International Business Machines Corporation

Cartic Ramakrishnan

Daniel Gruhl

Alfredo Alba

Clemens Drews

Neal Ryan Lewis

Pablo N Mendes

Meenakshi Nagarajan

Sheng Hua Bao

Linda Ha Kato

Christian Benjamin Kau

Anni Rosa Coden

Joseph F Terdiman

Shenghua Bao

Pathirage D Perera

Ramani Ranjan Routray

Ying Chen

Chad Eric DeLuca

Chris Kau

Anshu Narendra Jain

Pathirage D S U Perera

Maia A Smith

Tanveer Fathima Syeda-Mahmood

William Scott Spangler

Kenneth Lee Clarkson

Jacques Labrie

Ronald Fagin

Bruce Light Hillsberg

John Christian Unterholzner

Abhinandan Kelgere Ramesh

Pathirage Dinindu Sujan Udayanga Perera

Nakul Chakrapani

Vivek Krishnamurthy

Claire Abu-Assal

Embodiments relate to a system, program product, and method for use with a physical computing device to process a data access request. The requested data is encrypted with two keys, including a physical device authentication key and a transient key. Access to the data requires authentication on both the device level and situational level. Device situational data is monitored, which includes selectively enabling access to the requested data and de-activation of the transient key in response to a change in the monitored situational data. The transient key de-activation removes access to the requested data.

Transient management of data encryption and authentication

A mechanism is provided in a data processing system for training a computer implemented model. The mechanism determines an operation for which the computer implemented model is to be trained. The mechanism performs a statistical analysis of an enterprise dataset for an enterprise to generate one or more statistical distributions of cases and features correlated with the operation for which the computer implemented model is to be trained. The mechanism selects a subset of cases in the enterprise dataset for annotation based on the one or more statistical distributions of cases and features. The mechanism annotates the selected subset of cases to generate an annotated training dataset. The mechanism trains the computer implemented model, using the annotated training dataset, to perform the operation.

Cognitive data preparation for deep learning model training

At least some embodiments are directed to a computer-implemented method that comprises receiving original input text that includes a term, comparing a definition of the term to definitions of multiple candidate replacement terms to generate a set of candidate replacement terms, and substituting each of the candidate replacement terms in the set for the term in the original input text to produce a plurality of modified input texts. The method also comprises determining the grammatical accuracy of each of the plurality of modified input texts, comparing meanings of the modified input texts to a meaning of the original input text, and modifying the set of candidate replacement terms based on the determinations of grammatical accuracy and the comparisons of the meanings. The method still further comprises ranking the modified set of candidate replacement terms using one or more criteria, and displaying the ranking on a display.

Identifying linguistic replacements to improve textual message effectiveness

A method, a system, and a computer program product are provided. A machine learning model is generated to process adverse event information and produce multiple corresponding medical codes associated with the adverse event information, wherein the multiple medical codes are semantically and hierarchically related in a medical taxonomy. The machine learning model includes multiple parallel output layers, each of which is associated with a corresponding medical code. The machine learning model is trained with training data elements, each of which includes adverse event information mapped to respective multiple medical codes, wherein results from each of the output layers adjusts the machine learning model. After completing the training, information pertaining to an adverse event is applied to the machine learning model to determine the corresponding multiple medical codes within the medical taxonomy.

Architecture for machine learning model to leverage hierarchical semantics between medical concepts in dictionaries

The embodiments relate to generating hierarchical patterns based on a corpus of text. The corpus is analyzed, which includes extracting a set of features of the corpus. A set of grammatical patterns are generated based on the extracted features. The set of grammatical patterns includes at least one grammatical pattern generated from an internal pattern and at least one grammatical pattern generated from an external pattern. The grammatical patterns of the set are organized into a hierarchy and/or are ranked. The hierarchy and/or ranking are visually displayed.

