CN111124347B

CN111124347B - Method and device for forming interaction engine cluster through aggregation

Info

Publication number: CN111124347B
Application number: CN201911220476.5A
Authority: CN
Inventors: 原利鹏; 张伟萌; 戴帅湘
Original assignee: Hangzhou Suddenly Cognitive Technology Co ltd
Current assignee: Hangzhou Suddenly Cognitive Technology Co ltd
Priority date: 2019-12-03
Filing date: 2019-12-03
Publication date: 2023-05-26
Anticipated expiration: 2039-12-03
Also published as: CN111124347A

Abstract

The invention discloses a method and a device for forming an interaction engine cluster by aggregation, wherein the method comprises the following steps: selecting at least two historical tasks to form a historical task set; aiming at each historical task of the historical task set, acquiring an interaction engine associated with the historical task in a voice assistant system of the terminal to form an interaction engine set; judging whether at least two interaction engines in the interaction engine set have the same or corresponding slots in slots defined by the history tasks associated with the interaction engines respectively, and if so, aggregating the at least two interaction engines into an interaction engine cluster so as to facilitate the voice assistant system to use the interaction engine cluster. According to the method, the interaction engine cluster is generated based on the history dialogue, flexible tasks or association relations among the interaction engines which are more in line with the personalized demands of the user are provided, and further the voice assistant system intelligently executes the tasks in a mode which is more in line with the demands of the user and does not cause information redundancy to the user.

Description

Method and device for forming interaction engine cluster through aggregation

Technical Field

The embodiment of the invention relates to the technical field of information processing, in particular to a method and a device for forming an interaction engine cluster by aggregation.

Background

The voice assistant system allows the user to request execution of a specific task through voice instructions, for example, the user makes a ticket booking request, the voice assistant system can determine that the user intends to book a ticket, and then complete slot filling of the ticket booking task, for example, the departure place is Beijing, the destination is Shanghai, the date is 2019, 10 months 1 day, and the like.

The voice assistant system in the prior art can provide a certain correlation between tasks or interaction engines for executing tasks, for example, correlating an air ticket booking task and a weather inquiry task in advance, and correlating the air ticket booking interaction engines or the weather inquiry interaction engines.

One application scenario for the above-mentioned association is: when a user requests another task, the voice assistant system may fill the slots of the other task with slot fill content for the other task associated with the other task. For example, the user asks "what weather is then after booking an air ticket", the voice assistant system fills the city slots and date slots in the weather inquiry task with the filling content of the destination slots and date slots in the booking task associated with the weather inquiry task, without asking the user again "ask you want to inquire about the weather of which city" and/or "ask you want to inquire about the weather of which day", which provides the user with a degree of convenience in voice interaction. However, in the above scenario, if the user modifies the destination of the ticket from Shanghai to capital, the existing voice assistant system cannot infer that the user has a need to query the capital weather, and at this time, the user needs to actively query "how weather is that or" how weather is that "to capital" to be able to learn the capital weather, which is not good for the user experience.

Another application scenario for the above-mentioned association relationship is: after a certain task is executed, other tasks related to the task are automatically executed, for example, after the task of booking an air ticket is completed, the voice assistant system can automatically execute the weather inquiry task and provide the execution result of the weather inquiry task for a user, for example, the weather on the sea on the trip destination is gust, and the user can take rain gear. For the above scenario, whether the user needs or not, the user is provided with the execution result of other tasks, and the increase of the information provision may cause trouble to the user, for example, in some cases, the user does not need weather information, and at this time, the user needs to filter the information provided by the voice assistant system to obtain the ordering result of the air ticket.

In addition, the association relationship between tasks or interaction engines in the prior art is preset and fixed, which obviously cannot meet the personalized requirements of users.

In summary, how to provide a flexible task or association relationship between interaction engines, so that the voice assistant system can intelligently execute the task in a manner that better meets the user's requirements and does not bring information redundancy to the user, is a problem to be solved.

Disclosure of Invention

Aiming at the problems in the prior art, the invention provides a method and a device for forming an interaction engine cluster by aggregation.

The invention provides a method for forming an interaction engine cluster by aggregation, which comprises the following steps:

step 301: selecting at least two historical tasks to form a historical task set;

step 302: for each historical task in the historical task set, acquiring an interaction engine associated with the historical task in a voice assistant system of the terminal, and forming an interaction engine set by the interaction engine associated with each historical task in the historical task set; wherein each interaction engine is associated with at least one task, the interaction engine defining at least one slot for each task associated therewith;

step 303, determining whether at least two interaction engines in the interaction engine set have the same or corresponding slots in the slots defined for their associated history tasks, and if so, aggregating the at least two interaction engines into an interaction engine cluster, so as to facilitate the voice assistant system to use the interaction engine cluster.

The invention provides a device for forming an interaction engine cluster in an aggregation way, which comprises:

The historical task set generating unit is used for selecting at least two historical tasks to form a historical task set;

the system comprises an interaction engine set generating unit, a processing unit and a processing unit, wherein the interaction engine set generating unit is used for acquiring an interaction engine associated with each history task in a voice assistant system of a terminal aiming at each history task in the history task set, and the interaction engine associated with each history task in the history task set forms an interaction engine set; wherein each interaction engine is associated with at least one task, the interaction engine defining at least one slot for each task associated therewith;

and the aggregation unit is used for judging whether the slots defined by at least two interaction engines in the interaction engine set aiming at the history tasks associated with the interaction engines have the same or corresponding slots, and if so, aggregating the at least two interaction engines into an interaction engine cluster so as to facilitate the voice assistant system to use the interaction engine cluster.

The invention also provides a computer device characterized in that it comprises a processor and a memory in which a computer program is stored which is executable on the processor and which, when executed by the processor, implements a method as described above.

The invention also provides a computer readable storage medium, characterized in that it has stored therein a computer program executable on a processor, which when executed implements a method as described above.

The invention also provides a voice assistant system characterized by comprising the device as described above.

The invention also provides a terminal characterized by comprising the device or the voice assistant system.

According to the method and the device, the interaction engine cluster is generated based on the historical dialogue of the user and the voice assistant system, so that flexible tasks or association relations among interaction engines which are more in line with the personalized demands of the user are provided, and the voice assistant system can intelligently execute the tasks in a mode which is more in line with the demands of the user and does not cause information redundancy to the user.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a voice assistant system in one embodiment of the invention.

FIG. 2 is a method of voice interaction based on an interaction engine cluster in one embodiment of the invention.

FIG. 3 is a method of voice interaction based on an interaction engine cluster in another embodiment of the invention.

FIG. 4 is a method of voice interaction based on an interaction engine cluster in another embodiment of the invention.

FIG. 5 is a method of aggregation to form an interaction engine cluster in one embodiment of the invention.

FIG. 6 is a method of aggregation to form an interaction engine cluster in another embodiment of the invention.

FIG. 7 is a method of building an interaction engine cluster in one embodiment of the invention.

FIG. 8 is a method of building an interaction engine cluster in another embodiment of the invention.

FIG. 9 is an apparatus for voice interaction based on an interaction engine cluster in one embodiment of the invention.

FIG. 10 is an apparatus for voice interaction based on an interaction engine cluster in another embodiment of the present invention.

FIG. 11 is an apparatus for voice interaction based on an interaction engine cluster in another embodiment of the present invention.

FIG. 12 is an apparatus for aggregating to form clusters of interaction engines in one embodiment of the invention.

FIG. 13 is an apparatus for building an interaction engine cluster in one embodiment of the invention.

FIG. 14 is an apparatus for building an interaction engine cluster in another embodiment of the invention.

FIG. 15 is a method of constructing or aggregating clusters of interaction engines and voice interactions based on the clusters of interaction engines in one embodiment of the invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the embodiments of the present invention will be described in further detail with reference to the accompanying drawings. The embodiments of the present invention and the specific features of the embodiments are detailed descriptions of the technical solutions of the embodiments of the present invention, but not limited to the technical solutions of the descriptions of the present invention, and the embodiments of the present invention and the technical features of the embodiments of the present invention may be combined with each other without conflict.

1. Voice assistant system

FIG. 1 illustrates a block diagram of a voice assistant system that may be implemented on a standalone device or across multiple devices. In some embodiments, a portion of the modules, units, or functions of the voice assistant system belong to a server, and the remaining portion of the modules, units, or functions belong to a terminal that can communicate with the server over one or more networks. In some embodiments, some of the modules of the voice assistant system, such as the processing module 102, the interaction engine 112, may belong to both the server and the terminal, with the remaining modules belonging to one of the server or the terminal.

The voice assistant system mainly comprises: a human-computer interaction interface 101, a processing module 102, a database 103 and the like. The processing module includes n interaction engines 112, where n is a positive integer greater than or equal to 1, and each interaction engine 112 may include a semantic understanding module 201, a dialogue management and control module 202, a dialogue generation module 203, and a command execution module 204. The processing module 102 is connected to the human-computer interaction interface 101, and may receive data input by a user through the human-computer interaction interface, and output interaction data, such as dialogue data, task execution process and results, fed back to the user through the human-computer interaction interface.

In some embodiments, the interaction engine 112 may include one interaction main engine and/or at least one interaction sub-engine. Wherein the interaction host engine is the default engine of the voice assistant system. The interaction sub-engine on the server can be the voice assistant system itself, can be generated by the local training of the terminal and uploaded through a proprietary interface, and the interaction sub-engine in the terminal can be the voice assistant system itself, can be generated by the local training of the terminal and loaded in the voice assistant system, and can be downloaded from the server and loaded in the voice assistant system by the terminal.

In some embodiments, each interaction engine (interaction main engine, interaction sub-engine) is capable of performing at least one task, i.e. each interaction engine may be associated with at least one task, the tasks that different interaction sub-engines are capable of performing may be the same or different. The interaction engine defines at least one slot for each task with which it is associated. The terminal may download one or more interaction sub-engines from the server according to the user's needs and load the tasks into the voice assistant system of the terminal, for example, the weather interaction sub-engine may perform tasks as weather query, the slots defined for the weather query task are city and date, the user may download the weather interaction sub-engine and interact with the weather interaction sub-engine to query city weather forecast, the task performed by the music interaction sub-engine is music play, the slots defined for the music play task include singer name and song name, and the user may download the music interaction sub-engine and interact with the music interaction sub-engine to play music according to the user's requirements.

The slots defined by the interaction engine for the task associated with the interaction engine at least comprise basic slots and can also comprise extended slots, for example, the slots defined by the interaction engine B1 for the task weather query task comprise cities and dates, the slots defined by the interaction engine B2 for the task weather query task comprise cities, dates and times, the cities and dates slots are basic slots, and the time slots are extended slots.

The user may interact with the voice assistant system by voice or text, etc., the voice assistant system determines the user's intent (i.e., determines a task) according to the user's instructions, determines key knowledge data corresponding to each slot associated with the user's intent, and populates the key knowledge data into the corresponding slot. The voice assistant system then performs the task based on the filled slot or slots.

In some embodiments, the interaction host engine determines, based on the user instructions only, interaction sub-engines that are capable of handling the user instructions that do not themselves perform a particular task. Thus, determining the user intent in the above-described process may be accomplished by an interaction host engine in the voice assistant system, the interaction host engine further operable to select one or more interaction sub-engines to process the above-described user instructions based on the determined user intent, the slot filling and task execution in the above-described process being performed by the one or more interaction sub-engines determined by the interaction host engine.

2. Interactive engine cluster

At least two interaction engines are included in one interaction engine cluster. In some embodiments, the at least two interaction engines are at least two interaction sub-engines. Wherein the tasks associated with any one of the interaction engine clusters are different from the tasks associated with other interaction engines in the interaction engine clusters, the tasks associated with any one of the interaction engine clusters and the tasks associated with at least one other interaction engine in the interaction engine clusters have at least one same or corresponding slot, or the tasks belonging to the interaction engine cluster in any one of the interaction engine clusters are different from the tasks belonging to the interaction engine cluster in other interaction engine clusters, and the tasks belonging to the interaction engine cluster in any one of the interaction engine clusters and the tasks belonging to the interaction engine cluster in other interaction engine clusters have at least one same or corresponding slot

Wherein, the same slot refers to a slot with the same slot name and the same key knowledge data should be filled in the specific dialogue context, and the corresponding slot refers to a slot with different slot names and the same key knowledge data should be filled in the specific dialogue context.

An example of one interaction engine cluster is a trip interaction engine cluster, which includes an A1 interaction engine and a B1 interaction engine, where the task that the A1 interaction engine can execute is reserved for an air ticket, the slot of the task includes a departure place, a destination, and a date, where the destination is a fill-in-necessary slot, the departure place is a fill-out-selection slot, for the fill-in-selection slot, when the user instruction does not include corresponding key knowledge data, a default value is adopted to fill the slot, the default value of the place-of-departure slot is the city where the user is currently located, such as beijing, the date slot may be a fill-in-necessary slot or a fill-out-selection slot according to different settings of the A1 interaction engine, and when the user instruction does not include corresponding key knowledge data, the A1 interaction engine may return a clarity question "air ticket for what day you want to reserve". The task that the B1 interaction engine can execute is a weather query, and the slots of the task comprise cities and dates. Assuming that the user is more concerned about the weather condition of the destination of the current day of the trip when booking the air ticket, the date slot of the A1 interaction engine and the date slot of the B1 interaction engine are the same slot, the destination slot of the A1 interaction engine and the city slot of the B1 interaction engine are corresponding slots, and in this example, the correspondence of the same slot or corresponding slot of different interaction engines is one-to-one. However, the above assumption should not be considered as limiting the present invention, the user may also concern about the weather of the departure place and destination of the trip day, where the date slot of the A1 interaction engine and the date slot of the B1 interaction engine are the same slot, the destination slot of the A1 interaction engine and the city slot of the B1 interaction engine are corresponding slots, and at the same time, the departure place slot of the A1 interaction engine and the city slot of the B1 interaction engine are also corresponding slots, and in this example, the correspondence relationship of the same slots of different interaction engines is one-to-one, and the correspondence relationship of the corresponding slots is many-to-one.

The interaction engines in the interaction engine cluster may be in an inactive, active, suspended, etc. state.

The terminal can download the attribute information of the interaction engine cluster from the server, construct the interaction engine cluster locally according to the downloaded attribute information of the interaction engine cluster, and synchronize the attribute information of the constructed interaction engine cluster to other terminals of the user.

The terminal can locally aggregate and generate an interaction engine cluster according to the history dialogue record of the user and the voice assistant system, and synchronize attribute information of the generated interaction engine cluster to other terminals of the user, and/or upload the attribute information to a server for other users to download.

