing a network whose content is extracted directly from subjects' responses--and whose predictive powers in classic expert systems environments is therefore a direct challenge to propositional-AI notions of human expertise and reasoning. Within the main AI symbolic tradition, as I am defining it, it was simply inconceivable that a complex cognitive task, like controlling a fighter plane in real time, given input of a range of discrete sources of information from instruments, could be other than a matter for constraints and rules over coded expertise. There was no place there for a purely associative component based on numerical strengths of association or (importantly for his Pathfinder networks) on an overall statistical measure of clustering that establishes the Pathfinder network from the subject-derived data in the first place. Its challenge to traditional AI can be guaged from John McCarthy's classic response to any attempt to introduce statistical notions into 1970's AI: "Where do all these numbers COME FROM?". The Pathfinder example is highly relevant here, not only for its direct challenge to a core area of old AI, where it felt safe, as it were, but because the clustering behind Pathfinder networks was in fact very close, formally, to the clump theory behind the early IR work such as Sparck Jones (1966/1986) and others. Schvaneveldt and his associates later applied the same Pathfinder networks to commercial IR after applying them to lexical resources like LDOCE. There is thus a direct algorithmic link here between the associative methodology in IR and its application in an area that took AI on directly in a core area. It is Schvaneveldt's results on knowledge elicitation by these methods from groups like pilots, and the practical difference such structure make in training, that constitute their threat to propositionality here. This is no unique example of course: even in older AI one thinks of Judea Pearl's long advocacy (1985) of weighted networks to model beliefs, which captured (as did fuzzy logic and assorted forms of Connectionism since) the universal intuition that beliefs have strengths, and that these seem continuous in nature and not merely one of a set of discrete strengths, and that it is very difficult indeed to combine any system expressing this intuition with central AI notions of machine reasoning. Background: Information Extraction (IE) as a task and the adaptivity problem. In this paper, I am taking IE as a paradigm, naive though it still is, of an information processing technology separate from IR; formally separate, at least, in that one returns documents or document parts, and the other linguistic or data-base structures, although one must always bear in mind that virtually all IE search rests on a prior IR retrieval of relevant documents or paragraphs.