International Business Machines Corporation

Ruhi Sharma Mittal

Vijay Ekambaram

Prasenjit Dey

Ravindranath Kokku

Nitin Kumar Gupta

Nitendra Rajput

Hima Patel

Lokesh Nagalapatti

Ramasuri Narayanam

Ashish Mittal

Sambaran Bandyopadhyay

Shashank Mujumdar

Yedendra Shrinivasan

Dipyaman Banerjee

Sonia Soubam

Yedendra B Shrinivasan

A method, computer program, and computer system are provided for collecting and annotating data based on user preference. Unlabeled data corresponding to one or more entries within a dataset is received. Pseudo-labeled data is generated based on the unlabeled data. Based on one or more quality metrics, each entry from among the pseudo-labeled data is determining to be included within a final dataset. A user is prompted for annotations corresponding to entries of the pseudo-labeled data included within the final dataset. A determination is made as to whether additional data is needed based on comparing the final dataset to the one or more quality metrics, and the additional information is collected if the final dataset does not meet the quality metrics.

Annotating and collecting data-centric AI quality metrics considering user preferences

Methods, systems, and computer program products for automatically detecting outliers in federated data are provided herein. A computer-implemented method includes obtaining local outlier-related data from multiple client systems within a federated learning environment; detecting one or more federated learning environment-level outliers from at least a portion of the multiple client systems by processing at least a portion of the obtained local outlier-related data using one or more artificial intelligence models; determining at least one calibration parameter for detecting federated learning environment-level outliers based at least in part on the one or more detected federated learning environment-level outliers; and outputting the at least one determined calibration parameter to at least a portion of the multiple client systems within the federated learning environment.

Automatically detecting outliers in federated data

Automatically selecting data for machine learning datasets is provided. The method comprises receiving an input dataset and user-specified data quality metrics. The input dataset is matched to a subset of candidate datasets in a repository according to schema characteristics. A second subset of candidate datasets having a distance from the input dataset above a specified threshold is selected from the first subset of candidate datasets. The second subset of candidate datasets are merged into a merged dataset. Top ranked samples above a specified second threshold are identified from the merged dataset based on the user-specified data quality metrics. The input dataset, augmented with the top ranked samples, is returned to the user.

Automatically selecting relevant data based on user specified data and machine learning characteristics for data integration

Techniques for generating machine learning training data which corresponds to one or more downstream tasks are disclosed. In one example, a computer implemented method comprises generating one or more synthetic data instances for training a machine learning model, and determining a value of respective ones of the one or more synthetic data instances with respect to at least one task. One or more additional synthetic data instances for training the machine learning model are generated based at least in part on the values of the respective ones of the one or more synthetic data instances.

Generating task-specific training data

One embodiment provides a method, including: receiving a sample set for training a machine-learning model, wherein the sample set includes a plurality of classes, wherein classes within the plurality of classes have an imbalance in a number of samples; creating an enlarged minority class by generating new samples from the samples within the minority class and adding the new samples to the minority class; selecting subset samples from both the samples within the enlarged minority class and the majority class; weighting each of the subset samples based upon user input defining goals for attributes of a training sample set to be used in training the machine-learning model; and generating, using the neural network, the training sample set by re-running the selecting in view of the weighting.

Training sample set generation from imbalanced data in view of user goals

An exemplary method includes obtaining a group of classification labels and corresponding confidence values for at least one object identified within an image using a computer-based object recognition technique; generating a conversation, to resolve ambiguity among the classification labels, the generating including iteratively performing the following when (i) each of the confidence values is below a threshold value or (ii) two or more of the confidence values are above the threshold value: using a wordweb to identify properties that distinguish between a first one and a second one of the labels, ranking the properties; selecting the property having the highest rank to generate a question, and filtering at least one of the first and second label based on user input received in response to the question; and when only one of the confidence values exceeds the threshold value, classifying the object using the label corresponding to the one confidence value.

Image object recognition through multimodal conversation templates

Methods, systems, and computer program products for image object disambiguation resolution are provided herein. An example of a method includes: obtaining a group of classification labels and corresponding confidence values for an object in an image; using a wordweb to determine one or more properties that distinguish between at least a first one of the classification labels and at least a second one of the classification labels within the group; selecting a first property from the properties to generate a question based on information indicating a level of prior knowledge of the user with each of the properties and each of the one or more labels; assigning a belief score to an answer; and determining whether to present at least a second question to verify the first answer based on a comparison of the belief score to a belief threshold value.