The attribute information of the interaction engine cluster includes at least one of: the cluster name of the interaction engine cluster, the names of at least two interaction engines included in the interaction engine cluster, and the same or corresponding slots that the at least two interaction engine clusters can perform (or associate) tasks. Table 1 shows attribute information of one interaction engine cluster.

Table 1 attribute information of interaction engine cluster

The attribute information of the interaction engine cluster may further include tasks associated with the interaction engine, for example, the attribute information of the travel interaction engine cluster may also be as described in table 2.

Table 2 attribute information of interaction engine cluster

When the tasks associated with the interaction engine are multiple, and only some of the tasks belong to a certain interaction engine cluster, the attribute information of the interaction engine is used for indicating which one or more of the tasks belong to the interaction engine cluster. For example, the tasks associated with the A1 interaction engine include air ticket booking and aviation news, wherein the air ticket booking belongs to a trip interaction engine cluster, and the tasks associated with the B1 interaction engine include weather inquiry and alarm clock setting, wherein the weather inquiry belongs to the trip interaction engine cluster, and at this time, attribute information of the trip interaction engine cluster is also shown in table 2.

When the interaction engine cluster is used, the interaction engine included in the interaction engine cluster only executes the tasks belonging to the interaction engine cluster in the associated tasks, and the tasks not belonging to the interaction engine cluster cannot be executed. For example, in the above example, when using the trip interaction engine cluster, the air ticket booking task associated with the A1 interaction engine and the weather query task associated with the B1 interaction engine are performed, and the aviation news task associated with the A1 interaction engine and the alarm clock setting task associated with the B1 interaction engine are not performed.

When the interaction engine cluster is used, if any interaction engine (hereinafter referred to as a first interaction engine) in the interaction engine cluster is activated, for example, the first interaction engine receives a user instruction, the first interaction engine is activated, or an interaction master engine determines a user intention based on the user instruction, selects an interaction sub-engine capable of processing the user intention based on the user intention and sends the user instruction to the interaction sub-engine, at the moment, the interaction sub-engine is activated, the voice assistant system activates other interaction engines (hereinafter referred to as at least one second interaction engine) in the interaction engine cluster, and performs instant synchronization on a first slot and/or a second slot of at least one second task associated with the at least one second interaction engine according to key knowledge data filled by the first slot of the first task associated with the first interaction engine in the interaction engine cluster, and at least one second task associated with the first slot and/or the second slot of the at the same time, and provides all the interaction engines to the user to execute the interaction engine simultaneously. Wherein a first task associated with a first interaction engine and at least one second task associated with at least one second interaction engine belong to the interaction engine cluster.

The instant synchronization is specifically implemented by, for example, based on a new task instruction of a user, when a first slot of a first task associated with a first interaction engine in an interaction engine cluster is changed from unfilled to filled with first critical knowledge data, immediately filling the first critical knowledge data into a first slot and/or a second slot of a second task associated with a second interaction engine, and/or, based on a cluster task modification instruction of a user, when the critical knowledge data filled in the first slot of the first interaction engine associated with the first task in the interaction engine cluster is changed from the first critical knowledge data to the second critical knowledge data, immediately modifying the first slot and/or the critical knowledge data filled in the second slot of the second interaction engine associated with the second task from the first critical knowledge data to the second critical knowledge data.

The instant synchronization is specifically that, when a first slot of a first task executed by a first interaction engine in an interaction engine cluster is changed from unfilled to filled first critical knowledge data, the first critical knowledge data is filled into a first slot and/or a second slot of a second task associated with a second interaction engine in real time, when a corresponding relation between the first interaction engine and a corresponding slot of the second interaction engine is multiple pairs of time slots, the filling is additional filling, and/or, when a user's cluster task modification instruction is based, for example, when critical knowledge data filled in the first slot of the first interaction engine in the interaction engine cluster is changed from the first critical knowledge data to the second critical knowledge data, if a corresponding relation between the first interaction engine and the corresponding slot of the second interaction engine is one-to-one, the filled critical knowledge data of the second task associated with the second interaction engine is modified from the first critical knowledge data to the second critical knowledge data, and/or the corresponding relation between the second interaction engine and the second interaction engine is modified to the second critical knowledge data, and if the corresponding relation between the first interaction engine and the second interaction engine is multiple pairs of critical knowledge data, the first interaction engine is modified to the second critical knowledge data, and the corresponding relation between the first interaction engine and the second interaction engine is modified to the second critical knowledge data.

For example, when the corresponding relationship between the corresponding slots of the first interaction engine and the second interaction engine is one-to-one, taking the date slot of the A1 interaction engine and the date slot of the B1 interaction engine as the same slot, and the destination slot of the A1 interaction engine and the city slot of the B1 interaction engine as the corresponding slots as examples, assuming that the new task instruction of the user is an "ticket from tomorrow to seas", the voice assistant system determines that the first interaction engine is the A1 interaction engine, sends the instruction to the A1 interaction engine, the A1 interaction engine fills the slot of the scheduled task of the ticket based on the instruction, fills the departure slot as the Beijing place where the user is currently located, fills the destination slot as seas, and fills the date slot as 8 months 24 days, at this time, the instant synchronization process is to fill the first key knowledge data into the city slot of the weather query task associated with the B1 interaction engine, fill the first key data into the weather query task associated with the weather query engine for 8 months 24 days, and then, provide the weather interaction engine and the weather interaction engine execution result of the B1 interaction engine to the user at the same time. Based on the task execution result, the user sends a cluster task modification instruction or goes to a pubic bar, at this time, the A1 interaction engine modifies the key knowledge data filled in the destination slot of the air ticket booking task from the first key knowledge data Shanghai to the second key knowledge data achievements, and the instant synchronization process is to modify the key knowledge data filled in the city slot of the weather inquiry task associated with the B1 interaction engine from the first key knowledge data Shanghai to the second key knowledge data achievements, wherein the modification is all modification. Subsequently, the task execution results of the A1 interaction engine and the B1 interaction engine are simultaneously provided to the user again.

When the corresponding relation between the corresponding slots of the first interaction engine and the second interaction engine is multiple pairs, taking the date slot of the A1 interaction engine and the date slot of the B1 interaction engine as the same slot, the destination slot of the A1 interaction engine and the city slot of the B1 interaction engine as the corresponding slots, the departure slot of the A1 interaction engine and the city slot of the B1 interaction engine as the corresponding slots as examples, assuming that the new task instruction of the user is 'reservation of a tomorrow to the air ticket of the sea', the voice assistant system determines that the first interaction engine is the A1 interaction engine, sends the instruction to the A1 interaction engine, the A1 interaction engine fills the slots of the reservation task of the air ticket based on the instruction, filling a departure place slot as Beijing of a current place of a user, filling a destination slot as Shanghai, filling a date slot as 8 months and 24 days, wherein at the moment, the instant synchronization process is to fill Beijing of first key knowledge data into urban slots of weather inquiry tasks associated with a B1 interaction engine, additionally fill Shanghai of the first key knowledge data into the urban slots of weather inquiry tasks associated with the B1 interaction engine, namely fill Beijing and Shanghai of the key knowledge data in the urban slots, fill 8 months of the first key data into the date slots of the weather inquiry tasks associated with the B1 interaction engine for 24 days, and then simultaneously provide task execution results of the A1 interaction engine and the B1 interaction engine for the user. Based on a task execution result, a user sends a cluster task modification instruction or goes to a pubic bar, at this time, the A1 interaction engine modifies key knowledge data filled in a destination slot of an air ticket booking task from the first key knowledge data Shanghai to the second key knowledge data dubbing, and the instant synchronization process is to modify the first key knowledge data Shanghai in the key knowledge data filled in a city slot of a weather query task associated with the B1 interaction engine to the second key knowledge data dubbing, namely, a partial modification mode is adopted, and the key knowledge data filled in the modified city slot is Beijing and dubbing. Subsequently, the task execution results of the A1 interaction engine and the B1 interaction engine are simultaneously provided to the user again.

3. Cluster task mode of operation

The voice assistant system has two modes of operation, a single task mode of operation and a cluster task mode of operation, respectively. The voice assistant system determines a first task such as air ticket reservation based on a user instruction, further determines a first interaction engine such as an A1 interaction engine, executes the first task in a single task working mode, and in a cluster task working mode, interaction engine clusters included in the voice assistant system are all in an available state, and when the first interaction engine belongs to at least one interaction engine cluster, the instruction is processed based on the at least one interaction engine cluster, namely the voice assistant system executes not only the first task but also other tasks belonging to the interaction engine cluster, such as weather inquiry tasks in table 2.

FIG. 2 illustrates a method of voice interaction based on an interaction engine cluster for a terminal with a voice assistant system, comprising the steps of:

step 201, obtaining a user instruction;

step 202, determining a first task based on the instruction, determining a first interaction engine based on the first task, and sending the instruction to the first interaction engine so that the first interaction engine fills the slot of the first task based on the instruction; wherein the first interaction engine is associated with the first task;

Step 203, judging whether the voice assistant system is in a cluster task operation mode, if yes, executing

steps

205a, 205b, 205c and 206, and if no, executing step 204;

step 204, receiving a first task execution result from the first interaction engine, and providing the first task execution result to a user;

step 205a, determining an interaction engine cluster to which the first interaction engine belongs, wherein the interaction engine cluster comprises the first interaction engine and at least one second interaction engine;

step 205b, activating the at least one second interaction engine;

step 205c, performing instant synchronization on the first slot position of at least one second task associated with at least one second interaction engine in the interaction engine cluster and/or the second slot position corresponding to the first slot position based on the key knowledge data filled by the first interaction engine on the first slot position of the first task;

step 206, receiving a first task execution result from the first interaction engine, receiving at least one second task execution result from the at least one second interaction engine, and providing the first task execution result and the at least one second task execution result to a user at the same time.

The above-described methods may be performed by a voice assistant system or by an apparatus included in the voice assistant system.

The voice assistant system of the terminal comprises at least one interaction engine cluster, and specifically, the voice assistant system stores attribute information of the at least one interaction engine cluster. The voice assistant system may locally aggregate to generate the interaction engine clusters, such as methods 300 and/or 400 hereinafter, or may download attribute information of the interaction engine clusters from a server and locally construct to form the interaction engine clusters, such as methods 500 and/or 600 hereinafter.

The instruction is sent to the first interaction engine, i.e. the first interaction engine is activated, step 202.

In step 205b, the voice assistant system may activate the at least one second interaction engine by sending an activation message to the at least one second interaction engine.

The first task associated with the first interaction engine and the at least one second task associated with the at least one second interaction engine belong to the interaction engine cluster.

Preferably, if in step 205a, the first interaction engine does not belong to any interaction engine cluster, step 204 is performed.

Preferably, in step 205a, the interaction engine cluster to which the first interaction engine belongs is determined by means of a search, for example, the name of the first interaction engine is searched in attribute information of the interaction engine cluster in the terminal or the voice assistant system, if so, the corresponding interaction engine cluster is the interaction engine cluster described by the first interaction engine, step 205b is executed, and if not, the first interaction engine does not belong to the task interaction engine cluster, at this time, step 204 is executed.

Preferably, in step 205a, the first interaction engine and the interaction engine cluster to which the first task belongs are determined. For example, the instruction sent by the user is "reservation of a tomorrow to Shanghai air ticket", the first interaction engine is an A1 interaction engine, the first task is reservation of an air ticket, and in step 205a, it is determined that the travel interaction engine cluster to which the A1 interaction engine and the air ticket reservation task belong is obtained, but other interaction engine clusters to which the A1 interaction engine and the aviation news task belong are not determined.

Preferably, in step 205a, the interaction engine cluster to which the first interaction engine belongs is determined by determining the interaction engine cluster to which the first task belongs, e.g. looking up the first task in attribute information of the interaction engine cluster in the terminal or the voice assistant system.

In some embodiments, one or more of at least two interaction engines included in the interaction engine cluster are dominant interaction engines, the other interaction engines are slave interaction engines, when the voice assistant system is in the cluster task working mode, if the first interaction engine is the dominant interaction engine of the interaction engine cluster to which the first interaction engine belongs, when the first interaction engine is activated, other interaction engines in the interaction engine cluster are also activated, and instant synchronization of slots among the interaction engines is performed, if the first interaction engine is not the dominant interaction engine of the interaction engine cluster to which the first interaction engine belongs, only the first interaction engine is activated. For example, in an interaction engine cluster including an A1 interaction engine and a B1 interaction engine, the A1 interaction engine is the dominant interaction engine, when an instruction issued by a user is "reservation of a tomorrow to an air ticket in the sea", the first interaction engine is the A1 interaction engine, which is the dominant interaction engine, and therefore, both the A1 interaction engine and the B1 interaction engine are activated, and both the air ticket reservation and the weather inquiry task are performed, whereas when an instruction issued by a user is "how the weather in the sea today", the first interaction engine is the B1 interaction engine, which is not the dominant interaction engine, and therefore, only the B1 interaction engine is activated, the A1 interaction engine is not activated, and the air ticket reservation task is not performed. It is noted that the first interaction engine may be a master interaction engine in one interaction engine cluster and slave interaction engines in other interaction engine clusters.

The step 205a further includes: judging whether the first interaction engine is the dominant interaction engine of the interaction engine cluster, if yes, executing step 205b and subsequent steps, otherwise, executing step 204.

In some embodiments, after the

steps

204 and 206 are performed, the process returns to step 201 to continue receiving the instruction of the user.

In certain embodiments, the method shown in fig. 1 further comprises step 201a before performing step 202: judging the type of the instruction, and if the instruction is a new task instruction, executing the step 202 and the subsequent steps;

in step 201a, it is determined that the instruction is of the type, if the instruction is a single task modification instruction, step 207 is executed: and sending the single task modification instruction to a first interaction engine so that the first interaction engine modifies key knowledge data filled in the slots of the first task based on the instruction, acquiring a first task modification execution result from the first interaction engine, providing the first task modification execution result to a user, and returning to step 201.

In step 201a, the type of the instruction is determined, and if the instruction is a cluster task modification instruction, steps 208 to 209 are executed.

Step 208: determining a third interaction engine corresponding to the cluster task modification instruction, and sending the instruction to the third interaction engine so that the third interaction engine modifies a third slot of a third task associated with the third interaction engine based on the instruction; performing instant synchronization on the third slot of at least one fourth task associated with a fourth interaction engine in the interaction engine cluster and/or a fourth slot corresponding to the third slot based on the modified key knowledge data filled by the third slot of the third task;

step 209, receiving the third task execution result from the third interaction engine, receiving at least one fourth task execution result from the at least one fourth interaction engine, and providing the third task execution result and the at least one fourth task execution result to the user at the same time, and returning to step 201.

