Abstract: A method of joining a first dataset configured to store a set of data entries each identified by a respective key of a first type and a second dataset configured to store a second set of data entries identified by a respective key of a second type, the method comprising: selecting an intermediate mapping entity from a set of possible intermediate mapping entities, each mapping entity storing association between keys of the first type and keys of the second type; providing the selected intermediate mapping entity for use in joining the first data set with the second data set; wherein the step of selecting the intermediate mapping entity is based on the intersection weight between the first and second data sets via each of the intermediate mapping entities, wherein the intersection weight is the proportion of overlapping data entries between the first and second datasets.

Abstract: At least one social media channel includes a plurality of user messages, which are accessible via a communication network. At least some of the stored messages are retrieved from the at least one social media channel via the communication network, and a collection of frequently asked questions (FAQ) is generated or updated by analyzing the retrieved messages to form a plurality of topical issue clusters. Each topical issue cluster is associated with at least one topic parameter from among a plurality of topic parameters, each topic parameter relates to at least one of use, installation or maintenance of a product or service, and each topical issue cluster includes at least one issue identified by a community of users and at least one resolution of the issue identified by the community of users. The generated or updated FAQ is uploaded to a storage location accessible to the community of users.

Abstract: A method and system for automatically performing a discovery of topics within temporal ordered text document collections are provided. The method includes generating a bag of words vector for each text document collection using a predefined dictionary. The method also includes iteratively calculating, based on the generated bag of words vectors, for each text document collection, a hidden topic vector representing topics of the respective text document collection using a calculated hidden state vector memorizing a hidden state of all previous text document collections.

Abstract: Systems, methods, and computer program products are provided for auctioning advertising inventory for various applications, including for mobile applications. In various embodiments, a request for an advertisement is received at a first server from a client device. The request is sent, from the first server, to two or more remote servers. A bid value and an asset to be presented at the client device is received from each of the two or more remote servers. A highest bid is determined from the received bid values. Whether the asset associated with the highest bid is compliant based on the request is determined. When the asset is determined to be compliant, the asset is sent to the client device.

Abstract: A system described herein may provide techniques for receiving natural language inputs at a live dashboard. The natural language inputs may be analyzed to identify entities such as keywords and to determine an intent associated with each input. The identified entities may be mapped to various database elements and a formatted database request may be generated based on the mapping and the identified entities. The formatted database request may be sent to an associated database and a database response may be received. The data included in the database response may be provided by generating a report using report attributes associated with the identified search intent.

Abstract: An example method includes presenting a remote broadcast event by delivering content from a first user endpoint device to a plurality of user endpoint devices of a plurality of audience members, estimating reactions of the audience members, based on streams of data received from the plurality of user endpoint devices, grouping the audience members into a plurality of groups, based on the reactions, wherein each group of the plurality of groups is associated with a different reaction, and wherein each audience member who is a member of the each group was estimated to demonstrate a common reaction of the plurality of reactions, wherein the common reaction is associated with the each group, selecting, for a first group, a first audience member from the first group to be representative of the first group, and presenting, to the first user endpoint device, an image of the first audience member.

Abstract: The disclosure provides a data transmission device. The data transmission device includes a storage medium, a processor and a communication interface. The storage medium is configured to store a plurality of applications. The processor is coupled to the storage media and configured to determine a host application among the plurality of applications, where the host application includes a software development kit (SDK) program, and executes the SDK program of the host application to request behavior data of the plurality of applications. The communication interface is coupled to the processor and is configured to transmit the behavior data to a server.

Abstract: The present disclosure provides systems and methods for modifying and controlling access to private data objects. A system can maintain a node graph comprising nodes that each maintains an association with one or more neighbor nodes. The system can maintain private data objects that are generated based on a request from an organizer node, and include a list of authorized nodes and a setting to enable neighbors of authorized nodes to request inclusion in the list of authorized nodes. The system can receive a request for private data objects from a first node, and identify private data objects that include one or more neighbor nodes of the first node. The system can present a list of private data objects to the first node, and add the first node to the list of authorized nodes of the private data object subsequent to receiving a request from the first node.

