Abstract: A computer implemented method for managing wireless bands and/or wireless channels comprises receiving information on connected local devices for a plurality of wireless routers, wherein the plurality of wireless routers are using a same wireless band and/or wireless channel, and wherein each connected local device for which information is received utilizes one or more of the plurality of wireless routers for a network connection. The method further comprises calculating, for each of the plurality of wireless routers, a congestion quotient from the received information on the connected local devices, wherein the congestion quotient defines a likelihood of destructive interference at each respective wireless router. The method further comprises determining changes in wireless band and/or wireless channel for one or more wireless routers of the plurality of wireless routers, wherein the changes lower the respective congestion quotients for each of the respective one or more wireless routers.

Abstract: Aspects of the present disclosure relate to managing messages within a collaboration environment. A group collaboration channel in which messages are displayed from multiple users is accessed, where the channel includes an application that provides application messages to the channel. A transcript of the channel is analyzed to determine a set of topics a user has discussed. A subset of topics is selected and a user sentiment of each topic is determined. A machine learning classification model is trained based on the selected subset of topics and the user sentiment of each topic. A message is then received from the application. Prior to displaying the message at the channel, a management category for the message is determined based on the classification model. The management category is applied to the message and the display of the message is managed based on the management category.

Abstract: Based on analyzing social media content, a model is constructed representing development of a first user's posts and interactions with other users over time. Using the model, a likelihood is computed that a post of the first user represents constructive dialogue. Using the model, the likelihood, and the post, a trajectory of future posts is forecasted. Responsive to forecasting a negative trajectory, a remedial action to positively alter the trajectory is applied.

Abstract: Methods and systems of outputting feedback indicators for content are described. A processor may receive a request from a device. The request may be a request for content. The processor may generate feedback data that comprises a feedback indicator assigned to a segment of the content. The feedback indicator may be based on feedbacks from a plurality of users that consumed the segment of the content. The processor may output the content on the device. The processor may detect an occurrence of the segment during the output of the content. The processor may, in response to the occurrence of the segment, output the feedback indicator assigned to the segment on the device.

Abstract: From social media content using a processor and a memory, a class model is constructed that is representative of a type and a level of toxicity associated with the content. A first social media message is analyzed, using a processor and a memory, to compute a first toxicity score associated with the first social media message and assign the first social media message to a first class of the class model. A second social media message is analyzed, using a processor and a memory, to compute a second toxicity score associated with the second social media message and assign the second social media message to a second class of the class model. Based on the first toxicity score, the first class, the second toxicity score, and the second class, a toxicity trend is forecasted. In response to the toxicity trend being forecasted to increase, a remedial measure is applied to reduce the toxicity trend.

Abstract: As a conference call is occurring, network data for the call is analyzed to identify an occurrence of a disturbance in the call. By injecting diagnostic traffic into the call, a first node is identified having a lowest performance score from a set of performance scores. By rerouting network traffic around the first node, the disturbance is remediated.

Abstract: Merging two or more user groups for collaboration includes constructing a social graph of each user of a plurality of users within the two or more user groups based upon a user profile and a workflow associated with each user. A weighted adjacency matrix is calculated based on at least one user interaction between two or more users indicated by the social graph. A model inferring leading edge contributions and least significant edge contributions among the users is generated based upon the social graph and the weighted adjacency matrix. One or more prospective groups expected to be involved in a future collaboration are identified based upon the identified leading edge contributions and least significant edge contributions.

Abstract: A new data structure of a social clustered topic model comprising new data is constructed, the new data being generated from content of online interactions using a processor and a memory. A social media message is analyzed, using a processor and a memory, to compute an emoji probability score reflecting a degree of correspondence between an emoji present in the social media message and a meaning extracted from the new social clustered topic model. The social media message is modified automatically, when the emoji probability score is above a threshold probability score, by adding additional data to the social media message, the additional data including the meaning.

Abstract: Disclosed embodiments provide techniques for automatically synchronizing a visual presentation with a live presenter. Visual presentation slides are preprocessed to determine one or more topics for each slide. A topic index contains one or more topics corresponding to slides of the presentation. As a presenter provides a verbal presentation for corresponding slides, natural language processing analyzes the verbal presentation and creates one or more temporal verbal topic categories. The temporal verbal topic categories are used to search the topic index to find one or more slides that best match the current temporal verbal topic categories. In this way, the slides can automatically follow the discussion of the presenter, enabling improved presentations that can enhance the user experience, increase audience engagement, and improve the dissemination of information.

Abstract: Methods and systems for leveraging a topic divergence model to generate dynamic sidebar chat conversations are disclosed. A method includes: collecting, by a computing device, data from a first group chat; determining, by the computing device, a main topic of the first group chat and a current topic of the first group chat using the data collected from the first group chat; determining, by the computing device, a first divergence score representing a degree of divergence between the main topic of the first group chat and the current topic of the first group chat; determining, by the computing device, that the first divergence score exceeds a predetermined threshold; and in response to determining that the first divergence score exceeds the predetermined threshold, the computing device creating a second group chat.