The first task, the second task, the third task, and the fourth task described above are merely references to tasks, and do not mean that the tasks described above are necessarily identical or different. Specifically, the third interaction engine may be a first interaction engine, and correspondingly, the fourth interaction engine may be at least one second interaction engine, where the first task is the same as the third task, and the second task is the same as the fourth task; the third interaction engine may be a second interaction engine, and correspondingly, the fourth interaction engine may be another second interaction engine or a first interaction engine, where the third task is the same as the second task and the fourth task is the same as the first task or another second task.

The third task associated with the third interaction engine and the at least one fourth task associated with the at least one fourth interaction engine belong to the interaction engine cluster.

In some embodiments, the task execution results given by the interaction engine may be final state task execution results; in some embodiments, the task execution results given by the interaction engine may be transient task execution results, at which time the execution status of the corresponding task is marked as incomplete. After receiving the task confirmation instruction of the user, the interaction engine can convert the transient task execution result into a final task execution result, for example, the conversion of the task execution result is completed by communicating with a corresponding server. In

steps

204 and 206, the task execution result given by at least one interaction engine included in the interaction engine cluster is a transient task execution result, in step 204, the at least one interaction engine is a first interaction engine, and in step 206, the at least one interaction engine may be a first interaction engine and/or a second interaction engine. As described above, after

steps

204 and 206, returning to step 201, the user's instruction is continuously received, in step 201a, the type of the instruction is determined, and if the instruction is a task confirmation instruction, step 210 is executed: and sending the task confirmation instruction to the at least one interaction engine to trigger the at least one interaction engine to convert the transient task execution result into a final state task execution result, receiving the final state task execution result from the at least one interaction engine, providing the final state task execution result to a user, and returning to step 201.

For example, if the instruction sent by the user for the first time is "ticket scheduled tomorrow to Shanghai", in step 201a, if it is determined that the instruction is a new instruction, step 202 is executed, in step 202, the voice assistant system determines that the first task engine is an A1 interaction engine, sends the instruction to the A1 interaction engine, the A1 interaction engine fills the departure slot as the Beijing of the current location of the user based on the slot of the scheduled task of the instruction filling ticket, fills the destination slot as Shanghai, fills the date slot as 8 months 24 days, and if the working mode of the voice assistant system is a cluster task working mode, or if the working mode of the voice assistant system is a cluster task working mode and the first interaction engine is a dominant interaction engine of the travel interaction engine cluster to which the voice assistant system belongs, in step 205B 1 interaction engine in the travel interaction engine cluster is also activated, in this example, and if the corresponding relationship between the first interaction engine and the second interaction engine is one-to-one in-the above text, the voice assistant system fills the first key knowledge interaction engine to the relevant slot of the Beijing 1 interaction engine to the relevant task, and queries weather information in the first slot of the relevant slot of the first key slot of the Beijing 1 interaction engine to the relevant slot to the first slot of the relevant weather 1, and the weather query for the weather in the first month 1 is filled to the relevant slot in the relevant slot 1. Then, the method is executed to step 206, the task execution result provided to the user is the first task execution result of the A1 interaction engine, "for you searching for an air ticket from beijing to Shanghai by tomorrow CA232, the price is 500 yuan, whether to reserve" and the second task execution result of the B1 interaction engine, "Shanghai tomorrow cloudy, temperature 28-34 degrees", wherein the task execution result of the A1 interaction engine is the transient task execution result, at this time, returning to step 201, continuing to receive the user instruction, and the method of the invention is described by taking as an example that three different user instructions are continuously received after returning to step 201:

First kind: if the received instruction is "confirm reservation" after returning to step 201, the A1 interaction engine communicates with a server, such as a national navigation server, to complete ticket reservation, and converts the transient task execution result into a final task execution result.

Second kind: if the received command is "get-to-pub" after returning to step 201, in step 201a, it is determined that the command is a cluster task modification command, at this time, step 208 is executed, where the third interaction engine corresponding to the command is an A1 interaction engine, that is, the third interaction engine is a first interaction engine, the third task is the same as the first task, the third slot (destination slot) of the third task (air ticket reservation) is modified based on the command, the key knowledge data is modified from Shanghai to dumb, the fourth slot (city) of at least one fourth task (weather query) associated with at least one fourth interaction engine (B1 interaction engine) in the interaction engine cluster is synchronized in real time based on the modified key knowledge data (dumb) filled in the third slot of the third task, in step 209, the third interaction engine and the at least one fourth interaction engine respectively execute the related tasks, and return the task execution results to the voice assistant system, and the voice assistant system provides the third task execution results of the third interaction engine, namely, whether you search for an air ticket from Beijing to the capital in the tomorrow CA233, the price is 400 yuan, and whether you want to reserve, and the fourth task execution results of the at least one fourth interaction engine, namely, whether you get tomorrow cloudy, and the temperature is 22-26 degrees, to the user.

Third kind: if the received instruction is "how good the weather is" after returning to step 201, in step 201a, it is determined that the instruction is a cluster task modification instruction, at this time, step 208 is executed, where the third interaction engine corresponding to the instruction is a B1 interaction engine, that is, the third interaction engine is a second interaction engine, the third task is the same as the second task, the third slot (city slot) of the third task (weather query) is modified based on the instruction, the critical knowledge data is modified from Shanghai to complete, the fourth slot (destination slot) of at least one fourth task (ticket reservation) associated with at least one fourth interaction engine (A1 interaction engine) in the interaction engine cluster is synchronized in real time based on the modified critical knowledge data (complete) filled in the third slot of the third task, in step 209, the third interaction engine and at least one fourth interaction engine respectively execute the related tasks, and return the task execution results to the voice assistant system, and the voice assistant system provides the user with the third task execution results of the third interaction engine, namely, "get the best in the sky and cloudy, the temperature of 22-26 degrees" and the fourth task execution results of the at least one fourth interaction engine, "get the best in the sky and CA233 from Beijing to get the best in the sky, the price is 400 yuan, and whether to reserve the air ticket.

Similarly, if the first instruction sent by the user is "preset tomorrow to Shanghai air ticket", in step 201a, it is determined that it is a new instruction, step 202 is executed, it is determined in step 202 that the first task engine is an A1 interaction engine, and subsequent steps are continuously executed, if the working mode of the voice assistant system is not a cluster task working mode and/or the A1 interaction engine is not a dominant interaction engine, the method proceeds to step 204, and if the task execution result provided to the user is the first task execution result of the A1 interaction engine, "if you search for tomorrow's ticket from north to Shanghai, the price is 500 yuan, if reservation is to be made", step 201 is also returned to continuously receive the user instruction. The method of the present invention will be described below taking as an example that two different user instructions are continuously received after returning to step 201:

Second kind: if the received command is "get to all bar" after returning to step 201, in step 201a, it is determined that the command is a single task modification command, and step 207 is executed: the first interaction engine (A1 interaction engine) modifies the slot position (destination) of the first task type (reservation of air ticket) based on the instruction, executes the first task type, searches for an air ticket from beijing to the capital for tomorrow CA233, has a price of 400 yuan, and returns to step 201.

For example, when the user gives an instruction that the user does not specify the date of purchasing the air ticket, after step 205c, step 205d is further included: judging whether a cluster task clarity question is received from the first interaction engine, if so, executing a step 205e, otherwise, executing a step 206;

step 205e: providing the cluster task clarity question to a user, returning to step 201;

in step 201a, if the instruction is an instruction for the task clarity question response, step 205f is executed;

and step 205f, sending the instruction to a first interaction engine, so that the first interaction engine fills the slot of the first task according to the instruction for the clarity question response, and continuing to execute step 205c.

Similarly, when the user instruction is incomplete, and the first task associated with the first interaction engine cannot be executed, in step 203, when it is determined that the voice assistant system is not in the cluster task operation mode, step 204a is executed: judging whether a single-task clarity question is received from the first interaction engine, if so, executing a step 204b, otherwise, executing a step 204;

Step 204b: providing the single-task clarity question to a user, returning to step 201;

in step 201a, if the instruction is an instruction for the single task clarity question response, step 204c is performed;

and step 204c, sending the instruction to a first interaction engine, so that the first interaction engine fills the slot of the first task according to the instruction for the clarity question response, and continuing to execute step 204a.

If in step 205a, the first interaction engine does not belong to any interaction engine cluster, or it is determined that the first interaction engine is not the dominant interaction engine of the interaction engine cluster, step 204a is executed: and judging whether a single-task clarity question is received from the first interaction engine, if so, executing a step 204b, otherwise, executing a step 204.

When the number of the first interaction engine or the interaction engine clusters to which the first interaction engine and the first task belong in step 205b is multiple, the instructions are processed using the multiple interaction engine clusters, that is, sequentially according to the multiple interaction engine clusters.

When the number of the first interaction engine or the interaction engine clusters to which the first interaction engine and the first task belong in step 205b is multiple, selecting to use one of the interaction engine clusters according to a third preset rule, or using the multiple interaction engine clusters.

In some embodiments, the third preset rule is at least one of: the most frequently used interaction engine cluster, the most recently used interaction engine cluster, the interaction engine cluster with the highest user score, etc. The third preset rule may be the same as or different from the second preset rule mentioned below.

FIG. 3 illustrates a method of voice interaction based on an interaction engine cluster for a terminal with a voice assistant system, comprising the steps of:

step 201, obtaining a user instruction;

step 201a, judging the type of the instruction, if the instruction is a cluster task working mode changing instruction, executing step 211;

step 211, the voice assistant system modifies the working mode of the voice assistant system into a cluster task working mode according to the cluster task working mode changing instruction;

and the voice assistant system processes task instructions of a user by using the interaction engine cluster in the cluster task working mode, and simultaneously provides at least two task execution results provided by at least two interaction engines included in the interaction engine cluster for the user.

And processing task instructions of users by using the interaction engine clusters, specifically, when a first interaction engine is activated based on the task instructions, activating at least one second interaction engine in the interaction engine clusters to which the first interaction engine belongs, and filling at least one first slot of at least one second task associated with at least one second interaction engine in the interaction engine clusters and/or a second slot corresponding to the first slot based on key knowledge data filled in the at least one first slot of the first task associated with the first interaction engine.

And simultaneously providing at least two task execution results provided by at least two interaction engines included in the interaction engine cluster to a user, specifically, receiving the first task execution result from the first interaction engine, receiving at least one second task execution result from the at least one second interaction engine, and simultaneously providing the first task execution result and the at least one second task execution result to the user.

Through the method, the user can modify the working mode of the voice assistant system into the cluster task working mode according to the requirements of the user, so that the execution of the associated task is more in line with the requirements of the user, and the transmission of redundant information to the user is avoided.

Preferably, after step 211, step 201 is returned.

Preferably, after step 211, the method further comprises:

step 212, judging whether the last instruction of the working mode changing instruction is a new task instruction, if so, executing step 213a;

step 213a: and processing the last instruction of the user by using the interaction engine cluster, and simultaneously providing at least two task execution results provided by at least two interaction engines included in the interaction engine cluster for the user.

Step 213a specifically includes the following steps:

step 215, determining a first task based on the new task instruction, and determining a first interaction engine based on the first task, wherein the first interaction engine is associated with the first task; the new task instruction is sent to the first interaction engine, so that the first interaction engine fills the slot of the first task based on the new task instruction;

step 205b, activating the at least one second interaction engine;

step 205c, performing instant synchronization on a first slot of at least one second task associated with at least one second interaction engine in the interaction engine cluster and/or a second slot corresponding to the first slot based on key knowledge data filled in the first slot by the first interaction engine;

step 206, receiving the first task execution result from the first interaction engine, receiving at least one second task execution result from the at least one second interaction engine, and providing the first task execution result and the at least one second task execution result to a user at the same time.

The preferred embodiment hereinbefore described with respect to the method shown in fig. 2 is equally applicable to the method shown in fig. 3. And will not be described in detail herein.

The above method is described below with reference to specific examples. For example, after the user issues the instruction "schedule tomorrow to air ticket in Shanghai", immediately after the user gives the instruction "switch operation mode to cluster task operation mode", at this time, in step 211, the voice assistant system modifies its operation mode to cluster task operation mode, and acquires the last instruction "schedule tomorrow to air ticket in Shanghai", and determines that the last instruction is a new task instruction, and then step 215 is executed. In step 215, the first interaction engine is determined to be an A1 interaction engine, and the last instruction "reservation of a tomorrow to the air ticket in the open sea" is sent to the A1 interaction engine, after which steps 205a-216 are performed, so that the first task execution result and the at least one second task execution result can be provided to the user at the same time.

By the method, the user can modify the working mode of the voice assistant system at any time, and the user can automatically execute the last new task instruction according to the cluster voice interaction mode after modifying the working mode of the voice assistant system, so that the user does not need to re-send the new task instruction again under the condition that the user modifies the working mode after sending the new task instruction, and the voice conversation experience of the user is improved.

Preferably, when executing step 213a, step 214 is executed first: judging whether the voice assistant system provides the task execution result for the user aiming at the newly-built task instruction, and if so, executing the step 205a and the subsequent steps; if not, step 215 and subsequent steps are performed. That is, if the task execution result is already provided to the user for the new task instruction, step 215 may be skipped, that is, the first interaction engine does not need to be determined again to determine the first task, and the new task instruction does not need to be sent to the first interaction engine, so that the first interaction engine does not need to fill the slot of the first task based on the new task instruction again. It should be noted that, even if the task execution result for the new task instruction is provided to the user, in step 216, the voice assistant system still obtains the first task execution result from the first interaction engine and provides the first task execution result and the second task execution result to the user together, which is considered that, after the user gives the new task instruction and obtains the corresponding task execution result, the user finds that the task execution result does not conform to the expectation, and immediately modifies the working mode of the voice assistant system, at this time, the user may not carefully listen to or view the first task execution result, so after modifying the working mode of the voice assistant system, the first task execution result and the second task execution result provided before are provided to the user together, so as to better conform to the experience requirement of the user, avoid that the user still needs to relocate the first task execution result provided by the voice assistant system before modifying the working mode of the voice assistant system, and provide a smoother use experience for the user.

Preferably, in step 212, when it is determined that the previous instruction of the operation mode change instruction is a new task instruction, step 213 is executed: and judging whether the receiving time of the new task instruction meets the preset condition, and if so, executing step 213a.

