International Business Machines Corporation

Kavitha Srinivas

Achille Belly Fokoue-Nkoutche

Anastasios Kementsietsidis

Julian Timothy Dolby

Mihaela Ancuta Bornea

Songyun Duan

Aditya Anand Kalyanpur

Oktie Hassanzadeh

Ibrahim Abdelaziz

Octavian Udrea

Michael James Ward

Horst Cornelius Samulowitz

Achille Fokoue

Edith Gail Schonberg

Bishwaranjan Bhattacharjee

Martin J Hirzel

Guo Tong Xie

Min Wang

Brian Patrick Byrne

Wen Sun

Gang Hu

Avraham Ever Shinnar

Feifei Li

Louis Mandel

Kiran A Kate

Wangchao Le

Marcelo Arenas

Gonzalo Diaz

Larisa Shwartz

Genady Grabarnik

Alfio Massimiliano Gliozzo

James J Fan

Nagui Halim

Li Li Ma

Lingfei Wu

Patrick Dantressangle

Darrell Christopher Reimer

Shirin Sohrabi Araghi

Mark David Feblowitz

Harini Srinivasan

William Karol Lynch

Udayan Khurana

Kirill M Osipov

Cristina Cornelio

Bassem Makni

Aaron Stephen Jay Kershenbaum

Fabio Lorenzi

Aaron Kershenbaum

Veronika Thost

Maxwell Crouse

Brendan O'Sullivan

Sainyam Galhotra

Manik Bhandari

Wenting Zhao

A corpus of source code from a code database is accessed and a language prediction model is trained based on the corpus of source code. A given program is accessed and a completion of a given line of the given program is predicted by performing inferencing using the language prediction model and at least a portion of the given program. The given line is completed based upon the prediction.

Performing code completion using data flow

Techniques for computer software code analysis are disclosed. One or more data flows are generated, based on analyzing software code using static analysis. A data object is identified in the software code using the one or more data flows, the data object relating to a structured dataset. A correspondence between a code expression in the software code and a characteristic of the structured dataset is identified, based on analyzing one or more reads from and one or more writes to the data object using the one or more data flows. The code expression for the structured dataset is analyzed, based on the correspondence, including at least one of: (i) generating a software code recommendation engine based on the code expression and the structured dataset, or (ii) generating one or more lambda expressions for application to the structured dataset, based on the code expression.

Mining code expressions for data analysis

A processor may receive structured data. The structured data may include one or more columns and associated column names. The processor may analyze the structured data. Analyzing the structured data may include gathering a requisite set of keywords from the associated column names across all columns and/or a sample of column cells. The processor may access a corpus of documents. Each of the documents in the corpus may be associated with a respective keyword. The processor may search the corpus of documents based on the requisite set of keywords. The processor may summarize one or more documents associated with the requisite set of keywords.

Automatic domain annotation of structured data

A computer-implemented method is provided that includes accessing candidate text and a candidate pair including first and second phrases, substituting the first and second phrases into cause-effect patterns to generate variant sentences. An artificial intelligence model is leveraged to determine respective probabilities that the variant sentences are inferred from the candidate text, calculate a statistical measure of the respective probabilities, and assess the calculated statistical measure to ascertain whether the first and second phrases possess a causal relationship or non-causal relationship to one another. A knowledge base including one or more pairs of cause-effect phrase pairs is populated with the first and second phrases possessing the causal relationship. A computer system and a computer program product are also provided.

Causal knowledge identification and extraction

Generate, from a logical formula, a directed acyclic graph having a plurality of nodes and a plurality of edges. Assign an initial embedding to each mode and edge, to one of a plurality of layers. Compute a plurality of initial node states by using feed-forward networks, and construct cross-dependent embeddings between conjecture node embeddings and premise node embeddings. Topologically sort the DAG with the initial embeddings and node states. Beginning from a lowest rank, compute layer-by-layer embedding updates for each of the plurality of layers until a root is reached. Assign the embedding update for the root node as a final embedding for the DAG. Provide the final embedding for the DAG as input to a machine learning system, and carry out the automatic theorem proving with same.

