Abstract: Low latency search for nearest neighbors in a dataset containing a large number of entries is improved using an error correction code (ECC) for partitioning data into clusters and retrieval. During initialization and preprocessing a d-dimensional space with clusters corresponding to ECC codewords is specified. Entries in the dataset are embedded into this space and associated with respective codewords, each codeword specifying a cluster. An index associates the codewords, clusters, and entries. During a query of the dataset, a query entry is processed to determine a query embedding in the d-dimensional space. The query embedding is used as input for a list decoder of the ECC. The list decoder provides a set of nearest codewords, with those codewords representing a set of candidate clusters that may contain nearest neighbors. The dataset entries associated with the candidate clusters may then be searched to determine query results comprising specific entries.

Abstract: Techniques and a system to facilitate estimation of a quantum phase, and more specifically, to facilitate estimation of an expectation value of a quantum state, by utilizing a hybrid of quantum and classical methods are provided. In one example, a system is provided. The 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 include an encoding component and a learning component. The encoding component can encode an expectation value associated with a quantum state. The learning component can utilize stochastic inference to determine the expectation value based on an uncollapsed eigenvalue pair.

Abstract: According to the present invention, provided is a method for collecting filtered text data, comprising the steps in which: a computing device acquires first text data and records the first text data in a text data pool; the computing device acquires second text data; the computing device performs a calculation in a deep learning model by using the first text data and the second text data as input values and calculates a first feature vector corresponding to the first text data and a second feature vector corresponding to the second text data; and the computing device compares the degree of similarity between the first feature vector and the second feature vector, and records the second text data in the text data pool when the degree of similarity is less than a predetermined value.

Abstract: A risk-prediction-preparation module to generate a risk-prediction-model, is provided herein. The risk-prediction-preparation module includes accessing a data-storage of transactions to operate a group-by operation on transactions related to data-points, according to a logical-entity into entities. Then, clustering entities of a clean-financial dataset into clusters. Selecting data-points of: (a) entities from the clusters to a first dataset and (b) a preconfigured amount of entities randomly to a second dataset. Selecting all entities that have at least one ‘fraudulent’ data-points in at least one related data-point to add all the entities to the first dataset and the second dataset. Using vectorized and scaled extracted features for training a first machine-learning-model of fraud detection on the first dataset and training a second machine-learning-model of fraud detection on the second dataset to collect results.

Abstract: A method including generating, by a navigation service, a route for navigating from a route origin to a route destination using a private roads repository. The method includes identifying a ghost origin and a ghost destination of a ghost road along the route. The method includes sending, using an application programming interface of a base roads engine, a first request for a route from the ghost origin to the ghost destination. The method includes receiving, from the base roads engine in response to the first request, a replacement section from the ghost origin to the ghost destination. The method includes replacing, in the route, the ghost road with the replacement section to create an updated route including segments. The method includes generating an estimated travel time from the route origin to the route destination over the segments of the updated route. The method includes presenting the estimated travel time.

Abstract: A method for deployment of machine-learning based operators within a query is described. For this embodiment, a sequence of operators associated with a query is identified, which includes at least a first operator and at least a second operator. The second operator is configured to perform operations, in accordance with a machine learning (ML) component, on data received as input from execution of the first operator. Schemas associated with the machine learning component is retrieved along with schemas associated with other operators within the sequence. Compatibility between at least an output schema associated with the first operator and an input schema associated with the second operator associated with the ML component is determined. Thereafter, a portion of the sequence of operators including at least the second operator and another operator of the sequence of operators successive to the second operator may be stored within a data store for subsequent use.

Abstract: An example system includes a processor to execute a key point analysis to extract key points from a number of input sentences. The processor can compute directional relation scores between the extracted key points. The processor can construct a key point graph based on the extracted key points and the directional relation scores. The processor can generate a key point graph visualization based on the key point graph.

Abstract: A system and method of enhancing the trustworthiness of an artificial intelligence system include detecting whether a data element includes an existing data domain tag, processing the data element into a transformed data element, generating a data domain tag, where the data domain tag includes at least a data domain identifier and a timestamp, appending the data domain tag to the transformed data element, creating a signature for the transformed data element and the appended data domain tag using a private key, and creating another signature for the data domain tag using the private key.