Preferably, if it is determined that the receiving time of the new task instruction does not meet the preset condition, the step 201 is returned.

The preset condition is that the difference value between the receiving time of the new task instruction and the receiving time of the cluster task work mode change instruction is smaller than a preset value.

By the method, the situation that the previous new task instruction is executed in the cluster voice interaction mode unnecessarily is avoided, for example, a user sends out a new task instruction at ten am and obtains a task execution result, and a working mode modification instruction is sent out at three pm.

In step 201a, the voice assistant system may determine, for a keyword included in the user instruction, whether the instruction is a cluster task work mode change instruction.

The step 211 further includes: and sending a synchronous message to other terminals of the user, wherein the synchronous message is used for indicating a voice assistant system on the other terminals of the user to modify the working mode of the voice assistant system into a cluster task working mode.

4. Cluster name

FIG. 4 shows a method for voice interaction based on an interaction engine cluster, comprising the steps of:

step 201, obtaining a user instruction;

step 201a, judging the type of the instruction, if the instruction includes an interaction engine cluster enabling instruction, executing step 216; the interaction engine cluster enabling instruction carries a cluster name;

step 216, enabling the interaction engine cluster corresponding to the cluster name by the voice assistant system;

step 220, the voice assistant system processes the new task instruction based on the interaction engine cluster corresponding to the cluster name carried in the interaction engine cluster enabling instruction according to the new task instruction carried in the instruction;

and 206, simultaneously providing at least two task execution results provided by at least two interaction engines included in the interaction engine cluster to a user.

After step 216 and before step 220, step 217 is further included to determine whether the instruction includes a new task instruction, if yes, step 220 is executed, otherwise, step 201 is returned.

In step 217, if the instruction includes a new task instruction, further executing step 218, determining a first task based on the new task instruction, determining a first interaction engine associated with the first task based on the first task, determining whether the first interaction engine belongs to the interaction engine cluster, if yes, executing step 220, otherwise, executing step 221;

step 221 is: the user is prompted to newly create a task instruction that does not match the cluster name, and the process returns to step 201.

In step 217, if the instruction includes a new task instruction, further executing step 218, determining a first task based on the new task instruction, determining a first interaction engine associated with the first task based on the first task, determining whether the first interaction engine belongs to the interaction engine cluster, if yes, executing step 219, otherwise, executing step 221;

step 219: judging whether the first interaction engine is the dominant interaction engine of the interaction engine cluster, if so, executing step 220, otherwise, executing step 222;

Step 222 is to prompt the user to newly build the interaction engine non-dominant interaction engine associated with the task instruction, and return to step 201.

Preferably, step 220 includes the following steps.

Step 220a, a new task instruction carried in the instruction is sent to a first interaction engine in the interaction engine cluster, so that the first interaction engine fills the slot of a first task based on the instruction; the first interaction engine is associated with a first task, which is determined based on the new task instruction;

step 205b, activating at least one second interaction engine of the interaction engine cluster; the at least one second interaction engine is other interaction engines in the interaction engine cluster than the first interaction engine;

step 205c, performing instant synchronization on the first slot of at least one second task associated with at least one second interaction engine in the interaction engine cluster and/or the second slot corresponding to the first slot based on the key knowledge data filled in the first slot by the first interaction engine.

Step 206 specifically comprises: receiving the first task execution result from the first interaction engine, receiving at least one second task execution result from the at least one second interaction engine, and simultaneously providing the first task execution result and the at least one second task execution result to a user.

Preferably, step 220 comprises the steps of:

step 220a (as previously described, not described herein in detail);

step 205b (as previously described, not described herein in detail);

step 205c (as described above, not described here again);

step 205d: judging whether a clarity question is received from the first interaction engine, if so, executing a step 205e, otherwise, executing a step 206;

step 205f: and sending the instruction to a first interaction engine, so that the first interaction engine fills the slot of the first task according to the instruction for the clarity question response, and executing step 205c after step 205 f.

Step 206 (as previously described, not described here in detail).

In certain embodiments, in step 218, it is determined whether the first interaction engine and the first task belong to the interaction engine cluster.

In step 201a, if the user instruction is a new task instruction, step 202a is executed, and it is determined whether the previous instruction of the new task instruction is an interaction engine cluster enabling instruction, if yes, step 220 and subsequent steps are executed, or step 218 and subsequent steps are executed, otherwise, step 202 is executed. It should be noted that, the above user instruction is a new task instruction, which means that the user instruction includes only the new task instruction and does not include the interaction engine cluster enabling instruction, conversely, if the user instruction includes the interaction engine cluster enabling instruction, even if the user instruction includes the new task instruction at the same time, the user instruction is not classified as the new task instruction and executed in step 202a, but executed in step 216 according to the interaction engine cluster enabling instruction, that is, the judgment level of the interaction engine cluster enabling instruction is higher than that of the new task instruction.

The above describes that the interaction engine clusters are used in two manners of cluster task working mode and cluster name, through the former use of the interaction engine clusters, all the interaction engine clusters in the voice assistant system of the terminal are in a usable state, and the specific use of which interaction engine cluster or clusters depends on the task instruction given by the user, while through the latter use of the interaction engine clusters, only the interaction engine cluster corresponding to the cluster name can be used, in the two use manners, the use of the interaction engine cluster through the cluster name is preferentially considered, and the working mode of the voice assistant system is judged when the user does not give an interaction engine cluster enabling instruction. Through the two different modes, more convenient and richer interaction engine cluster use experience is provided for users.

5. Local aggregation to form interaction engine clusters

Referring to FIG. 5, a method 300 of aggregating clusters of interaction engines for a terminal with a voice assistant system in one embodiment of the invention is shown, comprising the steps of:

Preferably, in step 301, at least two history tasks are selected according to a first preset rule and a history dialogue record.

Preferably, in step 301, at least two history tasks are selected according to a first preset rule and a history dialogue record, specifically: selecting at least two historical tasks executed in a preset time in a historical dialogue record; and/or selecting at least two history tasks executed in a preset dialog turn number in the history dialog record; and/or selecting at least two history tasks continuously executed in the history dialogue record. The at least two history tasks that are continuously executed means that no other history task that does not belong to the at least two history tasks is executed between the at least two history tasks.

In step 303, after the at least two interaction engines are aggregated to form an interaction engine cluster, one or more interaction engines in the interaction engine cluster are determined to be dominant interaction engines according to the sequence of the at least two history tasks.

Referring to fig. 6, a method 400 of aggregating clusters of interaction engines for a terminal with a voice assistant system in one embodiment of the invention is shown, comprising the steps of:

Step 401, acquiring a history dialogue record of a user and a voice assistant system, and judging whether the history dialogue record has the following conditions: interrupting a dialogue related to a first task, entering a dialogue related to at least one second task, and restoring the dialogue related to the first task after the at least one second task is executed; if so, forming a historical task set, wherein the historical task set comprises the first task and at least one second task;

step 402: for each historical task in the historical task set, acquiring an interaction engine associated with the historical task in a voice assistant system of the terminal, and forming an interaction engine set by the interaction engine associated with each historical task in the historical task set; wherein each interaction engine is associated with at least one task, the interaction engine defining at least one slot for each task associated therewith;

step 403, determining whether at least two interaction engines in the interaction engine set have the same or corresponding slots in the slots defined for their associated history tasks, and if so, aggregating the at least two interaction engines into an interaction engine cluster, so as to facilitate the voice assistant system to use the interaction engine cluster.

Preferably, in step 401, it is determined whether the following situations exist in the history dialogue record: interrupting a dialogue related to a first task, entering a dialogue related to at least one second task, and recovering the dialogue related to the first task after the at least one second task is executed, wherein the method specifically comprises the following steps: analyzing the history dialogue record to obtain a corresponding history task execution time line and a history task execution state; judging whether a historical task closed loop exists or not based on the historical task execution time line and the historical task execution state; the history task closed loop is a history task sequence with a first task at the head and the tail nodes, at least one second task at the middle node, incomplete execution state of the head node and complete execution state of the at least one middle node. In the historical task sequence, the task execution state of the tail node can be complete or incomplete, and the invention is not limited to the task execution state.

In step 403, after aggregating the at least two interaction engines into one interaction engine cluster, the interaction engine associated with the first task is determined to be the dominant interaction engine.

The following applies to the methods 300 and 400 described above.

Preferably, in steps 303 and 403, based on the instruction of the user in the history dialogue record, it is determined whether at least two interaction engines in the interaction engine set have the same or corresponding slots in the slots defined for the corresponding history task.

Preferably, in steps 303 and 403, based on an instruction of a user in the history dialogue record, key knowledge data of at least one slot defined by an interaction engine associated with a history task corresponding to the instruction for the history task is determined, and based on the slot and the key knowledge data, it is determined whether at least two interaction engines in the interaction engine set have the same or corresponding slot in the slots defined by the interaction engine for the associated history task.

Preferably, in steps 303 and 403, the at least two interaction engines are aggregated to form an interaction engine cluster, specifically: and recording names of the at least two interaction engines in attribute information of the interaction engine cluster, and recording the same or corresponding slots in slots defined by the at least two interaction engines in the interaction engine set aiming at the associated historical task in the attribute information of the interaction engine cluster.

In steps 303 and 403, the voice assistant system uses the interaction engine cluster, specifically: when a first interaction engine in the interaction engine cluster is activated, at least one second interaction engine in the interaction engine cluster is activated, at least one first slot and/or a second slot of at least one second task associated with at least one second interaction engine in the interaction engine cluster is synchronized on the basis of key knowledge data filled into the at least one first slot of the first task associated with the first interaction engine, the second task associated with the second interaction engine has the at least one first slot and/or the second slot corresponding to the first slot, the first task execution result is received from the first interaction engine, the at least one second task execution result is received from the at least one second interaction engine, and the first task execution result and the at least one second task execution result are simultaneously provided to a user.

In steps 303 and 403, the voice assistant system uses the interaction engine cluster, specifically: the voice assistant system uses the interaction engine cluster when the first interaction engine determined based on the user instruction is in the cluster task work mode.

In steps 303 and 403, the voice assistant system uses the interaction engine cluster, specifically: the voice assistant system is in a cluster task working mode, and uses the interaction engine cluster when a first interaction engine and a first task determined based on a user instruction belong to the interaction engine cluster.

In steps 303 and 403, the voice assistant system uses the interaction engine cluster, specifically: and when the received user instruction comprises the cluster name of the interaction engine cluster, the interaction engine cluster is used.

In steps 303 and 403, the voice assistant system uses the interaction engine cluster, including in particular some or all of the steps 201-222 described previously.

The above methods 300 and 400 are described below in connection with specific examples. The method 300 will be specifically described first. For example, a historical dialog of the user with the voice assistant system at 10 am to 11 am on 24 days 8 months is recorded as follows:

the user: reservation of open day to Shanghai airline tickets.

Voice assistant system: two pm CA232 has reserved for your day an air ticket from beijing to Shanghai;

the user: how does the open sky Shanghai weather?

Voice assistant system: the cloudy of the Mingshanghai changes into fine.

The user: a sunny day of Zhou Jielun is played.

Voice assistant system: a sunny day of Zhou Jielun has been found for you, i.e. is to be played.

Taking step 301 as an example of selecting at least two history tasks executed within a preset time in the history dialogue record to further describe the above method, if the preset time is 1 hour, three history tasks are selected based on the history dialogue record: air ticket reservation, weather inquiry and song playing; as previously described, the interaction engine associated with the airline reservation task is assumed to be an A1 interaction engine, the interaction engine associated with the weather query task is a B1 interaction engine, and the interaction engine associated with song play is assumed to be a C1 interaction engine. In step 302, a set of interaction engines is formed including an A1 interaction engine, a B1 interaction engine, a C1 interaction engine. The slot defined by the A1 interaction engine for the ticket reservation includes departure place, destination and date, the slot defined by the B1 interaction engine for the weather inquiry task includes city and date, the slot defined by the C1 interaction engine for the song playing includes singer name and song name, then in step 303, based on the user's instruction in the history dialogue record, "reservation of the ticket from tomorrow to the Shanghai", the interaction engine associated with the history task corresponding to the instruction, namely, the A1 interaction engine for the ticket reservation defines the key knowledge data of the departure place slot as the current city of the user, namely, beijing, key knowledge data of the destination slot as the Shanghai and date slot is determined to be 8 months 25 days, based on the user's instruction in the history dialogue record, "how is the weather on tomorrow? The method comprises the steps of "", determining that key knowledge data of a city slot defined by at least two interaction engines in an interaction engine set for weather inquiry is the same or corresponding slot in slots defined by weather inquiry, determining that key knowledge data of a artist name slot defined by an interaction engine C1 interaction engine for song corresponding to a historical task, namely song playing, is the fine day of Zhou Jielun based on an instruction of a user in a historical dialogue record, namely, the fine day of Zhou Jielun, determining that key knowledge data of a city slot defined by at least two interaction engines in the interaction engine set for associated historical tasks is the fine day, determining that the city slot defined by an A1 interaction engine for a ticket preset definition and the city slot defined by a B1 interaction engine for a weather inquiry task are the corresponding slot, determining that the slot defined by an A1 interaction engine for a ticket preset definition and the slot defined by a B1 interaction engine for a song name slot defined by a song are the same or corresponding slot defined by a voice interaction engine, and determining that the city slot defined by an A1 interaction engine for a ticket preset definition and the city slot defined by a voice interaction engine for a voice interaction engine are the same or corresponding slot in a voice interaction engine, and a voice interaction engine A1 and a voice interaction engine are not formed in a subsequent interaction engine aggregation system.

The method 400 is described in detail below. For example, the history of the user's dialogue with the voice assistant system is recorded as follows:

the user: help me to set an alarm clock of 7 am on tomorrow.

Voice assistant system: an alarm clock of 7 points in the morning of tomorrow is helped.

The user: reservation of open day to Shanghai airline tickets.

Voice assistant system: for you search for tickets from Beijing to Shanghai at two pm in tomorrow CA 232;

the user: a sunny day of Zhou Jielun is played.

The user: how does the open sky Shanghai weather?

Voice assistant system: the cloudy of the Mingshanghai changes into fine.

The user: the reservation is confirmed.

Voice assistant system: two pm CA232 has reserved for your day to go from beijing to the Shanghai air ticket.

The user: querying nearby China banks.

Voice assistant system: there are 1 national banks within 1 km nearby, the address is construction road 88 number, and the phone is 12345678.

