Patent application number | Description | Published |
20080294982 | PROVIDING RELEVANT TEXT AUTO-COMPLETIONS - A processing device, such as, for example, a tablet PC, or other processing device, may receive non-textual language input. The non-textual language input may be recognized to produce one or more textual characters. The processing device may generate a list including one or more prefixes based on the produced one or more textual characters. Multiple text auto-completion predictions may be generated based on multiple prediction data sources and the one or more prefixes. The multiple text auto-completion predictions may be ranked and sorted based on features associated with each of the text auto-completion predictions. The processing device may present a predetermined number of best text auto-completion predictions. A selection of one of the presented predetermined number of best text auto completion predictions may result in a word, currently being entered, being replaced by the selected one of the predetermined number of best text auto completion predictions. | 11-27-2008 |
20080298691 | FLEXIBLE MQDF CLASSIFIER MODEL COMPRESSION - A memory footprint of an Modified Quadratic Discriminant Function (MQDF) pattern recognition classifier is reduced without resulting in unacceptable classification accuracy degradation. Covariance matrices for multiple classes are clustered into a smaller number of matrices where different classes share the same set of eigenvectors. According to another approach, different numbers of principal components are stored for different classes based on criteria such as class usage frequency, larger variation in writing, and the like, resulting in fewer principal components to be stored in memory. | 12-04-2008 |
20080300833 | COMBINER TRAINING AND EVALUATION WITH RANDOM DATA PARTITION - Available training data is randomly data partitioned to generate a training data set for training base modules of a pattern recognition system and a distinct tune data set for training the combiner of the system. The process can be repeated with randomly generated data sets each time. After the combiner is trained, it can be evaluated employing another random partitioning of the training data to a training data set for training new set of base modules and a distinct test data set for evaluating the combiner providing reliable statistical evaluation of the combined system without the need of a standalone test set. | 12-04-2008 |
20090119301 | SYSTEM AND METHOD FOR MODELING A SESSION-BASED SYSTEM WITH A TRANSACTION-BASED ANALYTIC MODEL - According to an embodiment of the present invention, a method for deriving an analytic model for a session-based system is provided. The method comprises receiving, by a model generator, client-access behavior information for the session-based system, wherein the session-based system comprises a plurality of interdependent transaction types. The method further comprises deriving, by the model generator, from the received client-access behavior information, a stateless transaction-based analytic model of the session-based system, wherein the derived transaction-based analytic model models resource requirements of the session-based system for servicing a workload. According to certain embodiments, the derived transaction-based analytic model is used for performing capacity analysis of the session-based system. | 05-07-2009 |
20090177829 | INTERRUPT REDIRECTION WITH COALESCING - An interrupt redirection and coalescing system for a multi-processor computer. Devices interrupt a processor or group of processors using pre-defined message address and data payloads communicated with a memory write transaction over a PCI, PCI-X, or PCI Express bus. The efficiency of processing may be improved by combining multiple interrupt notifications into a single interrupt message to a processor. For some interrupts on a multi-processor computer, such as those signaling completion of an input/output (I/O) operation assigned to a device, the efficiency of processing the interrupt may vary from processor to processor. Processing efficiency and overall computer system operation may be improved by appropriately coalescing interrupt messages within and/or across a plurality of queues, where interrupts are queued on the basis of which processor they target. | 07-09-2009 |
20090304296 | Compression of MQDF Classifier Using Flexible Sub-Vector Grouping - Systems and methods to compress MQDF data are disclosed herein. A plurality of eigenvectors is identified. Each eigenvector in the plurality of eigenvectors can correspond to a pattern to be recognized. Each eigenvector in the plurality of eigenvectors can be split into sub-vectors. The sub-vectors can then be grouped into one or more groups according to a location of the sub-vectors within each of the eigenvectors. Each group can be associated with location data of the sub-vectors in the group. At least one group can be compressed according to a codebook. The codebook can be identifiable via the location data. | 12-10-2009 |
20090326927 | ADAPTIVE GENERATION OF OUT-OF-DICTIONARY PERSONALIZED LONG WORDS - A system is provided, including a display unit, a memory unit, and a processor. The processor is configured to calculate a mutual information value between a first chunk and a second chunk, and to add a new word to a language unit when a condition involving the mutual information value is satisfied. The new word is a combination of the first chunk and the second chunk. The processor is also configured to add the new word into an n-gram store. The n-gram store includes a plurality of n-grams and associated frequency or count information. The processor is also configured to alter the frequency or count information based on the new word. | 12-31-2009 |
