Abstract: Data-parallel ensemble training using gradient boosted trees includes training an ensemble of trees. The training includes splitting a training dataset into several data portions. Each data portion is assigned to each thread group from a set of thread groups. The training further includes executing a stage, in which each thread group, in parallel, trains a respective ensemble of decision trees. Executing the stage includes performing, by each thread group, in parallel, machine learning operations for the respective ensemble of decision trees using the data portion assigned to each thread group. Further, each thread group validates, in parallel, the respective ensemble of decision trees using a data portion assigned to another thread group. Execution of the stage is repeated until a predetermined threshold is satisfied. Further, a prediction is inferenced using the ensemble of decision trees that is formed using the respective ensemble of trees from each of the thread groups.

Abstract: Methods and systems for managing an error correction mode at a first communications router. The first communication router transmits data packets to a second communications router and stores the first data packet in a local storage medium. When a delay inquiry message is received from the second communications router, the first communications router activates the error correction mode. When the error correction mode is activated, the first data packet is retransmitted to the second communications router and an error correction packet corresponding to the first data packet is also transmitted. When a back-to-normal message is received from the second communications router, the first communications router deactivates the error correction mode. The back-to-normal message indicates that the first communications router no longer needs to be in error correction mode.

Abstract: The present disclosure provides systems and methods which increase the throughput of a TCP-based communication between a first network node and a second network node. First, the first network node sent a first plurality of TCP segments to the second network node. Second, when the second network node receives a second plurality of TCP segments, which is all or part of the first plurality of the TCP segments, the second network node responds by sending one or more TCP acknowledgements to the first network node with the last sequence number of a last segment among all TCP segment within the second plurality of TCP segments. The present disclosure are able to increase the throughput of a TCP connection while decreasing its reliability.

Abstract: Embodiments for creating compact example subsets for intent classification in a conversational system are provided. A set of content used for training an intent classifier is received from a conversational corpus. Entries within the set of content are separated into a first subset and a second subset, and a cross-validation operation is performed on the first and second subsets to identify a correctly labeled portion and an incorrectly labeled portion of the set of content. A reduced content used for performing a final training of the intent classifier is formed by combining a first number of the entries from the correctly labeled portion and a second number of the entries from the incorrectly labeled portion of the set of content.

Abstract: Methods and systems of determining relevancy of electronic health records to medical analysis objectives. One system includes an electronic processor configured to access electronic health records and extract medical summary data items from the records. The electronic processor is also configured to determine a set of semantic vectors, where each semantic vector represents a medical summary data item. The electronic processor is also configured to determine a set of anatomical semantic vectors. The electronic processor is also configured to determine a similarity score for each medical summary data item. The electronic processor is also configured to receive a medical study and determine a relevancy score for each medical summary data item, the relevancy score representing a relevancy of each medical summary data item to the medical study. The electronic processor is also configured to generate and transmit a notification to a reviewer of the medical study.

Abstract: Aspects of the present disclosure relate to managing concurrent updates on data. A transactional update on at least one record within a data set is detected. A log is generated for the transactional update, the log including an image of the at least one record within the data set before the transactional update. A batch update is detected. A determination is made whether the batch update collides with the transactional update by referencing the log. In response to determining that the batch update collides with the transactional update, a collision policy is referenced. A collision action indicated in the collision policy is then issued.

Abstract: A combination of propagation operations and learning algorithms is applied, using a selected set of labeled conversational logs retrieved from a subset of a plurality of conversational logs, to a remaining corpus of the plurality of conversational logs to train an automated response system according to an intent associated with each of the conversational logs. The combination of propagation operations and learning algorithms may include defining the labels by a user for the selected set of the subset of the plurality of conversational logs; training a probabilistic classifier using the defined labels of features of the selected set, wherein the probabilistic classifier produces labeling decisions for the subset of conversational logs; weighting the features of the selected set in a model optimization process; and/or training an additional classifier using the weighted features of the selected set and applying the additional classifier to the remaining corpus.

Abstract: An embodiment for creating a dialog based on anticipated questions for database driven conversations is provided. The embodiment may include receiving content from a database. The embodiment may also include identifying one or more schemas, one or more entities, and relational data in the content. The embodiment may further include identifying a semantic type and a number of distinct entries for each entity. The embodiment may also include presenting choices for one or more query targets and one or more filtering conditions to a user. The embodiment may further include prompting the user for one or more annotations and one or more clarifying questions for each chosen query target and filtering condition. The embodiment may also include generating a plurality of modular phrases. The embodiment may further include combining the plurality of modular phrases into one or more sentences and paraphrases of the one or more sentences.

Abstract: A method of authoring a conversation service for a chatbot and a database includes receiving, from a user, a selection of a database, and connecting an authoring service of the chatbot to a table in the database; outputting, from the authoring service to the user, a question requesting a description of a subject matter of the table; receiving the description of the subject matter of the table; outputting to the user a question requesting an identification of a key column of the table that contains values that represent the subject matter of the table; receiving the identification of the key column of the table; and translating, by a natural language query service, the description of the subject matter of the table and the key column of the table into the conversation service, wherein the conversation service includes SQL statements suitable for querying the database table.