Abstract: Techniques for using contextual information relating to content presented by a television as part of a search query for an information search are presented. A search management component, at a given moment in time during presentation of television-related content by a communication device in or associated with a television, identifies contextual information associated with a section of the television-related content and generates a content identifier timestamp associated with the contextual information and the section of television-related content. A search component augments a search query using the contextual information to facilitate customization of a subset of search results based on the contextual information. The contextual information in the search query can facilitate disambiguating the search query or promoting a search result over another search result in the subset of search results, based on the contextual information, to facilitate customization of the subset of search results.

Abstract: A system, method, and machine-readable storage medium for resolving a candidate community are provided. In some embodiments, a method includes obtaining a candidate community and a neighbor set for the candidate community, the neighbor set including zero or more stable communities. The method also includes resolving the candidate community as being a new stable community if the neighbor set is empty. The method further includes resolving the candidate community as being part of a matching stable community if a hash value of the candidate community matches a hash value of one or more stable communities included in the neighbor set. The method also includes resolving the candidate community as being a new stable community if an entropy value is greater than a threshold, the entropy value being based on the candidate community and the neighbor set.

Abstract: The present technology addresses deficiencies associated with common practices for handling out of order data in a streaming data database. An aspect of the present technology is avoid storing out of order data in a snapshot but just store the out of order data as additional data linked to the temporal graph. The present technology receives out of order data and records a modification time for the data and a next modification time for the data that equals a timestamp of data previously stored in the database. If there is also data in the database for a time earlier than the timestamp of the out of order data, the earlier data is adjusted so that its next modification time matches the timestamp of the out of order data.

Abstract: A contextual hashtag generation method, system, and computer program product include receiving content from an online source, identifying a set of contextual indicators for the content, determining an entity-desired outcome for the content, and generating a hashtag for the content using the set of contextual indicators while maximizing the entity-desired outcome.

Abstract: Techniques for sectionalizing clinical documents are provided. In one set of embodiments, a computer system can, for each page of a clinical document: identify one or more section header candidates in the page and, for each section header candidate, attempt to classify the section header candidate as corresponding to one of a plurality of section types using a first classifier or a second classifier. The computer system can further partition the page into one or more sections based on corresponding section header candidates that have been successfully classified using either the first classifier or the second classifier, where the partitioning includes associating each section with a section type in the plurality of section types in accordance with the classification of the section's corresponding section header candidate. The computer system can then validate, for each section, the section's section type via an analysis of the body of the section.

Abstract: Systems and methods to identify document transitions between adjacent documents within document bundles are disclosed. Exemplary implementations may train a model: obtain training information including a first training bundle and corresponding document separation markers; determine page-specific feature information pertaining to individual pages of the first training bundle; determine, based on the obtained page-specific feature information, page-specific feature values for individual features of the individual pages of the first training bundle; generate, for the individual pages of the first training bundle, a page-specific feature vector; train the model, using the training document bundles, to determine whether the first page and the second page are part of different document. Systems and methods may utilize the trained model to identify document transitions between adjacent documents within document bundles.

Abstract: A method includes training a machine learning model using a current set of labeled data points. Each of the data points is multiple data records. A label of a data point indicates a classification of the data point. The training results in a trained machine learning model configured to classify a data point as representing a same entity or different entities. The method includes selecting a subset of unlabeled data points from a current set of unlabeled data points using classification results of the current set of unlabeled data points. The method includes providing the subset of unlabeled data points to a classifier and in response to providing receiving labels of the subset of unlabeled data points. The method may be repeated using the subset of labeled data points in addition to the current set of labeled data points as the current set of labeled data points.

Abstract: A system is provided for referral object processing for textual annotations. The system comprises a memory storing machine executable code and one or more processors coupled to the memory and configurable to execute the machine executable code to cause the one or more processors to parse a document to identify a reference identifier to an external object, the external object associated with information not contained in the document, retrieve the external object using the reference identifier, extract the information associated with the external object based on at least one data pattern detected in the external object, convert the extracted information into textual annotations associated with the reference identifier in the document, and enter the textual annotations to a corpus of content for the document so that the extracted information is associated with the reference in the document for the system.

Abstract: A data archiving method includes obtaining a data format of data to be processed, searching for a data category corresponding to the data format in a first database, extracting feature information of the data to be processed according to a first preset rule if the data category corresponding to the data format is found and storing the feature information in the first database according to a storage rule, and searching the data category corresponding to the data format in a second database according to a second preset rule if no data category corresponding to the data format is found, extracting feature information of the data to be processed, and storing the extracted feature information in the first database according to the storage rule. The first database stores data categories of different data formats of processed data. The second database stores feature information of multiple data categories.