20100027889 | CURVISE HANDWRITING RECOGNITION WITH HIERARCHICAL PROTOTYPE SEARCH - Various technologies and techniques are disclosed that improve cursive handwriting recognition. Cursive handwriting input is received from a user. The system performs a hierarchical prototype search as part of a recognition operation. A same space search is performed against a mixed database that has both print and cursive samples. A same space search is also performed against a cursive database that has only cursive samples. The results of these two same space searches are merged into a combined alternate list. The combined alternate list is then used as a constraint for the dynamic time warp searches that are performed against the mixed and cursive databases, respectively. The results of the dynamic time warp searches are also merged into a final combined alternate list, and the combined alternate list is used to make a recognition decision regarding the user's handwritten input. | 02-04-2010 |
20110007963 | COMBINER FOR IMPROVING HANDWRITING RECOGNITION - Various technologies and techniques are disclosed that improve handwriting recognition operations. Handwritten input is received in training mode and run through several base recognizers to generate several alternate lists. The alternate lists are unioned together into a combined alternate list. If the correct result is in the combined list, each correct/incorrect alternate pair is used to generate training patterns. The weights associated with the alternate pairs are stored. At runtime, the combined alternate list is generated just as training time. The trained comparator-net can be used to compare any two alternates in the combined list. A template matching base recognizer is used with one or more neural network base recognizers to improve recognition operations. The system provides comparator-net and reorder-net processes trained on print and cursive data, and ones that have been trained on cursive-only data. The respective comparator-net and reorder-net processes are used accordingly. | 01-13-2011 |
20110170779 | CURVISE HANDWRITING RECOGNITION WITH HIERARCHICAL PROTOTYPE SEARCH - Various technologies and techniques are disclosed that improve cursive handwriting recognition. Cursive handwriting input is received from a user. The system performs a hierarchical prototype search as part of a recognition operation. A same space search is performed against a mixed database that has both print and cursive samples. A same space search is also performed against a cursive database that has only cursive samples. The results of these two same space searches are merged into a combined alternate list. The combined alternate list is then used as a constraint for the dynamic time warp searches that are performed against the mixed and cursive databases, respectively. The results of the dynamic time warp searches are also merged into a final combined alternate list, and the combined alternate list is used to make a recognition decision regarding the user's handwritten input. | 07-14-2011 |
20110271070 | MEMORY USAGE SCANNING - A memory scanning system may scan memory objects to determine usage frequency by scanning each memory object using a mapping of the processes stored in memory. The scanning may be performed multiple times to generate a usage history for each page or unit of memory. In some cases, scanning may be performed at different frequencies to determine multiple classifications of usage. The mapping may create a detailed topology of memory usage, including multiple classifications of access frequency, as well as several other classifications. Based on the topology, the objects in memory may be copied to another storage medium or optimized for performance or power consumption. | 11-03-2011 |
20120144144 | DYNAMIC MEMORY ALLOCATION AND RELOCATION TO CREATE LOW POWER REGIONS - Memory objects may be allocated and re-allocated within a computer system to consolidate infrequently used memory objects to memory regions that may be operated at lower power. During initial allocation of memory objects, the objects may be placed into high power regions. During subsequent periodic analysis, memory objects in high power regions that are infrequently used may be relocated to lower power regions while memory objects in low power regions that are frequently used may be moved to the high power regions. Various heuristics or logic may be used to handle unmovable objects, shared objects, and other types of objects. | 06-07-2012 |
20130132616 | CHARACTERIZING DEVICE PERFORMANCE BASED ON USER-PERCEIVABLE LATENCY - A method and an apparatus for characterizing performance of a device based on user-perceivable latency. To characterize device performance, a value of a metric may be computed from latencies of operations performed by the device. In computing a value of a metric, latencies may be treated differently, such that some latencies perceivable by a user of the device may have a greater impact on the value of the metric than other latencies that either are not perceivable or are perceived by the user to a lesser degree. Such a performance metric based on user-perceivable latency facilitates identification of computing device that provide a desirable user experience. | 05-23-2013 |