Patent application number | Description | Published |
20090132087 | Learning Capture Points for Humanoid Push Recovery - A system and method is disclosed for controlling a robot having at least two legs, the robot subjected to an event such as a push that requires the robot to take a step to prevent a fall. In one embodiment, a current capture point is determined, where the current capture point indicates a location on a ground surface that is the current best estimate of a stepping location for avoiding a fall and for reaching a stopped state. The robot is controlled to take a step toward the current capture point. After taking the step, if the robot fails to reach a stopped state without taking any additional steps, an updated current capture point is determined based on the state of the robot after taking the step. The current capture points can be stored in a capture point memory and initialized based on a model of the robot. | 05-21-2009 |
20100161120 | Intelligent Stepping For Humanoid Fall Direction Change - A system and method is disclosed for controlling a robot having at least two legs that is falling down from an upright posture. An allowable stepping zone where the robot is able to step while falling is determined. The allowable stepping zone may be determined based on leg Jacobians of the robot and maximum joint velocities of the robot. A stepping location within the allowable stepping zone for avoiding an object is determined. The determined stepping location maximizes an avoidance angle comprising an angle formed by the object to be avoided, a center of pressure of the robot upon stepping to the stepping location, and a reference point of the robot upon stepping to the stepping location. The reference point, which may be a capture point of the robot, indicates the direction of fall of the robot. The robot is controlled to take a step toward the stepping location. | 06-24-2010 |
20100161126 | Humanoid Fall Direction Change Among Multiple Objects - A system and method is disclosed for controlling a robot having at least two legs, the robot falling down from an upright posture and the robot located near a plurality of surrounding objects. A plurality of predicted fall directions of the robot are determined, where each predicted fall direction is associated with a foot placement strategy, such as taking a step, for avoiding the surrounding objects. The degree to which each predicted fall direction avoids the surrounding objects is determined. A best strategy is selected from the various foot placement strategies based on the degree to which the associated fall direction avoids the surrounding objects. The robot is controlled to implement this best strategy. | 06-24-2010 |
20100161131 | Inertia Shaping For Humanoid Fall Direction Change - A system and method is disclosed for controlling a robot that is falling down from an upright posture. Inertia shaping is performed on the robot to avoid an object during the fall. A desired overall toppling angular velocity of the robot is determined. The direction of this velocity is based on the direction from the center of pressure of the robot to the object. A desired composite rigid body inertia of the robot is determined based on the desired overall toppling angular velocity. A desired joint velocity of the robot is determined based on the desired composite rigid body inertia. The desired joint velocity is also determined based on a composite rigid body inertia Jacobian of the robot. An actuator at a joint of the robot is then controlled to implement the desired joint velocity. | 06-24-2010 |
20100292838 | Machine Learning Approach for Predicting Humanoid Robot Fall - A system and method is disclosed for predicting a fall of a robot having at least two legs. A learned representation, such as a decision list, generated by a supervised learning algorithm is received. This learned representation may have been generated based on trajectories of a simulated robot when various forces are applied to the simulated robot. The learned representation takes as inputs a plurality of features of the robot and outputs a classification indicating whether the current state of the robot is balanced or falling. A plurality of features of the current state of the robot, such as the height of the center of mass of the robot, are determined based on current values of a joint angle or joint velocity of the robot. The current state of the robot is classified as being either balanced or falling by evaluating the learned representation with the plurality of features of the current state of the robot. | 11-18-2010 |
20110029130 | Systems and Methods for Controlling a Legged Robot Using a Two-Phase Disturbance Response Strategy - Systems and methods are presented that enable a legged robot to maintain its balance when subjected to an unexpected force. In the reflex phase, the robot withstands the immediate effect of the force by yielding to it. In one embodiment, during the reflex phase, the control system determines an instruction that will cause the robot to perform a movement that generates a negative rate of change of the robot's angular momentum at its centroid in a magnitude large enough to compensate for the destabilizing effect of the force. In the recovery phase, the robot recovers its posture after having moved during the reflex phase. In one embodiment, the robot returns to a statically stable upright posture that maximizes the robot's potential energy. In one embodiment, during the recovery phase, the control system determines an instruction that will cause the robot to perform a movement that increases its potential energy. | 02-03-2011 |