Abstract: A system for automated information support processing is disclosed. The system may receive a natural language information support request from a plain text input channel. The system may determine a first user intent based on the information support request. The system may compare the first user intent with a set of support rules. The system may determine a dispositioned outcome based on the set of support rules and the user intent. The system may return the dispositioned outcome.

Abstract: A method and apparatus for preparing an image on a cloud platform, and a device and a storage medium are provided. The method may include: receiving an image request for a target object, and obtaining a first storage backend of the target object according to the image request; obtaining a second storage backend of a Glance service in the cloud platform, and comparing the second storage backend with the first storage backend; and obtaining an image for the target object on the second storage backend in a cloning manner when the first storage backend is the same as the second storage backend. According to the method, when the first storage backend is the same as the second storage backend, the image for the target object is obtained on the second storage backend in the cloning manner, so that the downloading and uploading processes are omitted.

Abstract: Existing content such as books are reorganized and condensed as electronic books for display on a mobile computing device. The electronic books have a hierarchical structure. A user interacts with the content of the book by way of a touch screen of the mobile computing device. The navigation may be non-linear in nature and the book is reconstructed as a set of primary ideas, supporting ideas, stacks of cards for the supporting ideas, and individual cards comprising elements and commentary from or about the book.

Abstract: Domain name system (DNS) configuration during virtual private network (VPN) connection includes, by a VPN entry server, receiving from a client device, via a VPN tunnel between the entry server and the client device, a first request for first content that identifies a first external source for the first content, receiving, from an operative DNS server configured for the tunnel, an Internet Protocol (IP) address of a first VPN system exit server, in response to determining that the first content is unavailable via the first system exit server, identifying a second DNS server, such that a second VPN system exit server for obtaining the first content is available using the second DNS server, obtaining, from the second system exit server, the first content, wherein the second system exit server obtained the first content from the first external source, and transmitting, to the client device, via the tunnel, the first content.

Abstract: Embodiments disclosed herein describe systems and methods to generate statuses of higher level integrated logical clusters from one or more output files generated by the clusterwares of a plurality of lower level physical clusters. The systems and methods provide an interactive dashboard graphical user interface (GUI) rendering the statuses of the higher level logical clusters. In response to a drill down selection by the system administrator, the systems and methods may render resource level statuses of the plurality of resources forming the higher level logical cluster. Therefore, by using a single dashboard, a system administrator can seamlessly transition between statuses at different level of abstractions for multiple resources distributed across multiple geographically diverse datacenters. Furthermore, the dashboard allows the system administrator to perform failover operations.

Abstract: Provided are a computer program product, system, and method for pre-processing a table in a document for natural language processing (NLP). A graphical user interface (GUI) provides a representation of table items in a table in a document including a set of a main element comprising an entity whose value is to be extracted, a conditional element that refines the entity, and a value element comprising a value for the entity. Graphical controls are rendered in the GUI to enable a user to select an element from the table to be the main element, conditional element, and value element. The set of the main element, conditional element, and value element are updated with the user selected element to form a modified set. The modified set of the main element, conditional element, and the value element are provided to an NLP engine to perform natural language processing.

Abstract: Embodiments of the present disclosure relate to an article processing method, electronic device, and computer program product. The method includes: determining, based on content of a target article, a target article vector associated with the target article; acquiring a reference article vector set associated with a reference article set; and determining, based on a distance in an article vector space between the target article vector and a reference article vector in the reference article vector set, a reference article vector associated with the target article vector in the reference article vector set as an association article vector. By using the technical solution of the present disclosure, an association article associated with a target article can be accurately provided based on the target article selected by a user, so that reports on the target article and its association articles can be further provided to the user for analysis and selection.

Abstract: A system for generating a workload model for a target information handling system includes gathering workload data from a plurality of information handling systems, dividing the workload data into a plurality of workload data bins, identifying workload data characteristics for each workload data bin and identifying workload data sets that may be applicable to the target information handling system. A workload mix may be determined based on the target information handling systems and workload data characteristics. Real customer workload data including real-time or near real-time workload data may be used to check a workload model for accuracy before deploying the workload model to a target information handling system.

Abstract: The present disclosure relates to database technology and in particular to dynamically updating and customizing database partitions. A computer-implemented engine is disclosed for identifying and retrieving a number of data records applicable to generate a response to a request, the engine having access to at least two partitions. Partition statistics are generated indicating correlations between the data records and, based on that partition statistics, the data records having the strongest correlation with each other are relocated to partitions so that the number of partitions which have to be queried in order to generate a response to a data request is minimized. Furthermore, the computational load caused when generating responses is more equally distributed across the partitions.

