Abstract: A first seed concept term may be identified. The first seed concept term may be to train a cognitive computing system. The cognitive computing system may analyze the first seed concept term to generate a first set one or more concept terms that are candidates for being conceptually related to the first seed concept term. A first plurality of individual characters and the first seed concept term may be provided. A first user of a client computing device may be prompted to generate a second set of one or more concept terms that are conceptually related to the first seed concept term using one or more of the first plurality of individual characters.

Abstract: Vertical bipolar junction transistors (VBJTs), each with one or more resistors connected in a circuit in different circuit configurations, are disclosed. The VBJT has an emitter substructure that includes an emitter layer, a collector, an intrinsic base, one or more doped epitaxy regions, and one or more resistors. The intrinsic base, the doped epitaxy region(s), and the resistor(s) are stacked upon one another in a channel between the emitter layer and the collector. Various circuit configurations and structures are described including a common-collector circuit, a common-emitter circuit, and an emitter-degenerate circuit. Methods of making these configuration/structures are disclosed.

Abstract: A method and system for improving a machine learning task assignment process is provided to address productivity with respect to satisfaction. The method includes connecting hardware devices to a server system. Job related data associated with job roles for individuals is retrieved and associated with a time period. Work related items of the job related data are presented and selections for work related items are retrieved via selectors for the work items. Expected and actual satisfaction ratings for the work related items are received and analyzed in accordance with an order in which they are received. At least one work item is assigned to a user and a specialized memory repository is generated within a portion of a memory device. Results of the assignment are stored within the specialized memory repository. Self-learning software code for executing future task assignment processes is generated and modified based on reported satisfaction ratings.

Abstract: A computer-implemented method according to one embodiment includes simulating, for a predetermined time period, output of a first task that periodically runs within a system to create a simulated output, comparing the simulated output to actual output of the first task for the predetermined time period, and generating an alert in response to determining that the simulated output does not match the actual output for the predetermined time period.

Abstract: A method for manufacturing a semiconductor device includes forming a plurality of fins on a substrate. The plurality of fins each include a first portion having a first width, and a second portion having a second width greater than the first width. The method also includes forming a sacrificial layer on the substrate in a space between a first fin and a second fin of the plurality of fins, wherein the first fin and the second fin correspond to a vertical transistor. In the method, lower portions of the first and second fins are removed, and an epitaxial region is formed under remaining portions of the first and second fins. The sacrificial layer is removed from the space between the first fin and the second fin after forming the epitaxial region.

Abstract: A method, computer system, and a computer program product for optimizing a plurality of personality traits of a virtual agent based on a predicted customer satisfaction value is provided. The present invention may include identifying a customer. The present invention may also include retrieving a plurality of data associated with the customer. The present invention may then include analyzing the received plurality of data using a customer satisfaction prediction model. The present invention may further include generating a plurality of analyzed data from the customer satisfaction prediction model based on the analyzed plurality of data. The present invention may also include generating a plurality of personality traits for a virtual agent from the generated plurality of analyzed data.

Abstract: A deposition apparatus for depositing a material on a substrate is provided. The deposition apparatus has a processing chamber defining a processing space in which the substrate is arranged, an ultraviolet radiation assembly configured to emit ultraviolet radiation and a microwave radiation assembly configured to emit microwave radiation into an excitation space that can be the same as the processing space, and a gas feed assembly configured to feed a precursor gas into the processing space and a reactive gas into the excitation space. The ultraviolet radiation assembly and the microwave radiation assembly are operated in combination to excite the reactive gas in the excitation space. The material is deposited on the substrate from the reaction of the excited reactive gas and the precursor gas. A method for using the deposition apparatus to deposit a material on a substrate is provided.

Abstract: A fiber-optic switching system is provided which includes an optical fiber switch having first and second optical fiber portions and an electrically-controlled actuator. The first and second optical fiber portions are spaced apart with a gap between the portions that is sized to allow for light signal coupling between the optical fiber portions in a signal-passing state of the switch. The electrically-controlled actuator is coupled to transition the switch between the signal-passing state, where the light signal is allowed to pass between the optical fiber portions, and a signal-non-passing state, where the light signal is prevented from passing between the optical fiber portions. The electrically-controlled actuator includes an electroactive material exhibiting a physical change with change in an applied electrical field, where the physical change facilitates transitioning the optical fiber switch between the signal-passing and the signal-non-passing states.