20120035762 | Systems and Methods for Controlling a Legged Robot Based on Rate of Change of Angular Momentum - Systems and methods are presented that use the rate of change of a legged robot's centroidal angular momentum ({dot over (H)} | 02-09-2012 |
20120245734 | HUMANOID ROBOT PUSH RECOVERY ON LEVEL AND NON-LEVEL GROUND - A robot controller controls a robot to maintain balance in response to an external disturbance (e.g., a push) on level or non-level ground. The robot controller determines a predicted stepping location for the robot such that the robot will be able to maintain a balanced upright position if it steps to that location. As long as the stepping location predicted stepping location remains within a predefined region (e.g., within the area under the robot's feet), the robot will maintain balance in response to the push via postural changes without taking a step. If the predicted stepping location moves outside the predefined region, the robot will take a step to the predicted location in order to maintain its balance. | 09-27-2012 |
20120245735 | DAMAGE REDUCTION CONTROL FOR HUMANOID ROBOT FALL - A humanoid robot fall controller controls motion of a robot to minimize damage when it determines that a fall is unavoidable. The robot controller detects a state of the robot during the fall and determines a desired rotational velocity that will allow the robot to re-orient itself during the fall to land on a predetermined target body segment (e.g., a backpack). The predetermined target body segment can be specially designed to absorb the impact of the fall and protect important components of the robot. | 09-27-2012 |
20130179038 | Reverse Drive Assist for Long Wheelbase Dual Axle Trailers - A controller and control method assists a driver with backing up of a vehicle with an attached trailer. The vehicle has a front axle with steerable front wheels controlled by the driver and a rear axle with non-steerable rear wheels. The trailer has a front axle with non-steerable front wheels and a rear axle with steerable rear wheels controlled by a trailer steering controller. The controller receives an operator-controlled vehicle steering angle and a measured hitch angle. The controller determines a trailer steering angle based on the operator-controller vehicle steering angle and the measured hitch angle. The controller continuously controls the trailer (e.g., via a steering angle of the rear wheels) to maintain a trajectory with substantially no lateral slippage. | 07-11-2013 |
20140277941 | Jackknife Condition for Backing up of a Vehicle Attached to a Long Wheelbase Dual Axle Trailers - A jackknife warning condition controller and control method notifies a driver of a potential jackknife situation while backing up a vehicle with an attached trailer. The vehicle has a front axle with steerable front wheels controlled by the driver and a rear axle with non-steerable rear wheels. The trailer has a front axle with non-steerable front wheels and a rear axle with steerable rear wheels controlled by a trailer steering controller. The jackknife controller receives an operator-controlled vehicle steering angle and a measured hitch angle. The jackknife warning condition controller determines a directional jackknife warning condition and compares the measured hitch angle with the determined directional jackknife warning condition. If the measured hitch angle satisfies the directional jackknife warning condition then a notification is sent to the driver. | 09-18-2014 |
20140288705 | Humanoid Fall Damage Reduction - A robot controller controls a robot during a fall to reduce the damage to the robot upon impact. The robot controller causes the robot to achieve a tripod-like posture having three contact points (e.g., two hands and one foot) so that the robot motion in arrested with its center of mass (CoM) high above the ground. This prevents a large impact caused by the transfer of potential energy to kinetic energy during the fall. The optimal locations of the three contacts are learned through a machine learning algorithm. | 09-25-2014 |
20150202113 | WEARABLE SCISSOR-PAIRED CONTROL MOMENT GYROSCOPE (SP-CMG) FOR HUMAN BALANCE ASSIST - An apparatus for assisting the balance of a user is disclosed herein. The apparatus includes a first control moment gyroscope (CMG) and a second CMG configured as a scissor pair. The first CMG includes a first flywheel and a second flywheel, a first motor that rotates the first and second flywheels, a first gimbal supporting the first and second flywheel, and a first gimbal servo to rotate the first gimbal. The second CMG includes a third flywheel and a forth flywheel, a second motor that rotates the third and fourth flywheels, a second gimbal supporting the third and fourth flywheels, and a second gimbal servo to rotate the second gimbal. The apparatus additionally includes a gyroscope controller controlling the first gimbal servo and the second gimbal servo. | 07-23-2015 |