Abstract: Methods, systems, and computer-readable media are disclosed herein to provide rule-based reconciliation of records. Specifically, rules are utilized to reconcile one or more records and identify duplicates therein. Once duplicate records are identified, one or more ranking sets can be utilized to identify which of the duplicate records to write to the system.

Abstract: Clustering structured log data by key-values includes receiving, via a user interface, a request to apply an operator to cluster log messages according to values for keys associated with the request. At least a portion of each log message comprises structured machine data including a set of key-value pairs. The method further includes receiving a log message and determining whether to include the log message in a cluster based at least in part on an evaluation of values in the structured machine data of the log message for the keys associated with the request. The cluster is included in a set of clusters. Each cluster in the set is associated with a different combination of values for the keys associated with the request. The method further includes providing, via the user interface, information associated with the cluster.

Abstract: A storage device generates and stores a key stream and a value stream by extracting from data a plurality of keys and a plurality of values respectively corresponding to the plurality of keys. The storage device includes a controller and a non-volatile memory. The controller receives from a host information about an invalid key included in the key stream together with a compaction command, and performs a compaction operation on the key stream in response to the compaction command. The non-volatile memory stores the key stream and the value stream. The controller merges the key stream with another key stream based on the information about the invalid key in the compaction operation.

Abstract: Systems and methods for scaling application and/or storage system functions of a distributed storage system based on a heterogeneous resource pool are provided. According to one embodiment, the distributed storage system has a composable, service-based architecture that provides scalability, resiliency, and load balancing. The distributed storage system includes a cluster of nodes each potentially having differing capabilities in terms of processing, memory, and/or storage. The distributed storage system takes advantage of different types of nodes by selectively instating appropriate services (e.g., file and volume services and/or block and storage management services) on the nodes based on their respective capabilities. Furthermore, disaggregation of these services, facilitated by interposing a frictionless layer (e.g.

Abstract: Embodiments of the present disclosure provide for improved determination of complexities associated with event data objects. Such complexity determinations may be identified based at least in part on complexity score(s) generated from any number of augmented event codes. Embodiments identify augmented event codes from any of a myriad of sources, including various service encodings and participant history profiles for particular event entities. The augmented event codes may be processed in a myriad of ways, utilizing the entirety and/or portions thereof, to further enhance the accuracy of the event complexity determined therefrom. Based at least in part on the accurately generated complexity determination(s), one or more prediction-based actions may be initiated that correspond to the particular complexity determination(s), for example such that only the correct prediction-based action is initiated based at least in part on such a determined event complexity.

Abstract: A computer system displays a coalesced representation of a plurality of notifications in a notification-display user interface, wherein the coalesced representation, when selected, causes the computer system to concurrently display individual representations of at least two of the plurality of notifications in the notification-display user interface.

Abstract: A data intake and query system can generate local data enrichment objects and receive federated data enrichment objects from another data intake and query system. In response to receiving a query, the data intake and query system can determine whether the query is subquery of a federated query. If the query is a subquery, the data intake and query system can use the federated data enrichment objects to execute the query.

Abstract: A wearable electronic device may include a display, a camera, at least one first sensor, and at least one processor operatively coupled to the display, the camera, and the at least one first sensor.

Abstract: Based on a predetermined number of available processor sockets, a plurality of candidate matrix decompositions are identified, which correspond to a multiplication of matrices. Based on a first comparative relationship of a variation of first sizes of the plurality of candidate matrix decompositions along a first dimension and a second comparative relationship of a variation of second sizes of the plurality of candidate matrix decomposition sizes along a second dimension, a given candidate matrix decomposition is selected. Processing of the multiplication among the processor sockets is distributed based on the given candidate matrix decomposition.

Abstract: Systems and methods are provided for receiving a time series dataset from a monitored processor and group the dataset into a plurality of clusters. Using an unsupervised machine learning model, the system may combine a subset of the plurality of clusters by data signature similarities to form a plurality of motifs and combine the plurality of motifs into one or more shapelets. In some examples, the system may train a supervised machine learning model using the plurality of motifs and the one or more shapelets as input to the supervised machine learning model. The system can perform various actions in response to labelling the time series dataset, including predicting a second time series dataset, determining that a monitored processor corresponds with an overutilization at a particular time, or suggesting a reduction of additional utilization of the monitored processor.

