WO2024036804A1 - 意图指令的确定方法及装置、存储介质及电子装置 - Google Patents
意图指令的确定方法及装置、存储介质及电子装置 Download PDFInfo
- Publication number
- WO2024036804A1 WO2024036804A1 PCT/CN2022/134786 CN2022134786W WO2024036804A1 WO 2024036804 A1 WO2024036804 A1 WO 2024036804A1 CN 2022134786 W CN2022134786 W CN 2022134786W WO 2024036804 A1 WO2024036804 A1 WO 2024036804A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- intention
- instruction
- query vector
- home appliance
- intent
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 230000007704 transition Effects 0.000 claims abstract description 77
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 15
- 230000008859 change Effects 0.000 claims abstract description 8
- 239000013598 vector Substances 0.000 claims description 140
- 238000010586 diagram Methods 0.000 claims description 56
- 230000009471 action Effects 0.000 claims description 36
- 230000015654 memory Effects 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 16
- 230000007613 environmental effect Effects 0.000 claims description 16
- 230000008569 process Effects 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 description 10
- 238000012546 transfer Methods 0.000 description 8
- 230000005540 biological transmission Effects 0.000 description 6
- 238000013135 deep learning Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 238000013136 deep learning model Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005406 washing Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 241000475481 Nebula Species 0.000 description 1
- 238000004378 air conditioning Methods 0.000 description 1
- 238000004887 air purification Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 238000010438 heat treatment Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 239000004984 smart glass Substances 0.000 description 1
- 238000010408 sweeping Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B15/00—Systems controlled by a computer
- G05B15/02—Systems controlled by a computer electric
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/20—Pc systems
- G05B2219/26—Pc applications
- G05B2219/2642—Domotique, domestic, home control, automation, smart house
Definitions
- the present disclosure relates to the field of smart home technology, and specifically, to a method and device for determining intent instructions, a storage medium, and an electronic device.
- Embodiments of the present disclosure provide a method and device, a storage medium, and an electronic device for determining an intention instruction, so as to at least solve the problems of slow recognition speed and low recognition accuracy of existing intention recognition solutions in the related art.
- a method for determining an intention instruction including: generating a plurality of first intention nodes of the home appliance according to a state transition diagram of the home appliance, wherein the state transition diagram is used to indicate The transfer relationship between the states of the home appliance equipment, the plurality of first intention nodes are respectively used to store a plurality of first states of the home appliance equipment, and are used to instruct the home appliance equipment to change the second state at the current moment to The first intention instruction of the first state; use the association algorithm to calculate the state transition diagram to obtain the association relationship between the multiple first states, wherein the association relationship includes: the multiple first states The sequence relationship between them and the weight of multiple first states; in the case of receiving a voice instruction sent by the target object for controlling the home appliance device, determined according to the multiple first intention nodes and the association relationship.
- the first intention instruction corresponding to the voice instruction.
- an apparatus for determining intent instructions including: a generation module configured to generate a plurality of first intent nodes of the home appliance device according to the state transition diagram of the home appliance device, wherein, The state transition diagram is used to indicate the transition relationship between the states of the home appliance equipment.
- the plurality of first intention nodes are respectively used to store a plurality of first states of the home appliance equipment, and are used to indicate that the home appliance equipment will The second state at the current moment is changed to the first intention instruction of the first state;
- the calculation module is configured to use an association algorithm to calculate the state transition diagram to obtain the association relationship between the multiple first states, wherein,
- the association relationship includes: a sequence relationship between the plurality of first states and a weight of the plurality of first states; a determination module configured to receive a voice command sent by the target object for controlling the home appliance device.
- the first intention instruction corresponding to the voice instruction is determined according to the plurality of first intention nodes and the association relationship.
- a computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the above intended instructions when running. Determine the method.
- an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the above intention through the computer program. How to determine instructions.
- a plurality of first states of the home appliance device respectively used to store the home appliance device and a state transition diagram for indicating the transfer relationship between the states of the home appliance device are generated.