20160045385 | ADMITTANCE SHAPING CONTROLLER FOR EXOSKELETON ASSISTANCE OF THE LOWER EXTREMITIES - The control method for lower-limb assistive exoskeletons assists human movement by producing a desired dynamic response on the human leg. Wearing the exoskeleton replaces the leg's natural admittance with the equivalent admittance of the coupled system formed by the leg and the exoskeleton. The control goal is to make the leg obey an admittance model defined by target values of natural frequency, resonant peak magnitude and zero-frequency response. The control achieves these objectives objective via positive feedback of the leg's angular position and angular acceleration. The method achieves simultaneous performance and robust stability through a constrained optimization that maximizes the system's gain margins while ensuring the desired location of its dominant poles. | 02-18-2016 |
20160067061 | INTEGRAL ADMITTANCE SHAPING FOR AN EXOSKELETON CONTROL DESIGN FRAMEWORK - An assistive exoskeleton control system has a controller generating a positive assistance by shaping a closed loop integral admittance of a coupled human exoskeleton system to a desired assistance ratio A | 03-10-2016 |
Patent application number | Description | Published |
20120128239 | IMAGE QUALITY ASSESSMENT TO MERCHANDISE AN ITEM - Image-based features may be significantly correlated with click-through rates of images that depict a product, which may provide a more formal basis for the informal notion that good quality images will result in better click-through rates, as compared to poor quality images. Accordingly, an image assessment machine is configured to analyze image-based features to improve click-through rates for shopping search applications (e.g., a product search engine). Moreover, the image assessment machine may rank search results based on image quality factors and may notify sellers about low quality images. This may have the effect of improving the brand value for an online shopping website and accordingly have a positive long-term impact on the online shopping website. | 05-24-2012 |
20120221557 | METHOD AND SYSTEM FOR CLASSIFYING QUERIES TO IMPROVE RELEVANCE OF SEARCH RESULTS - Described herein are methods and systems for promoting item listings that satisfy a query based on the item listings being assigned to certain categories. A query to identify a set of item listings is processed where each item listing associated with an item or service being offered for sale and assigned to a leaf-level category. The scope of the query is determined based on a dictionary of queries or a length of the query when the query is not found in the dictionary. One or more categories are identified based on the scope of the query. A search results page is presented with the item listings from the identified one or more categories. | 08-30-2012 |
20120303615 | IMAGE-BASED POPULARITY PREDICTION - A machine may be configured to access an image of an item described by a description of the item. The machine may determine an image quality score of the image based on an analysis of the image. A request for search results that pertain to the description may be received by the machine, and the machine may present a search result that references the item's image, based on its image quality score. Also, the machine may access images of items and descriptions of items and generate a set of most frequent text tokens included in the item descriptions. The machine may identify an image feature exhibited by an item's image and determine that a text token from the corresponding item description matches one of the most frequent text tokens. A data structure may be generated by the machine to correlate the identified image feature with the text token. | 11-29-2012 |
20140164146 | IMAGE QUALITY ASSESSMENT TO MERCHANDISE AN ITEM - Image-based features may be significantly correlated with click-through rates of images that depict a product, which may provide a more formal basis for the informal notion that good quality images will result in better click-through rates, as compared to poor quality images. Accordingly, an image assessment machine is configured to analyze image-based features to improve click-through rates for shopping search applications (e.g., a product search engine). Moreover, the image assessment machine may rank search results based on image quality factors and may notify sellers about low quality images. This may have the effect of improving the brand value for an online shopping website and accordingly have a positive long-term impact on the online shopping website. | 06-12-2014 |
20150154503 | IMAGE-BASED POPULARITY PREDICTION - A machine may be configured to access an image of an item described by a description of the item. The machine may determine an image quality score of the image based on an analysis of the image. A request for search results that pertain to the description may be received by the machine, and the machine may present a search result that references the item's image, based on its image quality score. Also, the machine may access images of items and descriptions of items and generate a set of most frequent text tokens included in the item descriptions. The machine may identify an image feature exhibited by an item's image and determine that a text token from the corresponding item description matches one of the most frequent text tokens. A data structure may be generated by the machine to correlate the identified image feature with the text token. | 06-04-2015 |
Patent application number | Description | Published |