Abstract: A processor may receive a video including a plurality of video frames in sequence and a question regarding the video. For a video frame in the plurality of video frames, a processor may parse the video frame into objects and relationships between the objects, and create a subgraph of nodes representing objects and edges representing the relationships, where parsing and creating are performed for each video frame in the plurality of video frames, where a plurality of subgraphs can be created. A processor may create a hypergraph connecting subgraphs by learning relationships between the nodes of the subgraphs, where a hyper-edge is created to represent a relationship between at least one node of one subgraph and at least one node of another subgraph in the plurality of subgraphs. A processor may generate an answer to the question based on the hypergraph.

Abstract: Disclosed is an autonomous mining method of industrial big data based on model sets, which comprises the following steps: S1, building model sets and a mining engine based on domain knowledge and structural characteristics of multi-source heterogeneous data; S2, carrying out data sampling on the multi-source heterogeneous data, and counting the fault-tolerant estimation of random error variance; S3, mining data sets by using the mining engine, and determining the optimal fault-tolerant model of each sampled data sequence and the optimal fault-tolerant estimation of model parameters; S4, performing goodness-of-fit statistics calculation and VV&A test by using the optimal fault-tolerant model; S5, acquiring data model representation and connotation knowledge based on model clustering. The method can realize the automation of the mining process of big data, the integration of associated knowledge, the expansion of model sets, the integration of mining and modeling and the optimization of mining results.

Abstract: Some embodiments provide a program that determines a plurality of data objects. Each data object in the plurality of data objects includes a first attribute and a second attribute. The program further sorts values of the first attribute of the plurality of data objects. The program also sorts values of the second attribute of the plurality of data objects. The program further determines a first distance value based on the sorted values of the first attribute of the plurality of data objects. The program also determines a second distance value based on the sorted values of the second attribute of the plurality of data objects. The program further defines a plurality of clusters based on the sorted values of the first attribute of the plurality of data objects, the first distance value, the sorted values of the second attribute of the plurality of data objects, and the second distance value.

Abstract: Records can be matched by a graph neural network model performing entity resolution on the records, and representing each record as a respective node in a graph. Record matching explanations can be generated, each record matching explanation indicating a first set of attributes, and a first set of corresponding values, used for the matching at least two of the records. Nodes can be clustered into a plurality of clusters by aggregating the record matching explanations and, based on the record matching explanations, determining which of the records have high importance values, in the first set of values, that match. At least one cluster explanation can be generated, the cluster explanation indicating a second set of attributes, and a second set of values corresponding to the second set of attributes, used for the clustering the nodes. The record matching explanation and the cluster explanation can be output.

Abstract: Accordingly, embodiments herein disclose a method and apparatus for retrieving intelligent information from an electronic device (100). The method includes receiving, by the electronic device (100), an input from a user. Further, the method includes identifying, by the electronic device (100), at least one data item to generate at least one metadata tag. Further, the method includes automatically generating, by the electronic device (100), the at least one metadata tag related to the at least one data item based on a plurality of parameters. Further, the method includes providing, by the electronic device (100), at least one priority to the at least one metadata tag. Further, the method includes storing, by the electronic device (100), the at least one metadata tag at the electronic device (100).

Abstract: Domain Name System (DNS) security using process information is provided. An application accessing an internet service using a domain name is determined. Process information associated with the application along with an associated DNS query to identify an IP address associated with the domain name are identified. The process information and the associated DNS query to a DNS security service are sent. An action based on a response from the DNS security service is performed.

Abstract: A computer implemented method includes building a Positive Knowledge Base with directive words, designated verbs and designated objects. A Negative Knowledge Base with designated phrases and designated legal terms is built. Tasks and phrases from the Positive Knowledge Base and the Negative Knowledge Base are built. Regulations are received. Phrases from the regulations are weighted against the Positive Knowledge Base and the Negative Knowledge Base to isolate positive Maintenance Compliances. The positive Maintenance Compliances are matched to tasks to derive ranked Maintenance Compliances. The ranked Maintenance Compliances are supplied.