- the home appliance changes the second state at the current moment to multiple first intention nodes of the first intention instruction of the first state; uses an association algorithm to calculate the state transition diagram to obtain the relationship between the multiple first states.
- association relationship wherein the association relationship includes: the sequence relationship between the plurality of first states and the weight of the plurality of first states; after receiving the voice instruction sent by the target object for controlling the home appliance device
- the first intention instruction corresponding to the voice instruction is determined according to the plurality of first intention nodes and the association relationship; by generating the plurality of first intention nodes and the association relationship, an intention knowledge graph is generated to store Intention, so that when the intention comes, the intention is matched in the intention knowledge graph, and the corresponding intention is determined when the edge in the intention knowledge graph is hit.
- Figure 1 is a hardware structure block diagram of a computer terminal of a method for determining intent instructions according to an embodiment of the present application
- Figure 2 is a schematic diagram of the hardware environment of an optional method for determining intent instructions according to an embodiment of the present disclosure
- Figure 3 is a flowchart of an optional method for determining intent instructions according to an embodiment of the present disclosure
- Figure 4 is a schematic flowchart of an optional method for determining intent instructions according to an embodiment of the present disclosure
- Figure 5 is a state transition diagram of an optional home appliance according to an embodiment of the present disclosure.
- FIG. 6 is a structural block diagram of an optional device for determining intent instructions according to an embodiment of the present disclosure.
- FIG. 1 is a hardware structure block diagram of a computer terminal for a method for determining intent instructions according to an embodiment of the present application.
- the computer terminal may include one or more (only one is shown in Figure 1) processors 202 (the processor 202 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 204 for storing data.
- the above-mentioned computer terminal may also include a transmission device 106 for communication functions and an input and output device 108.
- the structure shown in Figure 1 is only illustrative, and it does not limit the structure of the above-mentioned computer terminal.
- the computer terminal may also include more or fewer components than shown in FIG. 1 , or have a different configuration with equivalent functions or more functions than shown in FIG. 1 .
- the memory 204 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the method for determining intent instructions in the embodiment of the present application.
- the processor 202 runs the computer program stored in the memory 204, thereby Execute various functional applications and data processing, that is, implement the above methods.
- Memory 204 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory.
- the memory 204 may further include memory located remotely relative to the processor 202, and these remote memories may be connected to the computer terminal through a network. Examples of the above-mentioned networks include but are not limited to the Internet, intranets, local area networks, mobile communication networks and combinations thereof.
- Transmission device 106 is used to receive or send data via a network.
- Specific examples of the above-mentioned network may include a wireless network provided by a communication provider of the computer terminal.
- the transmission device 106 includes a network adapter (Network Interface Controller, NIC for short), which can be connected to other network devices through a base station to communicate with the Internet.
- the transmission device 106 may be a radio frequency (Radio Frequency, RF for short) module, which is used to communicate with the Internet wirelessly.
- NIC Network Interface Controller
- a method for determining intent instructions is provided.
- This method of determining intent instructions is widely used in whole-house intelligent digital control application scenarios such as smart home, smart home, smart home device ecology, and smart residence (IntelligenceHouse) ecology.
- the above method for determining the intention instruction can be applied to a hardware environment composed of multiple terminal devices 102 and servers 104 as shown in FIG. 2 .
- the server 104 is connected to multiple terminal devices 102 through the network. It can be set to provide services (such as application services, etc.) for the terminals or clients installed on the terminals.
- a database can be set up on the server or independently of the server. , is set to provide data storage services for the server 104, cloud computing and/or edge computing services can be configured on the server or independently of the server, and is set to provide data computing services for the server 104.
- the above-mentioned network may include but is not limited to at least one of the following: wired network, wireless network.
- the above-mentioned wired network may include but is not limited to at least one of the following: wide area network, metropolitan area network, and local area network.
- the above-mentioned wireless network may include at least one of the following: WIFI (Wireless Fidelity, Wireless Fidelity), Bluetooth.
- the terminal device 202 may be, but is not limited to, a PC, a mobile phone, a tablet, a smart air conditioner, a smart hood, a smart refrigerator, a smart oven, a smart stove, a smart washing machine, a smart water heater, a smart washing equipment, a smart dishwasher, or a smart projection device.
