Fujitsu Corporation

Mehdi Bahrami

Wei-Peng Chen

Takuki Kamiya

Junhee Park

Lei Liu

Mukul Ranjan Prasad

Samir Parikh

Abhilash Ugaonkar

Oussama Chafiqui

Shrikanth Narayanaswamy Chandrasekaran

Manisha Mukherjee

Sarthak Arora

Yinlin Deng

Mehdi Assefi

According to an aspect of an embodiment, operations for deep parameter learning for code synthesis are provided. The operations may include receiving a source code file and generating an abstract syntax tree (AST). The operations may further include determining a set of classes, and functions/procedures from the computer-executable code and extracting metadata associated to each component. The operations may further include selecting a subset of functions for which descriptions in the extracted metadata satisfy filtering criteria and updating the computer-executable code by filtering lines of code (LoCs) corresponding to the subset of functions/procedures. The operations may further include generating a dataset of code features and respective metadata features that includes a deep connection between parameters and its usage based on the updated computer-executable code and the metadata generation task. The operations may further include training a language model on a sequence-to-sequence generation task, based on the dataset.

Deep parameter learning for code synthesis

A method may include obtaining a dataset that may include at least a first column and a second column, with a first title and a first value and a second title and a second value, respectively. The operations may further include determining a total similarity value between the first and second columns based on a metadata similarity value, a semantic similarity value, and/or a unit of measurement similarity value; adding the first and second columns to a cluster if the total similarity value is less than a threshold value; generating a new column to add to the cluster using a feature engineering function, where the new column may include a new title and a new value, the new value determined using the feature engineering function by acting on the first value and/or the second value. Further, the operations may include adding the new column to the dataset.

Unit prediction for custom feature engineering

According to an aspect of an embodiment, operations for code enrichment through metadata for code synthesis are provided. The operations include acquiring package data that include source code files and package metadata. The operations further include extracting additional metadata associated with software package and preparing metadata features based on the package metadata and the additional metadata. The operations further include identifying a set of target portions of a source code included in the source code files and updating one or more source code files using the metadata features. Such files are updated by performing at least one of a revision of existing code comments, and an addition of new code comments for the target portions. The operations further include generating a dataset of natural language (NL) text features and respective code features and training a language model on a sequence-to-sequence generation task.

Code enrichment through metadata for code synthesis

According to an aspect of an embodiment, operations for code enrichment for training language models on tasks related to computer programming are provided. The operations include receiving source code data including a computer-executable code and a natural language (NL) text. The operations further include determining blocks of code from the computer-executable code. The operations further include extracting a set of features related to components of the source code data from the blocks of code. The extraction is performed by parsing the blocks of code using Abstract Syntax Tree (AST) data of the blocks of code. The operations further include revising the AST data. The operations further include updating the source code data based on the revised AST data and generating a dataset of NL and abstracted code features as training data based on the updated source code data and further training a language model on a sequence-to-sequence generation task.

Code enrichment for training language models relating to computer programming

According to an aspect of an embodiment, operations include receiving a set of NL descriptors and a corresponding set of PL codes. The operations further include determining a first vector associated with each NL descriptor and a second vector associated with each PL code, using language models. The operations further include determining a number of a set of semantic code classes to cluster the set of PL codes into the set of semantic code classes, based on the number, the first vector, and the second vector. The operations further include training a multi-class classifier model to predict a semantic code class, from the set of semantic code classes, corresponding to an input NL descriptor. The operations further include selecting an intra-class predictor model based on the predicted semantic code class. The operations further include training the intra-class predictor model to predict a PL code corresponding to the input NL descriptor.

Code retrieval based on multi-class classification

A method comprises receiving vehicle data comprising information associated with a plurality of sensors of autonomous vehicle and segmenting the received vehicle data into non-public data and public data. The method further comprises partitioning the public data into a plurality of data partitions and generating a plurality of data levels of the public data. Each data level of the plurality of data levels is generated according to an access level of a plurality of access levels and includes one or more data partitions of the plurality of data partitions in an encrypted form. The method further comprises transmitting the generated plurality of data levels to a group of electronic devices. Each electronic device of the group of electronic devices retrieves, according to one of the plurality of access levels, at least a portion of the public data from the transmitted plurality of data levels.

Multi-level access control in sharing of vehicle data with devices

A method may include obtaining training code and extracting features from the training code. The extracted features of the training code may be mapped to natural language code vectors by a deep neural network. A natural language search query requesting source-code suggestions may be received, and the natural language search query may be mapped to a natural language search vector by the deep neural network. The method may include mapping the natural language search query to the natural language search vector in the same or a similar method as mapping the extracted features of the training code to natural language code vectors, and the natural language search vector may be compared to the natural language code vectors. Source code responsive to the natural language search query may be suggested based on the comparison between the natural language search vector and the natural language code vectors.