20100098407 | METHOD AND SYSTEM FOR PROTECTION SWITCHING IN ETHERNET PASSIVE OPTICAL NETWORKS - One embodiment provides a system that performs protection switching in an Ethernet passive optical network (EPON), which includes an optical line terminal (OLT) and at least one optical network unit (ONU). The system is configured with at least one redundant component for the OLT and/or ONUs, wherein the redundant component can be optical or electrical, and can be a port, line card or link. The system provides protection by detecting a failure, and switching automatically to the redundant components to reduce service disruption time. The protection switching comprises: preserving the existing configuration over the loss of at least one of a multiple-point control protocol (MPCP) message; an operations, administration and maintenance (OAM) message; and a signal on the physical layer. The system recovers from the failure without performing ONU discovery. | 04-22-2010 |
20100111523 | EPON WITH POWER-SAVING FEATURES - One embodiment provides a system for power saving in an Ethernet Passive Optic Network (EPON). The system includes an optical line terminal (OLT), an optical network unit (ONU), a traffic-detection module configured to detect status of traffic to and from the ONU, and a power-management module configured to place the ONU in sleep mode based on the detected traffic status. The ONU includes an optical transceiver that includes an optical transmitter configured to transmit optical signals to the OLT and an optical receiver configured to receive optical signals from the OLT. | 05-06-2010 |
20100329668 | Method and System for Optical Performance Monitoring in Ethernet Passive Optical Networks - One embodiment provides a system that tests optical performance in an Ethernet passive optical network (EPON), which includes an optical line terminal (OLT) and at least one optical network unit (ONU). The system configures an ONU with a circular queue that contains test frames for testing optical performance. The OLT then notifies the ONU to transmit test frames at a specified data rate for a specified duration. After receiving test frames at the OLT, the system measures frame loss and/or bit error rate based on the received test frames. | 12-30-2010 |
20110058813 | Ethernet Passive Optical Network Over Coaxial (EPOC) - Embodiments of the present invention exploit the existing capabilities of the Ethernet Passive Optical Network (EPON) MAC layer, designed for fiber optics communications, to provide a low cost MAC layer with upper layer connectivity over a hybrid fiber coaxial (HFC) network. In particular, embodiments allow for the EPON MAC to be used end-to-end (i.e., from an optical line terminal (OLT) to a coaxial network unit (CNU)) in a HFC network, thereby fully leveraging the packet processing capabilities, QoS functions, and management features of the EPON MAC. Furthermore, embodiments enable unified provisioning and management for both fiber and coaxial network units in a HFC network. | 03-10-2011 |
20120257892 | Method for Sub-Rating an Ethernet Passive Optical Network (EPON) Medium Access Control (MAC) Based Communication Link - Embodiments allow for EPON MAC traffic to be sub-rated according to available bandwidth and/or transmission capacity of the physical medium. EPON MAC traffic sub-rating is performed in the PHY layer. The PHY layer performing the sub-rating may be located anywhere in the communication path supporting the end-to-end EPON MAC link. Because the sub-rating is performed in the PHY layer, the EPON MAC layer at either end of the EPON MAC link is unaware of the sub-rating being performed, and thus continues to operate as it would normally according to the IEEE standard that it implements. | 10-11-2012 |
20130202286 | Ethernet Passive Optical Network Over Coaxial (EPOC) Protection Switching - Protection switching methods, systems, and architectures are provided for hybrid Ethernet Passive Optical Network (EPON)-Ethernet Passive Optical Network Over Coaxial (EPOC) networks. Protection switching embodiments enable protection of the EPON portion and/or the EPOC portion of the hybrid network. In embodiments, protection switching may be initiated by an Optical Line Terminal (OLT), a coaxial media converter (CMC), or an optical network unit (ONU)/coaxial network unit (CNU) in the hybrid network. | 08-08-2013 |
20130202293 | Systems and Methods for Ethernet Passive Optical Network Over Coaxial (EPOC) Power Saving Modes - Systems and methods for Ethernet Passive Optical Network Over Coaxial (EPOC) power saving modes are provided. The EPOC power savings modes allow an EPOC coaxial network unit (CNU) to enter a sleep mode based on user traffic characteristics. The sleep mode may include powering down one or more module of the EPOC CNU, including radio frequency (RF) transmit/receive circuitry and associated circuitry. In embodiments, the EPOC CNU may enter sleep mode based on instructions from an optical line terminal (OLT) or based on its own determination. Embodiments further include systems and methods that allow the EPOC CNU to maintain synchronization with a servicing coaxial media converter (CMC) when it enters a sleep mode. | 08-08-2013 |
20130202304 | Ethernet Passive Optical Network Over Coaxial (EPOC) Physical Layer (PHY) Link Up and Tuning - Example registration procedures for Ethernet Passive Optical Network Over Coax (EPOC) are provided. The procedures are compliant with the Ethernet Passive Optical Network (EPON) standard. The procedures cover initial coaxial media converter (CMC) registration; physical layer discovery and link up over a coaxial link of an EPOC network; and medium access control (MAC)-level discovery and link up over an EPOC network. As such, the procedures can he used to fully bring air EPOC network to user traffic readiness. In addition, the procedures, or a variation thereof, can be used to enable periodic maintenance of the coaxial link of the EPOC network, thereby maintaining adequate communication conditions over the coaxial link. | 08-08-2013 |