- smart TV smart clothes drying rack, smart curtains, smart audio and video, smart sockets, smart audio, smart speakers, smart fresh air equipment, smart kitchen and bathroom equipment, smart bathroom equipment, smart sweeping robot, smart window cleaning robot, smart mopping robot, Smart air purification equipment, smart steamers, smart microwave ovens, smart kitchen appliances, smart purifiers, smart water dispensers, smart door locks, etc.
- Figure 3 is a flow chart of a method for determining an intention instruction according to an embodiment of the present disclosure. The process includes the following steps:
- Step S202 Generate multiple first intent nodes of the home appliance device according to the state transition diagram of the home appliance device, where the state transition graph is used to indicate the transition relationship between states of the home appliance device, and the multiple first intent nodes
- the nodes are respectively configured to store a plurality of first states of the home appliance device, and a first intention instruction for instructing the home appliance device to change the second state at the current moment to the first state;
- Step S204 Use an association algorithm to calculate the state transition diagram to obtain an association relationship between the multiple first states, where the association relationship includes: a sequential relationship between the multiple first states and The weight of multiple first states;
- Step S206 Upon receiving a voice instruction sent by the target object for controlling the home appliance, determine a first intention instruction corresponding to the voice instruction according to the plurality of first intention nodes and the association relationship. .
- a plurality of first states of the home appliance device respectively used to store the home appliance device and a plurality of first states used to indicate the state of the home appliance device are generated according to the state transition diagram of the home appliance device used to indicate the transfer relationship between the states of the home appliance device.
- the home appliance changes the second state at the current moment to multiple first intention nodes of the first intention instruction of the first state; uses an association algorithm to calculate the state transition diagram to obtain the relationship between the multiple first states.
- association relationship wherein the association relationship includes: the sequence relationship between the plurality of first states and the weight of the plurality of first states; after receiving the voice instruction sent by the target object for controlling the home appliance device
- the first intention instruction corresponding to the voice instruction is determined according to the plurality of first intention nodes and the association relationship; by generating the plurality of first intention nodes and the association relationship, an intention knowledge graph is generated to store Intention, so that when the intention comes, the intention is matched in the intention knowledge graph, and the corresponding intention is determined when the edge in the intention knowledge graph is hit; the adoption of the above technical solution solves the problem of slow recognition speed of existing intention recognition solutions in related technologies. , the problem of low recognition accuracy; achieved the technical effect of improving the speed and recognition accuracy of intent recognition.
- the above-mentioned step S206 in the case of receiving a voice command from the target object, determine the first intention command corresponding to the voice command according to the plurality of first intention nodes and the association relationship, which can be done by the following The solution is implemented, including: generating an intent knowledge graph according to the plurality of first intent nodes and the association relationship, wherein the intent knowledge graph includes a plurality of first intent query vectors, and the first intent query vector is used to Indicate the association between the two first intention nodes; obtain the environmental parameters of the home appliance at the current moment, and generate a second intention query vector according to the voice command and the environmental parameters; in the intention knowledge graph The first intent query vector corresponding to the second intent query vector is matched to determine the first intent instruction corresponding to the voice instruction.
- an intent knowledge graph is first generated based on the first intention node and the association relationship.
- the intent knowledge graph stores the association relationship between each first intention node.
- Each first intention indication graph The intention query vector, that is, each edge indicates an intention; then obtain the environmental parameters of the home appliance at the current moment, thereby generating a second intention query vector based on the voice command and environmental parameters; match the second intention query vector in the intention knowledge graph
- the first intention query vector is used to query the intention corresponding to the voice command.
- the home appliance device After generating the intent knowledge graph, the home appliance device will also store the personal intent knowledge graph locally so that the home appliance device can perform edge computing and provide users with faster services; at the same time, the home appliance device will also regularly push the intent knowledge graph to
- the cloud summarizes everyone's habits into one super knowledge, so that it can make cold start recommendations for other new users. That is, for new users, there is no historical data and the system cannot make recommendations for them, so it can collect everyone's Habits, thereby predicting the habits of new users and recommending them to provide users with a better user experience.