Abstract: A method, system, and computer program product for optical inspection of objects. The method projects an optical test line on a device under test. A frame is captured of the optical test line projected onto the device under test. The method provides a reference line for the device under test and compares the reference line and the optical test line within the frame. The method generates a visual quality determination based on the comparison of the reference line and the optical test line.

Abstract: In one general embodiment, a computer-implemented method includes determining attributes of a plurality of objects in an object-based storage environment, grouping the objects into clusters based on similarities of the attributes using a clustering algorithm, storing indication of which objects are in which clusters, detecting occurrence of a trigger event associated with an object, determining to which cluster the object belongs using the clustering algorithm in response to detecting the trigger event, and performing a deduplication process for the object against other objects in the cluster to which the object belongs. In another general embodiment, a computer program product includes a computer readable storage medium having program instructions embodied therewith. The computer readable storage medium is not a transitory signal per se. The program instructions are readable and/or executable by a computer to cause the computer to perform the foregoing method.

Abstract: In some aspects, a method may include initializing a first array and a second array with a random voltage value, passing a forward pass by pulsing an input voltage value from an input of the first array and an input of the second array, and reading output voltage values at an output of the first array and an output of the second array. The method may further include passing a backward pass into the inputs of both of the first and second arrays, and reading voltage values at the inputs of the first and second arrays. The method may further include updating, with the first array, a first matrix update on the first array, updating, with the second array, a first matrix update on the second, and updating, with the second array, a second matrix update on the second array.

Abstract: The present invention may include a method for automatic correction and migration of a plurality of bulk data. The method may identify sources, where the sources include the plurality of bulk data. The method may identify a subset of the bulk data, where the subset represents one or more records selected from the plurality of bulk data. The method may upload the subset to an enterprise application. The method may determine an error list from the enterprise application based on the uploaded subset. The method may roll back the subset from the enterprise application and based on the determined error list, the method may match one or more rules to a one or more errors in the error list.

Abstract: A system, method, and computer program product provides contextual contracts across various enterprise blockchain applications. In an aspect of the present disclosure, a method is disclosed including receiving a contextual contract for addition to a blockchain. The contextual contract includes at least one state dependent term. The method further includes appending a block to the blockchain based on the received contextual contract and in response to appending the block, activating a ledger sensor for the contextual contract. The ledger sensor is configured to monitor at least one of the blockchain and at least one other blockchain for transactions that correspond to the at least one term. The method further includes receiving from the ledger sensor data about transactions that corresponding to the at least one term and finalizing the contextual contract by appending a block to the blockchain based on the received data and the contextual contract.

Abstract: A first fin field effect transistor (FinFET) has an internal source/drain (S/D) with a facetted face that is connected to a dielectric side of a first RRAM. A second FinFET and RRAM structure are also disclosed. In some embodiments, an electrode contact side of each RRAM is connected in common to form a 2T2R device. The locations of one or more electrode points on the diamond-shaped, facetted surface of the bottom electrode accurately position electric fields through the dielectric to accurately and repeatably locate where the filaments/current paths are formed (or reset) through the RRAM dielectric. Material selection and accurate thickness of the RRAM dielectric determine the voltage at which the filaments/current paths are formed (or reset).

Abstract: A method includes dividing a fraction of a floating point result into a first portion and a second portion. The method includes outputting a first normalizer result based on the first portion during to a first clock cycle. The method includes storing a first segment of the first portion during to the first clock cycle. The method includes outputting a first rounder result based on the first normalizer result during to the first clock cycle. The method includes outputting a second normalizer result based on the second portion during to a second clock cycle. The method includes outputting a second rounder result based on the second normalizer result and the first segment during to the second clock cycle.

Abstract: Alerting a service provider of failed network connectivity includes determining, by a modem device, a loss of connectivity of the modem device with a network. The modem device sends a first notification of the loss of connectivity to a service provider using a first communication. A second notification of a type of the loss of connectivity and a resolution time for the loss of connectivity is received using a second communication. The second communication includes an out-of-band communication. The second communication is sent by the service provider and received by a first device associated with a user.

Abstract: A spatial light modulator (SLM) is provided that includes an optical resonator (i.e., pixel) having nanoscale size. The optical resonator having nanoscale size includes a phase-change material such as, for example, a GeSbTe alloy, sandwiched between silicon nitride cladding layers. The phase-change material can undergo a crystalline-to-amorphous phase transition which is characterized by a large change in optical properties of the resonator.