20130294775 | EPON With Power-Saving Features - A system and method for reducing power consumption in a Passive Optic Network (PON). The system comprises an optical line terminal (OLT), an optical network unit (ONU), a traffic-detection module configured to detect status of traffic to and from the ONU, and a power-management module configured to place the ONU in sleep mode based on the detected traffic status. The ONU includes transmitting and receiving components that are selectively powered down during the sleep mode based on a type of traffic in the ONU. | 11-07-2013 |
20140079399 | Time to Time-Frequency Mapping and Demapping for Ethernet Passive Optical Network over Coax (EPoC) - Embodiments include, but are not limited to, systems and methods for enabling Orthogonal Frequency Division Multiple Access (OFDMA) in the upstream in an Ethernet Passive Optical Network over Coax (EPoC) network. Embodiments include systems and methods for translating Ethernet Passive Optical Network (EPON) upstream time grants to OFDMA resources represented by individual subcarriers of an upstream OFDMA frame. In an embodiment, the translation of EPON upstream time grants to OFDMA resources ensures that Coaxial Network Units (CNUs) sharing an OFDMA frame do not use overlapping subcarriers within the frame. Embodiments further include systems and methods for timing upstream transmissions by the CNUs in order for the transmissions to be received within the same upstream OFDMA frame at a Fiber Coax Unit (FCU). Embodiments further include systems and methods for re-generating a data burst from OFDMA resources for transmission from the FCU to an Optical Line Terminal (OLT). | 03-20-2014 |
20140133856 | Multiple Downstream Modulation Profiles for Ethernet Passive Optical Network over Coax (EPoC) - Embodiments provide systems and methods for supporting the use of multiple downstream modulation profiles in an Ethernet Passive Optical Network over Coax (EPoC) network. This includes, at the Fiber Coax Unit (FCU), processing downstream traffic to determine its intended destination Coaxial Network Unit (CNU) and using a customized downstream modulation profile for the traffic based on its intended destination CNU. In addition, with the downstream modulation profile used for the downstream traffic varying in time, a downstream map indicating upcoming downstream modulation profiles in the downstream traffic is sent along with the downstream traffic from the FCU. A CNU can read the downstream map to determine upcoming downstream modulation profiles in the downstream traffic and can decide to decode a given transmitted modulation profile in the downstream traffic when the transmitted modulation profiles matches one or more downstream modulation profiles associated with the CNU. | 05-15-2014 |
20150304037 | Ethernet Passive Optical Network Over Coaxial (EPOC) - Embodiments of the present invention exploit the existing capabilities of the Ethernet Passive Optical Network (EPON) MAC layer, designed for fiber optics communications, to provide a low cost MAC layer with upper layer connectivity over a hybrid fiber coaxial (HFC) network. In particular, embodiments allow for the EPON MAC to be used end-to-end (i.e., from an optical line terminal (OLT) to a coaxial network unit (CNU)) in a HFC network, thereby fully leveraging the packet processing capabilities, QoS functions, and management features of the EPON MAC. Furthermore, embodiments enable unified provisioning and management for both fiber and coaxial network units in a HFC network. | 10-22-2015 |
20160099780 | Time to Time-Frequency Mapping and Demapping for Ethernet Passive Optical Network over Coax (EPoC) - Embodiments include, but are not limited to, systems and methods for enabling Orthogonal Frequency Division Multiple Access (OFDMA) in the upstream in an Ethernet Passive Optical Network over Coax (EPoC) network. Embodiments include systems and methods for translating Ethernet Passive Optical Network (EPON) upstream time grants to OFDMA resources represented by individual subcarriers of an upstream OFDMA frame. In an embodiment, the translation of EPON upstream time grants to OFDMA resources ensures that Coaxial Network Units (CNUs) sharing an OFDMA frame do not use overlapping subcarriers within the frame. Embodiments further include systems and methods for timing upstream transmissions by the CNUs in order for the transmissions to be received within the same upstream OFDMA frame at a Fiber Coax Unit (ECU). Embodiments further include systems and methods for regenerating a data burst from OFDMA resources for transmission from the ECU to an Optical Line Terminal (OLT). | 04-07-2016 |
Patent application number | Description | Published |
20080294604 | XQUERY JOIN PREDICATE SELECTIVITY ESTIMATION - A method for estimating a selectivity of a join predicate in an XQuery expression is provided. The method provides for determining a first sequence size of a first sequence in the join predicate, determining a second sequence size of a second sequence in the join predicate, determining a type of comparison operator used between the first sequence and the second sequence, estimating the selectivity of the join predicate based on the first sequence size, the second sequence size, and the type of comparison operator used, selecting an execution plan for the XQuery expression based on the selectivity of the join predicate estimated, and executing the XQuery expression using the execution plan selected. | 11-27-2008 |
