Abstract: An approach is provided that receives a document and a document type of the document. The document type identifies a document category to which the received document belongs. A set of linguistic metrics are retrieved that correspond to the document type. A quality of the received document is automatically determined based on a set of linguistic features found in the document as compared to the retrieved set of linguistic metrics. The document is then ingested into a corpus that is utilized by a question-answering (QA) system. The ingestion of the document is based on the determined quality.

Abstract: Systems and methods for generating a natural language domain corpus to train a machine learning natural language understanding process. A base utterance expressing an intent and an intent profile indicating at least one of categories, keywords, concepts, sentiment, entities, or emotion of the intent are received. Machine translation translates the base utterance into a plurality of foreign language utterances and back into respective utterances in the target natural language to create a normalized utterance set. Analysis of each utterance in the normalized utterance set determines respective meta information for each such utterance. Comparison of the meta information to the intent profile determines a highest ranking matching utterance within the normalized utterance set. A set of natural language data to train a machine learning natural language understating process is created based on further natural language translations of the highest ranking matching utterance.

Abstract: In some examples, a document can be received and parsed to identify sentences of the document. A plurality of domain natural language classifiers (NLCs) trained based on domain training data associated with a respective domain of a plurality of domains can be programmed to classify each identified sentence to determine a sentence confidence score for each identified sentence. A plurality of document confidence scores for the document can be determined based on sentence confidence scores determined by the plurality of domain NLCs. Each document confidence score can characterize a relevance of the document to the respective domain of the plurality of domains. Document similarity data identifying at least one document among documents associated with a corresponding domain of the plurality of domains can be generated based on an evaluation of the document confidence scores for the given document and document confidence scores for the documents associated with the corresponding domain.

Abstract: Systems and methods for generating a natural language domain corpus to train a machine learning natural language understanding process. A base utterance expressing an intent and an intent profile indicating at least one of categories, keywords, concepts, sentiment, entities, or emotion of the intent are received. Machine translation translates the base utterance into a plurality of foreign language utterances and back into respective utterances in the target natural language to create a normalized utterance set. Analysis of each utterance in the normalized utterance set determines respective meta information for each such utterance. Comparison of the meta information to the intent profile determines a highest ranking matching utterance within the normalized utterance set. A set of natural language data to train a machine learning natural language understating process is created based on further natural language translations of the highest ranking matching utterance.

Abstract: In some examples, a document can be received and parsed to identify sentences of the document. A plurality of domain natural language classifiers (NLCs) trained based on domain training data associated with a respective domain of a plurality of domains can be programmed to classify each identified sentence to determine a sentence confidence score for each identified sentence. A plurality of document confidence scores for the document can be determined based on sentence confidence scores determined by the plurality of domain NLCs. Each document confidence score can characterize a relevance of the document to the respective domain of the plurality of domains. Document similarity data identifying at least one document among documents associated with a corresponding domain of the plurality of domains can be generated based on an evaluation of the document confidence scores for the given document and document confidence scores for the documents associated with the corresponding domain.

Abstract: An approach is provided that receives a document and a document type of the document. The document type identifies a document category to which the received document belongs. A set of linguistic metrics are retrieved that correspond to the document type. A quality of the received document is automatically determined based on a set of linguistic features found in the document as compared to the retrieved set of linguistic metrics. The document is then ingested into a corpus that is utilized by a question-answering (QA) system. The ingestion of the document is based on the determined quality.

Abstract: A method of voice-enabling an application for command and control and content navigation can include the application dynamically generating a markup language fragment specifying a command and control and content navigation grammar for the application, instantiating an interpreter from a voice library, and providing the markup language fragment to the interpreter. The method also can include the interpreter processing a speech input using the command and control and content navigation grammar specified by the markup language fragment and providing an event to the application indicating an instruction representative of the speech input.

Abstract: The present invention can include a speech enrollment system including an ordered stack of grammars and a recognition engine. The ordered stack of grammars can include an application grammars layer, a confusable grammar layer, a personal grammar layer, a phrase enrolled grammar layer, and an enrollment grammar layer. The recognition engine can return recognition results for speech input by processing the input using the ordered stack of grammars. The processing can occur from the topmost layer in the stack to the bottommost layer in the stack. Each layer in the stack can includes exit criteria based upon a defined condition. When the exit criteria is satisfied, a result can be returned based upon that layer and lower layers of the ordered stack can be ignored.

Abstract: A method for testing an automated interactive media system. The method can include establishing a communication session with the automated interactive media system. In response to receiving control and/or media information from the automated interactive media system, pre-recorded control and/or media information can be propagated to the automated interactive media system. The pre-recorded control and/or media information can be recorded in real time.

Abstract: A method for extending supported voice markup. The method can include a step of identifying a reference implementation (RI) for a software component that interprets voice-based markup. The RI can define a manner that the software component interprets voice-based markup. At least one plug-in can be identified that contains an extension to the RI. At runtime, the RI can be dynamically modified in accordance with the at least one plug-in. The software component can interpret voice-based markup documents based upon the modified reference implementation.