- the above step: matching the first intention query vector corresponding to the second intention query vector in the intention knowledge graph to determine the first intention instruction corresponding to the voice instruction includes the following steps: determining Whether the home appliance device has a preamble action, wherein the preamble action is used to indicate the previous action performed by the home appliance device before receiving the voice command; in the case where the home appliance device has a preamble action , determine the first intention query vector corresponding to the second intention query vector in the intention knowledge graph according to the preamble action and the second intention query vector; obtain the first intention query vector corresponding to the first intention query vector The first intention instruction stored in the first intention node.
- the second The pre-order action of merging the intent query vector queries the first intent query vector corresponding to the second intent query vector in the intent knowledge graph, and obtains the first intent instruction stored in the first intent node corresponding to the first intent query vector, To achieve faster query of intent instructions.
- the method further includes: in the case where the first intention query vector corresponding to the second intention query vector is not matched , input the second intention query vector into the preset network model for generalization processing; when the second intention query vector after generalization hits the first intention query vector in the intention knowledge graph , obtain the first intention instruction stored in the first intention node corresponding to the first intention query vector; the second intention query vector after the generalization process misses the first intention query vector in the intention knowledge graph In the case of , continue processing the second intention query vector through the preset network model to obtain a second intention instruction corresponding to the second intention query vector.
- the preset network model can be the GCN deep learning network model.
- the second intention query vector after generalization is matched in the intention knowledge graph. If the second intention query vector after generalization hits the If the first intention query vector in the intention knowledge graph is found, the first intention instruction stored in the first intention node corresponding to the first intention query vector is obtained; if there is a miss, the second intention instruction is continued to be processed through the default network model.
- the intention query vector is used to obtain the second intention instruction corresponding to the second intention query vector, that is, the intention is identified.
- GCN graph deep convolutional network can better model graph data.
- the sequence calculated using the correlation algorithm can be turned into a triplet and can be directly put into GCN for training, without the need for additional data processing, saving time.
- GCN calculates is directly the prediction of the graph data structure.
- the generated nodes can either become nodes in the intent knowledge graph, or can be combined with two components (model component + knowledge graph component) for generalized intent recognition, and the effect is better than Traditional deep learning networks such as CNN.
- the method further includes: according to The second intention instruction generates a second intention node; and an intention knowledge graph is generated again according to the second intention node, the plurality of first intention nodes and the association relationship.
- the home appliance device After obtaining the second intention instruction through the preset network model processing, in order to update the intention knowledge graph and make the intention knowledge graph more and more perfect, the home appliance device also needs to generate a second intention node according to the second intention instruction, and generate a second intention node according to the second intention knowledge graph.
- the method further includes: obtaining the information before the home appliance device executes the first intention instruction.
- the third state, and the fourth state after the home appliance device executes the first intent command is determined according to the first intention command; according to the state transition diagram of the home appliance device, the third state and the third state The fourth state determines whether the first intention instruction is executable; if it is determined that the first intention instruction is executable, the home appliance is controlled to execute the first intention instruction.
- the third state of the home appliance before executing the first intention instruction that is, the state at this time
- the fourth state is to predict the state after executing the first intention command. According to the state transition diagram, the third state and the fourth state of the home appliance, it is determined whether the first intention command is executable. If it is executable, the home appliance is controlled. Execute the first intention instruction.
- the above step: determining whether the first intention instruction is executable according to the state transition diagram of the home appliance, the third state and the fourth state includes the following solution: in the state transition diagram Acquire a plurality of fifth states that have a transfer relationship with the third state, wherein the transfer relationship is used to indicate that the state of the home appliance is transferred from the third state to the fifth state; in the plurality of fifth states, If the fourth state is included in the plurality of fifth states, it is determined that the first intention instruction is executable; if the fourth state is not included in the plurality of fifth states, it is determined that the first intention instruction is executable; Intent command is not executable.