Source code retrieval

A method may include obtaining machine-readable source code. The method may include parsing the source code for one or more code descriptions and identifying a section of the source code corresponding to each of the code descriptions. The method may include determining a description-code pair including a first element representing the code description and a second element representing the section of the source code corresponding to the code description. The method may include generating an augmented programming language corpus based on the description-code pair, the one or more code descriptions, and the source code. The method may include receiving a natural language search query for source-code recommendations, identifying source code from the augmented programming language corpus responsive to the natural language search query, and responding to the natural language search query with the identified source code.

Semantic code search based on augmented programming language corpus

A method may include providing access to a first application programming interface (API) provided by a first party and a second API provided by a second party. The method may also include collecting a first set of API data sources related to the first API and a second set of API data sources related to the second API. The method may additionally include using a deep learning model to predict a privacy trustworthiness level for the first API and the second API, and disabling access to the first API based on the privacy trustworthiness level of the first API being below a threshold level.

Privacy trustworthiness based API access

Operations may include obtaining source code describing an optimization problem. The operations may include identifying problem parameters associated with the optimization problem such that a specialized computing system may be enabled to solve the optimization problem. The operations may include extracting one or more first parameters of the problem parameters from the source code. The operations may include identifying one or more second parameters of the problem parameters that are not included in the source code. A user may be prompted via a GUI for input relating to the one or more second parameters. The operations may include compiling the extracted first parameters and the user-provided second parameters as input parameters of the specialized computing system. The operations may include providing the input parameters to the specialized computing system such that the specialized computing system is able to solve the optimization problem.

Community-oriented, cloud-based digital annealing platform

According to an aspect of an embodiment, operations may include receiving a first input corresponding to a selection of a combinatorial optimization problem and receiving a set of datapoints as input. The operations may further include generating a first quadratic unconstrained binary optimization (QUBO) formulation based on an objective function for the combinatorial optimization problem and the received set of datapoints. The operations may further include selecting a first privacy setting and encoding the first QUBO formulation based on the selected privacy setting to generate a second QUBO formulation. The operations may further include submitting the generated second QUBO formulation to an optimization solver machine and receiving a first solution of the second QUBO formulation. The operations may further include decoding the first solution to produce a second solution and publishing an output of the combinatorial optimization problem on a user device based on the second solution.

Data privacy protection on cloud-based optimization systems

A method may include obtaining input data for an application programming interface (API), and encrypting the input data for the API using a public key of a provider of the API. The method may also include transmitting, to an API management server, an API request that invokes the API, where the API request includes an API call for the API and the encrypted input data. The API request may be in a format such that the API management server is capable of performing API management services based on the API call but unable to decrypt the encrypted input data with the public key.

Secure API flow

A method of automated ticket resolution comprises training and testing feature-specific classifier models using ticket database records. The feature-specific classifier models include machine-learning-based classification models related to features of a ticket system. The method includes training and testing feature-specific solution models using resolved ticket solution database records. The feature-specific classifier models include machine-learning-based solution models related to the features. The method includes receiving a ticket inquiry including a ticket indicative of an issue related to the features, generating a problem statement representative of the issue using the tested classifier models, and communicating the problem statement to the tested solution models. The method includes predicting a solution to the problem statement by using the tested solution models. The solution includes directions to resolve the ticket. The method includes implementing the solution in the system to resolve the issue based on certainty characteristics of the solution and recover a system if required.

Automated machine-learning-based ticket resolution for system recovery

A method may include obtaining one or more software-repository packages. A programming-language function may be extracted from the one or more software-repository packages. A curation resource associated with the programming-language function may be identified. The curation resource may include descriptive information related to the programming-language function. The method may include generating a code description corresponding to the programming-language function based on the curation resource. A function-comment pair that includes the programming-language function and the generated code description may be determined. A programming language corpus that includes the one or more software-repository packages may be generated and augmented by the function-comment pair. The method may include training a machine learning model using the programming language corpus.

Programming language corpus generation

According to an aspect of an embodiment, operations include creating object-oriented software platform by performing a textual analysis of a documentation corpus associated with a set of APIs. The operations further include generating a set of API call objects for an API endpoint of a first API of the set of APIs and constructing a set of natural language descriptors corresponding to the set of API call objects based on the textual analysis. The operations further include generating a set of business objects. Each business object encapsulates conditions applicable on a set of input/output parameters associated with a corresponding API call object of the set of API call objects. The operations further include constructing a software package that encapsulates the set of API call objects, the set of natural language descriptors, and the set of business objects.

Growing community of inventors

Santa Clara, CA, United States of America

Mehdi Bahrami