International Business Machines Corporation

Achille Belly Fokoue-Nkoutche

Kavitha Srinivas

Anastasios Kementsietsidis

Mihaela Ancuta Bornea

Julian Timothy Dolby

Ping Zhang

Sanjoy Dey

Oktie Hassanzadeh

Songyun Duan

Aditya Anand Kalyanpur

Robert G Farrell

Mohammad Sadoghi Hamedani

Michael James Ward

Ibrahim Abdelaziz

Katherine Shen

Ryan A Musa

John A Gunnels

Mustafa Canim

Isabelle Marie Rouvellou

Louis Degenaro

Vijay Anand Saraswat

Maria Chang

Edith Gail Schonberg

Meinolf Sellmann

Guo Tong Xie

Jianying Hu

Min Wang

Brian Patrick Byrne

Wen Sun

Gang Hu

Hoi Yeung Chan

Mark Holmes Linehan

Spyros Kotoulas

Judah M Diament

Feifei Li

Kartik Talamadupula

Kshitij P Fadnis

Pavan Kapanipathi Bangalore

Michael John Witbrock

Samuel Monk Weber

Wangchao Le

Cristina Cornelio

Bassem Makni

Muhammad Mustafa Rafique

Heng Luo

Charles A Kerr, Jr

Arvind Rajpurohit

Veronika Thost

Maxwell Crouse

Arun Kwangil Iyengar

Larisa Shwartz

William Scott Spangler

Genady Grabarnik

Alfio Massimiliano Gliozzo

James J Fan

Li Li Ma

Lingfei Wu

Nagui Halim

Rosario A Uceda-Sosa

Chulaka Gunasekara

Michael Moser

Daniela Bourges-Waldegg

Marcel Graf

Alix M Lacoste

Alexander Gray

Mariano Rodriguez Muro

Kirill M Osipov

Guilherme Augusto Ferreira Lima

Sundar Saranathan

Yann Duponchel

Aaron Stephen Jay Kershenbaum

Aaron Kershenbaum

Nicholas Mattei

Avinash Balakrishnan

Aldo Pareja

Marcelo Machado

Brendan O'Sullivan

Gonzalo Ignacio Diaz Caceres

Siva Sankalp Patel

Provided are techniques for a search using an overlay graph mapping to source knowledge graphs. A plurality of overlay graphs are generated, where each overlay graph comprises entities represented by nodes and relations represented by edges, and where the entities and the relations map to a subset of entities and relations in a plurality of source knowledge graphs. A search request comprising an entity and a relation is received. An overlay graph is selected from the plurality of overlay graphs based on the entity and the relation. The search request is issued against the overlay graph, where the search request is translated to knowledge graph specific queries, and where the knowledge graph specific queries are issued against the plurality of source knowledge graphs. Search results are received from the plurality of source knowledge graphs. The search results are used to respond to the search request.

Search using an overlay graph mapping to source knowledge graphs

A method, apparatus and computer program product for establishing a logical connection between an indirect utterance and a transaction is described. An indirect utterance is received from a user as an input to a conversational system. The indirect utterance is parsed to a first logical form. A first set of predicates and terms is mapped from the first logical form to a first subgraph in a knowledge graph. A second set of predicates and terms is mapped from a second logical form belonging to a transaction to a second subgraph of the knowledge graph. A best path in the knowledge graph between the first subgraph and the second subgraph is searched for while transforming the first logical form using the node and edge labels along the best path to generate an intermediate logical form. A system action is performed for a transaction if a graph structure of the intermediate logical form matches the graph structure of the logical form of the transaction above a threshold.

Establishing a logical connection between an indirect utterance and a transaction

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

A method for assigning weights to a knowledge graph includes extracting information from a knowledge graph. The information including entities extracted from nodes of the knowledge graph and relations extracted from edges of the knowledge graph. A shortest path generator receives the extracted entities and relations, and potential assigned weights from a heuristic data repository. Weights for the edges of the knowledge graph are determined. The weights are assigned to the edges of the knowledge graph.

Interpretable knowledge contextualization by re-weighting knowledge graphs

A method of improving computing efficiency of a computing device for language-independent problem solving and reasoning includes receiving a query from a user, which is decomposed into one or more sub-queries arranged according to a tree structure. The one or more sub-queries are executed in a knowledge base. The results of the executed one or more sub-queries are received and composed into a query response. The query response is transmitted to the user.

Problem manipulators for language-independent computerized reasoning

Systems, computer-implemented methods, and computer program products that can facilitate encoding a tree data structure into a vector based on a set of constraints are provided. According to an embodiment, a 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 comprise a constraint former that can form a set of constraints based on a first tree data structure and a vector encoder that can encode the first tree data structure into a vector based on the set of constraints.

