International Business Machines Corporation

Thomas John Eggebraaten

Mark Gregory Megerian

Marie Louise Setnes

Richard Joseph Stevens

Jeffrey Wayne Tenner

Eric W Will

Shannon Everett Wenzel

Roberto Delima

Corville O Allen

John Edward Petri

Richard D Dettinger

Kevin Spencer Barker

Joel C Dubbels

Patrick Michael Wildt

Adam Clark

Robert J Bestgen

William Craig Rapp

David Alan Wall

Janice R Glowacki

Travis Michael Drucker

Patrick Jerome Fleming

Jesse R Gorzinski

Sue Sandra Kelling

Daniel Paul Kolz

Michael D Pfeifer

Fernando Jose Suarez Saiz

Terrence Ross O'Brien

Jeffrey Bernard Nowicki

Douglas Robert Tollefson

Jeffrey James Cornell

Method and apparatus for performing feature engineering using negative inferences are provided. One example method generally includes identifying a plurality of concepts and analyzing a corpus of documents to determine a first co-occurrence rate for a first concept and a second concept in the plurality of concepts. The method further includes analyzing the corpus of documents to determine a second co-occurrence rate for the second concept and at least a third concept of a set of concepts related to the first concept and determining an inverse relationship between the second concept and the third concept. The method further includes generating test data for training a machine learning model including a negative inference between the second concept and the third concept and training the machine learning model using the test data.

Negative inferences in machine learning treatment selection

A computing system receives a plurality of medical perspectives for each medical treatment option of a plurality of medical treatment options. A machine learning model assigns a weight to each medical perspective for each medical treatment option by determining how often a care provider has agreed with each perspective within the plurality of medical perspectives. The computing system receives a medical treatment option ranking for each of the medical perspectives. The computing system generates a score for each medical perspective as applied to each medical treatment option based on a combination of the weight and the ranking for each medical perspective. The computing system aggregates the scores across the plurality of medical perspectives into an overall score for each medical treatment option.

Customized aggregate scoring using machine learning

Embodiments provide for reduced user input requests by identifying predefined diagnosis paradigms; creating a synthetic diagnosis paradigm via a machine learning process based on prior selections of action plans recommended by the predefined diagnosis paradigms and values entered therefor, wherein the synthetic diagnosis paradigm identifies the action plans to treat the plurality of conditions based on a subset of the attribute inputs used by the predefined diagnosis paradigms; generating a graphical user interface (GUI) to prompt input for values for the subset of attribute inputs; in response to receiving the values for the subset of attribute inputs, identifying at least one condition according to the predefined and synthetic diagnosis paradigms; and displaying the action plans in the GUI in association with the synthetic and predefined diagnosis paradigms according to evaluations of the action plans based on the respective logical structures.

Reducing input requests in response to learned user preferences

Mechanisms are provided which configure a medical treatment recommendation system to implement a medical condition independent treatment recommendation model. The medical condition independent treatment recommendation model operates on medical condition independent scoring features that are independent of any specific medical condition. The mechanisms configure the medical treatment recommendation system to receive a medical condition cartridge providing at least one medical condition specific evaluation feature that is specific to a medical condition, and process patient information based on a combination of an application of the at least one medical condition specific evaluation feature, and an execution of the medical condition independent treatment recommendation model of the medical treatment recommendation system, to generate a treatment recommendation for a patient medical condition associated with the patient. The mechanisms then output the treatment recommendation for the patient.

Medical condition independent engine for medical treatment recommendation system

Managing answers in a question-answering environment is disclosed. Managing answers in the question-answering environment can include sorting, based on a set of answer categories for a subject matter, a first set of answers into a first answer category and a second set of answers into a second answer category. Managing answers in the question-answering environment can include determining, using the subject matter, a first category sequence including the first answer category and the second answer category, and establishing, based on the first category sequence, a first answer sequence established from a portion of the first set of answers from the first answer category and a portion of the second set of answers from the second answer category.

Answer management in a question-answering environment

Disclosed aspects relate to generating rationales for treatment options. A set of preference scores that indicates a first preference score for a first treatment option of a set of treatment options may be received. A rank-order that indicates a first ranking for the first treatment option may be received. The set of preference scores may be analyzed with respect to the rank-order to determine a relationship between the first preference score and the first ranking for the first treatment option. Based on the relationship between the first preference score and the first ranking, a set of rationale data for the first treatment option may be generated with respect to the first rank. Based on a user profile for a user, the set of rationale data may be configured for the user. The set of rationale data which is configured for the user may be provided.