- the state transition diagram indicates the state transition relationship of the home appliance, that is, it indicates what state the home appliance can be converted into in the current state. Therefore, first obtain the transition relationship with the third state in the state transition diagram, and it can be obtained from If it is determined that the plurality of fifth states converted from the third state include the fourth state, the first intention instruction is deemed to be executable, otherwise the first intention instruction is deemed to be unexecutable, and Home appliances are prohibited from executing the first intention instruction to prevent damage to home appliances.
- Figure 4 is a schematic flowchart of an optional method for determining intent instructions according to an embodiment of the present disclosure. As shown in Figure 4, it has the following steps:
- Step S302 Start intent query in the intent knowledge graph
- Step S304 Determine whether the generated intention query vector hits the edge in the intention knowledge graph. If it hits, execute step S306. If it does not hit, execute step S312;
- Step S306 Obtain the edge hit by the intent query vector
- Step S308 Determine the intended instruction for edge storage
- Step S310 Execute the intended instruction
- Step S312 Input the intent query vector into the GCN deep learning model for generalization processing, and query the generalized intent query vector in the intent knowledge graph. If there is a hit, execute step S306. If there is no hit, execute step S314. ;
- Step S314 Perform intent recognition through the GCN deep learning model to obtain new intent instructions, and store the new intent instructions into the intent knowledge graph to update the intent knowledge graph;
- Step S316 Determine whether the new intended instruction can be executed. If it can be executed, execute step S310; if it cannot be executed, execute step S318;
- Step S318 Exit the program.
- the intent knowledge graph is generated to store the intent, so that the intent can be quickly queried. If the generated intent query vector cannot be matched in the intent knowledge graph, it means that the intent knowledge graph is not stored.
- the GCN deep learning model First generalize the intent query vector, and then perform a match. If the match is successful, the intent is obtained. If the match fails, the intent is identified through the GCN deep learning model, a new intent instruction is generated, and the new intent instruction is stored in the intent knowledge in real time. graph to update the intent knowledge graph, and finally determine whether the new intent command can be executed. If it is not executable, the execution of the command is prohibited to avoid damage to home appliances.
- the above technical solution is adopted to solve the existing intent problems in related technologies.
- the recognition scheme has the problem of slow recognition speed and low recognition accuracy; it has achieved the technical effect of improving the intention recognition speed and recognition accuracy.
- Figure 5 is a state transition diagram of an optional home appliance according to an embodiment of the present disclosure, as shown in Figure 5:
- Figure 5 gives an example of a state transition diagram of a home appliance, which is used to reflect the transfer relationship between the various states of the home appliance.
- the air conditioning equipment needs to be turned on first, and then it can perform operations such as heating and cooling, mode adjustment, etc. No matter what state it is in, you can shut down the computer.
- state transition diagram is drawn based on the properties of the device itself, so each model of machine has a unique state transition diagram.
- home appliances will automatically generate a complete set of relationships, namely the intention knowledge graph.
- Each node in the intention knowledge graph is a clear intention or a clear executable electrical operation.
- the algorithm calculates the sequence and frequency of these actions, and then uses some tools such as pyspark to convert these relationships and nodes into an intent knowledge graph, and stores the intent knowledge graph into a knowledge graph database with large capacity and good performance. , such as nebula database.
- each node stores the relationship with other nodes.
- the node itself stores the attributes corresponding to the intention when it occurs, such as the time, space, user gender, age, etc. at the time of occurrence.
- the relationship between nodes It shows the order in which intentions occur.
- the home appliance When an intention occurs, the home appliance generates an intention query vector based on the environment variables and received instructions at the time, and queries it in the intention knowledge graph. If the query hits, the corresponding node storage is obtained. Intention instruction, and execute the action corresponding to the instruction; if it misses, the intention query vector is input into the GCN deep learning network for judgment.
- the GCN deep learning network can generalize the intent query vector, and the generalized intent query vector can match the existing intent node in the intent knowledge graph, the corresponding intent instruction will be executed. If it cannot hit, GCN deep learning will be used. Judgment is made in the network, the new intention is identified, and the intention is also stored in the intention knowledge graph, and it is judged whether the intention is executable. If it is executable, that is, the action is risk-free, the action execution module is directly controlled. Action is executed, otherwise the action will not be executed to avoid causing damage to home appliances.