Capturing the global structure of logical formulae with graph long short-term memory

To improve the technological process of programming a computer using a dynamic programming language, generate a first portion of training data which maps types in the dynamic programming language to corresponding functions and methods by performing information retrieval on documentation libraries in the dynamic programming language and/or generate a second portion of training data which maps program variables to the corresponding functions and methods by performing data flow analysis on a plurality of pre-existing programs written in the dynamic programming language. Train a neural network on the first and/or second portions of training data to infer unknown types in the dynamic programming language. Carry out inference with the trained neural network to infer the unknown types. Facilitate programming in the dynamic programming language based on the inferred unknown types. Optionally, execute a resulting program.

Type inference in dynamic languages

Embodiments are disclosed for a method. The method includes identifying a prefix updated by a searcher of a machine learning model. The machine learning model is configured to generate source code in a programming language. The method also includes determining whether the prefix violates a semantic correctness property of the programming language. Additionally, the method includes instructing the searcher, in response to the determination, to prune the prefix from a set of prefixes under consideration by the searcher.

Reducing semantic errors in code generated by machine learning models

A method, system, and computer program product are disclosed. The method includes extracting at least one identifier from a formula in a document and extracting text passages in the document that contain the identifier(s). The method also includes selecting an identifier and extracted text passages containing the identifier, as well as generating identifier-passage pairs for the selected text passages and the identifier. Further, the method includes submitting the identifier-passage pairs to a question answering (QA) model, which generates candidate answers from the selected text passages. A definition of the identifier is then selected from the candidate answers.

Question answering approach to semantic parsing of mathematical formulas

Embodiments relate to a system, program product, and method for employing feature engineering to improve classifier performance. A first machine learning (ML) model with a first learning program is selected. The first selected ML model is operatively associated with a first structured dataset. First features in the first dataset directed at performance of the selected ML model are identified. A second structured dataset is assessed with respect to the identified features in the first dataset, and new features in the second dataset are identified, where the new features are semantically related to the identified features in the first dataset. The first dataset is dynamically augmented with the identified new features in the second dataset. The dynamically augmented first dataset is applied to the selected ML model to subject an embedded learning algorithm of the selected ML model to training using the augmented first dataset.

Knowledge aided feature engineering

A schema for a dataset is identified by identifying a dataset comprising data and relationships between data pairs. An original schema is identified for the dataset. This original schema comprises an organizational structure. An initial fit between the dataset and the original schema is determined. The initial fit quantifying a conformity of the data in the dataset to the organizational structure of the original schema. A plurality of additional schemas are identified. Each additional schema is a distinct organizational schema. The dataset is partitioned into a plurality of subsets. Each subset comprises a modified fit quantifying a modified conformity of subset data in each subset to one of the original schema and the additional schemas. The modified fit is greater than the original fit.

Determining the schema of a graph dataset

Techniques for code analysis are provided. User code is received, and an import statement is identified in the user code. A first empty object is generated based on the import statement, and the first empty object is named based on the name of an import reference included in the import statement. A knowledge graph is generated based at least in part on the first empty object.

Generation of knowledge graphs based on repositories of code

Systems and methods for indexing and accessing code snippets in repositories. A program graph index is maintained for code snippets within a repository with documents that have at least one code snippet. The program graph index includes a program graph indicating a relationship between program elements within each source code snippet within the documents. A user provided code snippet is received and a target program graph indicating a relationship between program elements within the user provided code snippet is determined and compared to each respective program graph. Based on the comparison, an identified set of documents within the repository of documents is determined that have code snippets with respective program graphs that are also at least a sub-tree of the target program graph. At least one document in the identified set of documents is presented to a user.

Indexing and accessing source code snippets contained in documents

A processor may analyze one or more sections of code within a program. The processor may compute a context for each of the one or more sections of code. The processor may generate one or more search terms that are respectively associated with each of the one or more sections of code. The processor may automatically display reference data to a user.

Automatic identification of reference data

A system for identifying a schema for storing graph data includes a database containing a graph dataset of data and relationships between data pairs and a list of storage methods that each are a distinct structural arrangement of the data and relationships from the graph data set. An analyzer module collects statistics for the graph dataset, and a data classification module uses the collected statistics to calculate metrics describing the data and relationships in the graph dataset, uses the calculated metrics to group the data and relationships into a plurality of graph dataset subsets and associates each graph dataset subset with one of the plurality of storage methods. The resulting group of storage methods associated with the plurality of graph dataset subsets includes a unique storage method for each graph dataset subset. The data and relationships in each graph dataset subset are arranged in accordance with associated storage methods.

Method and apparatus for identifying the optimal schema to store graph data in a relational store

Growing community of inventors

Rye, NY, United States of America

Kavitha Srinivas