Abstract: The concepts, systems and methods described herein are directed towards a method for detection and quantification of influence. The system is provided to including: a network sampling processor, a narrative discovery processor, and an influence quantification processor. The network sampling processor is configured to sample information on one or more social media networks. The narrative discovery processor is configured to: receive sampled information from the network sampling processor, and in response thereto identify a narrative related to a subset of information sampled by the network sampling processor. The influence quantification processor is configured to: receive information related to the narrative and to process the information via a network causal inference process to quantify influence of the narrative on the one or more social media networks.

Abstract: Methods, systems and computer program products implementing hierarchical classification techniques are disclosed. A hierarchical classification system receives training data including labeled transaction records. The system determines tag sequences from the training data. The system clusters the tag sequences into clusters. The system determines a cluster-level classifier that is trained to predict a cluster for an input transaction record. The system determines a respective cluster-specific classifier for each cluster. The system trains the cluster-specific classifier to predict a label of entity of interest for an input transaction record, given a particular cluster. Upon receiving a test transaction record, the system first applies the cluster-level classifier to determine a particular cluster for the test transaction record, and then determines a label of entity of interest of the test transaction record by applying a cluster-specific classifier of that particular cluster.

Abstract: In some examples, active directory management and remediation may include obtaining data associated with active directories, and generating, based on the obtained data, a data model that includes a plurality of active directory concepts. A graph may be generated and include the active directory concepts as nodes and relationships between the nodes. The graph may be analyzed to identify user to group memberships. A membership score and a reasoning for the membership score may be determined for each identified user to group membership of the user to group memberships. The membership score and the reasoning for the membership score may be analyzed for each identified user to group membership. Further, based on the analysis of the membership score and the reasoning for the membership score, the at least one of the user to group memberships may be remediated.

Abstract: Correlating information associated with playback of video segments is disclosed. Information associated with playback of a first video segment is obtained. Information associated with playback of a second video segment is obtained. The information associated with playback of the first video segment is correlated with the information associated with playback of the second video segment based at least in part on at least one of a client application identifier, a video session identifier, and a set of timestamps. Output is provided based at least in part on the correlating.

Abstract: A non-transitory computer-readable storage medium storing a program that causes a computer to execute a process, the process including determining whether first answer information has been cited in a generation process of second answer information for a first inquiry based on history information, stored in a storage device, indicating a citation histories regarding pieces of answer information, and generating training data associating the first inquiry with the first answer information when the first answer information has been cited in a generation process.

Abstract: Systems and methods for managing content provenance are provided. A network system accesses a document of a plurality of documents to be analyzed. The network system extracts text fragments from the document including a first fragment and a second fragment. A determination is made whether each of the text fragments match an entry in a hash table. Based on a first fragment not matching any entries in the hash table, the network system creates a new entry in the hash table, whereby the first fragment is used to generate a key in the hash table. Based on a second fragment matching an entry of the hash table, the network system associates the document with a key of the matching entry in the hash table, whereby the associating comprising updating the hash table with an identifier of the document.

Abstract: A method may include receiving a request to save the exception to the first rule; in response to the request, updating the user interface, wherein: the first portion is updated to include a parameter of a second rule to exclude a subset of meetings from the ECDS; the second portion is updated to include an impact metric that indicates a percentage of meetings excluded from the ECDS based on an application of the second rule to the ECDS; and the third portion is updated to include selectable elements for creating an exception to the second rule; and a fourth portion of the user interface includes an indication of a number of meetings excluded from the ECDS based on application of the first rule with any exceptions to the first rule.

Abstract: Methods for asset type fingerprinting are provided. An example method involves an asset tracking system failing to identify an asset type fingerprint and requesting an asset type fingerprint from an asset data analysis system. The asset data analysis system generates a proposed asset type fingerprint, links the asset type fingerprint to a set of signal definitions that indicate how to decode data messages from the asset. The asset data analysis system transmits the proposed asset type fingerprint to the asset data analysis system, which then decodes at least some data messages from the asset thus obtaining asset information. The method may be used to identify vehicle types and decode proprietary or non-standard data messages via a vehicle's diagnostic port.

