Abstract: Certain aspects of the present disclosure provide techniques for enumerating relationships in an object graph using traversal context objects that are decoupled from the implementation details of an object graph and the objects stored therein. An example method generally includes receiving a request to traverse an object graph, the object graph including a plurality of objects. A root context object is generated, and a traversal context object associated with a root node of the object graph is generated. The traversal context object is linked to the root context object and added to a traversal data structure. The graph is traversed by adding and removing traversal context objects to and from the traversal data structure until a terminating event occurs.

Abstract: A method that predicts business income from user transaction data. A multinomial classifier is trained, using a vector of features from data related to a historical transaction and a label associated with the historical transaction, to generate a probability that the historical transaction belongs to a specific classification with respect to income. Data related to a new transaction is split into a set of unigrams. A new vector of features is generated from the data related to the new transaction. The new vector includes a set of values that correspond and are assigned to the set of unigrams. A classification with respect to income is determined for the new transaction by applying the multinomial classifier to the new vector. The new transaction is labeled with the classification. One or more fields of a form that is maintained by an online service is populated using the classification.

Abstract: A method including receiving an incoming call from a calling device of a caller and determining identification information for the calling device. The method also includes receiving voice audio data of the caller from the calling device, converting the voice audio data to caller phones, and identifying a customer account associated with the identification information. The method further includes obtaining user phones for multiple candidate users associated with the identified customer account, comparing the caller phones to the user phones for the multiple candidate users, and determining the identity of the caller based on the comparison.

Abstract: A method and system identify characteristics of transaction description strings. The method and system extracts time data and location data from transaction description strings. The method and system generate estimated time data and location data for transaction strings that lack time data and location data by analyzing the time data and location data extracted from other transaction description strings. The method and system generate a user profile based on the estimated time data and estimated location data.

Abstract: Capturing dependencies between variables using a variable agnostic object is disclosed. A system is configured to obtain an indication of a first dependency of a first variable to a second variable via a programming interface and depict the first dependency, the first variable, and the second variable in a first instance of a variable agnostic object in a source code. The system is also configured to obtain an indication of a second dependency of a third variable to a fourth variable via the programming interface and depict the second dependency, the third variable, and the fourth variable in a second instance of the variable agnostic object in the source code. The system is also configured to compile the source code to generate a computer-executable program capturing the first dependency and the second dependency based on the first instance and the second instance of the variable agnostic object.

Abstract: A method may including obtaining a voice transcript corpus and a chat transcript corpus, extracting voice transcript sentences from the voice transcript corpus and chat transcript sentences from the chat transcript corpus, encoding, by a series of neural network layers, the voice transcript sentences to generate voice sentence vectors, encoding, by the series of neural network layers, the chat transcript sentences to generate chat sentence vectors, determining, for each voice sentence vector, a matching chat sentence vector to obtain matching voice-chat vector pairs, and adding, to a parallel corpus, matching voice-chat sentence pairs using the matching voice-chat vector pairs. Each of the matching voice-chat sentence pairs may include a voice transcript sentence and a matching chat transcript sentence. The method may further include training a disfluency remover model using the parallel corpus.

Abstract: The disclosed embodiments relate to a system that provides a selective encryption technique that encrypts all of the fields in a profile, and selectively enables consumers of the profile information to decrypt specific fields in the profiles. This is accomplished by encrypting each field in the profile using a randomly generated symmetric key, and then encrypting the symmetric key for each field with public keys belonging to individuals who are authorized to access each field. These encrypted public keys are stored in a header of the profile to enable individuals to use their corresponding private keys to decrypt symmetric keys for the specific fields that they are authorized to access.

Abstract: Systems and methods for identifying one or more master tables of a datalake are described. A system may obtain a plurality of computer readable tables of a datalake (with each computer readable table including one or more features). The system may also group the plurality of computer readable tables into a plurality of groups based on a number of features of each computer readable table of the plurality of computer readable tables. The system may further generate, for each of one or more groups of the plurality of groups, one or more neighborhoods based on a similarity of features between computer readable tables of the group. The system may also identify, for each neighborhood, one or more master tables from the one or more computer readable tables of the group. The system may further provide an indication of one or more master tables identified in the datalake.

Abstract: A method for automatic anonymous visitor identity resolution using machine learning, which includes generating a visitor histogram set from visitor events of a visitor event stream that include a visitor identifier and an internet protocol address, filtering a set of user identifiers into a candidate set of user identifiers based on the internet protocol address, obtaining one or more user histogram sets generated from user events that include user identifiers from the candidate set of user identifiers, and mapping the visitor identifier to a user identifier of the candidate set of user identifiers using a machine learning model and a histogram similarity matrix generated from the visitor histogram set, the one or more user histogram sets, and a set of histogram similarity functions. The method further includes presenting a response based on the mapping of the visitor identifier to the user identifier.

Abstract: Systems and methods may be used to generate and use intent predictions to enhance user experience. The intent predictions may describe the data required to resolve a user request included in a user input (e.g., question, search query, and the like) submitted by a user. The intent predictions may be generated using a machine learning model that comprises a model framework for extracting features and classifying user inputs into intent classes based on the extracted features. The intent predictions may be integrated into an information service to improve business metrics including contact rate, transfer rate, helpful rate, and net total promoter score.