- each query needs to determine the intention to include environmental attributes, which can be any obtainable user attribute values, such as gender, age, the room in which the query is made, the weather, temperature, humidity, etc. at that time.
- environmental attributes can be any obtainable user attribute values, such as gender, age, the room in which the query is made, the weather, temperature, humidity, etc. at that time.
- a plurality of first states of the home appliance device respectively used to store the home appliance device and a state transition diagram for indicating the transfer relationship between the states of the home appliance device are generated according to the state transition diagram of the home appliance device.
- the home appliance changes the second state at the current moment to multiple first intention nodes of the first intention instruction of the first state; uses an association algorithm to calculate the state transition diagram to obtain the relationship between the multiple first states.
- association relationship wherein the association relationship includes: the sequence relationship between the plurality of first states and the weight of the plurality of first states; after receiving the voice instruction sent by the target object for controlling the home appliance device
- the first intention instruction corresponding to the voice instruction is determined according to the plurality of first intention nodes and the association relationship; by generating the plurality of first intention nodes and the association relationship, an intention knowledge graph is generated to store Intention, so that when the intention comes, the intention is matched in the intention knowledge graph, and the corresponding intention is determined when the edge in the intention knowledge graph is hit; the adoption of the above technical solution solves the problem of slow recognition speed of existing intention recognition solutions in related technologies. , the problem of low recognition accuracy; achieved the technical effect of improving the speed and recognition accuracy of intent recognition.
- the method according to the above embodiments can be implemented by means of software plus the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is Better implementation.
- the technical solution of the present disclosure can be embodied in the form of a software product in essence or that contributes to the existing technology.
- the computer software product is stored in a storage medium (such as ROM/RAM, disk, CD), including several instructions to cause a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of various embodiments of the present disclosure.
- This embodiment also provides a device for determining intent instructions.
- the device is configured to implement the above embodiments and preferred implementations. What has already been described will not be described again.
- the term "module” may be a combination of software and/or hardware that implements a predetermined function.
- the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.
- Figure 6 is a structural block diagram of an optional device for determining intent instructions according to an embodiment of the present disclosure.
- the device includes:
- the generation module 52 is configured to generate a plurality of first intention nodes of the home appliance according to a state transition diagram of the home appliance, wherein the state transition diagram is used to indicate a transition relationship between states of the home appliance, and the plurality of The first intention node is respectively used to store a plurality of first states of the home appliance device, and a first intention instruction used to instruct the home appliance device to change the second state at the current moment to the first state;
- the calculation module 54 is configured to use an association algorithm to calculate the state transition diagram to obtain an association relationship between the multiple first states, wherein the association relationship includes: an association relationship between the multiple first states. Sequence relationship and the weight of multiple first states;
- the determination module 56 is configured to, upon receiving a voice instruction sent by the target object for controlling the home appliance, determine the first intention node corresponding to the voice instruction based on the plurality of first intention nodes and the association relationship. One intention command.
- a plurality of first states of the home appliance device respectively used to store the home appliance device and a plurality of first states used to indicate the state of the home appliance device are generated according to the state transition diagram of the home appliance device used to indicate the transition relationship between the states of the home appliance device.
- the home appliance changes the second state at the current moment to multiple first intention nodes of the first intention instruction of the first state; uses an association algorithm to calculate the state transition diagram to obtain the relationship between the multiple first states.
- association relationship wherein the association relationship includes: the sequence relationship between the plurality of first states and the weight of the plurality of first states; after receiving the voice instruction sent by the target object for controlling the home appliance device
- the first intention instruction corresponding to the voice instruction is determined according to the plurality of first intention nodes and the association relationship; by generating the plurality of first intention nodes and the association relationship, an intention knowledge graph is generated to store Intention, so that when the intention comes, the intention is matched in the intention knowledge graph, and the corresponding intention is determined when the edge in the intention knowledge graph is hit; the adoption of the above technical solution solves the problem of slow recognition speed of existing intention recognition solutions in related technologies. , the problem of low recognition accuracy; achieved the technical effect of improving the speed and recognition accuracy of intent recognition.
