Abstract: A natural language processing system that trains task models for particular natural language tasks programmatically generates additional utterances for inclusion in the training set, based on the existing utterances in the training set and the existing state of a task model as generated from the original (non-augmented) training set. More specifically, the training augmentation module 220 identifies specific textual units of utterances and generates variants of the utterances based on those identified units. The identification is based on determined importances of the textual units to the output of the task model, as well as on task rules that correspond to the natural language task for which the task model is being generated. The generation of the additional utterances improves the quality of the task model without the expense of manual labeling of utterances for training set inclusion.

Abstract: A natural language processing system that trains task models for particular natural language tasks programmatically generates additional utterances for inclusion in the training set, based on the existing utterances in the training set and the existing state of a task model as generated from the original (non-augmented) training set. More specifically, the training augmentation module 220 identifies specific textual units of utterances and generates variants of the utterances based on those identified units. The identification is based on determined importances of the textual units to the output of the task model, as well as on task rules that correspond to the natural language task for which the task model is being generated. The generation of the additional utterances improves the quality of the task model without the expense of manual labeling of utterances for training set inclusion.

Abstract: A computer-implemented method for automated document review and quality control may include (1) dividing a set of documents to be reviewed for relevancy into sets of control documents, training documents, quality-control documents, and review documents, (2) calculating, based on a set of reviews performed by a group of reviewers on the set of quality-control documents, an effective speed score for each reviewer in the group of reviewers, (3) assigning, based on the effective speed score, the set of control documents and the set of training documents to the group of reviewers, (4) training a predictive model using a set of training reviews performed by the group of reviewers on both the set of training documents and the set of control documents, and (5) using the predictive model to evaluate the set of review documents. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: A system and method for providing access to a unified message store logically storing computer telephony messages is disclosed. A multiplicity of heterogeneous storage objects corresponding to computer telephony messages are stored in a unified message store. An application software layer exchanging the heterogeneous storage objects with a computer telephony server is exported. The computer telephony server is interfaced via a container subsystem defining encapsulated methods including organization strategy methods and storage strategy methods. A unified inbox generating an indicator for at least one of the computer telephony messages is provided. The unified inbox exports user controls for accessing the corresponding heterogeneous storage objects stored in the unified message store.