Image object disambiguation resolution using learner model based conversation templates

A computer-implemented wardrobe packing method, system, and a computer program product, include selecting a plurality of items stored in a memory to pack into a container, generating a packing order of the plurality of items based on packing preference data and item data stored in the memory, creating and displaying a video of step-by-step packing instructions for the plurality of items in the packing order, monitoring and comparing an actual packing process to a generated packing order, and controlling the displaying of a step of the step-by-step packing instructions, in response to the monitoring and comparing.

Providing notifications in video streams to control video playback

Methods, systems, and computer program products for image object disambiguation resolution are provided herein. An example of a method includes: initiating a conversation for resolving ambiguity among a group of labels corresponding to an object in an image; using a wordweb to determine properties that distinguish between at least one first label and at least one second label within the group; selecting a first property from the properties to generate a question, wherein said selecting is based at least in part on a learner model of a user including information indicating a level of prior knowledge of the user with each of the properties and each of the one or more labels; assigning a belief score to an answer; and determining whether to present at least a second question to verify the first answer based on a comparison of the belief score to a threshold value.

A system for performing implicit assessment of a pupil is provided. The system receives a learning content from a content provider, the learning content having checkpoints at different stages of the learning content. The system associates assessment instructions with each checkpoint of the learning content. The system renders the received learning content for presentation to a pupil and captures the behaviors of the pupil during the presentation of the learning content. The system performs an assessment of the pupil at each checkpoint encountered during the rendering of the learning presentation. The assessment at a checkpoint includes using the checkpoint's associated assessment instructions to assign a score based on the captured behaviors. The system then produces an overall evaluation based on the assessments performed at the encountered checkpoints.

Computationally derived assessment in childhood education systems

An exemplary method includes initiating a conversation with a user for resolving ambiguity among a group of labels corresponding to an object in an image, wherein each label is associated with a confidence value; iteratively performing, when each of the confidence values is below a first threshold value, or two or more of the confidence values are above the first threshold value: using a wordweb to identify properties that distinguish between a first one of the labels and a second one of the labels, ranking the identified one or more properties; selecting the property having the highest rank to generate a question, and filtering at least one of the first label and the second label based on user input received in response to the question; and when only one of the confidence values exceeds the first threshold value, classifying the object using the label corresponding to the one confidence value.

One embodiment provides a method, including: utilizing at least one processor to execute computer code that performs the steps of: receiving a plurality of student profiles, wherein each of the plurality of student profiles comprises a homework style comprising a pattern of activities performed by a reference student when the reference student is writing by hand, doing homework related to an identified topic; obtaining sensor information, from at least one sensor of a wearable device, the sensor information corresponding to activities performed by a student wearing the wearable device; identifying, using the sensor information, a series of activities performed by the student, wherein the identifying comprises comparing the sensor information to a plurality of models, wherein each model corresponds to an activity; determining if the series of activities corresponds to a homework style of a reference student having a profile similar to that of the student; and providing, if the series of activities do not correspond to the homework style of the reference student having a similar profile, a notification of a deviation from the homework style to a user. Other aspects are described and claimed.

Handwritten homework monitoring using wearable device

A computer-implemented wardrobe packing method, system, and a computer program product, include selecting a plurality of items stored in a memory to pack into a container, generating a packing order of the plurality of items based on packing preference data and item data stored in the memory, creating and displaying a video of step-by-step packing instructions for the plurality of items in the packing order, monitoring and comparing an actual packing process to a generated packing order, and controlling said displaying of step of the step-by-step packing instructions, in response to said monitoring and comparing.

System, method and computer program product for packing

Data representing an image is received by an image recognition system. An image recognition system generates an image classification output distribution for a plurality of image features based on analysis of the data representing the image and training data stored for the image recognition system. One or more filters are applied to the image classification output distribution to obtain an updated image classification output distribution. A highest confidence value is selected from the updated image classification output distribution. A selected image feature associated with the highest confidence value is identified from the plurality of image features. Information associated with the selected image feature is obtained from a database and communicated to the user's device by the image recognition system.

Growing community of inventors

Bengaluru, India

Ruhi Sharma Mittal