Hierarchical target centric pattern generation

Disclosed aspects relate to analytics engine selection management. A set of reference data may be ingested by a first analytics engine to compile a first set of characteristic data. The set of reference data may be ingested by a second analytics engine to compile a second set of characteristic data. The first set of characteristic data may be compiled for the first analytics engine. The second set of characteristic data may be compiled for the second analytics engine. A set of distinct attributes related to the first and second analytics engines may be determined based on the first and second sets of characteristic data. An analytics engine selection operation may be executed.

Analytics engine selection management

Embodiments relate to a system, program product, and method for use with a physical computing device to process a data access request. The associated data is encrypted with a key pair that includes both a persistent key and a transient key. Both keys require authentication to access the requested data. The transient key is subject to real-time monitoring, with changes in situational data selectively affecting the validity of the transient key, and selectively changing the physical state of the physical computing device.

Mechanisms are provided that implement a drug-adverse event causality evaluation engine to identify human resource selections based on a readability of unstructured text within an individual case safety report (ICSR) and a confidence value of the ICSR. The drug-adverse event causality evaluation engine receives the ICSR from a cognitive system. The drug-adverse event causality evaluation engine analyzes the ICSR to determine a readability value of the ICSR. The drug-adverse event causality evaluation engine determines whether or not an assessment, by a human reviewer, of the ICSR is required based on a combination of the readability value of the ICSR and the confidence value. The drug-adverse event causality evaluation engine outputs an indication of whether human reviewer assessment is required.

Human resource selection based on readability of unstructured text within an individual case safety report (ICSR) and confidence of the ICSR

One embodiment provides generating a similarity matrix corresponding to an input collection including initializing, by a processor, a working set as a collection of a multiple items. Until the similarity matrix converges: receiving a seed for similarity for at least one pair of items of the multiple items, and obtaining a similarity value for all other item pairs using a Naive Triangle Inequality process. The similarity is generated with obtained similarity values.

Efficient semi-supervised concept organization accelerated via an inequality process

A method, a system, and a computer program product are provided. A training set of adverse event text fragments assigned to medical codes is analyzed to determine first text fragments having frequently occurring medical code assignments and second text fragments having infrequently occurring medical code assignments. The training set is modified to undersample the first text fragments and to oversample the second text fragments such that the text fragments of the modified training set correspond to a substantially uniform assignment of the medical codes. At least one machine learning model is generated and trained with the modified training set. Some parameters of the at least one machine learning model are updated based on errors detected during the training. After completing the training, an adverse event text fragment is applied to the at least one machine learning model to assign at least one medical code.

Automatic classification of adverse event text fragments

A method, a computing system and a computer program product are provided. A computing system identifies elements within a collection of medical documents. The elements include patients, adverse events and medical drugs. The medical documents are analyzed by the computer system to determine associations between the identified medical drugs and corresponding identified adverse events. The identified elements and the determined associations may be encoded as features by the computing system. The computing system identifies portions of the medical documents as containing the identified elements and the determined associations. The computing system generates a classification model based at least on the encoded features associated with the identified portions for identifying medical case safety reports within medical documents. The classification model is applied to a new document to determine a classification of the new document with respect to a medical case safety report.

Cognitive framework to identify medical case safety reports in free form text

A method, system and computer program product for disambiguating meaning of terms. A natural language processing system obtains a first corpora of words used in a first sense and a second corpora of the same words that are used in a second sense. Each of these corpora of words may be associated with different linguistic domains. The natural language processing system generates a first and a second set of patterns using both the first and second corpora of words, respectively. A question passage is then received by the natural language processing system. The natural language processing system examines a word of interest in the question passage. The user is then notified that the word of interest is being used in the first sense or the second sense in response to identifying the word of interest in only the first set of patterns or the second set of patterns, respectively.

Disambiguation of the meaning of terms based on context pattern detection

Performing an operation comprising extracting, from an input comprising unstructured electronic text, a plurality of feature values for a plurality of features defined in a feature vector, identifying, based on a machine learning (ML) model applied to the plurality of feature values, a portion of the electronic text corresponding to an adverse event, and annotating the portion of the electronic text with an indication of the identified adverse event.

Growing community of inventors

San Jose, CA, United States of America

Cartic Ramakrishnan