- the determination module 56 is further configured to generate an intention knowledge graph according to the plurality of first intention nodes and the association relationship, wherein the intention knowledge graph includes a plurality of first intention nodes.
- the first intention query vector is used to indicate the association between two first intention nodes; obtain the environmental parameters of the home appliance device at the current moment, and generate it according to the voice command and the environmental parameters a second intention query vector; matching the first intention query vector corresponding to the second intention query vector in the intention knowledge graph to determine the first intention instruction corresponding to the voice instruction.
- an intent knowledge graph is first generated based on the first intention node and the association relationship.
- the intent knowledge graph stores the association relationship between each first intention node.
- Each first intention indication graph The intention query vector, that is, each edge indicates an intention; then obtain the environmental parameters of the home appliance at the current moment, thereby generating a second intention query vector based on the voice command and environmental parameters; match the second intention query vector in the intention knowledge graph
- the first intention query vector is used to query the intention corresponding to the voice command.
- the home appliance device After generating the intent knowledge graph, the home appliance device will also store the personal intent knowledge graph locally so that the home appliance device can perform edge computing and provide users with faster services; at the same time, the home appliance device will also regularly push the intent knowledge graph to
- the cloud summarizes everyone's habits into one super knowledge, so that it can make cold start recommendations for other new users. That is, for new users, there is no historical data and the system cannot make recommendations for them, so it can collect everyone's Habits, thereby predicting the habits of new users and recommending them to provide users with a better user experience.
- the determination module 56 is also configured to determine whether the home appliance device has a preceding action, wherein the preceding action is used to indicate the previous action performed by the home appliance device before receiving the voice instruction; In the case where the home appliance device has a preceding action, a first intention query vector corresponding to the second intention query vector is determined in the intention knowledge graph according to the preceding action and the second intention query vector. ; Obtain the first intention instruction stored in the first intention node corresponding to the first intention query vector.
- the second The pre-order action of merging the intent query vector queries the first intent query vector corresponding to the second intent query vector in the intent knowledge graph, and obtains the first intent instruction stored in the first intent node corresponding to the first intent query vector, To achieve faster query of intent instructions.
- the generation module 52 is also configured to, after generating the second intention query vector according to the voice instruction and the environmental parameter, if the first intention query vector corresponding to the second intention query vector is not matched.
- the second intent query vector is input into the preset network model for generalization processing; after the generalization processing, the second intent query vector hits the first position in the intent knowledge graph.
- an intent query vector obtain the first intent instruction stored in the first intent node corresponding to the first intent query vector; when the second intent query vector after the generalization process does not hit the intent knowledge graph In the case of the first intention query vector, continue to process the second intention query vector through the preset network model to obtain a second intention instruction corresponding to the second intention query vector.
- the preset network model can be the GCN deep learning network model.
- the second intention query vector after generalization is matched in the intention knowledge graph. If the second intention query vector after generalization hits the If the first intention query vector in the intention knowledge graph is found, the first intention instruction stored in the first intention node corresponding to the first intention query vector is obtained; if there is a miss, the second intention instruction is continued to be processed through the default network model.
- the intention query vector is used to obtain the second intention instruction corresponding to the second intention query vector, that is, the intention is identified.
- the generation module 52 is also configured to continue processing the second intention query vector through the preset network model to obtain the second intention instruction corresponding to the second intention query vector, according to the first
- the second intention instruction generates a second intention node; and an intention knowledge graph is generated again according to the second intention node, the plurality of first intention nodes and the association relationship.
- the home appliance device After obtaining the second intention instruction through the preset network model processing, in order to update the intention knowledge graph and make the intention knowledge graph more and more perfect, the home appliance device also needs to generate a second intention node according to the second intention instruction, and generate a second intention node according to the second intention knowledge graph.
- the determination module 56 is further configured to obtain the home appliance device to execute the first intention instruction after determining the first intention instruction corresponding to the voice instruction according to the plurality of first intention nodes and the association relationship.