Abstract: An imaging mass spectrometry unit (1) executes predetermined analysis on each of a plurality of micro areas set in a two-dimensional measurement region on a sample or a three-dimensional measurement region in a sample to acquire spectrum data. A clustering execution section (21) classifies spectrum data for a plurality of measurement points obtained for a reference sample into any of a plurality of clusters. A clustering model information storage section (22) stores a clustering model obtained by clustering processing. A segmentation execution section (23) classifies respective spectral data for a plurality of measurement points obtained for a sample other than a reference sample using a clustering model, and a spatial distribution image creation section (24) creates a segmentation image obtained by partitioning a two-dimensional or three-dimensional image into a plurality of small regions on the basis of a result of the classification.

Abstract: A computer-implemented method, comprising: using a normalization service of a first application virtual compute instance hosted in and executing in a cloud computing facility, accessing a plurality of digitally stored data records of a transaction processing system, each of the data records comprising a plurality of data values for a corresponding plurality of attributes; using the normalization service, executing a data cleaning operation on the data records by programmatically calling two or more data cleaning functions, each of the data cleaning functions corresponding to a particular attribute among the plurality of attributes; using the normalization service, creating and storing a plurality of clusters in memory of the first application virtual compute instance by clustering the data records based on exact matches of the data values of a first attribute and creating and storing references to the data records in the clusters; using the normalization service, joining two or more clusters among the plurali

Abstract: Messages, such as text messages, emails, and social media posts, are a popular way to communicate, however, messages can be overwhelming. In order to better manage messages, messages may be grouped, such as those having a common topic. A recipient may then address the grouped messages as a group. However, certain messages may have a common topic but have an attribute that warrants maintaining the message as individual messages, excluded from the group. For example, the message may be from someone who rarely communicates with the recipient, a celebrity, an important figure at work or in politics, a family member, or other attribute that would warrant exclusion from a common topic group.

Abstract: Methods, systems, and computer programs are presented for displaying a customized news stream. One method includes an operation for identifying dwell times spent by users while accessing a first plurality of items, each item belonging to one media type from a plurality of media types (e.g., news articles, videos, slide shows, etc.). In addition, the method includes operations for determining statistical parameters for each media type based on the identified dwell times, and for detecting a news corpus having a second plurality of items. A priority for each item in the news corpus is determined based on the media type of the item, the corresponding statistical parameters for the media type of the item, and the profile of a user. The news stream is sent to the user for presentation on a display, the news stream being sorted based on the priority of the items in the news corpus.

Abstract: Methods, systems, and media for lot classification are disclosed. In one example, a classification system for identifying lot listings receives a description for a listing in a publication system, identifies a string in the listing, identifies a quantity word or digit in the string, and converts an identified quantity word into digit form. A normalized string is tokenized to produce tokens, the tokenizing of the normalized string including splitting the normalized string into a series of substrings using a sequence of delimiters. For each substring, an additional split is performed by separating any digit from any other adjacent character, unless that character is another digit, and maintaining an internal character order of each split substring to produce a flattened list of tokenized tokens.

Abstract: A building management system includes one or more memory devices configured to store instructions thereon, that, when executed by one or more processors, cause the one or more processors to receive a publication by an agent on an agent communication channel, the publication comprising timeseries data, identify, based on the publication, an object entity of an entity database associated with the agent, wherein the entity database includes one or more object entities and relationships between the one or more object entities and one or more data entities, identify a data entity related to the object entity based on a relationship of the relationships relating the object entity and the data entity, and ingest the timeseries data into the data entity.

Abstract: A resource integration method includes the following steps: a receiving module receives access information from a guest operating system on the host device; the access information is used to determine whether the frame rate is lower than a frame rate threshold; when the receiving module determines that the frame rate is lower than the frame rate threshold, the receiving module transmits an external resource request signal to the receiving module; after the receiving module receives the external resource request signal, a resource management module (which is located in the bridge module) selects an optimal external device from a specific category (among a plurality of categories in a candidate list), and a calculation operation or a storage operation corresponding to the specific category is transmitted to the optimal external device for calculation or storage by the bridge module.

Abstract: Devices, systems and methods are provided for managing records. One method involves constructing a fuzzy match graph data structure with edges between different field values for a fuzzy matched field of a plurality of records in accordance with a matching rule, assigning unique identifiers to different subgraphs of field values within the fuzzy match graph data structure, determining a match key value associated with a respective record based on a respective unique identifier assigned to a respective subgraph of field values including a respective field value for the fuzzy matched field of the respective record, constructing a heterogenous graph data structure having edges between the respective match key values associated with the respective records of the plurality of records, identifying a subset of related records using the heterogenous graph data structure, and performing one or more actions on the subset of related records at a database system.

