Xerox Corporation

Palo Alto Research Center Inc.

Other

Conduent Business Services, LLC

Daniel Davies

Julia A Craig

Rong Zhou

Glen W Petrie

Leigh L Klotz, Jr

Robert S Bauer

Dan S Bloomberg

Ying Di Zhang

Ramesh Nagarajan

Eric Saund

Daniel Gureasko Bobrow

Fritz F Ebner

Ajay Raghavan

Craig Eldershaw

Juan Liu

Ronald M Kaplan

Bhaskar Saha

Margaret Helen Szymanski

Raymond J Clark

Dale T Platteter

Robert Eugene Weltman

Isaak Rivshin

Steven C Bagley

Kyle D Dent

Jesse Vig

David Garcia

Harish Arsikere

Michael W Marchetti

Kelly L Coupe

Chuanjiang Luo

Luke R Plurkowski

Edward E Ewer

Wayland R Hicks

A computing system framework and method for configuration thereof are provided. A plurality of processing modules are accessed. Each processing module includes a plurality of processing nodes and each processing node is associated with an intra-module port and an inter-module port. A plurality of intra-module networks are formed. Each intra-module network includes connections between at least a portion of the processing nodes in one of the processing modules via the associated intra-module ports. An enclosed shape of the processing modules is formed by connecting at one inter-module port on each processing module to one inter-module port on an adjacent processing modules. A cable is linked between one of the inter-module ports of one processing module of the enclosed shape to an inter-module port of another processing module of a different group of interconnected processing modules.

Computing system framework and method for configuration thereof

A computing system framework and method for configuration thereof are provided. A plurality of processing modules is accessed. Each processing module includes a plurality of processing nodes and each processing node is associated with an intramodule port and an intermodule port. The processing modules are connected in a ring via intermodule connections between at least a portion of the intermodule ports of the processing modules. A network switch is arranged in a center of the ring of processing modules and connections are formed between the network switch and at least one of the processing modules by connecting every Sth processing module to the network switch, connecting every Sth and Sth&#x2212;1 processing modules to the network switch, or by connecting every Sth and Sth&#x2212;r processing modules to the network switch. S is a number of steps between the processing modules.

Embodiments of the present invention provide a system for automatically extracting conversational structure from a voice record based on lexical and acoustic features. The system also aggregates business-relevant statistics and entities from a collection of spoken conversations. The system may infer a coarse-level conversational structure based on fine-level activities identified from extracted acoustic features. The system improves significantly over previous systems by extracting structure based on lexical and acoustic features. This enables extracting conversational structure on a larger scale and finer level of detail than previous systems, and can feed an analytics and business intelligence platform, e.g. for customer service phone calls. During operation, the system obtains a voice record. The system then extracts a lexical feature using automatic speech recognition (ASR). The system extracts an acoustic feature. The system then determines, via machine learning and based on the extracted lexical and acoustic features, a coarse-level structure of the conversation.

System for automatic extraction of structure from spoken conversation using lexical and acoustic features

A system and method for network switching is provided. A plurality of processing modules is accessed and each processing module includes a plurality of processing nodes. Each processing node is associated with an intra-module port and an inter-module port. At least a portion of the processing nodes are connected within each processing module via the intra-module ports. A ring of the processing modules is formed via inter-module connections between a portion of the inter-module ports of the processing modules. One of the processing nodes of at least a portion of the processing modules is connected with a network switch.

System and method for network switching

One embodiment of the present invention provides a computing system. The computing system includes a processor and a computer-readable storage medium. The computer-readable storage medium stores instructions which when executed by the processor cause the processor to perform a method. The method comprises obtaining from a first data structure one or more pointers to corresponding elements in a second data structure. A respective vertex of a graph corresponds to an element in the first data structure and the graph supports multiple edge types between respective vertex pairs. The method further comprises obtaining from the second data structure a respective edge type associated with a respective vertex and a respective successor vertex of the edge type and enumerating a respective successor vertex of an edge type of a vertex from the second data structure based on a pointer in an element in the first data structure associated with the vertex.

Efficient representations of graphs with multiple edge types

A system and method for making and using a computing system framework with unified storage, processing, and network switching fabrics are provided. Processing nodes, either physical or virtual, are associated with intra-module ports, inter-module ports, and local storage spaces. A plurality of processing nodes are linked through intra-module ports to form processing modules. A plurality of the processing modules are connected through inter-module ports to form the computing system. Network switch can be incorporated into intra-module or inter-module connections. Several inter-module connection schemes, which can be adapted to use with existing network packet routing algorithms, are disclosed. Each processing node needs only to keep track of the states of its directly connected neighbors, obviating the need for a high-speed connection to the rest processing nodes within the system. Dedicated network switching equipment can be flexibly employed and network capacity grows naturally as processing nodes are added.

Computing system framework with unified storage, processing, and network switching fabrics incorporating network switches and method for making and using the same

One embodiment of the present invention provides a system for dynamically assigning vertices to processors to generate a recommendation for a customer. During operation, the system receives graph data with customer and product vertices and purchase edges. The system traverses the graph from a customer vertex to a set of product vertices. The system divides the set of product vertices among a set of processors. Subsequently, the system determines a set of product frontier vertices for each processor. The system traverses the graph from the set of product frontier vertices to a set of customer vertices. The system divides the set of customer vertices among a set of processors. Then, the system determines a set of customer frontier vertices for each processor. The system traverses the graph from the set of customer frontier vertices to a set of recommendable product vertices. The system generates one or more product recommendations for the customer.