Abstract: A system and method for segmenting text by receiving a machine learning (ML) model for language processing, receiving segmentable text including properly joined segments, separating properly joined segments of the text into separate segments, generating positive segment pairs including properly joined segments, generating negative sentence fragment pairs including a first segment and a second segment, where the first segment and the second segment are not properly joined, and training the ML model using the positive segment pairs and the negative segment pairs, and a contrastive self-supervised learning framework training objective loss function.

Abstract: A method, apparatus, and system are provided for resolving conflicts between training data conflicts by retrieving independent training data sets, each comprising a plurality of intents and end-user utterances for use in training one or more classifiers to recognize a corresponding intent from one or more of the end-user utterances, providing a first test end-user utterance associated with a first intent from the first independent training data set to the one or more classifiers to select an output intent generated by the one or more classifiers; identifying a first conflict when the first intent does not match the output intent, and automatically generating, by the system, one or more conflict resolution recommendations for display and selection by an end user to resolve the first conflict.

Abstract: A method and apparatus are provided for automatically generating and processing first and second concept vector sets extracted, respectively, from a first set of concept sequences and from a second, temporally separated, concept sequences by performing a natural language processing (NLP) analysis of the first concept vector set and second concept vector set to detect changes in the corpus over time by identifying changes for one or more concepts included in the first and/or second set of concept sequences.

Abstract: Data-parallel ensemble training using gradient boosted trees includes training an ensemble of trees. The training includes splitting a training dataset into several data portions. Each data portion is assigned to each thread group from a set of thread groups. The training further includes executing a stage, in which each thread group, in parallel, trains a respective ensemble of decision trees. Executing the stage includes performing, by each thread group, in parallel, machine learning operations for the respective ensemble of decision trees using the data portion assigned to each thread group. Further, each thread group validates, in parallel, the respective ensemble of decision trees using a data portion assigned to another thread group. Execution of the stage is repeated until a predetermined threshold is satisfied. Further, a prediction is inferenced using the ensemble of decision trees that is formed using the respective ensemble of trees from each of the thread groups.

Abstract: A method and apparatus are provided for recommending concepts from a first concept set in response to user selection of a first concept Ci by performing a natural language processing (NLP) analysis comparison of vector representations of user concepts contained in written content authored by the user and candidate concepts in a first concept set to determine a similarity measure for each candidate concept, and to select therefrom one or more of the candidate concepts for display as recommended concepts which are related to the user concepts contained in written content authored by the user based on the similarity measure between each candidate concept and each user concept.

Abstract: A first observation window in a first time series is identified. The first observation window is preceded by a first portion of the first time series. A neural network is trained using the first portion of the first time series and the first observation window, and weights are extracted from the middle layers of the neural network. A first feature vector is generated based on the weights. A second observation window in a second time series is identified, where the second observation window is preceded by a first portion of the second time series. A second feature vector associated with the second observation window is determined. The second feature vector is based at least in part on the first set of weights. A similarity between the first and second observation windows is determined based on comparing the first feature vector and the second feature vector.

Abstract: Embodiments provide for driving intent expansion via anomaly detection by ranking, according to anomaly scores, a plurality of historic utterances that have been associated by a classifier with a given intent of a plurality of predefined intents; identifying a given utterance from the plurality of historic utterances having a given anomaly score greater than an anomaly threshold; in response to verifying that the given utterance is associated with the given intent, adding the given utterance to a training dataset as a positive example for the given intent; and in response to verifying that the given utterance is not associated with the given intent, adding the given utterance to the training dataset as a complement example for the given intent. A complement example for one intent may be added as a positive example for a different intent. The training dataset may be used to train or retrain an intent classifier.

Abstract: A computer-implemented method, cognitive intelligence system and computer program product adapt a relational database containing multiple data types. Non-text tokens in the relational database are converted to a textual form. Text is produced based on relations of tokens in the relational database. A set of word vectors is produced for the tokens based on the text. A cognitive intelligence query expressed as a structured query language (SQL) query may be applied to the relational database using the set of word vectors. The form of non-text tokens is one of a numeric value, an SQL type, an image, a video, a time series, latitude and longitude, or chemical structures. A single word embedding model may be applied over one or more tokens in the text. A plurality of sets of preliminary word vectors are computed by applying more than one embedding model over all tokens in the text. The preliminary word vector sets are merged to form the set of word vectors.

Abstract: From a first attribute-value pair in a record, new data comprising a first token is created. From each token using a processor and a memory, new data including a corresponding vector is computed. From the record, a target row is selected, wherein a target attribute-value pair in the target row includes a value requiring correction. Using a similarity measure, a set of most similar rows to the target row is determined, wherein each row in the set of most similar rows to the target row has a corresponding similarity measure above a threshold similarity measure and wherein each row in the set of most similar rows includes the target attribute. From values corresponding to the target attribute in the set of most similar rows, a replacement value is determined. The value requiring correction in the target row is replaced with the replacement value.

Abstract: A distributed computing system is configured to compute operational data for a video advertisement delivery system. Cloud-based resource are used to calculate operational parameters such as geographical data, unique advertisement delivery instances and segments of consumers that received the video advertisements.

Abstract: Structured and semi-structured databases and files are processed using natural language processing techniques to impute data for null value tokens in database records from other records that have non-null values for the same attributes. Vector embedding techniques are used, including, in some cases, appropriately tagging null value tokens to reduce or eliminate their undue impact on semantic vectors generating using a neural network.