Abstract: A server may include a communication circuit, at least one processor electrically connected to the communication circuit, and at least one memory electrically connected to the at least one processor. The at least one memory stores instructions for receiving, from a first external device, first data related to a user input that requests performance of a task, receiving, second data associated with a plurality of application programs, determining a domain related to the task based on at least some of the first data and at least some of the second data, determining a likely intention of a user providing the user input, based on at least some of the first data, and determining a sequence of states of the first external device based on the domain and the likely intention, providing the sequence of states to the first external device.

Abstract: Systems and methods to identify a filter set in a keyword query are described. The system receives a query from a client machine. The system identifies filter sets based on the query and a based on rules. The filter sets include a first filter set that includes a first filter. The rules are utilized to associate at the least one keyword from the query to the first filter. The system further scores the filter sets based on probabilities to generate scores. The probabilities describe occurrences of attribute-value pairs in listings that respectively describe items that were previously transacted on a network-based marketplace. The system further identifies the first filter set from the filter sets based on the scores, generates a user interface including search results that are identified based on the identified first filter set, and communicates the user interface, over the network, to the client machine.

Abstract: An object of the present invention is to provide a reliable supplier evaluation system for correctly evaluating suppliers. A supplier evaluation system 117 according to the present invention is a system for evaluating suppliers based on information relating to a transaction conducted between a buyer and the supplier. The supplier evaluation system 117 is provided with a reception function for receiving communication information 134 and e-commerce information conducted between a buyer and a supplier and that exists in a company system such as that of a buyer, a conversion function for converting the communication information and the e-commerce information into a predetermined format, a database 133 for storing keyword information, and an analysis function 127 for comparing keyword information 138 registered in advance in a storage device with the converted communication information and e-commerce information, and determining a cause of a transaction result of the e-commerce information.

Abstract: Techniques and tools are described for producing time-series data using faceted search based on document creation time. For example, index information can be created from events (e.g., by adding meta-data and indexing the events as documents). In addition, index information can be created that maps document creation time to time ranges. Search queries can then be executed (e.g., comprising ad-hoc filters to filter on the meta-data), and search results can be faceted on the time ranges to produce time-series data. The time-series data can be graphed to display trends of activity (e.g., trends of events based on user activity).

Abstract: Disclosed are methods, systems, and non-transitory computer-readable record mediums for processing a chatroom based on a level of use. A chatroom processing method including verifying a level of use by a user with respect to each chatroom of chatrooms included in a chatroom list, selecting at least one chatroom from the chatroom list as a chatroom to be managed based on the level of use, and processing the chatroom to be managed to be distinguished from remaining chatrooms in the chatroom list may be provided.

Abstract: A computer-implemented method of answering questions comprises: receiving, by one or more processors, a query; based on the query, generating, by the one or more processors, a matrix; based on the matrix, modifying, by the one or more processors, a dynamic memory; based on the matrix, determining, by the one or more processors, a first response from the dynamic memory; based on the matrix, determining, by the one or more processors, a second response from a database; based on the first response and the second response, determining, by the one or more processors, a third response; and in response to the query, providing, by the one or more processors, the third response.

Abstract: Methods, systems, and computer program products for dynamically analyzing diagnostic operations data via machine learning techniques are provided herein. A computer-implemented method includes defining aspects of machine learning techniques to be performed in connection with diagnostic operation data analysis, including: defining dynamic analysis granularity selection functionality based on time constraints and the level of progress of the analysis; defining dynamic data classification identifier type selection functionality based on the time constraints and the level of progress of the analysis; and defining dynamic ordering of data classification identifiers during runtime based on data classification preferences, information pertaining to system workload, and information pertaining to requested analysis compliance parameters.

Abstract: Example embodiments described herein pertain to a geospatial interface system configured to cause display of geospatial data within a graphical user interface at a client device, receive data points from multiple data sources, unify the data points, and present the unified data points as interactive graphical elements within the graphical user interface, in a presentation layer separate from the geospatial data. In example embodiments, the geospatial interface system may be or include a group of one or more server machines configured to provide one or more geospatial data display, indexing, and management services. A client device may accordingly display a graphical user interface generated by the geospatial interface system.