The historical task execution time line obtained based on the analysis of the historical dialog records is as follows: the alarm clock sets up- > ticket reservation- > song play- > weather inquiry- > ticket reservation- > map inquiry, the historical task execution state is: and determining that a history task closed loop exists in the history dialogue record based on the history task execution time line and the history task execution state, wherein a head node and a tail node of the history task closed loop are reserved for a first task, namely an air ticket, the history task closed loop is provided with two middle nodes, a corresponding second task is song playing and weather inquiring, and the execution state of the head node is incomplete and the execution state of the middle nodes is complete. At this time, the generated history task set includes: air ticket reservation, song play, and weather inquiry. In this example, the historical task set generated by the method 400 is the same as the historical task set generated by the method 300 in the previous example, and thus, the execution of the steps 402 and 403 is similar to the steps 302 and 303, and will not be repeated here.

In certain embodiments, prior to step 301 or 401, the following steps are performed:

step 201: receiving an instruction of a user;

step 201a, judging the type of the instruction, and if the instruction is an interaction engine cluster aggregation instruction, executing steps 301-303 and/or 401-403.

In the embodiment in which steps 301 to 303 and steps 401 to 403 are performed, the steps 301 to 303 may be performed first, or steps 401 to 403 may be performed first, without limiting the order of the execution of the steps.

After step 303 or 403, step 304 or 404 is performed, said step 304 or 404: and providing a first interface for the user so that the user names the interaction engine cluster through the first interface, and recording the cluster name input by the user in attribute information of the interaction engine cluster by the voice assistant system.

Setting corresponding use level for the interaction engine cluster. In step 303 or step 403, after the at least two interaction engines are aggregated to form one interaction engine cluster, the usage level of the interaction engine cluster is marked as the lowest, and the usage level of the interaction engine cluster is recorded in attribute information thereof.

In some embodiments, in step 301 or 401, a plurality of historical task sets are generated according to the historical dialog records, statistics are performed on the generated plurality of historical task sets, and the use level of the interaction engine cluster is marked according to the frequency of the historical task sets in the statistics result. For example, 3 historical task sets are generated, namely (air ticket reservation, song play and weather inquiry), (restaurant empty table inquiry and taxi) and (air ticket reservation, song play and weather inquiry), respectively, and the statistical result is (air ticket reservation, song play and weather inquiry) twice, (restaurant empty table inquiry and taxi), the voice assistant system marks a higher use level for the interaction engine cluster generated based on the task set (air ticket reservation, song play and weather inquiry) and a lower use level for the interaction engine cluster generated based on the task set (restaurant empty table inquiry and taxi).

In certain embodiments, after step 303 or 403, step 305 or 405 is performed: and providing a second interface for the user, so that the user modifies the use level of the interaction engine cluster through the second interface, and recording the modified use level of the interaction engine cluster in attribute information of the interaction engine cluster. The user may modify the usage level of the interaction engine cluster labels by the voice assistant system according to his needs.

In certain embodiments, in step 302 or 402, for each historical task of the set of historical tasks, an interaction engine associated therewith in the voice assistant system of the terminal, in particular, an interaction engine that performs the historical task, is obtained.

In some embodiments, in step 302 or 402, for each historical task of the set of historical tasks, an interaction engine associated with the historical task in a voice assistant system of the terminal is obtained, specifically, at least one interaction engine capable of executing the historical task in the interaction engines included in the voice assistant system of the terminal is obtained, and one interaction engine is selected from the at least one interaction engine as the interaction engine associated with the historical task according to a second preset rule.

In some embodiments, the second preset rule is at least one of: the most frequently used interaction engine, the most recently used interaction engine, the interaction engine with the highest user score, the interaction engine with the most frequently adopted task execution results by the user, and the like.

In certain embodiments, the step 206 further comprises: updating the use level of the interaction engine cluster based on the use of the interaction engine cluster. The use of the interaction engine cluster refers to the use of the interaction engine cluster triggered by a user sending an instruction comprising a cluster name corresponding to the interaction engine cluster, or the use of the interaction engine cluster triggered by a user sending an instruction when the voice assistant system is in a cluster task working mode.

In some embodiments, after step 303 or 403, step 306 or 406 is performed to upload the interaction engine cluster to a server for other users to download, specifically, upload attribute information of the interaction engine cluster.

In some embodiments, after step 303 or 403, step 307 or 407 is performed to provide a third interface to the user, so that the user sets the attribute of the interaction engine in the interaction engine cluster through the third interface, for example, sets the dominant interaction engine of the interaction engine cluster, and records the attribute of the interaction engine in the attribute information of the interaction engine cluster from the interaction engine.

In some embodiments, after step 303 or 403, step 308 or 408 is performed to send a synchronization message to the other terminal of the user, where the synchronization message carries the aggregated interaction engine cluster, and specifically, the synchronization message carries attribute information of the interaction engine cluster.

The present invention is not limited to the order of execution of steps 304-308, and the five steps may be executed in any order, and similarly, the present invention is not limited to the order of execution of steps 404-408, and the five steps may be executed in any order.

6. Building interaction engine clusters

As previously described, the voice assistant system of the terminal may also obtain or download attribute information of the interaction engine cluster from the server and construct the interaction engine cluster locally.

Referring to FIG. 7, a method 500 of building an interaction engine cluster for a terminal having a voice assistant system in one embodiment of the invention is shown, comprising the steps of:

step 501, receiving attribute information of a first interaction engine cluster sent by a server, wherein the attribute information of the interaction engine cluster comprises names of at least two interaction engines included in the interaction engine cluster;

step 502, determining whether the voice assistant system of the terminal includes all the interaction engines included in the interaction engine cluster attribute information, and if the voice assistant system does not include the fifth interaction engine in all the interaction engines, executing step 503;

Step 503, acquiring a task associated with the fifth interaction engine;

step 504, determining whether the voice assistant system of the terminal includes a sixth interaction engine associated with the task, and if so, executing step 506;

step 506, replacing a fifth interaction engine in the attribute information of the first interaction engine cluster with a sixth interaction engine to form attribute information of a second interaction engine cluster;

and step 508, constructing an interaction engine cluster based on the attribute information of the second interaction engine cluster so as to facilitate the voice assistant system to use the interaction engine cluster.

Preferably, the attribute information of the first interaction engine cluster in step 501 may be obtained by locally aggregating the attribute information of the first interaction engine cluster by other users at the terminal thereof according to the foregoing methods 300 and/or 400 and uploading the attribute information of the first interaction engine cluster to the server, or may be preset attribute information of the interaction engine cluster by the server.

When the tasks associated with the fifth interaction engine are multiple, in step 503, the task associated with the fifth interaction engine and belonging to the first interaction engine cluster is acquired, and the above-described acquisition process may be completed based on the first interaction engine cluster attribute information (for example, table 2).

Preferably, before step 501, the method comprises the steps of

Step 201: receiving an instruction of a user;

step 201a, determining the type of the instruction, and if the instruction is a first interaction engine cluster downloading instruction, executing step 501.

Preferably, the voice assistant system uses the interaction engine cluster, specifically: when a first interaction engine in the interaction engine cluster is activated, at least one second interaction engine in the interaction engine cluster is activated, at least one first slot and/or a second slot of at least one second task associated with at least one second interaction engine in the interaction engine cluster is synchronized on the basis of key knowledge data filled into the at least one first slot of the first task associated with the first interaction engine, the second task associated with the second interaction engine has the at least one first slot and/or the second slot corresponding to the first slot, the first task execution result is received from the first interaction engine, the at least one second task execution result is received from the at least one second interaction engine, and the first task execution result and the at least one second task execution result are simultaneously provided to a user.

Preferably, the interaction engine cluster downloading instruction includes a cluster name of the interaction engine cluster.

Preferably, in step 502, if the voice assistant system of the terminal includes all interaction engines included in the first interaction engine cluster attribute information, step 509 is performed.

Preferably, in step 504, if the voice assistant system of the terminal does not include a sixth interaction engine associated with the task, steps 507 and 509 are performed.

Preferably, in step 504, if the voice assistant system of the terminal includes a sixth interaction engine associated with the task, step 505 is executed, and step 505 is executed to determine whether the slots defined by the sixth interaction engine for the task include all slots defined by the fifth interaction engine for the task, and if so, step 506 is executed.

Preferably, in step 505, if the slots defined by the sixth interaction engine for the task do not include all slots defined by the fifth interaction engine for the task, then step 507 and step 509 are executed;

step 507, downloading the fifth interaction engine from the server;

step 509, constructing an interaction engine cluster based on the first interaction engine cluster attribute information, so that the voice assistant system uses the interaction engine cluster.

Preferably, step 508 further comprises: and sending a synchronization message to other terminals of the user, wherein the synchronization message carries the constructed interaction engine cluster, and specifically, the synchronization message carries the attribute information of the second interaction engine cluster.

Step 509 further includes sending a synchronization message to other terminals of the user, where the synchronization message carries the constructed interaction engine cluster, and specifically, the synchronization message carries attribute information of the first interaction engine cluster.

Preferably, the synchronization message carries download instruction information besides the attribute information of the first interaction engine cluster, where the download instruction information includes a name of a fifth interaction engine, and instructs the other terminals to download the fifth interaction engine.

Based on the above, after receiving the synchronization message, the other terminals determine whether the synchronization message carries download instruction information, if the synchronization message carries download instruction information, then download the fifth interaction engine according to the download instruction information, and construct an interaction engine cluster, and if the synchronization message does not carry download instruction information, then the other terminals directly construct an interaction engine cluster according to the interaction engine cluster attribute information carried in the synchronization message, such as the first interaction engine cluster attribute information or the second interaction engine cluster attribute information. This is because the voice assistant systems of the different terminals of the user include the same interactive engine before downloading the cluster of interactive engines, and since the terminal performing the above-described method 500 has already performed the determinations in steps 502, 504 and 505, the other terminals do not need to repeatedly perform the above-described determination process, and the operations of the other terminals can be simplified according to whether the instruction information is downloaded in the synchronization message.

Step 507 is specifically to immediately download the fifth interaction engine from the server; or adding the fifth interaction engine in the waiting downloading list, periodically judging whether the terminal is connected to the wifi network, and if so, executing downloading according to the waiting downloading list.

Before step 509, step 507a is performed to determine whether the downloading of the fifth interaction engine is completed, and if yes, step 509 is performed.

In step 507a, the fifth interaction engine download is determined to be complete by determining that the wait for download list is empty.

The above method is described below with reference to specific examples. Assume that the first interaction engine cluster attribute information received in step 501 is as shown in table 2. In step 502, assuming that only the A1 interaction engine is included in the voice assistant system of the terminal and the B1 interaction engine is not included, the method proceeds to step 503, where the task associated with the fifth interaction engine, i.e., the B1 interaction engine, is acquired, and in this example, assuming that only one of the tasks associated with the B1 interaction engine, i.e., the weather query, is performed, it is determined in step 504 whether the voice assistant system of the terminal includes the sixth interaction engine associated with the task, and in this embodiment, assuming that the B2 interaction engine associated with the weather query is included in the voice assistant system of the terminal, step 505 is performed, the slot defined by the B1 interaction engine for the weather query task includes a city and a date, if the slots defined by the B2 interaction engine for the weather query task include city, date and time, in step 505, it is determined that the slots defined by the sixth interaction engine for the task include all slots defined by the task of the fifth interaction engine, at this time, the method is executed to step 506, where the fifth interaction engine, that is, the B1 interaction engine, in the attribute information of the first interaction engine cluster is replaced by the sixth interaction engine, that is, the B2 interaction engine, to form attribute information of the second interaction engine cluster, and in step 508, an interaction engine cluster is constructed based on the attribute information of the second interaction engine cluster, where the constructed interaction engine cluster includes an A1 interaction engine and a B2 interaction engine.

By the method, after the interaction engine cluster attribute information is downloaded from the server, the interaction engine cluster attribute information can be localized according to the interaction engine locally included in the voice assistant system of the terminal, so that unnecessary downloading of the interaction engine capable of executing the same task is avoided.

Referring to FIG. 8, a method 600 of building an interaction engine cluster for a terminal having a voice assistant system in one embodiment of the invention is shown, comprising the steps of:

step 601, receiving first interaction engine cluster configuration information sent by a server, wherein the first interaction engine cluster configuration information comprises tasks associated with interaction engines included in a first interaction engine cluster;

step 602, for each task included in the first interaction engine cluster configuration information, selecting an interaction engine associated with the task from a voice assistant system of the terminal, and putting the interaction engine into an interaction engine set;

after the interaction engine set includes the interaction engine associated with each task, step 603, an interaction engine cluster is constructed based on the interaction engines in the interaction engine set, so that the voice assistant system can use the interaction engine cluster.

Preferably, before step 601, the method comprises the steps of

Step 201: receiving an instruction of a user;

step 201a, determining the type of the instruction, and if the instruction is a download instruction of the second interaction engine cluster, executing step 601.

Preferably, for each interaction engine in the interaction engine cluster, the interaction engine cluster configuration information includes tasks associated therewith. The server generates the configuration information of the interaction engine cluster after receiving the interaction engine cluster (specifically, the attribute information of the interaction engine cluster) uploaded by the terminal, wherein the generation process of the configuration information of the interaction engine cluster is to replace the interaction engine name in the attribute information of the interaction engine cluster with the task associated with the interaction engine, for example, aiming at the attribute information of the interaction engine cluster shown in table 1, or delete the interaction engine name in the attribute information of the interaction engine cluster, for example, aiming at the attribute information of the interaction engine cluster shown in table 2. Preferably, the above process of generating the configuration information of the interaction engine cluster is applicable to a case that the interaction engine included in the interaction engine cluster is associated with only one task.

Preferably, for each interaction engine in the interaction engine cluster, the interaction engine cluster configuration information includes a task belonging to the interaction engine cluster in tasks associated with the interaction engine. The generating process of the interaction engine cluster configuration information is to delete the interaction engine name in the interaction engine cluster attribute information. Preferably, the above process of generating the configuration information of the interaction engine cluster is applicable to a case that at least one interaction engine included in the interaction engine cluster is associated with a plurality of tasks.

Preferably, when receiving the interaction engine cluster attribute information uploaded by the terminal, the server judges whether the interaction engine cluster attribute information contains the task associated with the interaction engine, if so, the server deletes the interaction engine name in the interaction engine cluster attribute information to generate interaction engine cluster configuration information, and if not, the server replaces the interaction engine name in the interaction engine cluster attribute information with the task associated with the interaction engine to generate interaction engine cluster configuration information.