Abstract: Embodiments relate to an intelligent computer platform to construct ground truth data. Annotations are extracted from one or more data elements and characteristic data associated with the extracted annotations is dynamically identified. One or more models are created with the identified characteristic data. Domain activity impact is measured and the accuracy of the annotation data is calculated. A policy is generated from the calculated accuracy, wherein the policy includes assigning a score to the annotation data as metadata. Compliance activities are orchestrated based on the generated policy. One or more annotations are selectively re-assigned to the data in compliance with the orchestration, and ground truth data is created from the selective re-assignment.

Abstract: Operating a data processing and storage apparatus to perform continuous backup monitoring for a collection of stored file system objects. In response to changes respective to a number of the file system objects, a set of Data Management Application Program Interface (DMAPI) events is created. For each of the events, accessory file system object information relative to a file system object subject to the change is determined. The events are grouped into an event group and a unique group identifier and an overall group size count are assigned to each of the events included in the event group. The events are sent together with the respective accessory file system object information, the respective group identifier and the respective group size to a number of backup clients registered as applications. The backup clients are to process the events together with the respective accessory file system object information, the respective group identifier and the respective group size into a number of backup requests.

Abstract: Provided are a computer program product, moveable robot block, and method for a moveable robot block deployed to form a barrier and sense environmental conditions. A command is received to couple to a location in a coordinate system comprising the barrier formed of a plurality of moveable robot blocks. Movement motors are controlled to cause the moveable robot block to move to the location in the coordinate system to couple to the barrier according to the command. An environmental sensor senses an environmental condition related to water sensed external to the moveable robot block when the moveable robot block is coupled to the barrier. The environmental condition is transmitted to the assembly management system over the network.

Abstract: A computer-implemented method for modifying video-based communications produced during a conference call, is disclosed. The computer-implemented method can include monitoring a plurality of images transmitted via a video feed of a device connected to the conference call. The computer-implemented method can include identifying a first unwanted image component transmitted via the video feed. The computer-implemented method can include actively modifying the video feed by removing the first unwanted image component from the video feed.

Abstract: Provided are techniques for storing and retrieving restricted datasets to and from a cloud network with non-restricted datasets. A request to dump datasets from one or more storage devices to the cloud storage is received, wherein the datasets include restricted datasets and non-restricted datasets, and wherein the restricted datasets are in a format that is not supported by the cloud storage. The restricted datasets are identified. The restricted datasets are converted to a format that is supported by the cloud storage to generate converted datasets. The converted datasets and the non-restricted datasets are dumped to one container in the cloud storage.

Abstract: An apparatus includes an antenna array package cover comprising a radiating surface, a mating surface disposed opposite the radiating surface, and an array of antenna array sub-patterns wherein each antenna array sub-pattern comprises at least one antenna element. The antenna array package also includes an array of sub-pattern interface packages mated to the mating surface of the antenna array package cover. Each sub-pattern interface package of the array of sub-pattern interface packages comprises a package carrier, a sub-pattern integrated circuit electrically and mechanically coupled to the package carrier, and a set of interface lines corresponding to the antenna elements of the antenna array sub-pattern that corresponds to the sub-pattern interface package. Methods for mounting the above apparatus into a host circuit are also disclosed herein.

Abstract: Provided are a computer program product, system, and method for managing read and write requests from a host to tracks in storage cached in a cache. A determination is made whether track format table support information for a track indicates that a track format table was previously determined to have or not have the track format code for track format metadata. Track format metadata for the track is rebuilt to determine whether the track format table includes a track format code for the rebuilt track format metadata when the track format table support information indicates that the track format table was previously determined to have a track format code for the track. The track format metadata is not rebuilt when the track format table support information indicates that the track format table was previously determined to not have a track format code for the track.

Abstract: A website navigation system has an analysis system which receives a request for an Internet web page from a client device. The analysis system receives web page data associated with the web page from the Internet and performs a data analysis process to organize the web page data for use in a virtual conversation with the user in order to present the web page in an audible format. The analysis system identifies separate elements of the web page from the web page data and extracts information from the separate elements based on the web page data. The analysis system groups the separate elements into categories based on the extracted information and sorts the groups of separate elements based on usage statistics. The analysis system then generates a prompt for being output to the user by the client device as audible output based on the sorted groups of separate elements.