Encoding and decoding tree data structures as vector data structures

One embodiment of the invention provides a method for natural language processing (NLP). The method comprises extracting knowledge outside of text content of a NLP instance by extracting a set of subgraphs from a knowledge graph associated with the text content. The set of subgraphs comprises the knowledge. The method further comprises encoding the knowledge with the text content into a fixed size graph representation by filtering and encoding the set of subgraphs. The method further comprises applying a text embedding algorithm to the text content to generate a fixed size text representation, and classifying the text content based on the fixed size graph representation and the fixed size text representation.

Infusing knowledge into natural language processing tasks using graph structures

Mechanisms are provided to implement a genomic database curation (GDC) system. The GDC system generates a ground truth database based on a training subset of datasets from an uncurated large scale genomic database, and label metadata for the training subset. The GDC system trains at least one classification engine of the GDC system based on the training subset and the ground truth database at least by performing a machine learning operation on the at least one classification engine. The GDC system automatically applies the at least one trained classification engine on the uncurated large scale genomic database to generate an automatically curated large scale genomic database. A meta-classifier engine generates an output specifying at least one of significant gene signatures or gene pathways for at least one of diseases or drug agents based on the automatically curated large scale genomic database.

Identifying gene signatures and corresponding biological pathways based on an automatically curated genomic database

Embodiments of the present invention are directed to a computer-implemented method for generating a framework for analyzing adverse drug reactions. A non-limiting example of the computer-implemented method includes receiving to a processor, a plurality of drug chemical structures. The non-limiting example also includes receiving, to the processor, a plurality of known drug-adverse drug reaction associations. The non-limiting example also includes constructing, by the processor, a deep learning framework for each of a plurality of adverse drug reactions based at least in part upon the plurality of drug chemical structures and the plurality of known adverse-drug reaction associations.

Identifying chemical substructures associated with adverse drug reactions

Embodiments of the present invention disclose a method, a computer program product, and a computer system for predicting drug and disease interactions. A computer identifies one or more drug similarity measures between one or more drugs and one or more disease similarity measures between one or more diseases. In addition, the computer identifies one or more interactions between the one or more drugs and the one or more diseases, then calculates one or more drug-disease feature vectors based on the one or more interactions, the one or more drug similarity measures, and the one or more disease similarity measures. Furthermore, the computer calculates a first probability indicating whether a first drug of the one or more drugs will interact with a first disease of the one or more diseases based on a model, wherein the model is trained based on the one or more drug-disease feature vectors.

Predicting interactions between drugs and diseases

A method, system, and computer program product for a data modelling platform to engage with a user via a user interface is provided. A predictive data model is displayed based on the user query. Provenance and evidence is provided based on a user selected result. The ground truth dataset is modified in response to receiving a user action via the user interface.

Data curation on predictive data modelling platform

A semantic embedding model using geometrical set-centric approach to capture both ABox and TBox representational models is disclosed. The model transforms a semantic-rich knowledge graph into a set of overlapping, disjoint, and/or subsumed n-dimensional spheres that captures and represents semantics embedded in the knowledge graph.

Set-centric semantic embedding

A method, computer system, and a computer program product for determining a composite similarity metric for data of a first data type is provided. The present invention may include providing a plurality of similarity metrics for the first data type. The present invention may also include providing a metric quantifying correlation of entities belonging to the first data type and entities belonging to a second data type. The present invention may then include developing a first regression model to predict values of the provided metric quantifying correlation of entities belonging to the first data type and entities belonging to the second data type using the provided plurality of similarity metrics. The present invention may further include calculating the composite similarity metric from a plurality of first regression coefficients in the developed first regression model.

Aggregating similarity metrics

Mechanisms are provided that implement a plurality of heterogeneous causality models and a metaclassifier for predicting a likelihood of causality between a drug and an adverse event (AE). The plurality of heterogenous causality models process drug information to generate a plurality of risk predictions for a drug and AE pair. The risk predictions include a risk score or a risk label indicating a probability of the AE occurring with use of the drug. The plurality of heterogenous causality models provide risk predictions, associated with the drug and AE pair, to a metaclassifier which generates a single causality score value indicative of a probability of causality between the drug and the AE, of the drug and AE pair, based on an aggregation of the risk predictions from the plurality of heterogenous causality models. The metaclassifier outputs the single causality score value in association with information identifying the drug and AE pair.

Evaluating drug-adverse event causality based on an integration of heterogeneous drug safety causality models

Growing community of inventors

White Plains, NY, United States of America

Achille Belly Fokoue-Nkoutche