Development of Voice and Other Interaction Applications

Abstract

Among other things, a developer of an interaction application for an enterprise can create items of content to be provided to an assistant platform for use in responses to requests of end-users. The developer can deploy the interaction application using defined items of content and an available general interaction model including intents and sample utterances having slots. The developer can deploy the interaction application without requiring the developer to formulate any of the intents, sample utterances, or slots of the general interaction model.

1 . A machine-implemented method comprising:
 providing a user interface comprising a text entry area and an invocable element that, when, invoked, causes an open-ended slot to be added to a phrase in the text entry area,   the open-ended slot configured to be filled by clauses that include actions to be interpreted by natural language processing, such that, when the open-ended slot is filled with two or more different possible specific utterances of end-users, the phrase including the filled open-ended slot corresponds to two or more different respective intents;   in response to a user interaction in the user interface, adding code representing the phrase and the open-ended slot to an interaction model as a new sample utterance; and   incorporating the interaction model into an interaction application, wherein the interaction application is configured to use the interaction model to map a received utterance to the new sample utterance, identify a portion of the received utterance that corresponds to the open-ended slot, and parse the portion of the received utterance using natural language processing, to determine an intent corresponding to the received utterance.   
     
     
         2 . The method of  claim 1  in which the invocable element comprises a clickable graphic icon. 
     
     
         3 . The method of  claim 1  comprising providing a second user interface comprising a list of phrase variations corresponding to a second intent. 
     
     
         4 . The method of  claim 3  in which the second user interface comprises a second invocable element that, when invoked, causes a new phrase variation corresponding to the intent to be added to the interaction model. 
     
     
         5 . The method of  claim 1  in which the new sample utterance comprises an abstract characterization of requests, such that the received utterance cannot be mapped directly to specific content without parsing and interpretation of the portion of the received utterance filling the open-ended slot. 
     
     
         6 . The method of  claim 1  comprising providing a second user interface comprising a second text entry area configured to receive spoken response phrases to be spoken by voice assistant devices. 
     
     
         7 . The method of  claim 6  in which the second user interface comprises one or more second invocable elements that, when invoked, initiate a process in which respective audio effects are added to a highlighted portion of a spoken phrase in the second text entry area. 
     
     
         8 . The method of  claim 6  in which the second user interface comprises a second invocable element that, when invoked, causes display of a code representation of a spoken phrase in the second text entry area, wherein the code representation comprises code indicating one or more audio effects added to the spoken phrase. 
     
     
         9 . The method of  claim 8  in which the code representation comprises speech synthesis markup language code. 
     
     
         10 . The method of  claim 1 , comprising storing an association between a speech response to a request from an end-user and a specific prosody value. 
     
     
         11 . A machine-implemented method comprising
 receiving a first utterance comprising a first phrase and a first slot expression,   receiving a second utterance comprising the first phrase and a second slot expression that is different from the first slot expression, the first utterance and the second utterance having been derived by an assistant platform from requests of end-users of interaction assistants, and   applying the first utterance and the second utterance to a sample utterance in an interaction model to determine a first intent corresponding to the first utterance and a second intent corresponding to the second utterance, wherein the second intent is different from the first intent   the interaction model comprising, in code representing the interaction model, the sample utterance, wherein the sample utterance comprises the first phrase and an open-ended slot to which the first slot expression and the second slot expression are mapped,   wherein applying the first utterance and the second utterance to the sample utterance in the interaction model comprises parsing the first slot expression and the second slot expression using natural language processing to determine the first intent and the second intent.   
     
     
         12 . The method of  claim 11  in which parsing the first slot expression and the second slot expression using natural language processing is secondary natural language processing, the secondary natural language processing distinct from first natural language processing performed on the first phrase and the first slot expression together and on the first phrase and the second slot expression together by the assistant platform. 
     
     
         13 . The method of  claim 11  in which the natural language processing comprises at least one of key word extraction or sentiment analysis on the first slot expression and the second slot expression. 
     
     
         14 . The method of  claim 11  in which at least one of the first slot expression or the second slot expression comprises a verb. 
     
     
         15 . A machine-implemented method comprising
 accessing an interaction model;   determining that a sample utterance does not correspond to an existing sample utterance pattern of the interaction model;   based on determining that the sample utterance does not correspond to an existing sample utterance pattern, generating a new sample utterance pattern and providing a recommendation to add the new sample utterance pattern to the interaction model; and   adding the new sample utterance pattern to the interaction model,   wherein the new sample utterance pattern comprises a phrase and an open-ended slot,   wherein the new sample utterance pattern encompasses two or more different possible specific utterances, and   wherein the sample utterance corresponds to the new sample utterance pattern having a specific clause inserted in the open-ended slot.   
     
     
         16 . The method of  claim 15  in which providing the recommendation comprises:
 breaking down the sample utterance into individual words; and 
 generating the new sample utterance pattern by adjoining the open-ended slot to a set of one or more initial words of the sample utterance or to a set of one or more final words of the sample utterance. 
 
     
     
         17 . The method of  claim 16  in which generating the new sample utterance pattern comprises selecting the set of one or more initial words or the set of one or more final words based on a comparison between a first potential sample utterance pattern based on the set of one or more initial words and a second potential sample utterance pattern based on the set of one or more final words. 
     
     
         18 . The method of  claim 16  comprising selecting the set of one or more initial words or the set of one or more final words based on parts of speech of the set of one or more initial words or the set of one or more final words. 
     
     
         19 . The method of  claim 15  comprising providing the recommendation using a machine learning model trained to recognize correlations between sample utterance patterns and vertical industries. 
     
     
         20 . The method of  claim 15 , wherein the phrase includes a portion of the sample utterance.