Rationale generation management

Mechanisms are provided for extending a medical condition base cartridge (MCBC) based on user modifications of the MCBC. The mechanisms generate a MCBC for a medical condition at least by performing cognitive processing of natural language content in a corpus to extract patient attributes and treatment information associated with the medical condition. Information stored in the MCBC is presented to a user via at least one interface and user input is received, where the user input indicates at least one of a modification of, addition to, or deletion of, a portion of the information presented to generate an extension to the MCBC. The extension to the MCBC is stored as an expert medical condition cartridge (EMCC) in association with the MCBC in a repository. The EMCC is input to a medical cognitive system to perform a medical cognitive operation on patient information based on the EMCC.

Extending medical condition base cartridges based on SME knowledge extensions

Systems and computer program products to, responsive to receiving a case by a deep question answering (deep QA) system, identify, in a corpus of information, a first variable for which a value was not specified in the case, compute an importance score for the first variable based on a concept in the corpus, wherein the concept is associated with the first variable, and upon determining that the importance score exceeds an importance threshold, determine that specifying a value for the first variable increases a confidence score of a response returned by the deep QA system beyond a confidence threshold.

Scoring attributes in a deep question answering system based on syntactic or semantic guidelines

Methods to, responsive to receiving a case by a deep question answering (deep QA) system, identify, in a corpus of information, a first variable for which a value was not specified in the case, compute an importance score for the first variable based on a concept in the corpus, wherein the concept is associated with the first variable, and upon determining that the importance score exceeds an importance threshold, determine that specifying a value for the first variable increases a confidence score of a response returned by the deep QA system beyond a confidence threshold.

Systems and computer program products to perform an operation comprising: identifying a first attribute of a source code in a deep question answering system, computing an influence score for the first attribute based on a rule in the source code used to compute a confidence score for each of a plurality of candidate answers generated by the deep question answering system, computing an importance score for the first attribute based at least in part on the computed influence score, and upon determining that the importance score exceeds a predefined threshold, storing an indication that the first attribute is an important attribute relative to other attributes specified in the source code.

Scoring attributes in deep question answering systems based on algorithmic source code influences

Methods to perform an operation comprising: identifying a first attribute of a source code in a deep question answering system, computing an influence score for the first attribute based on a rule in the source code used to compute a confidence score for each of a plurality of candidate answers generated by the deep question answering system, computing an importance score for the first attribute based at least in part on the computed influence score, and upon determining that the importance score exceeds a predefined threshold, storing an indication that the first attribute is an important attribute relative to other attributes specified in the source code.

System, method, and computer program product to identify candidate values to provide to a deep question answering (QA) system as part of a case, by receiving a case, wherein the case includes a plurality of documents for evaluation by the deep QA system, evaluating the plurality of documents using natural language processing (NLP) to identify one or more concepts reflected by text content within the plurality of documents in the case, wherein the plurality of documents includes a plurality of distinct values for at least a first one of the concepts, selecting, from the plurality of distinct values, a candidate value for the first concept to provide to the deep QA system to process the case, and prior to submitting the case to the deep QA system, returning at least the candidate value selected for the first concept to present in a user interface.

Verification of natural language processing derived attributes

Managing confidence data in a question-answering environment is disclosed. Managing confidence data can include sorting, based on a set of answer categories for a subject matter, a first set of a plurality of answers into a first answer category. The first set can correspond to at least one of a third set of a plurality of confidence scores and the second set can correspond to at least one of a fourth set of the plurality of confidence scores. Managing confidence data can include classifying confidence scores of the third set into one of a plurality of confidence buckets using a first threshold and determining a fifth set of a plurality of thresholds using the plurality of confidence scores. Managing confidence data can include classifying unclassified confidence scores of the third set into one of the plurality of confidence buckets using the fifth set of the plurality of thresholds.

Answer category data classifying using dynamic thresholds

Method, system, and computer program product to analyze a plurality of candidate answers identified as responsive to a question presented to a deep question answering system, by computing a first feature score for a first feature of an item of evidence, of a plurality of items of evidence, the first feature score being based on at least one attribute of the first feature, the item of evidence relating to a first candidate answer, of the plurality of candidate answers, and computing a merged feature score for the first candidate answer by applying the first feature score to a second feature score for a second feature of the item of evidence.

Growing community of inventors

Rochester, MN, United States of America

Thomas John Eggebraaten