Preferably, in the step 602, the interaction engine associated with the task is selected from the voice assistant system of the terminal, and put into the interaction engine set, specifically:

step 602a, judging whether the voice assistant system of the terminal comprises an interaction engine associated with the task, if so, executing step 602b; otherwise, execute step 602c and subsequent steps;

step 602b, putting the interaction engine into an interaction engine set;

step 602c, obtaining a first selectable interaction engine list of the task, providing the first selectable interaction engine list for a user, and returning to step 201;

in step 201a, if the instruction is a first download instruction, step 602d is executed; the first downloading instruction carries an interaction engine selected by a user for the task from the first selectable interaction engine list;

And step 602d, downloading the interaction engine indicated in the first downloading instruction, and putting the interaction engine into an interaction engine set.

The step 602c of obtaining the first selectable interaction engine list of the task may specifically be sending a first selectable interaction engine list obtaining request to the server, where the task is carried, and receiving the first selectable interaction engine list returned by the server.

The interaction engine indicated in the first downloading instruction in the step 602d is specifically: immediately downloading the interaction engine indicated in the first download instruction from the server; or adding an interaction engine indicated in the first downloading instruction in a waiting downloading list, periodically judging whether the terminal is accessed to a wifi network, and if the terminal is accessed to the wifi network, executing downloading according to the waiting downloading list.

Preferably, in step 603, by determining whether the number of interaction engines included in the interaction engine set is equal to the number of interaction engines included in the first interaction engine configuration set, if so, it may be determined that the interaction engine set includes an interaction engine associated with each task.

Preferably, the first selectable interaction engine list of the task acquired from the server in step 602c specifically includes: the name of the interaction engine is selected.

Preferably, the first selectable interaction engine list of the task obtained from the server in step 602c further includes: and selecting recommendation degree of the interaction engine.

Preferably, if the number of the interaction engines associated with the task in the voice assistant system of the terminal is multiple, one of the interaction engines is selected according to a fourth preset rule, and the selected interaction engine is put into the interaction engine set.

In some embodiments, the fourth preset rule is at least one of: the most frequently used interaction engine, the most recently used interaction engine, the interaction engine with the highest user score, the interaction engine with the most frequently adopted task execution results by the user, and the like.

If at least one interaction engine in the interaction engine clusters uploaded by other users defines an expansion slot for the task associated with the interaction engine, the interaction engine cluster configuration information further comprises a second selectable interaction engine list of the task for the task, wherein the slot defined for the task comprises the slot defined for the task by the at least one interaction engine. Specifically, after the server generates the configuration information of the interaction engine cluster according to the foregoing process, based on the attribute information of the interaction engine cluster, for each interaction engine in the interaction engine cluster, whether the slot defined for the associated task includes an extended slot of the task is judged, and if so, a second optional interaction engine list of the task is added in the configuration information of the interaction engine cluster. Preferably, the associated task belongs to the interaction engine cluster.

step 602a, if the configuration information of the interaction engine cluster includes a second selectable interaction engine list of the task, determining whether the voice assistant system of the terminal includes an interaction engine in the second selectable interaction engine list, and if so, executing step 602b; if the interaction engine in the second selectable interaction engine list is not included, executing step 602e and subsequent steps; if the interaction engine cluster configuration information does not include the second selectable interaction engine list of the task, judging whether the voice assistant system of the terminal includes an interaction engine associated with the task, and if so, executing step 602b; otherwise, step 602c and subsequent steps are performed.

Step 602b, putting the interaction engine into an interaction engine set.

Step 602e and subsequent steps are:

step 602e, providing the second selectable interaction engine list to the user, and returning to step 201;

in step 201a, if the instruction is a second download instruction, step 602f is executed; the second downloading instruction carries an interaction engine selected by the user for the task from the second selectable interaction engine list;

And step 602f, downloading the interaction engine indicated in the second downloading instruction, and putting the interaction engine into an interaction engine set.

Step 602c and subsequent steps are:

Preferably, the first selectable interaction engine list and the second selectable interaction engine list may include a preset number of interaction engines. The interaction engines included in the first and second selectable interaction engine lists are selected based on user scores, number of downloads, etc.

Preferably, the step 603 further includes: and sending a synchronous message to other terminals of the user, wherein the synchronous message carries the constructed interaction engine cluster, and specifically, the synchronous message carries the attribute information of the constructed interaction engine cluster.

Preferably, if the interaction engine is downloaded in the process of constructing the interaction engine cluster, the synchronization message also carries download indication information, where the download indication information includes the name of the interaction engine downloaded in the process of constructing the interaction engine cluster.

Based on the above, after receiving the synchronization message, the other terminals download the interaction engine according to the download instruction if the download instruction information is carried in the synchronization message, and construct the interaction engine cluster, if the download instruction is not carried in the synchronization message, the other terminals construct the interaction engine cluster directly according to the interaction engine cluster attribute information. This is because the voice assistant systems of the different terminals of the user include the same interaction engine before downloading the interaction engine cluster, and because the terminal performing the method 600 described above has already performed the determination in step 602, the other terminals do not need to perform the determination in step 602 any more.

The above method is described below with reference to specific examples. Assuming that the attribute information of the first interaction engine cluster uploaded by other terminals is shown in table 2, the server generates first interaction engine cluster configuration information based on the attribute information of the first interaction engine, wherein the first interaction engine cluster configuration information comprises an air ticket reservation task and a weather inquiry task. In step 601, the terminal downloads the first interaction engine cluster configuration information. Assuming that the voice assistant system of the terminal only includes an A1 interaction engine associated with the air ticket scheduled task and does not include an interaction engine associated with the weather inquiry task, in step 602, the A1 interaction engine is added to the interaction engine set for the air ticket scheduled task, the method executes to step 602c for the weather inquiry task, the terminal obtains a first selectable interaction engine list of the task, including a B1 interaction engine and a B2 interaction engine, and provides the user with the first download instruction, the user instructs to download the B1 interaction engine, then step 602d is executed, the terminal downloads the B1 interaction engine and places it into the interaction engine set, and in step 603, an interaction engine cluster is constructed based on the interaction engines in the interaction engine set.

The interaction engine associated with the weather query, which is included in the first interaction engine cluster attribute information uploaded by the other terminals, is a B2 interaction engine, and since the slot defined by the B2 interaction engine for the weather query task includes not only two basic slots, namely, a city slot and a date slot, but also one extended slot, namely, a time slot, the first interaction engine configuration information generated by the server includes a second selectable interaction engine list of the weather query task, where the second selectable interaction engine list includes the B2 interaction engine and the B3 interaction engine, and the slot defined by the B3 interaction engine for the weather query task includes two basic slots, namely, a city slot and a date slot, and also includes two extended slots, namely, a time slot and a region slot. Then in step 602a, for the air ticket reservation task, add the A1 interaction engine to the interaction engine set, for the weather inquiry task, the method will execute to step 602e, the terminal provides the user with the above second selectable interaction engine list, and downloads the interaction engine according to the second download instruction given by the user, so as to construct the interaction engine cluster in step 603.

By the method, the terminal downloads the configuration information of the interaction engine cluster from the server, and based on the configuration information of the interaction engine cluster, the interaction engine locally included in the voice assistant system is preferentially selected to construct the interaction engine cluster, so that unnecessary downloading of a plurality of interaction engines capable of executing the same task is avoided.

In the foregoing embodiments, the instruction type is determined in step 201a, and in particular, the voice assistant system may determine, for the keywords and the dialogue context included in the user instruction, and the task execution result provided to the user, the operation mode of the voice assistant system, and other factors, which type the instruction is specifically of.

The method for performing voice interaction based on the interaction engine cluster, the method for forming the interaction engine cluster by aggregation and the method for constructing the interaction engine cluster can be partially or completely combined. Referring to fig. 15, a method of forming an interaction engine cluster and performing voice interaction based on the interaction engine cluster according to the present invention is shown, and in other embodiments, the method of forming an interaction engine cluster and performing voice interaction based on the interaction engine cluster may also include all or part of the steps in the foregoing method embodiments. The invention is not limited in this regard.

The invention also provides a device for voice interaction based on the interaction engine cluster, referring to fig. 9, the device comprises:

a user instruction acquisition unit that acquires an instruction of a user;

the task execution control unit is used for determining a first task based on the instruction, determining a first interaction engine based on the first task, and sending the instruction to the first interaction engine so that the first interaction engine fills the slot position of the first task based on the instruction; wherein the first interaction engine is associated with the first task;

The work mode control unit is used for judging whether the voice assistant system is in a cluster task work mode, if so, triggering the task execution control unit, and if not, triggering the task execution result receiving unit;

the task execution control unit is further configured to determine an interaction engine cluster to which the first interaction engine belongs, where the interaction engine cluster includes the first interaction engine and at least one second interaction engine, activate the at least one second interaction engine, and synchronize, based on key knowledge data filled by the first interaction engine on a first slot of the first task, a first slot of at least one second task associated with the at least one second interaction engine in the interaction engine cluster and/or a second slot corresponding to the first slot in real time;

the task execution result receiving unit is configured to receive a first task execution result from the first interaction engine, provide the first task execution result to a user, or receive the first task execution result from the first interaction engine, receive at least one second task execution result from the at least one second interaction engine, and provide the first task execution result and the at least one second task execution result to a user at the same time.

Preferably, the device further comprises a storage unit, configured to store at least one interaction engine cluster, and specifically, store attribute information of the at least one interaction engine cluster.

Preferably, if the task execution control unit determines that the first interaction engine does not belong to any interaction engine cluster, the task execution result receiving unit is triggered to receive a first task execution result from the first interaction engine.

Preferably, the task execution control unit is further configured to determine the first interaction engine and an interaction engine cluster to which the first task belongs.

Preferably, the task execution control unit is further configured to determine, in response to the trigger of the working mode control unit, an interaction engine cluster to which the first interaction engine belongs, determine whether the first interaction engine is a dominant interaction engine of the interaction engine cluster, and if yes, activate the at least one second interaction engine, and perform instant synchronization on a first slot and/or a second slot of at least one second task associated with the at least one second interaction engine in the interaction engine cluster based on key knowledge data that the first interaction engine fills the first slot of the first task; otherwise, triggering the task execution result receiving unit.

The task execution result receiving unit is further configured to trigger the user instruction acquiring unit after the first task execution result is provided to the user, or after the first task execution result and the at least one second task execution result are provided to the user at the same time.

The user instruction acquisition unit is further used for responding to the trigger of the task execution result receiving unit and continuously receiving the user instruction.

Preferably, the device further includes an instruction classifying unit, configured to determine a type of the user instruction acquired by the user instruction acquiring unit, and trigger the task execution control unit if the instruction is a new task instruction. The task execution control unit is further used for responding to the trigger of the instruction classification unit when the instruction is a new task instruction, determining a first task based on the instruction, determining a first interaction engine based on the first task, and sending the instruction to the first interaction engine so that the first interaction engine fills the slot of the first task based on the instruction.

The instruction classifying unit is further configured to: and if the instruction is a single task modification instruction, triggering the task execution control unit. The task execution control unit is further used for responding to the triggering of the instruction classification unit when the instruction is a single task modification instruction, and sending the instruction to the first interaction engine so that the first interaction engine modifies key knowledge data filled in the slots of the first task based on the instruction.

The task execution result receiving unit is further configured to obtain a first task modification execution result from the first interaction engine, provide the first task modification execution result to a user, and trigger the user instruction obtaining unit.

The instruction classifying unit is further configured to: and if the instruction is a cluster task modification instruction, triggering the task execution control unit. The task execution control unit is further configured to determine a third interaction engine corresponding to the cluster task modification instruction according to the triggering of the instruction classification unit when the instruction is the cluster task modification instruction, and send the instruction to the third interaction engine, so that the third interaction engine modifies a third slot of a third task associated with the third interaction engine based on the instruction; and carrying out instant synchronization on the third slot position of at least one fourth task associated with a fourth interaction engine in the interaction engine cluster and/or a fourth slot position corresponding to the third slot position based on the modified key knowledge data filled by the third slot position of the third task.

The task execution result receiving unit is further configured to receive the third task execution result from the third interaction engine, receive at least one fourth task execution result from the at least one fourth interaction engine, and provide the third task execution result and the at least one fourth task execution result to a user at the same time, so as to trigger the user instruction obtaining unit.

The task execution result receiving unit receives at least one task execution result from at least one interaction engine, which may be a first interaction engine and/or a second interaction engine, as a transient task execution result. And the task execution result receiving unit triggers the user instruction acquisition unit after providing the task execution result to the user. And triggering the instruction classifying unit after the user instruction acquiring unit acquires the instruction of the user again.

The instruction classifying unit is further configured to: and if the instruction is a task confirmation instruction, triggering the task execution control unit. The task execution control unit is further configured to send the task confirmation instruction to the at least one interaction engine in response to the trigger of the instruction classification unit when the instruction is a task confirmation instruction, so as to trigger the at least one interaction engine to convert the transient task execution result into a final state task execution result.

The task execution result receiving unit is further configured to receive a final task execution result from the at least one interaction engine, provide the final task execution result to a user, and trigger the user instruction obtaining unit.

The task execution control unit is further configured to determine whether a cluster task clarity question is received from the first interaction engine after the immediate synchronization, if yes, provide the cluster task clarity question to a user, and trigger the user instruction acquisition unit, otherwise trigger the task execution result receiving unit to receive a first task execution result from the first interaction engine and receive at least one second task execution result from the at least one second interaction engine.

The user instruction acquisition unit is further used for responding to the trigger of the task execution control unit and continuously receiving the user instruction.

The instruction classifying unit is further configured to: if the instruction is an instruction for responding to the clarity question of the cluster task, triggering the task execution control unit; the task execution control unit is further configured to send an instruction to a first interaction engine in response to a trigger of the instruction classification unit when the instruction is an instruction for the task clarity question response of the cluster, so that the first interaction engine fills a slot of the first task according to the instruction for the task clarity question response of the cluster, and immediately synchronizes a first slot and/or a second slot of at least one second task associated with at least one second interaction engine in the interaction engine cluster based on key knowledge data filled by the first slot of the first task by the first interaction engine.

The task execution control unit is further configured to determine, when it is determined that the voice assistant system is not in the cluster task working mode, whether to receive a single task clarity question from the first interaction engine, if yes, provide the single task clarity question to a user, and trigger the user instruction acquisition unit, otherwise trigger the task execution result receiving unit to receive a first task execution result from the first interaction engine.

The instruction classifying unit is further configured to: if the instruction is an instruction for responding to the single-task clarity question, triggering the task execution control unit; the task execution control unit is further used for responding to the triggering of the instruction classification unit when the instruction is the instruction responding to the single task clarity question, and sending the instruction to a first interaction engine so that the first interaction engine fills the slot position of the first task according to the instruction responding to the single task clarity question.

