Discussion of probabilistic information retrieval (IR) systems challenges the probability ranking principle in IR from the perspective of (1) signal detection-decision theory and (2) utility theory. Calibration, certainty, and independent assessment are discussed in terms of the relevance of documents, and standard retrieval policies are analyzed.…

Discussion of information-retrieval effectiveness focuses on the use of multiple query or document representations, or multiple retrieval techniques, and combining the retrieval results (also known as data fusion). Experimental results with the SMART System (developed at Harvard and Cornell Universities for over 38 years) indicate combining…

Describes SENTINEL, a prototype information-retrieval system that is a fusion of multiple information-retrieval technologies, integrating n-grams, a vector space model, and a neural network training rule. Discusses three-dimensional visualization capability, precision and recall, mathematical representation of a document, query building, and…

Presents a unified mathematical definition for the classical models of information retrieval and identifies a mathematical structure behind relevance feedback. Highlights include vector information retrieval; probabilistic information retrieval; and similarity information retrieval. (Contains 118 references.) (Author/LRW)

Proposes a novel approach to automatically retrieve keywords and then uses genetic algorithms to adapt the keyword weights. Discusses Chinese text retrieval, term frequency rating formulas, vector space models, bigrams, the PAT-tree structure for information retrieval, query vectors, and relevance feedback. (Author/LRW)

Discusses user preferences for the representation, interpretation, and measurement of the relevance or usefulness of retrieved documents. Highlights include measurement of user judgments on documents; distance between rankings; measures of retrieval effectiveness; normalized performance measure; and the relationship of distance-based measures to…

Describes an approach that combines concepts from information retrieval, fuzzy set theory, and genetic programing to improve weighted Boolean query formulation via relevance feedback. Highlights include background on information retrieval systems; genetic algorithms; subproblem formulation; and preliminary results based on a testbed. (Contains 12…

Discussion of probabilistic information retrieval systems highlights the probability ranking principle and discusses when the standard retrieval policy is optimal. Topics discussed include calibration and refinement; independent assessment of relevance by the inquirer; certainty about the computed probabilities of relevance; and confidence and…

Discussion of document ranking methods to calculate the conceptual distance between a Boolean query and a document focuses on the Knowledge-Based Extension Boolean Model which evaluates weighted queries and documents effectively and avoids problems of previous methods. Semantic networks are discussed, and is-a hierarchies are explained. (21…

Examines the relationship between recall and precision in information retrieval performance and explains why a trade-off between them is unavoidable under certain conditions. Topics discussed include the theoretical behavior of precision and recall; mathematics of the recall curve; an example using a test document collection; and two-stage…

Presents a relevance-feedback strategy that improves the effectiveness of information-retrieval systems based on back-propagation of the relevance of retrieved documents using an algorithm developed in a neural approach. Describes a neural information-retrieval model and reports results obtained with the algorithm in three different environments.…

Discussion of information retrieval focuses on a general method for the construction of measures of similarity taking into account the rank of presentation of the document, called ordered similarity measures. Presents an experimental evaluation used to quantify the filtering impact of a system and discusses recall and precision. (Author/LRW)

Discussion of hypertext browsing proposes a filtering algorithm which restricts the amount of information made available to the user by calculating the set of most relevant hypertext nodes for the user, utilizing the user profile and data clustering technique. An example is provided of an optimal cluster of relevant data items. (Author/LRW)

Presents a preliminary empirical test of a maximum likelihood approach to using relevance data for training information retrieval parameters. Discusses similarities to language models; the unification of document-oriented and query-oriented views; tests on data sets; algorithms and scalability; and the effectiveness of maximum likelihood…

Discusses genetic algorithms in information retrieval, especially for relevance feedback, and evaluates the efficacy of a genetic algorithm with various order-based fitness functions for relevance feedback in a test database. Compares results with the Ide dec-hi method, one of the best traditional methods. (Contains 56 references.) (Author/LRW)