Abstract: A system and method for generating dynamic, non-subjective hashtags using a corpus-based approach includes capturing social media outputs from a user and from cohorts of the user to create a user corpora and a user cohort corpora, respectively, storing the user corpora and the user cohort corpora in a computer readable storage device coupled to the computing system, deriving a hashtag collation model by analyzing the user corpora and the user cohort corpora, the hashtag collation model being stored on the computer readable storage device, generating a best matched hashtag using the hashtag collation model, wherein the hashtag collation model infers the optimal hashtag from the user corpora and user cohort corpora based on a usage of one or more terms within the user corpora and the user cohort corpora, and adjusting the best matched hashtag over time as the user corpora and the user cohort corpora change over time.

Abstract: An embodiment of the invention may include a method, computer program product and system for redirecting data within a network. The embodiment may include receiving data for aggregation. The received data originates from one or more computing devices and flows through one or more edge devices. The received data comprises associated metadata. The embodiment may include identifying a topology for the received data based on the associated metadata. The embodiment may include determining whether the received data can be aggregated at an earlier edge device. The embodiment may include determining an aggregation location for subsequent data. The subsequent data is affiliated with the received data. The embodiment may include determining a network route for utilization by the subsequent data. The embodiment may include propagating the determined aggregation location and network route. The embodiment may include transmitting the received data to the determined aggregation location.

Abstract: A computer-implemented method controls communication channels in a communication network used for an electronic meeting. One or more processors collect a set of participant data related to a set of participants of an electronic meeting. One or more processors ascertain one or more sets of predicted sub-topics of the electronic meeting and then determine, based on the set of participant data and the one or more sets of predicted sub-topics, a set of leader readiness factors that identifies participants that are best qualified to lead the electronic meeting when particular sub-topics arise during the electronic meeting. In response to detecting that the particular sub-topic has arisen during the electronic meeting, a hardware communication controller opens a communication channel on the communication network that is dedicated for use by a best qualified participant to transmit information to other participants of the electronic meeting.

Abstract: According to embodiments of the present invention, techniques are provided for producing energy from one or more energy production sources. Energy consumption and energy production sources for an entity in a geographic location are analyzed. Sensor data based on the analysis of the entity is collected, wherein the sensor data includes crowdsourced sensor data. A model is generated based on the collected sensor data and the analysis of the entity, wherein the model includes optimization criteria for energy production. The model is utilized to determine an energy production source and energy production for the entity. Occurrence of a triggering event is detected, and the energy production source is adjusted based on the detected triggering event. The model is updated based on newly collected sensor data.

Abstract: A method for validating an engagement activity of a contract-bound party with respect to a contractual term is provided. The method may include generating first and second latent class models, including first and second sets of behavioral classes, respectively. The first model is generated based on a representative population of individuals, respective contractual activities of the individuals, and the contractual term. The second model is generated based on an actual population of individuals, respective contractual activities of the individuals, and the contractual term. The party is classified into a class of each set of classes, respectively, based on detection of the engagement activity. An inequivalency score is determined for the models based on the classifications of the party into the classes of each model, and responsive to determining that the score exceeds a predetermined threshold, a breach of the contractual term is detected, and a remedial action is performed.

Abstract: Modification of a user skill profile based upon social presence includes receiving a document indicating skill competencies of a user, and analyzing the document to determine a first model. The first model includes skill competencies indicated within the document. A social media profile associated with the user is received. The social media profile is analyzed to determine a second model. The second model includes current skills competencies of the user indicated by the social media profile. The first model and the second model is compared to determine missing skill competencies, the missing skill competencies including current skill competencies in the second model that are missing from the document. The document is modified to include the missing skill competencies.

Abstract: A computer determines a cognitive overload of a user interacting with a visual display based on eye tracking data. The visual display includes content of the dashboard and graphical elements of the content of the dashboard. The computer adjusts the visual display by modifying the content based on determining the cognitive overload.

Abstract: A method, a device and a computer program product for processing a segment are proposed. In the method, a property of at least one of a first segment and a second segment in a segment set is obtained. The segment set includes a plurality of segments belonging to at least one conversation. The second segment occurs after the first segment. A boundary feature of at least one of the first segment and the second segment is determined based on the property. The boundary feature indicates whether there is a boundary of a conversation after the first segment.

Abstract: Methods, systems, and computer readable media are provided for identifying photos such as selfies in which the individuals in the photo are in imminent danger. One or more profiles are generated, each profile associated with a corresponding first user of a networked system and pertaining to problematic conditions for image capture. An image of a second user is captured via an image capture device. The captured image of the second user is uploaded to the networked system. The captured image of the second user is compared to the generated one or more profiles to determine a score value for the captured image of the second user. A score value may be generated based upon a combination of scores from one or more of geolocation information, metadata analysis, and image processing and machine learning. The upload of the captured image of the second user is rejected in response to the score value satisfying a threshold indicating a presence of problematic conditions during capturing of the image of the second user.

Abstract: Audio waveform data can be received from a plurality of client devices. The audio waveform data for each client device can be generated by sampling at least one portion of an audio stream received by the respective client device. The audio waveform data received from the plurality of client devices can be compared. Based on the comparison, which of the plurality of client devices are located in a same location can be determined. Further, based on the comparison, which of the plurality of client devices located in the same location are to mute their respective input audio transducers can be determined, and an indicator indicating to the client devices to mute their respective input audio transducers can be communicated to the client devices that are determined to mute their respective input audio transducers.