Abstract: An exemplary method includes forming a sacrificial layer along sidewalls of an array of trenches that are indented into a substrate, depositing a fill layer over the sacrificial layer, and then creating an array of gaps between the fill layer and the substrate by removing the sacrificial layer along the sidewalls of the trenches, while maintaining a structural connection between the substrate and the fill layer at the floors of the trenches. The method further includes covering the substrate, the fill layer, and the gaps with a cap layer that seal fluid-tight against the substrate and the fill layer. The method further includes indenting a first reservoir and a second reservoir through the cap layer, and into the substrate and the fill layer, across the lengths of the array of gaps, so that the array of gaps connects the first reservoir in fluid communication with the second reservoir.

Abstract: Embodiments of the present disclosure relate to a method for processing alerts. According to an embodiment of the present disclosure, a set of alerts matching a metric template are identified from received alerts during a period of time. A plurality of variable values are acquired from the set of alerts based on the metric template. The plurality of variable values are normalized according to a normalization rule of the metric template. A severity level for the set of alerts is determined based on the normalized variable values. In response to the severity level exceeding a certain threshold, an abstract alert including information related to the set of alerts is generated.

Abstract: A big data analytics system includes an event monitoring controller that detects a user targeted event from an initial digital query data flow. An event conversion controller determines at least one contextual element associated with the at least one user targeted event, and automatically converts the at least one contextual element into a search criteria. An external data query builder controller automatically executes an external search query based on the search criteria. A data fusing controller automatically fuses the at least one contextual element generated from the at least one user targeted event, together with external search results generated from the external search query and the initial query data flow to generate a fused data file.

Abstract: Data-analysis-based validation of product review data and linking to product record data are provided to facilitate forwarding product-related guidance. The processing includes one or more processors obtaining record data documenting, at least in part, a product's specific supply chain history, and receiving by the processor(s) review data for the product. Based on data analysis, the processor(s) authenticates the review data of the product to establish verified review data, and links the verified review data to the record data documenting, at least in part, the product's supply chain history. The processor(s) provides to a user product-related guidance based, at least in part, on the verified review data for the product that has been linked to the record data documenting, at least in part, the product's supply chain history.

Abstract: Management of available energy among multiple unmanned aerial vehicles (UAVs) is provided by identifying tasks to be completed by the multiple UAVs, and determining energy requirements of one or more UAVs of the multiple UAVs to facilitate completing one or more tasks of the tasks to be completed by the multiple UAVs. Further, the approach includes identifying an energy sharing approach for completion of the task(s) by the UAV(s) where one or more other UAVs of the multiple UAVs transfer energy in transit to the UAV(s) to facilitate completion of the task(s). In operation, the multiple UAVs may be detachably coupled, and the approach may include implementing the energy sharing approach by transferring energy from the other UAV(s) to the UAV(s) to facilitate completion of the task(s), for instance, prior to decoupling of the other UAV(s) from the UAV(s).

Abstract: Tamper-respondent assemblies are provided which include a circuit board, an enclosure assembly mounted to the circuit board, and a pressure sensor. The circuit board includes an electronic component, and the enclosure assembly is coupled to the circuit board to enclose the electronic component within a secure volume. The enclosure assembly includes an enclosure with a sealed inner compartment, and a structural material within the sealed inner compartment of the enclosure. The structural material within the enclosure inhibits deflection of the enclosure due to a pressure differential between pressure of the sealed inner compartment and pressure around, at least in part, the enclosure. The pressure sensor senses pressure within the sealed inner compartment of the enclosure to facilitate identifying a pressure change indicative of a tamper event.

Abstract: The system, method, and computer program product described herein provide unified real-time rule analytics to users through the use of an analytics logic editor that allows a user to construct an analytic logic rule unit that may be used on both edge and cloud devices. The user may select a data source, transform, rule condition, and action using an interface of the analytics logic editor to construct an analytics logic rule unit that may be deployed to both edge and cloud devices in real-time without the need to separately program each device. The analytics logic rule unit may be installed and executed by the edge and cloud device in real-time.

