International Business Machines Corporation

Jeffrey Scott Boston

Stephen Peter Wood

Jennifer Ceil Lai

Shimei Pan

Mercan Topkara

Anni Rosa Coden

Zhen Wen

Jie Lu

Matthew R Arnold

Michael Desmond

Donald Philip Pazel

Zvi P Weiss

A method optimizes a training of a machine learning system. A conflict detection system discovers a conflict between a first training data and a second training data for a machine learning system, where the first training data and the second training data are ground truths that describe a same type of entity, and where the first training data and the second training data have different labels. In response to discovering the conflict between the first training data and the second training data for the machine learning system, an oracle adjusts the different labels of the first training data and the second training data. The machine learning system is then trained using the first training data and the second training data with the adjusted labels.

Discovering and resolving training conflicts in machine learning systems

Methods, systems and computer program products for improving ground truth quality for modeling are provided. Aspects include receiving a plurality of data inputs, wherein each of the plurality of data inputs has an associated label. Aspects also include training a model based on the plurality of data inputs. Aspects also include generating a plurality of vector representations corresponding to the plurality of data inputs based on the model. Aspects also include clustering the plurality of vector representations into one or more clusters. Aspects also include identifying at least one anomalous data input based on the one or more clusters. The at least one anomalous data input can be a data input of the plurality of data inputs that is mislabeled, contributes to an ambiguous class structure or is an outlier. Aspects also include outputting a notification that provides an indication of the at least one anomalous data input.

Ground truth quality for machine learning models

A method includes generating, as executed by a processor on a computer, a plurality of topic-specific user knowledge models for each user of a plurality of users, each topic-specific user knowledge model representing a level of knowledge possessed by a respective user on a single topic from a set of globally defined topics shared among the plurality of users, generating a plurality of topic-specific expert knowledge models, each topic-specific expert knowledge model representing an aggregate level of knowledge possessed by a plurality of expert users on a single topic from a set of globally defined topics shared among the plurality of users, comparing the topic-specific user knowledge model of the first user with the topic-specific expert knowledge model for a respective topic to determine a distance between a user knowledge level and an aggregate expert knowledge level for the topic.

Systems, methods, and computer program products for expediting expertise

A system includes a user model module that generates a plurality of topic-specific user knowledge models for each user of a plurality of users, each topic-specific user knowledge model representing a level of knowledge possessed by a respective user on a single topic from a set of globally defined topics shared among the plurality of users, a expertise model building module that generates a plurality of topic-specific expert knowledge models, each topic-specific expert knowledge model representing an aggregate level of knowledge possessed by a plurality of expert users on a single topic from a set of globally defined topics shared among the plurality of users, and a processor of a computer that executes instructions for comparing the topic-specific user knowledge model of the first user with the topic-specific expert knowledge model for a respective topic to determine a distance between a user knowledge level and an aggregate expert knowledge level for the topic.

A system including a global topic model building module that builds a global topic model based on a set of data that is updated according to an activity of each user of a plurality of users, a user model building module that builds a plurality of user models, each user model being built based on an activity of a respective user, an expertise model building module that builds an expertise model for the topic based on the activity of at least one user of the plurality of users, the expertise model for the topic setting a target level of knowledge for a first user of the plurality of users, a processor to compare a user model of the first user with the expertise model for the topic; and an expertise assessment and learning recommendation module that recommends an activity associated with the set of data to the first user based on the comparison.

A method including generating a global topic model based on a set of data that is updated according to an activity of each user of a plurality of users, the global topic model including a topic representation for a topic, generating a plurality of user models, each user model being generated based on the activity of a respective user, generating an expertise model for the topic based on the activity of at least one user of the plurality of users, the expertise model for the topic setting a target level of knowledge for a first user of the plurality of users, comparing a user model of the first user with the expertise model for the topic, the comparing being performed by a processor of a computer system, and recommending an activity associated with the set of data to the first user based on the comparison.

Automatically generating a meeting digest of a set of meetings is provided. A set of topics of interest to the parties to the set of meetings is detected utilizing a user model associated with a user that is based on at least one of communications, relationships, and roles of the parties to the set of meetings. Topic-related content associated with the set of topics of interest to the parties is extracted from meeting data corresponding to the set of meetings. Then, the meeting digest of the set of meetings is generated using the topic-related content associated with the set of topics of interest to the parties extracted from the meeting data corresponding to the set of meetings.

Automatically generating a personalized digest of meetings

A first method manages a display screen window of an application program that is halted and includes the following steps: in responsive to an application program being halted, (a) copying the informational content of a window of the halted application program; (b) creating a second window, referred to as a shadow window, to be in registration with the window of the halted application program such that the window of the halted application program is hidden from the view of an observer of the display screen; (c) painting the second window with the copy of the informational content of the halted application program; and (d) in response to a message to re-paint the second window, re-painting the second window with the copy of the informational content of the halted application program. A second method is responsive to an application program being halted for (a) copying the window or windows of the halted application program; (b) intercepting a window re-paint message that is directed to the halted application program; and (c) re-painting the display screen window of the halted application program with that portion of the copy that corresponds to the window of the halted application program. Both methods may also invalidate obscured portions of the application window, such as by setting the corresponding portion of the copy to some predetermined color.

Growing community of inventors

Wappingers Falls, NY, United States of America

Jeffrey Scott Boston