Capturing and Leveraging Signals Reflecting BOT-to-BOT Delegation

Abstract

A technique is described herein for capturing signals that indicate when any calling BOT delegates control to a called BOT, or when a calling BOT is preconfigured to contact a called BOT (e.g., as conveyed by a manifest file associated with the calling BOT). The technique can leverage these signals to facilitate the selection of BOTs. For example, the technique can use the signals to improve searches performed by a search engine and/or recommendation engine. The technique can also use the signals to generate metadata items that describe the properties of the available BOTs.

1 - 20 . (canceled) 
     
     
         21 . A computer-implemented method comprising:
 accessing a signal data store having BOT delegation signals corresponding to BOT delegation instances where calling conversational BOTs, when performing tasks having subtasks, delegated control to called conversational BOTs to perform the subtasks during previous natural language conversations with users;   during a current natural language conversation with a user, receiving current user input expressing a particular intent relating to a particular subtask;   choosing a selected conversational BOT that matches the particular intent expressed by the current user input based at least on a particular BOT delegation feature, derived from the signal data store, indicating that control was previously delegated to the selected conversational BOT to perform the particular subtask as part of a particular task performed by another conversational BOT; and   outputting an identification of the selected conversational BOT in response to the current user input.   
     
     
         22 . The computer-implemented method of  claim 21 , the called conversational BOTs being requested by users during the previous natural language conversations. 
     
     
         23 . The computer-implemented method of  claim 21 , the called conversational BOTs being selected by the calling conversational BOTs during the previous natural language conversations. 
     
     
         24 . The computer-implemented method of  claim 21 , further comprising:
 invoking the selected conversational BOT in response to an activation event.   
     
     
         25 . The computer-implemented method of  claim 21 , each BOT delegation signal including at least:
 a first data item that identifies a specific calling conversational BOT;   a second data item that identifies a specific called conversational BOT; and   a third data item that identifies a specific intent that the specific called conversational BOT was called on to fulfill by the specific calling conversational BOT during a course of a specific previous natural language conversation involving the specific calling conversational BOT, the specific called conversational BOT, and a specific human user.   
     
     
         26 . The computer-implemented method of  claim 21 , further comprising:
 mapping the called conversational BOTs into respective BOT vectors in a semantic space;   mapping the current user input into a current user input vector in the semantic space; and   choosing the selected conversational BOT based at least on similarity of the current user input vector to a selected BOT vector for the selected conversational BOT.   
     
     
         27 . The computer-implemented method of  claim 26 , further comprising:
 determining the respective BOT vectors based at least on a plurality of features relating to the previous natural language conversations.   
     
     
         28 . The computer-implemented method of  claim 27 , the plurality of features employed to determine the respective BOT vectors including BOT delegation features indicating when control was delegated to the called conversational BOTs to perform the subtasks as part of the tasks being performed by the calling conversational BOTs. 
     
     
         29 . The computer-implemented method of  claim 28 , further comprising weighting the BOT delegation instances based on relative age by discounting older BOT delegation instances relative to newer BOT delegation instances. 
     
     
         30 . The computer-implemented method of  claim 28 , the plurality of features employed to determine the respective BOT vectors including at least one user-specific feature associated with prior interactions of a current user with the called conversational BOTS. 
     
     
         31 . A computing device comprising:
 a processing device; and   a storage device having logic which, when executed by the processing device, causes the computing device to:   access a signal data store having BOT delegation signals identifying BOT delegation instances where calling conversational BOTs, when performing tasks having subtasks, delegated control to called conversational BOTs to perform the subtasks during previous natural language conversations with users;   receive, over a network, current user input expressing a particular intent relating to a particular subtask;   choose a selected conversational BOT that matches the particular intent expressed by the current user input based at least on a particular BOT delegation feature, derived from the signal data store, indicating that control was previously delegated to the selected conversational BOT to perform the particular subtask as part of a particular task performed by another conversational BOT; and   send an identification of the selected conversational BOT over the network in response to the current user input.   
     
     
         32 . The computing device of  claim 31 , wherein the current user input identifies a reference conversational BOT, and the particular BOT delegation feature indicates that the reference conversational BOT identified by the current user input previously delegated control to the selected conversational BOT. 
     
     
         33 . The computing device of  claim 31 , wherein the current user input comprises a query describing the particular subtask. 
     
     
         34 . The computing device of  claim 31 , wherein the logic, when executed by the processing device, causes the processing device to populate the signal data store by:
 access one or more natural language message streams associated with the calling conversational BOTs and detect references to the called conversational BOTs in the one or more natural language message streams.   
     
     
         35 . The computing device of  claim 31 , wherein the logic, when executed by the processing device, causes the processing device to populate the signal data store by:
 access calls or commands exchanged directly between the calling conversational BOTs and the called conversational BOTs.   
     
     
         36 . The computing device of  claim 31 , wherein the logic, when executed by the processing device, causes the processing device to:
 invoke the selected conversational BOT to perform the particular subtask.   
     
     
         37 . A computer-readable storage medium storing computer-readable instructions which, when executed by hardware processor, cause the hardware processor to perform acts comprising:
 accessing a signal data store having BOT delegation signals associated with BOT delegation instances where calling conversational BOTs, when performing tasks having subtasks, delegated control to called conversational BOTs to perform the subtasks during previous natural language conversations with users;   receiving current user input expressing a particular intent relating to a particular subtask;   choosing a selected conversational BOT that matches the particular intent expressed by the current user input based at least on a particular BOT delegation feature, wherein the particular BOT delegation feature is based on a particular BOT delegation signal indicating that control was previously delegated to the selected conversational BOT to perform the particular subtask as part of a particular task performed by another conversational BOT; and   outputting an identification of the selected conversational BOT in response to the current user input.   
     
     
         38 . The computer-readable storage medium of  claim 37 , the acts further comprising:
 choosing multiple selected conversational BOTs that match the particular intent according to a diversity criterion.   
     
     
         39 . The computer-readable storage medium of  claim 38 , wherein the diversity criterion ensures that vectors representing the multiple selected conversational BOTs are mutually separated from one another by at least a prescribed distance. 
     
     
         40 . The computer-readable storage medium of  claim 39 , the acts further comprising:
 determining the vectors based at least on BOT delegation features indicating when control was delegated to the called conversational BOTs to perform the subtasks as part of tasks being performed by the calling conversational BOTs.