System and method for improved parallel search on bipartite graphs using dynamic vertex-to-processor mapping

A system and method for making and using a computing system with unified storage, processing, and network switch fabrics are provided. Processing nodes, either physical or virtual, are associated with intra-module ports, inter-module ports, and local storage spaces. A plurality of processing nodes are linked through intra-module ports to form processing modules. A plurality of the processing modules are further connected through inter-module ports to form the computing system. Several inter-module connection schemes are described, each of which can be adapted to use with existing network packet routing algorithms. Each processing node need only to keep track of the states of its directly connected neighbors, obviating the need for a high-speed connection to the rest processing nodes within the system. As a result, dedicated network switching equipment is not needed and network capacity grows naturally as processing nodes are added.

Computing system with unified storage, processing, and network switch fabrics and method for making and using the same

A method () is provided for classifying an image as being one of a daytime image or a nighttime image. The method includes: obtaining an image to be analyzed (); determining a first parameter (n) representative of a first amount of pixels in the image that are sufficiently red or yellow (); determining a second parameter (n) representative of a second amount of pixels in the image that are sufficiently light (); and classifying an image as one of a daytime image or a nighttime image based upon the first and second parameters ().

Day night classification of images using thresholding on HSV histogram

One embodiment of the present invention provides a computing system, which identifies, for graph clustering, first and second sets of vertices in a graph. A respective vertex in the first set of vertices is an exemplar. A respective vertex in the second set of vertices is not an exemplar and not associated with a vertex identifier. The system assigns a vertex identifier of a first vertex in the first set to a second vertex in the second set based on type and strength of relationship between the vertices. The vertices represent elements in a data set. The vertex identifiers of the first and second vertices do not change before convergence. The system allocates the first vertex and the second vertex to a first vertex cluster associated with the vertex identifier, and detects convergence based on modularity associated with vertex clusters representing information of interest in the data set.

Efficient detection of information of interest using greedy-mode-based graph clustering

One embodiment of the present invention provides a computing system. The computing system includes a processor and a memory. The memory stores instructions which when executed by the processor cause the processor to perform a method. The method comprises identifying a first vertex and a second vertex in a graph as candidates for a link between the first and the second vertices. The first and second vertices represent elements in a data set. The method further comprises determining a link indicator based on a first link count, which indicates a sum of respective weights of links of a vertex, associated with the first vertex and a second link count associated with the second vertex. The method also includes determining whether to connect the first and second vertices with a link based on the link indicator's conformity with a link threshold, which determines whether to connect two vertices using a link.

Efficient link management for graph clustering

One embodiment of the present invention provides a computing system. The computing system includes a processor and a computer-readable storage medium. The computer-readable storage medium stores instructions which when executed by the processor cause the processor to perform a method. The method comprises assigning a vertex identifier of a first vertex to a second vertex in a graph based on an edge between the first vertex and the second vertex. The vertices represent elements in a data set, the edge represents a type and strength of relationship between the vertices. The method further comprises provisionally assigning each vertex to a vertex cluster associated with the vertex identifier, and converging the assignments based on a measure of cluster quality associated with all the vertex clusters. Each vertex cluster represents information of interest in the data set. One measure of clustering quality is called 'modularity.'

Efficient detection of information of interest using mode-based graph clustering

One embodiment of the present invention provides a system for generating a product recommendation. During operation, the system receives graph data indicating vertices and edges of the graph. The vertices represent customers and products and the edges represent purchases. The system then receives a query of the graph to determine a product recommendation. Next, the system generates a finite-state machine (FSM) based on the query, executes the query, and determines whether a current state of the FSM is a traversal state. In response to the current state being a traversal state, the system generates a traversal FSM. The system then searches the traversal FSM for a nearest future traversal state, generates a bitmask for the future traversal state, and utilizes the generated bitmask when executing the future traversal state to generate the product recommendation.

High-performance graph analytics engine making recommendations using a finite state machine/FSM, bitmasks, and graphs with edges representing purchases, and vertices representing customers and products

A method for determining a marking material usage estimate for a print job to be performed by a printing system is provided. The method is implemented in a computer system having one or more processors configured to execute one or more computer program modules. The method includes receiving compressed data of an image of the print job, wherein the image has a plurality of image pixels each having an image pixel intensity value, wherein the image pixels are transformed into the compressed data using a compression scheme; obtaining marking material usage statistics by partially decompressing the compressed data so as to obtain average pixel intensity values of the image pixels in the image; and determining the marking material usage estimate for the print job using the obtained marking material usage statistics.

System and method for estimating marking material usage for a printing system

A system which generates a floor plan of a building includes: a camera which obtains a series of images as it is moved, each image represented by a first data set representing color and a second 3D data set representing depth. A processor generates the floor plan from the image data, defined by polylines that represent structures of the building and polygons that represent an area which has been observed by the camera. Suitably, the processor: performs a local matching sub-process which is operative to align two adjacent images with one another; performs a global matching sub-process in which key images and registered to one another; finds a 2D subset of points from the image data associated with each image, corresponding to a plane defined therethrough; determines the polylines based on said subset of points; and defines the polygons based on the polylines and a determined pose of the camera.

Growing community of inventors

Palo Alto, CA, United States of America

Daniel Davies