Abstract: A device holder includes a base, a hinge, and an arm. The base includes a mounting feature configured to affix the device holder to a portable computer device, and the hinge is connected to the base. The arm is connected to the hinge and opposes the base. The arm is rotatable with respect to the base to change the device holder between a stowed configuration, when the arm is positioned alongside the base, and a deployed configuration, when the first arm is positioned away from the base. In addition, the device holder is configured to hold a peripheral device of the portable computer device when the device holder is in the deployed configuration.

Abstract: Embodiments for preemptive substitution of resources in a disaggregated computing environment. Failure patterns and mitigation actions are analyzed for specific failures of respective resources within the disaggregated computing environment. Responsive to determining a failure threshold has been reached for a first resource of a first type of the respective resources, a mitigation action is performed according to the analyzed failure patterns. A result of the mitigation action is determined and the result is used to improve the failure pattern analyzation.

Abstract: An intelligent computer platform to introduce adversarial training to natural language processing (NLP). An initial training set is modified with synthetic training data to create an adversarial training set. The modification includes use of natural language understanding (NLU) to parse the initial training set into components and identify component categories. As input is presented, a classifier evaluates the input and leverages the adversarial training set to identify the intent of the input. An identified classification model generates accurate and reflective response data based on the received input.

Abstract: A method, system, and program product for selecting software is provided. The method includes retrieving audio data during a call with a subject matter expert (SME). The audio data is converted into a data training set and documents of the SME are converted into a document training set. Canonical documents generated by authors are analyzed and specified code is extracted from the text data training set and document training set. Attributes of individuals are detected. The attributes are compared with specified data and the canonical documents and it is determined that the individuals are requesting information associated with the text data. The information is provided to the individuals via the canonical documents or the documents of the SME and it is determined if a matched set of data exists between the attributes, the specified data, and the canonical documents. A digital action associated with results of the determination is executed.

Abstract: A storage port is enabled for security. The storage port performs Input/Output (I/O) in plaintext on a path between the storage port and a host port, in response to determining that an audit mode indicator has been enabled to allow I/O even if authentication or security association negotiation between the storage port and the host port cannot be completed successfully. Concurrently with performing of I/O in plaintext on the path, the storage port enables encryption of data for I/O on the path.

Abstract: A computer-implemented method includes: receiving, by a computing device, a database-level auto_purge value for a database having a plurality of tables, wherein the database-level auto_purge value represents a value associated with the entire database for when data is set to be purged or deleted from the plurality of tables in the database; setting, by the computing device, the database-level auto_purge value for the database; setting, by the computing device, table-level auto_purge values for the plurality of tables included in the database based on the database-level auto_purge value; and automatically purging data in the plurality of tables based on the table-level auto_purge values.

Abstract: Method, system and computer program product are disclosed for an error recovery process. One method comprises: program instructions are executable by the processor to perform; determining a component of the computer system related to the error event; determining a quality of service requirement for the component; determining an error recovery process for the error event based on the quality of service requirement; and executing the error recovery process for the error event.

Abstract: A host port is enabled for security. The host port performs Input/Output (I/O) in plaintext on a path between the host port and a storage port, in response to determining that an audit mode indicator has been enabled to allow I/O even if authentication or security association negotiation between the host port and the storage port cannot be completed successfully. Concurrently with performing of I/O in plaintext on the path, the host port enables encryption of data for I/O on the path.

Abstract: Technical solutions are described for controlling an artificial intelligent gaming device. For example, a computer-implemented method includes identifying an electronic opponent profile for an opponent. The computer-implemented method also includes selecting, from a profile repository, a first set of artificial intelligence profiles, where machines trained using artificial intelligence profiles from the first set of artificial intelligence profiles have previously defeated the opponent. The computer-implemented method also includes selecting, from the profile repository, a second set of artificial intelligence profiles, where machines trained using artificial intelligence profiles from the second set of artificial intelligence profiles have previously lost to the opponent. The computer-implemented method also includes generating a current artificial intelligence profile based on the first set of artificial intelligence profiles and the second artificial intelligence profiles.

Abstract: Methods, systems, and computer program products for alerting users to media events of interest using social media analysis are provided herein. A computer-implemented method includes collecting user media preferences indicating events of interest to a given user; buffering at least a portion of media content associated with the collected user media preferences of the given user; monitoring social media for one or more events of interest to the given user based on the collected user media preferences of the given user; identifying at least one event that satisfies one or more predefined interest criteria for the given user based on the monitored social media contributions of other parties and the collected user media preferences of the given user; and notifying the given user of the at least one event of interest.