Abstract: Systems, computer program products, and methods are described herein for detection and classification of intrusion using machine learning techniques. The present invention is configured to electronically receive, from a computing device of a user, an indication that the user has initiated a first resource interaction; retrieve information associated with the first resource interaction, wherein the information comprises at least one or more parameters associated with the first resource interaction; initiate a machine learning model on the one or more parameters associated with the first resource interaction; and classify, using the machine learning model, the first resource interaction into one or more classes, wherein the one or more classes comprises one or more access types.

Abstract: An apparatus for the generation of electronic media is disclosed. The apparatus includes at least a processor and a memory communicatively connected to the at least a processor. The memory instructs the processor to receive an entity profile from the user, wherein the entity profile comprises a plurality of entity data. The memory instructs the processor to classify the plurality of entity data into a plurality of entity categories. The memory instructs the processor to generate development data as a function of the classification. The memory instructs the processor to generate electronic media as a function the development data. The memory instructs the processor to display the electronic media using a display device.

Abstract: Examples of wearable systems and methods can use multiple inputs (e.g., gesture, head pose, eye gaze, voice, totem, and/or environmental factors (e.g., location)) to determine a command that should be executed and objects in the three-dimensional (3D) environment that should be operated on. The wearable system can detect when different inputs converge together, such as when a user seeks to select a virtual object using multiple inputs such as eye gaze, head pose, hand gesture, and totem input. Upon detecting an input convergence, the wearable system can perform a transmodal filtering scheme that leverages the converged inputs to assist in properly interpreting what command the user is providing or what object the user is targeting.

Abstract: This present application describes one or more embodiments of methods, non-transitory computer-readable media, and systems provide developer tools for generating and providing visualizations of data densities for various portions of a computer application. For example, the systems can determine a data density that reflects a ratio or an amount of data presented within a (portion of a) display window relative to the entire (portion of the) display window. The systems can further provide a visual representation of a data density for display on a client device, along with indications or suggestions for how to improve (e.g., reduce or increase) the data density for better comprehensibility (e.g., upon distribution of the application). In certain embodiments, the systems can generate suggestions based on an application type of a computer application in development and/or based on a function associated with a particular (portion of a) display window.

Abstract: A continuous query identified by an archived relation of an application may be received. A physical query plan may be generated for the continuous query and/or a child operator plan may be identified. A buffer operator may be inserted between the child operator and its parent operator and/or the event information from the child operator may be stored in the buffer operator. A query operator may be identified in the physical query plan. It may be determined that the query operator identified in the physical query plan is a same type as a second query operator in a second continuous query concurrently executing. Further, a combined query plan may be generated based at least in part on the query operator identified in the physical query plan being the same type as the second query operator in the second continuous query.

Abstract: Various embodiments of systems and methods for providing virtualized (e.g., serverless) broker clusters for a data streaming service are disclosed. A data streaming service uses a front-end proxy layer and a back-end broker layer to provide virtualized broker clusters, for example in a Kafka-based streaming service. Resources included in a virtualized broker cluster are monitored and automatically scaled-up, scaled-down, or re-balanced in a way that is transparent to data producing and/or data consuming clients of the data streaming service.

Abstract: A building management system includes one or more memory devices configured to store instructions thereon, that, when executed by one or more processors, cause the one or more processors to receive a publication by an agent on an agent communication channel, the publication comprising timeseries data, identify, based on the publication, an object entity of an entity database associated with the agent, wherein the entity database includes one or more object entities and relationships between the one or more object entities and one or more data entities, identify a data entity related to the object entity based on a relationship of the relationships relating the object entity and the data entity, and ingest the timeseries data into the data entity.

Abstract: Using a natural language (NL) latent presentation in the automated conversion of source code from a base programming language (e.g., C++) to a target programming language (e.g., Python). A base-to-NL model can be used to generate an NL latent representation by processing a base source code snippet in the base programming language. Further, an NL-to-target model can be used to generate a target source code snippet in the target programming language (that is functionally equivalent to the base source code snippet), by processing the NL latent representation. In some implementations, output(s) from the NL-to-target model indicate canonical representation(s) of variables, and in generating the target source code snippet, technique(s) are used to match those canonical representation(s) to variable(s) of the base source code snippet. In some implementations, multiple candidate target source code snippets are generated, and a subset (e.g., one) is selected based on evaluation(s).