Abstract: A method and system for testing voice applications, such as VoiceXML applications, is provided. The system provides a run-time simulation environment for voice applications that simulates and automates user interaction. A user simulation script is provided in a customized mark-up language. The voice application is processed to derive a nominal output of the voice application. The user simulation script is processed to generate a simulated output for the voice application corresponding to the nominal output. The user simulation script is also processed to generate a simulated input for the voice application corresponding to a pre-determined user input to the voice application. The simulated inputs and outputs are executed in conjunction with the voice application in real-time to test the application.

Abstract: The present invention can include a speech processing method for providing dictation capabilities to a voice server. The method can include a step of establishing a real-time voice communication session involving a voice interface. Speech for the communication session can be streamed to a remotely located voice server. A real-time stream of text can be received from the voice server. The stream of text can include text that has been speech-to-text converted by the voice server from the streamed speech. The voice server can use a MRCP based non-halting interface to receive the real-time stream of speech and a delivery interface to deliver real-time text to a designated endpoint.

Abstract: A method for testing an automated interactive media system. The method can include establishing a communication session with the automated interactive media system. In response to receiving control and/or media information from the automated interactive media system, pre-recorded control and/or media information can be propagated to the automated interactive media system. The pre-recorded control and/or media information can be recorded in real time.

Abstract: A method for testing an automated interactive media system. The method can include establishing a communication session with the automated interactive media system. In response to receiving control and/or media information from the automated interactive media system, pre-recorded control and/or media information can be propagated to the automated interactive media system. The pre-recorded control and/or media information can be recorded in real time.

Abstract: The present invention discloses a system and a method for authenticating a user based upon a spoken password processed though a standard speech recognition engine lacking specialized speaker identification and verification (SIV) capabilities. It should be noted that the standard speech recognition grammar can be capable of acoustically generating speech recognition grammars in accordance with the cross referenced application indicated herein. The invention can prompt a user for a free-form password and can receive a user utterance in response. The utterance can be processed through a speech recognition engine (e.g., during a grammar enrollment operation) to generate an acoustic baseform. Future user utterances can be matched against the acoustic baseform. Results from the future matches can be used to determine whether to grant the user access to a secure resource.

Abstract: A method (200) of building a grammar with baseforms generated dynamically from user utterances can include the steps of recording (205) a user utterance, generating (210) a baseform using the user utterance, creating or adding to (215) a grammar rule using the baseform, and binding (230) the grammar rule in a grammar document of a voice extensible markup language program. Generating a baseform can optionally include introducing a new element to VoiceXML with attributes that enable generating the baseform from a referenced recording such as the user utterance. In one embodiment, the method can be used to create (235) a phonebook and a grammar to access the phonebook by repeatedly visiting a form containing the grammar rule with attributes that enable generating the baseform from the referenced recording.

Abstract: A method (200) and a system (100) for coordinated streaming use a single Real Time Protocol (RTP) producer (130) for handling multiple audio services (110). The method can include the steps of assigning (202) a RTP producer to handle multiple audio objects, and maintaining (204) a service for each object in accordance with a delivery schedule. RTP packets can be sent in accordance with the delivery schedule for complying with real-time requirements of a media rendering client thereby providing continuous real-time service delivery. The method can further include determining a wait time and updating the delivery schedule in view of the wait time. In one arrangement, the RTP producer can sleep for a pre-specified interval, and upon wake, prioritizes service delivery based on an audio object's wait time.

Abstract: The present invention can include a method of allocating an interpreter module within an application program. The application program can create one or more interpreter module instances. The method also can include updating a property descriptor of the interpreter module instance and directing the interpreter module instance to allocate speech and audio resources. Content then can be loaded into the interpreter module instance and run.

Abstract: The present invention provides a solution that unifies independent data sources into a single data source by hiding one of the data sources in metadata. The non-hidden source, which can reference the hidden source, can be processed by a primary XML processing agent. This agent can initially ignore syntax and semantics of content associated with the hidden source, which is contained within the metadata. A secondary XML processing agent can transform the hidden source into a readable form, which is consumable by the primary processing agent. The primary XML processing agent can receive this readable form as a content stream. The primary XML processing agent can be unaware that the content stream originates from the unified source, which permits the unified source to represent multiple sources while preserving “rules” that typically prohibit multiple data types to be included in a single source.

Abstract: A method and system for testing voice applications, such as VoiceXML applications, is provided. The system provides a run-time simulation environment for voice applications that simulates and automates user interaction. A user simulation script is provided in a customized mark-up language. The voice application is processed to derive a nominal output of the voice application. The user simulation script is processed to generate a simulated output for the voice application corresponding to the nominal output. Conditional logic may be applied to the nominal output to generate a simulated input in response thereto. The user simulation script is specified in a customized mark-up language having a set of one or more conditional tags and an internal variable for the nominal output of the voice application.