Abstract: A system and method for generating context vectors for use in storage and retrieval of documents and other information items. Context vectors represent conceptual relationships among information items by quantitative means. A neural network operates on a training corpus of records to develop relationship-based context vectors based on word proximity and co-importance using a technique of "windowed co-occurrence". Relationships among context vectors are deterministic, so that a context vector set has one logical solution, although it may have a plurality of physical solutions. No human knowledge, thesaurus, synonym list, knowledge base, or conceptual hierarchy, is required. Summary vectors of records may be clustered to reduce searching time, by forming a tree of clustered nodes. Once the context vectors are determined, records may be retrieved using a query interface that allows a user to specify content terms, Boolean terms, and/or document feedback.
Abstract: A systolic array of processing elements is connected to receive weight inputs and multiplexed data inputs for operation in two dimension convolution mode, or fully-connected neural network mode, or in cooperative, competitive neural network mode. Feature vector or two-dimensional image data is retrieved from external data memory and is transformed via input look-up table to input data for the systolic array. The convoluted image or outputs from the systolic array are scaled and transformed via output look-up table for storage in the external data memory. The architecture of the system allows it to calculate convolutions of any size within the same physical systolic array, merely by adjusting the programs that control the data flow.
Abstract: A method for performing a variety of expert system functions on any continuous-state feedforward neural network. These functions include decision-making, explanation, computation of confidence measures, and intelligent direction of information acquisition. Additionally, the method converts the knowledge implicit in such a network into a set of explicit if-then rules.
Abstract: An automated real estate appraisal system (100) and method generates estimates of real estate value using a predictive model such as a neural network (908). The predictive model (908) generates these estimates based on learned relationships among variables describing individual property characteristics (905) as well as general neighborhood characteristics at various levels of geographic specificity (906). The system (100) may also output reason codes indicating relative contributions (1009) of various variables to a particular result, and may generate reports (701) describing property valuations, market trend analyses, property conformity information, and recommendations regarding loans based on risk related to a property.
October 19, 1992
Date of Patent:
November 1, 1994
Allen Jost, Jennifer Nelson, Krishna Gopinathan, Craig Smith
Abstract: A method for generating context vectors for use in a document storage and retrieval system. A context vector is a fixed length list of component values generated to approximate conceptual relationships. A context vector is generated for each word stem. The component values may be manually determined on the basis of conceptual relationships to word-based features for a core group of word stems The core group of context vectors are used to generate the remaining context vectors based on the proximity of a word stem to words and the context vectors assigned to those words. The core group may also be generated by initially assigning each core word stem a row vector from an identity matrix and then performing the proximity based algorithm. Context vectors may be revised as new records are added to the system, based on the proximity relationships between word stems in the new records.
Abstract: A systolic array of processing elements is connected to receive weight inputs and multiplexed data inputs for operation in feedforward, partially-- or fully-connected neural network mode or in cooperative, competitive neural network mode. Feature vector or two-dimensional image data is retrieved from external data memory and is transformed via input look-up table to input data for the systolic array that performs a convolution with kernal values as weight inputs. The convoluted image or neuron outputs from the systolic array are scaled and transformed via output look-up table for storage in the external data memory.