Preferably, the task execution control unit is further configured to determine whether a single task clarity question is received from the first interaction engine if it is determined that the first interaction engine does not belong to any interaction engine cluster, or determine that the first interaction engine is not a dominant interaction engine of the interaction engine cluster, if yes, provide the single task clarity question to a user, and trigger the user instruction obtaining unit, otherwise trigger the task execution result receiving unit to receive a first task execution result from the first interaction engine.

The task execution control unit is further configured to select to use one of the interaction engine clusters or use the plurality of interaction engine clusters according to a third preset rule when the first interaction engine or the first interaction engine and the interaction engine clusters to which the first task belongs are multiple.

The invention also provides a device for voice interaction based on the interaction engine cluster, referring to fig. 10, the device comprises:

a user instruction acquisition unit that acquires an instruction of a user;

the instruction classification unit is used for judging the type of the user instruction, and triggering the working mode control unit if the instruction is a cluster task working mode changing instruction;

The working mode control unit is used for modifying the working mode of the voice assistant system into a cluster task working mode according to the cluster task working mode changing instruction;

The device also comprises a task execution control unit and a task execution result receiving unit;

the instruction classification unit is further used for judging whether the last instruction of the working mode changing instruction is a new task instruction, and if so, triggering the task execution control unit;

the task execution control unit is used for processing the last instruction of the user by using an interaction engine cluster;

and the task execution result receiving unit is used for simultaneously providing at least two task execution results provided by at least two interaction engines included in the interaction engine cluster to a user.

The task execution control unit processes the last instruction of the user by using an interaction engine cluster, specifically, determines a first task based on a new task instruction in response to the trigger of the instruction classification unit when the last instruction is the new task instruction, and determines a first interaction engine based on the first task, wherein the first interaction engine is associated with the first task; the new task instruction is sent to the first interaction engine, so that the first interaction engine fills the slot of the first task based on the new task instruction; determining an interaction engine cluster to which the first interaction engine belongs, wherein the interaction engine cluster comprises the first interaction engine and at least one second interaction engine; activating the at least one second interaction engine; and carrying out instant synchronization on the first slot position of at least one second task associated with at least one second interaction engine in the interaction engine cluster and/or the second slot position corresponding to the first slot position based on key knowledge data filled in the first slot position of the first task by the first interaction engine.

The task execution result receiving unit is configured to receive the first task execution result from the first interaction engine, receive at least one second task execution result from the at least one second interaction engine, and provide the first task execution result and the at least one second task execution result to a user at the same time.

Preferably, the device further comprises a storage unit, configured to store at least one interaction engine cluster, and in particular, store attribute information of the at least one interaction engine cluster.

The same unit in the device as the one shown in fig. 9 has the same function as the unit in fig. 9, and the preferred embodiment of any unit of the device shown in fig. 9 is also applicable to the same unit in the device in the embodiment. For example, the task execution control unit in this embodiment has the same function as that of the task execution control unit described in fig. 9, and the preferred embodiment of the task execution control unit shown in fig. 9 is also applicable to the task execution control unit in this embodiment, as is the case for other units having the same names, and is not described here again.

Preferably, the task execution control unit is configured to process the last instruction of the user by using an interaction engine cluster, specifically: responding to the trigger of the instruction classification unit when the last instruction is a new task instruction, receiving a first message from the task execution result receiving unit, wherein the first message carries an instruction for judging whether the new task instruction provides a task execution result for a user or not, if so, determining an interaction engine cluster to which the first interaction engine belongs, activating the at least one second interaction engine, and carrying out instant synchronization on the first slot position and/or the second slot position of at least one second task associated with the at least one second interaction engine in the interaction engine cluster based on key knowledge data filled in the first slot position of the first task by the first interaction engine; otherwise, determining a first task based on the new task instruction, determining a first interaction engine based on the first task, and sending the new task instruction to the first interaction engine so that the first interaction engine fills the slots of the first task based on the new task instruction, determining an interaction engine cluster to which the first interaction engine belongs, activating the at least one second interaction engine, and performing instant synchronization on the first slots and/or the second slots of at least one second task associated with the at least one second interaction engine in the interaction engine cluster based on key knowledge data filled by the first interaction engine on the first slots of the first task;

The task execution result receiving unit is further configured to send the first message to the task execution control unit.

Preferably, the task execution control unit is configured to determine whether a receiving time of a new task instruction meets a preset condition in response to a trigger of the instruction classification unit when the previous instruction is the new task instruction, and if so, process the previous instruction of the user by using an interaction engine cluster.

The task execution control unit is further configured to trigger the user instruction acquisition unit if it is determined that the receiving time of the new task instruction does not meet a preset condition.

The device also comprises a synchronization unit which is used for sending a synchronization message to other terminals of the user and indicating a voice assistant system on the other terminals of the user to modify the working mode of the voice assistant system into a cluster task working mode.

The invention also provides a device for voice interaction based on the interaction engine cluster, referring to fig. 11, the device comprises:

the user instruction acquisition unit acquires an instruction of a user, wherein the instruction carries a new task instruction;

the instruction classification unit is used for judging the type of the user instruction, and triggering the task execution control unit if the instruction comprises an interaction engine cluster enabling instruction;

The task execution control unit is used for enabling the interaction engine cluster corresponding to the cluster name, and processing the new task instruction based on the interaction engine cluster corresponding to the cluster name carried in the interaction engine cluster enabling instruction according to the new task instruction carried in the instruction;

and the task execution result receiving unit is used for simultaneously providing at least two task execution results provided by at least two interaction engines included in the interaction engine cluster for a user.

The same unit name as in the device of the two embodiments (the embodiments corresponding to fig. 9 to 10) in the device of the present embodiment also has the functions of the unit of the two embodiments, and the preferred embodiment regarding any unit of the device of the two embodiments also applies to the same unit in the device of the present embodiment. For example, the task execution control unit in this embodiment has the same function as that of the task execution control units in the foregoing two embodiments, and the preferred embodiments of the task execution control units in the foregoing two embodiments are also applicable to the task execution control units in this embodiment, and the same is true for other units with the same names, which are not described herein again.

Preferably, the instruction classifying unit is further configured to determine whether the instruction includes a new task instruction, and if yes, trigger the task execution control unit, and if not, trigger the user instruction obtaining unit.

The task execution control unit is further configured to respond to a trigger when the instruction classification unit includes a new task instruction in the instruction, and process the new task instruction based on the interaction engine cluster corresponding to the cluster name according to the new task instruction carried in the instruction.

The task execution control unit is further configured to respond to a trigger when the instruction classification unit includes a new task instruction in the instruction, determine a first task based on the new task instruction, determine a first interaction engine associated with the first task based on the first task, determine whether the first interaction engine belongs to the interaction engine cluster, if yes, process the new task instruction based on the interaction engine cluster corresponding to the cluster name according to the new task instruction carried in the instruction, otherwise prompt a user that the new task instruction is not matched with the cluster name, and trigger the user instruction acquisition unit.

The task execution control unit is further configured to, in response to a trigger when the instruction classification unit includes a new task instruction in the instruction, determine a first task based on the new task instruction, determine a first interaction engine associated with the first task based on the first task, determine whether the first interaction engine belongs to the interaction engine cluster, if the first interaction engine belongs to the interaction engine cluster, further determine whether the first interaction engine is a dominant interaction engine of the interaction engine cluster, if the first interaction engine is the dominant interaction engine of the interaction engine cluster, process the new task instruction based on an interaction engine cluster corresponding to the cluster name according to the new task instruction carried in the instruction, if the new task instruction is not the dominant interaction engine, prompt a user of an interaction engine non-dominant interaction engine associated with the new task instruction, and trigger the user instruction acquisition unit; if the first interaction engine does not belong to the interaction engine cluster, prompting a user that a new task instruction is not matched with the cluster name, and triggering the user instruction acquisition unit.

The task execution control unit is configured to process a new task instruction based on an interaction engine cluster corresponding to the cluster name according to the new task instruction carried in the instruction, and specifically includes: the new task instruction is sent to the first interaction engine, so that the first interaction engine fills the slot of the first task based on the instruction; the first interaction engine is associated with a first task, which is determined based on the new task instruction; activating at least one second interaction engine of the cluster of interaction engines; the at least one second interaction engine is other interaction engines in the interaction engine cluster than the first interaction engine; and carrying out instant synchronization on the first slot position of at least one second task associated with at least one second interaction engine in the interaction engine cluster and/or the second slot position corresponding to the first slot position based on key knowledge data filled in the first slot position of the first task by the first interaction engine.

And the task execution result receiving unit is used for receiving the first task execution result from the first interaction engine, receiving at least one second task execution result from the at least one second interaction engine and simultaneously providing the first task execution result and the at least one second task execution result for a user.

And the task execution control unit is further used for judging whether a clarity question is received from the first interaction engine after the instant synchronization, if so, providing the clarity question for a user, and triggering the user instruction acquisition unit.

The instruction classifying unit is further configured to: if the instruction is the instruction aiming at the clarity question response, triggering the task execution control unit; the task execution control unit is further configured to send the instruction to a first interaction engine in response to the trigger of the instruction classification unit when the instruction is the instruction for the clarity question response, so that the first interaction engine fills the slot of the first task according to the instruction for the clarity question response, and timely synchronizes the first slot and/or the second slot of at least one second task associated with at least one second interaction engine in the interaction engine cluster based on key knowledge data filled by the first interaction engine for the first slot of the first task.

The instruction classifying unit is further configured to determine, when the user instruction is a new task instruction, whether a previous instruction of the new task instruction is an interaction engine cluster enabling instruction, and if yes, trigger the task execution control unit.

The task execution control unit is used for responding to the trigger of the instruction classification unit when the last instruction is an interaction engine cluster enabling instruction, and processing the new task instruction according to the mode.

In another embodiment, the device for performing voice interaction based on the interaction engine cluster comprises all or part of the units of the device in the three embodiments.

The present invention also provides an apparatus for aggregating to form an interaction engine cluster, see fig. 12, comprising the following units:

Preferably, the historical task set generating unit is configured to select at least two historical tasks according to a first preset rule and a historical dialogue record.

Preferably, the history task set generating unit is configured to select at least two history tasks executed in a preset time in the history dialogue record; and/or selecting at least two history tasks executed in a preset dialog turn number in the history dialog record; and/or selecting at least two history tasks which are adjacently performed in the history dialogue record. The at least two history tasks being adjacently performed means that no other history task is performed between the at least two history tasks.

Preferably, the aggregation unit is further configured to determine, after aggregating the at least two interaction engines to form an interaction engine cluster, one or more interaction engines in the interaction engine cluster as dominant interaction engines according to the sequence of the at least two historical tasks.

The invention also provides a device for forming the interaction engine cluster by aggregation, which comprises the following units:

the history task set generating unit acquires a history dialogue record of the user and the voice assistant system and judges whether the history dialogue record has the following conditions: interrupting a dialogue related to a first task, entering a dialogue related to at least one second task, and restoring the dialogue related to the first task after the at least one second task is executed; if so, forming a historical task set, wherein the historical task set comprises the first task and at least one second task;

Preferably, the historical task set generating unit determines whether the historical dialog records have the following conditions: interrupting a dialogue related to a first task, entering a dialogue related to at least one second task, and recovering the dialogue related to the first task after the at least one second task is executed, wherein the method specifically comprises the following steps: analyzing the history dialogue record to obtain a corresponding history task execution time line and a history task execution state; judging whether a historical task closed loop exists or not based on the historical task execution time line and the historical task execution state; the history task closed loop is a history task sequence with a first task at the head and the tail nodes, at least one second task at the middle node, incomplete execution state of the head node and complete execution state of the at least one middle node.

Preferably, the aggregation unit is further configured to determine, after aggregating the at least two interaction engines to form one interaction engine cluster, an interaction engine associated with the first task as a dominant interaction engine.

The following applies to the means for aggregating to form clusters of interaction engines in the two embodiments described above.

And the aggregation unit is used for judging whether the slots defined by at least two interaction engines in the interaction engine set aiming at the corresponding historical tasks have the same or corresponding slots or not based on the instructions of the users in the historical dialogue record.

The aggregation unit is configured to determine key knowledge data of at least one slot defined by an interaction engine associated with a history task corresponding to an instruction in the history dialogue record for the history task based on the instruction of the user, and determine, based on the slot and the key knowledge data, whether at least two interaction engines in the interaction engine set have the same or corresponding slots in the slots defined by the interaction engine for the associated history task.

The aggregation unit aggregates the at least two interaction engines into an interaction engine cluster, specifically: and recording names of the at least two interaction engines in attribute information of the interaction engine cluster, and recording the same or corresponding slots in slots defined by the at least two interaction engines in the interaction engine set aiming at the associated historical task in the attribute information of the interaction engine cluster.

The voice assistant system uses the interaction engine cluster, specifically: when a first interaction engine in the interaction engine cluster is activated, at least one second interaction engine in the interaction engine cluster is activated, at least one first slot and/or a second slot of at least one second task associated with at least one second interaction engine in the interaction engine cluster is synchronized on the basis of key knowledge data filled into the at least one first slot of the first task associated with the first interaction engine, the second task associated with the second interaction engine has the at least one first slot and/or the second slot corresponding to the first slot, the first task execution result is received from the first interaction engine, the at least one second task execution result is received from the at least one second interaction engine, and the first task execution result and the at least one second task execution result are simultaneously provided to a user.

The voice assistant system uses the interaction engine cluster, specifically: the voice assistant system uses the interaction engine cluster when the first interaction engine determined based on the user instruction is in the cluster task work mode.

The voice assistant system uses the interaction engine cluster, specifically: the voice assistant system is in a cluster task working mode, and uses the interaction engine cluster when a first interaction engine and a first task determined based on a user instruction belong to the interaction engine cluster.

The voice assistant system uses the interaction engine cluster, specifically: and when the received user instruction comprises the cluster name of the interaction engine cluster, the interaction engine cluster is used.

The device for aggregating to form the interaction engine cluster in the above two embodiments may include all or part of the units of the device for performing voice interaction based on the interaction engine cluster in the three embodiments described before the above two embodiments, which are not described herein.

The units with the same names in the devices for aggregating and forming the interaction engine clusters in the two embodiments may have the functions in the two embodiments at the same time.

The apparatus for aggregating to form an interaction engine cluster further comprises:

a user instruction acquisition unit that acquires an instruction of a user;

the instruction classification unit is used for judging the type of the user instruction, and triggering the historical task set generation unit if the instruction is an interaction engine cluster aggregation instruction.

