Abstract: A target set of texts, for training and/or evaluating a text classification model, is augmented using insertions into a base text within the original target set. In an embodiment, an expanded text, including the base text and an insertion word, must satisfy one or more inclusion criteria in order to be added to the target set. The inclusion criteria may require that the expanded text constitutes a successful attack on the classification model, the expanded text has a satisfactory perplexity score, and/or the expanded text is verified as being valid. In an embodiment, if a number of expanded texts added into the target set is below a threshold number, insertions are made into an expanded text (which was generated based on the base text). Inclusion criteria are evaluated against the doubly-expanded text to determine whether to add the doubly-expanded text to the target set.

Abstract: Techniques are disclosed for augmenting data sets used for training machine learning models and for generating predictions by trained machine learning models. These techniques may increase a number (and diversity) of examples within an initial training dataset of sentences by extracting a subset of words from the existing training dataset of sentences. The extracted subset includes no stopwords and fewer content words than found in the initial training dataset. The remaining words may be re-ordered. Using the extracted and re-ordered subset of words, the dataset generation model produces a second set of sentences that are different from the first set. The second set of sentences may be used to increase a number of examples in classes with few examples.

Abstract: A computing network includes nodes of different work groups. Nodes of a work group are dedicated to transactions of the work group. If a node of a first work group is predicted to have an idleness window, a second work group may borrow the node to execute a transaction of the second work group. At least a subset of steps of the transaction may be categorized into a step group. Trees of a transaction may be categorized into one or more tree groups. A node is selected for executing a transaction, if the predicted idleness duration of the node is sufficient relative to the predicted runtime of the transaction, the step group, and/or tree group. A credit system is maintained. A first work group transfers a credit to a second work group when borrowing a node of the second work group for executing a transaction of the first work group.

Abstract: A computing network includes nodes of different work groups. Nodes of a work group are dedicated to transactions of the work group. If a node of a first work group is predicted to have an idleness window, a second work group may borrow the node to execute a transaction of the second work group. At least a subset of steps of the transaction may be categorized into a step group. Trees of a transaction may be categorized into one or more tree groups. A node is selected for executing a transaction, if the predicted idleness duration of the node is sufficient relative to the predicted runtime of the transaction, the step group, and/or tree group. A credit system is maintained. A first work group transfers a credit to a second work group when borrowing a node of the second work group for executing a transaction of the first work group.

Abstract: A computing network includes nodes of different work groups. Nodes of a work group are dedicated to transactions of the work group. If a node of a first work group is predicted to have an idleness window, a second work group may borrow the node to execute a transaction of the second work group. At least a subset of steps of the transaction may be categorized into a step group. Trees of a transaction may be categorized into one or more tree groups. A node is selected for executing a transaction, if the predicted idleness duration of the node is sufficient relative to the predicted runtime of the transaction, the step group, and/or tree group. A credit system is maintained. A first work group transfers a credit to a second work group when borrowing a node of the second work group for executing a transaction of the first work group.

Abstract: A target set of texts, for training and/or evaluating a text classification model, is augmented using insertions into a base text within the original target set. In an embodiment, an expanded text, including the base text and an insertion word, must satisfy one or more inclusion criteria in order to be added to the target set. The inclusion criteria may require that the expanded text constitutes a successful attack on the classification model, the expanded text has a satisfactory perplexity score, and/or the expanded text is verified as being valid. In an embodiment, if a number of expanded texts added into the target set is below a threshold number, insertions are made into an expanded text (which was generated based on the base text). Inclusion criteria are evaluated against the doubly-expanded text to determine whether to add the doubly-expanded text to the target set.

Abstract: Techniques are described for identifying successful adversarial attacks for a black box reading comprehension model using an extracted white box reading comprehension model. The system trains a white box reading comprehension model that behaves similar to the black box reading comprehension model using the set of queries and corresponding responses from the black box reading comprehension model as training data. The system tests adversarial attacks, involving modified informational content for execution of queries, against the trained white box reading comprehension model. Queries used for successful attacks on the white box model may be applied to the black box model itself as part of a black box improvement process.

Abstract: A system provides automatic, end-to-end tagging of email messages. While a message is being composed at a sending email client, the server may receive email information that is used as an input to a predictive model. The model identifies tags that are available to a specific user group or email list that apply to the email message. These predicted tags are sent back to the email client, where they may be embedded in the email message with other user-defined tags. As the message is passed through the email server, the system may use any changes made to the predicted tags to retrain the model. When the message is received at a second email client, the receiver may further edit the tags, and any changes may again be used to retrain the model.

Abstract: An artificial neural network (ANN) determines a conversation snippet sentiment score based on content of the conversation snippet and contextual attributes associated with the conversation snippet. Contextual attributes may include, for example, a role within an organizational hierarchy of a user participating in the conversation snippet. Information representing the content is input into a hidden layer sequence of the ANN; information representing the contextual attributes is input into another hidden layer sequence of the ANN. Additionally or alternatively, a weighing engine determines a topical sentiment score by aggregating weighted conversation snippet sentiment scores.

Abstract: A system provides automatic, end-to-end tagging of email messages. While a message is being composed at a sending email client, the server may receive email information that is used as an input to a predictive model. The model identifies tags that are available to a specific user group or email list that apply to the email message. These predicted tags are sent back to the email client, where they may be embedded in the email message with other user-defined tags. As the message is passed through the email server, the system may use any changes made to the predicted tags to retrain the model. When the message is received at a second email client, the receiver may further edit the tags, and any changes may again be used to retrain the model.

Abstract: A system provides automatic, end-to-end tagging of email messages. While a message is being composed at a sending email client, the server may receive email information that is used as an input to a predictive model. The model identifies tags that are available to a specific user group or email list that apply to the email message. These predicted tags are sent back to the email client, where they may be embedded in the email message with other user-defined tags. As the message is passed through the email server, the system may use any changes made to the predicted tags to retrain the model. When the message is received at a second email client, the receiver may further edit the tags, and any changes may again be used to retrain the model.

Abstract: An artificial neural network (ANN) determines a conversation snippet sentiment score based on content of the conversation snippet and contextual attributes associated with the conversation snippet. Contextual attributes may include, for example, a role within an organizational hierarchy of a user participating in the conversation snippet. Information representing the content is input into a hidden layer sequence of the ANN; information representing the contextual attributes is input into another hidden layer sequence of the ANN. Additionally or alternatively, a weighing engine determines a topical sentiment score by aggregating weighted conversation snippet sentiment scores.