Abstract: Techniques for solving a multi-agent plan recognition problem are provided. In one example, a computer-implemented method comprises transforming, by a device operatively coupled to a processor, a problem model and an at least partially ordered sequence of observations into an artificial intelligence planning problem through a transform algorithm. The problem model can comprises a domain description from a plurality of agents and a durative action. Furthermore, at least one of the observations of the at least partially ordered sequence of observations can be a condition that changes over time. The computer-implemented method further comprises determining, by the device, plan information using an artificial intelligence planner on the artificial intelligence planning problem. The computer-implemented method further comprises translating, by the device, the plan information into information indicative of a solution to the artificial intelligence planning problem.

Abstract: The described implementations are operable to determine potential objects of interest to a user based on a blend of the user's long-term behavior and short-term interests. Long term user behavior may be determined for the user over a period of time and represented as continuous data. Short-term interest may be determined based on objects with which the user has recently interacted and attributes of those objects may be represented together as continuous data corresponding to the short-term user interest. The continuous data of the short-term interest and long-term user behavior may be blended to produce a user embedding. The user embedding may then be compared with objects to determine objects that are of potential interest to the user.

Abstract: Example implementations described herein are directed to a meta-data processing system that supports the creation and deployment of the Analytical Solution Modules in development of industrial analytics. The example implementations described herein can involve a first system configured to be directed to a data scientist for receiving flow and operator definitions to generate an analytics library, which is provided to a second system configured to be directed to a domain expert for applying the analytics library to generate analytics modules to be executed on data input to the second system.

Abstract: Embodiments also include a method for filtering and securing content of datasets in computer readable form designated for release to reduce discernable inferences therein. The method includes receiving a first dataset having first records associated with a quasi-identifier. The first records have respective first data values associated with the quasi-identifier. The method includes receiving a second dataset having second records associated with the quasi-identifier. The second records have respective second data values associated with the quasi-identifier. The method includes defining a first cluster having a first boundary based on a combination of the first dataset and the second dataset. The method includes replacing a first one of the first data values with the first boundary and a second one of the second data values with the first boundary.

Abstract: Methods and systems for estimating recall while screening an ordered list of bibliographic references are provided. According to one embodiment, a method includes: sorting a list of bibliographic references according to a sorting algorithm to produce a first list in order from most to least relevant; selecting, from the list, the most relevant reference, and displaying, to a human screener, information associated with the selected reference; and receiving the screener's judgment of the relevance of the selected reference. If sufficiently relevant, the selected reference is moved from the first list to a second list. The received indication is used to re-sort the remaining references in the first list. A statistical model is used to estimate the number of relevant references remaining. That estimate is displayed to the screener. The process ends when the screener determines, based on the displayed estimate, that a sufficient number of relevant references has been found.

Abstract: A database may contain representations of: (i) software packages managed by a software management tool, including publishers, titles, and categories associated with each, and (ii) a plurality of software activities, including descriptions and amounts associated with each. A server device may be configured to obtain classifications of the software activities that predict the publishers, titles, and categories of the software activities from the descriptions. The server device may further compare the software packages to the classifications in order to identify: (i) unmanaged software packages, and (ii) amounts associated with the software packages.

Abstract: In an implementation, a non-transitory machine-readable storage medium stores instructions that when executed by a processor, cause the processor to allocate classifier data structures to persistent memory, read a number of categories from a set of training data, and populate the classifier data structures with training data including training-based, category and word probabilities calculated based on the training data.

Abstract: Computer vision for unsuccessful queries and iterative search is described. The described system leverages visual search techniques by determining visual characteristics of objects depicted in images and describing them, e.g., using feature vectors. In some aspects, these visual characteristics are determined for search queries that are identified as not being successful. Aggregated information describing visual characteristics of images of unsuccessful search queries is used to determine common visual characteristics and objects depicted in those images. This information can be used to inform other users about unmet needs of searching users. In some aspects, these visual characteristics are used in connection with iterative image searches where users select an initial query image and then the search results are iteratively refined.

Abstract: A method, a computing system and a computer program product are provided. A model is generated and trained. The model is based on clinical data with outcomes from clinically-defined hierarchical metadata in a selected level of clinically-defined hierarchical metadata serving as an initial set of prediction targets. A score is determined for each of the prediction targets based on the generated model and the set of evaluation factors. The set of prediction targets, the generated model, and the scores for the set of prediction targets are updated until the updated scores for the updated set of prediction targets satisfy acceptance criteria. The updated generated model, using the updated set of prediction targets, is applied to predict one of a set of updated prediction targets of mutually exclusive outcome categories.