The aggregation unit is further configured to provide a first interface for a user, so that the user names the interaction engine cluster through the first interface, and records the cluster name input by the user in attribute information of the interaction engine cluster.

The device for forming the interaction engine cluster by aggregation further comprises a storage unit, wherein the storage unit is used for storing the interaction engine cluster formed by aggregation, and particularly storing attribute information of the interaction engine cluster formed by aggregation.

And the aggregation unit is further configured to aggregate the at least two interaction engines to form an interaction engine cluster, mark the use level of the interaction engine cluster as the lowest, and record the use level of the interaction engine cluster in attribute information thereof.

The apparatus for aggregating to form an interaction engine cluster further comprises: a statistics unit for performing statistics based on a plurality of historical task sets generated by the historical dialogue record;

The aggregation unit is further configured to mark a usage level of the interaction engine cluster according to a frequency of the historical task set in the statistics result.

The aggregation unit is further configured to provide a second interface for a user, so that the user modifies the usage level of the interaction engine cluster through the second interface, and records the usage level of the interaction engine cluster after modification in attribute information of the usage level.

The interaction engine set generating unit obtains, for each history task of the history task set, an interaction engine associated with the history task in a voice assistant system of the terminal, specifically, an interaction engine for executing the history task.

The interaction engine set generating unit acquires, for each history task of the history task set, an interaction engine associated with the history task in a voice assistant system of the terminal, specifically, acquires at least one interaction engine capable of executing the history task in the interaction engines included in the voice assistant system of the terminal, and selects one interaction engine from the at least one interaction engine as the interaction engine associated with the history task according to a second preset rule.

The aggregation unit is further used for updating the use level of the interaction engine cluster based on the use of the interaction engine cluster by the task execution control unit.

The apparatus for aggregating to form an interaction engine cluster further comprises: and the uploading unit is used for uploading the interaction engine cluster to a server for downloading by other users, and particularly uploading the attribute information of the interaction engine cluster.

The aggregation unit is further configured to provide a third interface for a user, so that the user sets an attribute of an interaction engine in the interaction engine cluster through the third interface, for example, sets a dominant interaction engine of the interaction engine cluster and a slave interaction engine of the interaction engine cluster, and records the attribute of the interaction engine in attribute information of the interaction engine cluster.

The apparatus for aggregating to form an interaction engine cluster further comprises: and the synchronization unit is used for sending a synchronization message to other terminals of the user, wherein the synchronization message carries the interaction engine cluster obtained by aggregation, and specifically, the synchronization message carries attribute information of the interaction engine cluster.

The invention also provides a device for constructing the interaction engine cluster, see fig. 13, comprising the following units:

The first acquisition unit is used for receiving attribute information of a first interaction engine cluster sent by a server, wherein the attribute information of the interaction engine cluster comprises names of at least two interaction engines included in the interaction engine cluster;

the construction unit is used for judging whether the voice assistant system of the terminal comprises all interaction engines included in the interaction engine cluster, and if the voice assistant system does not comprise a fifth interaction engine in all the interaction engines, acquiring a task associated with the fifth interaction engine; judging whether the voice assistant system of the terminal comprises a sixth interaction engine associated with the task, if the voice assistant system of the terminal comprises the sixth interaction engine associated with the task, replacing a fifth interaction engine in the attribute information of the first interaction engine cluster by the sixth interaction engine to form attribute information of a second interaction engine cluster, and constructing an interaction engine cluster based on the attribute information of the second interaction engine cluster so that the voice assistant system can use the interaction engine cluster.

The device for constructing the interaction engine cluster in this embodiment may include all or part of the units of the device for aggregating to form the interaction engine cluster and the device for performing voice interaction based on the interaction engine cluster in the foregoing five embodiments, which are not described herein again.

Preferably, the apparatus further comprises:

a user instruction acquisition unit that acquires an instruction of a user;

the instruction classifying unit is used for judging the type of the user instruction, and triggering the first acquiring unit if the instruction is a first interaction engine cluster downloading instruction.

Preferably, the construction unit is further configured to construct an interaction engine cluster based on the attribute information of the first interaction engine cluster if the voice assistant system includes all interaction engines included in the interaction engine cluster, so that the voice assistant system uses the interaction engine cluster.

Preferably, the construction unit is further configured to, if the voice assistant system of the terminal does not include the sixth interaction engine associated with the task, download the fifth interaction engine from the server, and construct an interaction engine cluster based on the attribute information of the first interaction engine cluster, so that the voice assistant system uses the interaction engine cluster.

Preferably, the construction unit is further configured to determine whether the slots defined by the sixth interaction engine for the task include all slots defined by the fifth interaction engine for the task if the voice assistant system of the terminal includes a sixth interaction engine associated with the task, and if so, replace the fifth interaction engine in the first interaction engine cluster attribute information with the sixth interaction engine to form second interaction engine cluster attribute information, and construct an interaction engine cluster based on the second interaction engine cluster attribute information, so that the voice assistant system uses the interaction engine cluster.

Preferably, the device further includes a synchronization unit, configured to send a synchronization message to other terminals of the user, where the synchronization message carries the constructed interaction engine cluster, and specifically, the synchronization message carries attribute information of the second interaction engine cluster.

Preferably, the building unit is further configured to download the fifth interaction engine from the server if the slots defined by the sixth interaction engine for the task do not include all slots defined by the fifth interaction engine for the task, and build an interaction engine cluster based on the attribute information of the first interaction engine cluster after the fifth interaction engine is downloaded, so that the voice assistant system uses the interaction engine cluster.

Preferably, the synchronization unit is further configured to send a synchronization message to other terminals of the user, where the synchronization message carries the constructed interaction engine cluster, and specifically, the synchronization message carries attribute information of the first interaction engine cluster.

Preferably, the building unit downloads the fifth interaction engine from the server, specifically: immediately downloading the fifth interaction engine from the server; or adding the fifth interaction engine in the waiting downloading list, periodically judging whether the terminal is connected to the wifi network, and if so, executing downloading according to the waiting downloading list.

Preferably, the construction unit judges whether the fifth interaction engine is downloaded, and if so, constructs an interaction engine cluster based on the attribute information of the first interaction engine cluster.

Preferably, the construction unit determines that the fifth interaction engine downloading is completed by determining that the waiting for downloading list is empty.

The invention also provides a device for constructing the interaction engine cluster, see fig. 14, comprising the following units:

the second acquisition unit is used for receiving first interaction engine cluster configuration information sent by the server, wherein the first interaction engine cluster configuration information comprises tasks associated with interaction engines included in the first interaction engine cluster;

and the construction unit is used for selecting an interaction engine associated with each task from the voice assistant system of the terminal aiming at each task included in the first interaction engine cluster configuration information, putting the interaction engine into an interaction engine set, and constructing an interaction engine cluster based on the interaction engines in the interaction engine set after the interaction engine associated with each task is included in the interaction engine set so as to facilitate the voice assistant system to use the interaction engine cluster.

The device for constructing an interaction engine cluster in this embodiment may include all or part of the units of the device for aggregating to form an interaction engine cluster, the device for performing voice interaction based on the interaction engine cluster, and the device for constructing an interaction engine cluster in the foregoing six embodiments, which are not described herein.

Preferably, the apparatus further comprises:

a user instruction acquisition unit that acquires an instruction of a user;

the instruction classifying unit is used for judging the type of the user instruction, and triggering the second acquiring unit if the instruction is a downloading instruction of the second interaction engine cluster.

Preferably, the construction unit is configured to: judging whether a voice assistant system of the terminal comprises an interaction engine associated with each task contained in the first interaction engine cluster configuration information, if so, putting the interaction engine into an interaction engine set; otherwise, acquiring a first selectable interaction engine list of the task, providing the first selectable interaction engine list for a user, and triggering the instruction classification unit;

the instruction classifying unit is used for triggering the constructing unit if the instruction is a first downloading instruction; the first downloading instruction carries an interaction engine selected by a user for the task from the first selectable interaction engine list;

The construction unit is used for responding to the triggering of the instruction classification unit when the instruction is a first downloading instruction, downloading the interaction engine indicated in the first downloading instruction and putting the interaction engine into the interaction engine set.

The construction unit obtains a first selectable interaction engine list of the task, which may specifically be sending a first selectable interaction engine list obtaining request to the server, where the task is carried with the first selectable interaction engine list, and receiving the first selectable interaction engine list returned by the server.

The construction unit downloads the interaction engine indicated in the first downloading instruction, specifically: immediately downloading the interaction engine indicated in the first download instruction from the server; or adding an interaction engine indicated in the first downloading instruction in a waiting downloading list, periodically judging whether the terminal is accessed to a wifi network, and if the terminal is accessed to the wifi network, executing downloading according to the waiting downloading list.

The construction unit may determine that the interaction engine set includes the interaction engine associated with each task by determining whether the number of interaction engines included in the interaction engine set is equal to the number of interaction engines included in the first interaction engine configuration set, if so.

The construction unit is further configured to, if there are multiple interaction engines associated with the task in the voice assistant system of the terminal, select one of the interaction engines according to a fourth preset rule, and put the selected interaction engine into the interaction engine set.

Preferably, the construction unit is further configured to: for each task included in the first interaction engine cluster configuration information, if the interaction engine cluster configuration information includes a second optional interaction engine list of the task, judging whether a voice assistant system of the terminal includes an interaction engine in the second optional interaction engine list, and if so, putting the interaction engine into an interaction engine set; if the interaction engine in the second selectable interaction engine list is not included, providing the second selectable interaction engine list for a user, and triggering the instruction classification unit; if the configuration information of the interaction engine cluster does not comprise a second optional interaction engine list of the task, judging whether a voice assistant system of the terminal comprises an interaction engine associated with the task, if so, putting the interaction engine into an interaction engine set, otherwise, providing the first optional interaction engine list for a user, and triggering the instruction classification unit;

The instruction classifying unit is used for triggering the constructing unit if the instruction is a first downloading instruction or a second downloading instruction; the first downloading instruction carries an interaction engine selected by a user for the task from the first selectable interaction engine list; the second downloading instruction carries an interaction engine selected by the user for the task from the second selectable interaction engine list;

the construction unit is used for responding to the triggering of the instruction classification unit when the instruction is a first downloading instruction, downloading the interaction engine indicated in the first downloading instruction and putting the interaction engine into an interaction engine set; and responding to the trigger of the instruction classifying unit when the instruction is a second downloading instruction, downloading the interaction engine indicated in the second downloading instruction, and putting the interaction engine into an interaction engine set.

The apparatus further comprises: and the synchronization unit is used for sending a synchronization message to other terminals of the user, wherein the synchronization message carries the constructed interaction engine cluster, and specifically, the synchronization message carries the attribute information of the constructed interaction engine cluster.

The seven devices in the seven embodiments described above may be freely combined, any of which may include all or part of the units in the remaining six.

The invention also provides a terminal characterized by comprising the device as described above or a voice assistant system as described above.

The voice assistant system or the terminal may include all or part of the means for aggregating to form the interaction engine cluster, the means for performing voice interaction based on the interaction engine cluster, and the means for constructing the interaction engine cluster in the foregoing seven embodiments.

Any combination of one or more computer readable media may be employed. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. The computer readable storage medium may include: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), a flash memory, an erasable programmable read-only memory (EPROM), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

Computer program code for carrying out operations of the present invention may be written in one or more programming languages or combinations thereof.

The above description is only an example for the convenience of understanding the present invention, and is not intended to limit the scope of the present invention. In the specific implementation, the person skilled in the art may change, increase, decrease the components of the apparatus according to the actual situation, and may change, increase, decrease or change the order of the steps of the method according to the actual situation on the basis of not affecting the functions implemented by the method.

While embodiments of the invention have been shown and described, it will be understood by those skilled in the art that: various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents, and modifications which are not to be construed as being within the scope of the invention.

Claims

1. A method of aggregating clusters of interaction engines for a terminal having a voice assistant system, the method comprising:

2. The method according to claim 1, wherein in step 303, after the at least two interaction engines are aggregated to form one interaction engine cluster, one or more interaction engines in the interaction engine cluster are determined to be dominant interaction engines according to the order of the at least two history tasks.

3. The method of claim 1, wherein the step of determining the position of the substrate comprises,

in step 303, based on the user's instruction in the history dialogue record, it is determined whether at least two interaction engines in the set of interaction engines have the same or corresponding slots in the slots defined for the corresponding history task.

4. The method of claim 1, wherein the step of determining the position of the substrate comprises,

in step 303, the at least two interaction engines are aggregated to form an interaction engine cluster, specifically: and recording names of the at least two interaction engines in attribute information of the interaction engine cluster, and recording the same or corresponding slots in slots defined by the at least two interaction engines in the interaction engine set aiming at the associated historical task in the attribute information of the interaction engine cluster.

5. The method according to any of claims 1-4, characterized in that after step 303, step 306 is performed, wherein the interaction engine cluster is uploaded to a server for download by other users.

6. An apparatus for aggregating clusters of interaction engines, the apparatus comprising:

7. The apparatus of claim 6, wherein the aggregation unit is further configured to determine one or more interaction engines in the interaction engine cluster as dominant interaction engines based on a sequencing of the at least two historical tasks after aggregating the at least two interaction engines to form one interaction engine cluster.

8. The apparatus of claim 6, wherein the aggregation unit is configured to determine whether at least two interaction engines in the set of interaction engines have the same or corresponding slots in slots defined for corresponding historical tasks based on instructions of a user in a historical dialog record.

9. The apparatus according to claim 6, wherein the aggregation unit aggregates the at least two interaction engines into one interaction engine cluster, specifically: and recording names of the at least two interaction engines in attribute information of the interaction engine cluster, and recording the same or corresponding slots in slots defined by the at least two interaction engines in the interaction engine set aiming at the associated historical task in the attribute information of the interaction engine cluster.

10. The apparatus according to any one of claims 6-9, wherein the apparatus further comprises: and the uploading unit is used for uploading the interaction engine cluster to a server for downloading by other users.

11. A computer device, characterized in that it comprises a processor and a memory, in which a computer program is stored which is executable on the processor and which, when executed by the processor, implements the method according to any of claims 1 to 5.

12. A computer readable storage medium, characterized in that it has stored therein a computer program executable on a processor, which when executed implements the method according to any of claims 1 to 5.

13. A voice assistant system comprising the apparatus of any of claims 6 to 10.

14. A terminal comprising the apparatus of any of claims 6 to 10 or comprising the voice assistant system of claim 13.