Abstract: Potential client/small business data and specialist profile data for multiple, and in various embodiments, hundreds, thousands, tens of thousands, hundreds of thousands, millions, tens of millions, or even hundreds of millions or more, potential clients/small businesses and specialists are used to train one or more matching models in an offline training environment using machine learning techniques. Once the one or more matching models are trained, the one or more matching models are used in an execution environment to process a given user's data and available specialist data to identify one or more available specialists determined to be a good match for the user. The user is then provided specialist recommendation data listing the matched available specialists and, in some cases, the user is provided one or more communication mechanisms for connecting with matched available specialists selected by the user.

Abstract: A method may include receiving a first definition of an object type from a first software component and a second definition of the object type from a second software component. The object type may be labeled by an ID. The method may further include storing, in a dynamic graph, a node labeled by the ID, and storing, in a type definition repository external to the dynamic graph, the first definition of the object type and the second definition of the object type. The method may further include receiving, from the first software component, a modified first definition of the object type. The method may further include replacing, in the type definition repository and using the ID, the first definition of the object type with the modified first definition, and transmitting, to the second software component, a message indicating a need to lookup, by the ID, the modified first definition.

Abstract: The disclosed embodiments provide a system that facilitates electronic communication among users. During operation, the system obtains financial data as a set of financial transactions. Upon receiving a request to transmit a message, the system obtains one or more recipient attributes from the request. Next, the system uses the financial transactions to identify a set of recipients that matches the one or more recipient attributes. The system then transmits the message to the set of recipients without revealing the PII for the set of recipients to a creator of the message.

Abstract: Aspects of the present disclosure provide techniques for providing a graphical user interface. Embodiments include displaying a text input field. Embodiments include receiving an input of at least a portion of a tag via the text input field. Embodiments include displaying, in response to the input and proximate to the text input field, a graphical representation of an existing tag that relates to the input. The graphical representation includes a type of the existing tag, the existing tag, and a colored section on a right side or a left side of the graphical representation having a color that is associated with the type of the existing tag in the computing application. Embodiments include receiving a selection of the graphical representation and displaying an instance of the graphical representation inside of the text input field.

Abstract: A method may include receiving an event from an event source. The event may correspond to event data. The event source may be a container executing an image. The image may correspond to image metadata including attributes describing the image. The method may further include combining the event data with the image metadata to obtain enriched data, detecting, using the enriched data, a deviation from a policy, and in response to detecting the deviation from the policy, performing an action to enforce the policy.

Abstract: This disclosure provides systems, methods and apparatuses for recommending items to users of a recommendation system. In some implementations, the recommendation system determines a plurality of contribution values based on interactions between a plurality of users and a plurality of items. Each of the plurality of contribution values represents a confidence level that a respective user prefers a respective item. The recommendation system further determines a capacity of each of the plurality of items. The capacity of each item represents a maximum number of users to which the item can be recommended. The recommendation system recommends one or more items of the plurality of items to each of the plurality of users based at least in part on the plurality of contribution values and the capacities of the plurality of items.

Abstract: A method including receiving a request for a report on a data set. The method also includes providing the report. The report includes a macro page having a subset of the data set. The method also includes converting the macro page into a primary tree data structure having levels. The method also includes buffering the primary tree data structure in a buffer to form a buffered tree data structure. The buffered tree data structure is buffered in a level order of the levels. The method also includes selecting a first micro page from the buffered tree data structure. The first micro page is configured for display on a user interface. The method also includes transmitting, to the user interface, the first micro page.

Abstract: A data management system manages secured data for a plurality of users. The data management system utilizes an access authorization system to authenticate users seeking access to the data management system. The access authorization system provides access tokens to authenticated users. The access tokens enable the authenticated users to access the data management system without again providing authentication data. The access authorization system includes, for each user, an access policy that governs whether the users can use the access tokens to access the data management system. The access tokens have a finite lifetime. If the users use the access tokens within the finite lifetime and if the users satisfy all of the access rules of the access policies, then the lifetime of the access tokens can be extended a finite number of times.

Abstract: Certain aspects of the present disclosure provide techniques for generating a replacement sentence with the same or similar meaning but a different sentiment than an input sentence. The method generally includes receiving a request for a replacement sentence and iteratively determining a next word of the replacement sentence word-by-word based on an input sentence. Iteratively determining the next word generally includes evaluating a set of words of the input sentence using a language model configured to output candidate sentences and evaluating the candidate sentences using a sentiment model configured to output sentiment scores for the candidates sentences. Iteratively determining the next word further includes calculating convex combinations for the candidate sentences and selecting an ending word of one of the candidate sentences as the next word of the replacement sentence. The method further includes transmitting the replacement sentence in response to the request for the replacement sentence.

Abstract: Certain aspects of the present disclosure provide techniques for performing inferences in a distributed computing environment. An example method generally includes receiving a request to perform an inference with respect to a set of input data. One or more client devices are selected for use in performing the inference with respect to the set of input data. A request to perform the inference is transmitted to the selected one or more client devices. The request generally includes an anonymized, vectorized version of the set of input data such that the selected one or more client devices are to perform the inference based on anonymized data. An inference response is received from each of the selected one or more client devices.