Abstract: Provided are a computer program product, system, and method for managing failover from a first processor node including a first cache to a second processor node including a second cache. Storage areas assigned to the first processor node are reassigned to the second processor node. For each track indicated in a cache list of tracks in the first cache for the reassigned storage areas, the first processor node adds a track identifier of the track and track format information indicating a layout and format of data in the track to a cache transfer list. The first processor node transfers the cache transfer list to the second processor node. The second processor node uses the track format information transferred with the cache transfer list to process read and write requests to tracks in the reassigned storage areas staged into the second cache.

Abstract: Utilizing social information for recommending an application includes providing an application recommendation system based on social characterizations, and responsive to a user searching for an application meeting a criteria, utilizing the application recommendation system by searching for applications meeting the criteria, characterizing the applications according to a social proximity factor to the user, and presenting the applications ordered by the social proximity factor.

Abstract: Multistage deterministic lateral displacement devices, methods of forming the devices, and methods of separating a fluid mixture including particles having three or more particle sizes generally include a first module and at least one additional module. Each module includes a condenser portion and a separate portion. The condenser portion is generally configured to focus a streamline of all particles to a center of a channel whereas the separator separates the streamline of all particles into two different streamlines. One of the streamlines focuses the largest particles in the fluid mixture along a sidewall of the channel and the other streamline of smaller particles is between opposing sidewalls that define the channel. Each additional module can be used to further separate the largest particles remaining in the fluid mixture from the smaller particles.

Abstract: An illustrative embodiment includes a method for solving a dynamical system. The method comprises: obtaining multidimensional snapshots representing respective discrete solutions of the dynamical system; storing the multidimensional snapshots within a snapshot tensor having an order of at least three; generating a basis for at least a subspace of a state space of the dynamical system at least in part by performing a decomposition of the snapshot tensor; deriving a reduced order model at least in part by using the basis to project the dynamical system from the state space onto the subspace; and solving the reduced order model of the dynamical system.

Abstract: Techniques for tailoring sampling rates for data from interactive digital properties on a feature-by-feature basis and collecting the data using the tailored sampling rates. Each feature may have an independent sampling rate irrespective of sampling rates assigned to other features. The independent sampling rates are determined based on at least one factor of predictive feature usage information based on historical feature usage information, predetermined rules, and current usage velocity of the feature. In some embodiments the independent sampling rate is influenced by the usage of an allocated resource provided to the digital property relative to a total allocation of that resource for a given time period. In some embodiments, the allocated resource is server calls to a digital data analytics server for the purposes of providing feature usage information from the interactive digital property for the performance of digital data analytics.

Abstract: Systems, computer-implemented methods, and computer program products that can facilitate creating and querying a knowledge base of identified topics, computer code actions, and parameters, are provided. According to an embodiment, a system can comprise a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can comprise a parameter component that can identify a result parameter in ones of one or more results, wherein the one or more results comprise topics and computer code actions. The computer executable components can further comprise a result component that can select a result of the one or more results based on a mapping of a query to the one or more results, the mapping being based on the result parameter identified in the result and a criterion.

Abstract: A metal is formed into an opening that is located in an interlayer dielectric (ILD) material that laterally surrounds a semiconductor fin of a partially fabricated vertical transistor and on a physically exposed topmost surface of the semiconductor fin. A patterned material stack of, and from bottom to top, a membrane and a doped amorphous semiconductor material layer is formed on the metal and a topmost surface of the ILD material. A metal induced layer exchange anneal is then employed in which the metal and doped semiconductor material change places such that the doped semiconductor material is in direct contact with the topmost surface of the semiconductor fin. The exchanged doped semiconductor material, which provides a top source/drain structure of the vertical transistor, may have a different crystalline orientation than the topmost surface of the semiconductor fin.

Abstract: An approach is provided in which the approach aggregates a set of pixel values from a bitmap image into a set of row sum values and a set of column sum values. The bitmap image is a pixelated representation of a document. The approach applies a localized Fourier transform to the set of row sum values and the set of column sum values to generate frequency representations of the set of row sum values and the set of frequency sum values. The approach decomposes the bitmap image into a set of image portions based on at least one separation location identified in the set of frequency representations, and sends the set of image portions to a text recognition system.