20090006314 | INDEX EXPLOITATION - Various embodiments of a computer-implemented method, computer program product, and data processing system are provided that generate an index plan that produces a superset of data comprising the query result. In some embodiments, a computer-implemented method, computer program product, and data processing system produce a maximal-index-satisfiable query tree. | 01-01-2009 |
20090006447 | BETWEEN MATCHING - Various embodiments of a computer-implemented method, computer program product, and data processing system are provided that identify a range filter in a mark-up language query. In response to receiving a query of at least one mark-up language document, the query comprising a plurality of singleton filters, at least one group of the plurality of singleton filters are identified. Each group of comprises at least two singleton filters, wherein each group is equivalent to a range filter having a start value and a stop value. The start value and stop value are based on at least two singleton filters of each group. A query plan is generated to process the query based on, at least in part, a range defined by the start value and the stop value of the at least two singleton filters of each group. | 01-01-2009 |
20090070303 | GENERALIZED PARTITION PRUNING IN A DATABASE SYSTEM - A system for executing a query on data that has been partitioned into a plurality of partitions is provided. The system includes providing partitioned data including one or more columns and the plurality of partitions. The partitioned data includes a limit key value associated with each column for a given partition. The system further includes receiving a query including a predicate on one of the one or more columns of the partitioned data; and utilizing the predicate on the one of the one or more columns in a pruning decision on at least one of the one or more partitions based on the limit key values associated with the plurality of partitions. | 03-12-2009 |
20100125447 | LANGUAGE IDENTIFICATION FOR DOCUMENTS CONTAINING MULTIPLE LANGUAGES - Multiple nonoverlapping languages within a single document can be identified. In one embodiment, for each of a set of candidate languages, a set of non-overlapping languages is defined. The document is analyzed under the hypothesis that the whole document is in one language and that part of the document is in one language while the rest is in a different, non-overlapping language. Language(s) of the document are identified based on comparing these competing hypotheses across a number of language pairs. In another embodiment, transitions between non-overlapping character sets are used to segment a document, and each segment is scored separately for a subset of candidate languages. Language(s) of the document are identified based on the segment scores. | 05-20-2010 |
20100125448 | AUTOMATED IDENTIFICATION OF DOCUMENTS AS NOT BELONGING TO ANY LANGUAGE - An “impostor profile” for a language is used to determine whether documents are in that language or no language. The impostor profile for a given language provides statistical information about the expected results of applying a language model for one or more other (“impostor”) languages to a document that is in fact in the given language. After a most likely language for a test document is identified, the impostor profile is used together with the scores for the test document in the various impostor languages to determine whether to identify the test document as being in the most likely language or in no language. | 05-20-2010 |
20110087668 | CLUSTERING OF NEAR-DUPLICATE DOCUMENTS - Documents likely to be near-duplicates are clustered based on document vectors that represent word-occurrence patterns in a relatively low-dimensional space. Edit distance between documents is defined based on comparing their document vectors. In one process, initial clusters are formed by applying a first edit-distance constraint relative to a root document of each cluster. The initial clusters can be merged subject to a second edit-distance constraint that limits the maximum edit distance between any two documents in the cluster. The second edit-distance constraint can be defined such that whether it is satisfied can be determined by comparing cluster structures rather than individual documents. | 04-14-2011 |
20130191111 | LANGUAGE IDENTIFICATION FOR DOCUMENTS CONTAINING MULTIPLE LANGUAGES - Multiple nonoverlapping languages within a single document can be identified. In one embodiment, for each of a set of candidate languages, a set of non-overlapping languages is defined. The document is analyzed under the hypothesis that the whole document is in one language and that part of the document is in one language while the rest is in a different, non-overlapping language. Language(s) of the document are identified based on comparing these competing hypotheses across a number of language pairs. In another embodiment, transitions between non-overlapping character sets are used to segment a document, and each segment is scored separately for a subset of candidate languages. Language(s) of the document are identified based on the segment scores. | 07-25-2013 |