International Business Machines Corporation

Stephan Joseph Roorda

Andrew Ronald Freed

James W Murdock, Iv

Peter Frederick Haggar

Mary Swift

Michael D Whitley

Richard L Stillwell

Dane A Johnson

Corville O Allen

Kevin B Haverlock

John Edward Petri

Richard Joseph Stevens

Mark Gregory Megerian

Martin Labsky

David Bacarella

Leah Rae Smutzer

Rebecca Rose James

Arunava Majumdar

Jasmeet Singh

Miroslav Vodolán

Frank Somich

Alexander McMullen

Dilan Orrino

Josef Ondrej

Pattern recognition by receiving a set multi-variable data records, each record including a plurality of variables, representing at least two of the plurality of variables as geometric shapes, defining a boundary enclosing the geometric shapes, configuring at least one geometric shape to move within the boundary, capturing a location of each of the geometric shapes within the boundary as a system state, one or more times, combining one or more system states as a system signature, providing a model trained to recognize patterns in system signatures, and recognizing a pattern in the system signature.

Multi-variable pattern recognition for predictive deep learning models

The present disclosure relates to a method for a dialog system. The method comprises: receiving a user input at the dialog system. A machine learning module may receive from the dialog system a dialog system response to the user input. In response to determining that the machine learning module is in a deactivated mode, selected one or more training data items of the dialog system response and the user input may be used for training the machine learning module. In response to determining that the machine learning module is in an active mode the trained machine learning module may estimate from the set of output data items and the user input a machine learning module response to the user input. The machine learning module response or the dialog system response may be provided.

Machine learning module for a dialog system

Embodiments include a method, system and computer program product for detecting impersonation attempts in social media messaging. Aspects include receiving, via a social media network, a message from a sender to a recipient and analyzing a content of the message to extract factual statements from the message. Aspects also include analyzing a profile of the recipient to extract facts from the profile and comparing each of the factual statements to the facts from the profile. Based on a determination that one of the factual statements are verifiable by at least one of the facts, aspects include assigning a likelihood score to the factual statements. Aspects further include calculating a legitimacy score for the message based at least in part on the likelihood score of each verified factual statement from the message and transmitting the legitimacy score and the message to the recipient.

Detection impersonation attempts social media messaging

Provided is a method, computer program product, and system for improving online behavior in social networking environments using predictive analytics. A processor may monitor communications and user action data of a plurality of users participating in a social networking environment. The processor may analyze the communications and the user action data to determine a user score related to a set of behavioral characteristics for each of the plurality of users. The processor may compare the user score for each of the plurality of users to a set of corrective action thresholds. In response to a first user score associated with a first user meeting one of the set of corrective action thresholds, the processor may implement a first corrective action related to improving behavior of the first user.

Online behavior using predictive analytics

A method, computer program product, and computer system, for receiving a first set of ground truth instances from a first source. A second set of ground truth instances may be received from a second source. The first and second sets of ground truth instances may be weighed differently based on a level of trust associated with each of the first and second sources. The weighted first and second sets of ground truth instances may be applied in a machine learning task executed by a computer.

System and method for defining and using different levels of ground truth

A method, computer program product, and computer system, for receiving a first set of ground truth instances from a first source. A second set of ground truth instances may be received from a second source. The first and second sets of ground truth instances may be weighted differently based on a level of trust associated with each of the first and second sources. The weighted first and second sets of ground truth instances may be applied in a machine learning task executed by a computer.

A mechanism is provided in a data processing system comprising a processor and a memory configured to implement a question and answer system (QA), for improving answers to an input question. An input question is received and parsed for a known set of critical information required to answer the input question. Responsive to one or more pieces of critical information failing to be identifiable in the input question, a user who submitted the input question is prompted for the one or more pieces of critical information via a graphical user interface (GUI).

Analyzing natural language questions to determine missing information in order to improve accuracy of answers

An approach is provided to analyze posts included in threads of one or more threaded discussions. The analysis identifies child posts that refer to a parent post that was selected a user of a user interface. A statement in the parent post is identified and that statement is analyzed using agreement data found in the child posts that refer to the parent post. The parent post and the analysis of the statement is displayed to the user at the user interface.

Annotating posts in a forum thread with improved data

An approach is provided to train a question answering (QA) system using answer keys based on forum content. In the approach, a question is selected from a post in a threaded discussion. An answer to the selected question is automatically identified from crowd-based sources, with the identified answer having a confidence level greater than a threshold. An answer key is built using the selected question and the identified answer. The QA system is automatically trained using the answer key.

Training a question/answer system using answer keys based on forum content

An approach is provided for evaluating subject matter experts (SMEs) in a question and answering (QA) system. In the approach, a number of responses are received at the QA system with each of the responses being a response to a common question and each of the responses being received from a SME. One of the responses is selected with the selected response being from one of the SMEs that is being evaluated. The approach evaluates the selected response by comparing the selected response to the responses received from the other SMEs. Based on the evaluation, the approach updates a SME confidence score that corresponds to the selected SME.

Evaluating expert opinions in a question and answer system

Mechanisms are provided for generating an answer for an input question when the answer is not directly present in a corpus of information. An input question is received from a computing device and analyzed to determine whether the input question is requesting an answer that is calculable. In response to a determination that the input question is requesting an answer that is calculable, one or more constituent data values are retrieved, from a corpus of information, for calculating the requested answer to the input question. A value corresponding to the requested answer is calculated based on the one or more retrieved constituent data values and is then output as the requested answer to the input question.

Determining answers in a question/answer system when answer is not contained in corpus

Mechanisms are provided for dynamically selecting a subset of candidate answer sources for use by a question and answer (QA) system. The QA system receives an input question for which an answer is sought and generates one or more queries based on the input question. The QA system selects a subset of candidate answer sources, from a group of candidate answer sources, based on a ranked listing of candidate answer sources. The ranked listing of candidate answer sources is generated based on characteristics of previous candidate answers generated by the QA system for previously input questions. The QA system applies the one or more queries to the subset of candidate answer sources and generates at least one candidate answer for the input question based on results of applying the one or more queries to the subset of candidate answer sources.

Growing community of inventors

Fuquay-Varina, NC, United States of America

Stephan Joseph Roorda