Abstract: A method, system, and program product for implementing intelligent cache preloading is provided. The method includes monitoring current usage of a system of record (SOR) system. Historical data associated with historical usage of the SOR system is retrieved and analyzed based on the current usage. A ranked list of data items configured to be loaded within a cache structure of the cache system is generated and currently requested data items from the SOR system are loaded into the cache structure via a throttling process. A malfunction associated with operation of the SOR system is detected and access to the currently requested data items is enabled. The currently requested data items are organized within the cache structure in accordance with an order of the ranked list and access to the currently requested data items is enabled, during the malfunction, in accordance with the order of the ranked list.

Abstract: An online system receives information describing events corresponding to actions associated with a third party system performed by an individual. The received information describes event types and times at which the events occurred. The online system generates nodes of a directed graph associated with the third party system, in which each node corresponds to an event type. For each event, a node count associated with a node corresponding to the event's type is incremented by the online system. Pairs of consecutively occurring events are identified based on times at which the events occurred and an edge describing each transition from one event to another is generated by the online system. The online system determines an edge count for each transition indicating a number of edges describing the transition as well as a sequential order of event types based on one or more node counts and one or more edge counts.

Abstract: One embodiment provides a system for facilitating a graph search engine. During operation, the system receives, by a server from a client computing device, a search request which includes a user-inputted graph. The system performs a search based on a structure of the user-inputted graph for a plurality of relevant graphs. The system orders the plurality of relevant graphs from a most relevant ranking to a least relevant ranking. The system returns, to the client computing device, the ordered plurality of relevant graphs for display on a user interface of the client computing device, thereby enhancing the search for relevant graphs by allowing the graph search engine to take as an input the user-inputted graph and return as an output the relevant graphs.

Abstract: A computer-implemented method, system and computer program product for improving query searches. After receiving a query from a user to conduct a content search, the query is analyzed for its semantic meaning and a categorized group of query tags and content tags in the central repository that is most semantically similar in meaning to the meaning of the query is identified. Furthermore, the content tags and query tags in the user's repository are analyzed to determine the interests of the user. The query may then be augmented to include one or more other terms of interest from the identified categorized group of query tags and content tags in the central repository that match the determined interests of the user within a threshold degree of relatedness, where these other terms of interest correspond to the content tags and query tags of the identified categorized group based on their assigned weight.

Abstract: Systems and methods including one or more processing modules and one or more non-transitory storage modules storing computing instructions configured to run on the one or more processing modules and perform acts of training one or more first models to recommend a first item after a user has had an interaction on the web site of the online retailer with a second item, determining static features common to both the first item and the second item, training a second model to determine whether to coordinate a display of any new item as one of one or more recommended items with any of a plurality of items, and coordinating the display of the new item as one of the one or more recommended items when the one or more of the plurality of items are displayed on the website of the online retailer based on the static features of the new item.

Abstract: Systems, methods, and computer program products for associating a media content clip(s) with other media content clip(s) having a different modality by determining first embedding vectors of media content items of a first modality, receiving a media content clip of a second modality, determining a second embedding vector of the media content clip of the second modality, ranking the first embedding vectors based on a distance between the embedding vectors and the second embedding vector, and selecting one or more of the media content items of the first modality based on the ranking, thereby pairing media content clips based on emotion.

Abstract: A network data transmission method is provided. A switch device receives one or more data flows, classifies each of the received data flows into one of two classes according to data features of the data flow by using a decision tree model established by a flow table pipeline of the switch device. If a data flow belongs to a first class, the switch device reports the data flow to a controller, so that the controller computes a transmission path for the data flow. If a data flow belongs to a second class, the switch device obtains a transmission path for the data flow according to local flow table information, and transmits the data flow according to the obtained transmission path. Data flows are classified and filtered by using a switch, so as to improve network transmission efficiency while ensuring bearing capability of a network control system.

Abstract: Methods and systems for text classification include identifying a text sample, classified according to a first class in a set of hierarchical graphs of classes, that is more similar to a second class in the set of hierarchical graphs of classes than to the first class according to a similarity metric. A word is deleted from the identified text sample that is associated with the second class. A word is added to the identified text sample that is associated with the first class. A text classification model is trained using the text sample as training data. Text classification is performed on an input text using the trained text classification model implemented on a processor device.