- the previous third state, and the fourth state after the home appliance device executes the first intent command is determined according to the first intent command; according to the state transition diagram of the home appliance device, the third state and the The fourth state determines whether the first intention instruction is executable; if it is determined that the first intention instruction is executable, the home appliance is controlled to execute the first intention instruction.
- the third state of the home appliance before executing the first intention instruction that is, the state at this time
- the fourth state is to predict the state after executing the first intention command. According to the state transition diagram, the third state and the fourth state of the home appliance, it is determined whether the first intention command is executable. If it is executable, the home appliance is controlled. Execute the first intention instruction.
- the determination module 56 is further configured to obtain a plurality of fifth states that have a transition relationship with the third state in the state transition diagram, where the transition relationship is used to indicate the state of the home appliance. Transfer from the third state to the fifth state; in the case where the fourth state is included in the plurality of fifth states, determine that the first intention instruction is executable; in the plurality of fifth states, If the fourth state is not included in the state, it is determined that the first intended instruction is not executable.
- the state transition diagram indicates the state transition relationship of the home appliance, that is, it indicates what state the home appliance can be converted into in the current state. Therefore, first obtain the transition relationship with the third state in the state transition diagram, and it can be obtained from If it is determined that the plurality of fifth states converted from the third state include the fourth state, the first intention instruction is deemed to be executable, otherwise the first intention instruction is deemed to be unexecutable, and Home appliances are prohibited from executing the first intention instruction to prevent damage to home appliances.
- Embodiments of the present disclosure also provide a computer-readable storage medium that stores a computer program, wherein the computer program is configured to execute the steps in any of the above method embodiments when running.
- the above-mentioned storage medium can be configured to store a computer program for performing the following steps:
- S1 Generate multiple first intent nodes of the home appliance device according to the state transition diagram of the home appliance device, where the state transition graph is used to indicate the transition relationship between states of the home appliance device, and the multiple first intent nodes Respectively used to store a plurality of first states of the home appliance, and a first intention instruction used to instruct the home appliance to change the second state at the current moment to the first state;
- S2 Use an association algorithm to calculate the state transition diagram to obtain an association relationship between the multiple first states, where the association relationship includes: a sequence relationship between the multiple first states and a plurality of first states. The weight of the first state;
- S3 Upon receiving a voice instruction sent by the target object for controlling the home appliance, determine the first intention instruction corresponding to the voice instruction based on the plurality of first intention nodes and the association relationship.
- the above-mentioned computer-readable storage medium may include but is not limited to: U disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM) , mobile hard disk, magnetic disk or optical disk and other media that can store computer programs.
- ROM read-only memory
- RAM random access memory
- mobile hard disk magnetic disk or optical disk and other media that can store computer programs.
- Embodiments of the present disclosure also provide an electronic device, including a memory and a processor.
- a computer program is stored in the memory, and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.
- the above-mentioned processor may be configured to perform the following steps through a computer program:
- S1 Generate multiple first intent nodes of the home appliance device according to the state transition diagram of the home appliance device, where the state transition graph is used to indicate the transition relationship between states of the home appliance device, and the multiple first intent nodes Respectively used to store a plurality of first states of the home appliance, and a first intention instruction used to instruct the home appliance to change the second state at the current moment to the first state;
- S2 Use an association algorithm to calculate the state transition diagram to obtain an association relationship between the multiple first states, where the association relationship includes: a sequence relationship between the multiple first states and a plurality of first states. The weight of the first state;
- S3 Upon receiving a voice instruction sent by the target object for controlling the home appliance, determine the first intention instruction corresponding to the voice instruction based on the plurality of first intention nodes and the association relationship.
- the above-mentioned electronic device may further include a transmission device and an input-output device, wherein the transmission device is connected to the above-mentioned processor, and the input-output device is connected to the above-mentioned processor.
- modules or steps of the present disclosure can be implemented using general-purpose computing devices, and they can be concentrated on a single computing device, or distributed across a network composed of multiple computing devices. They may be implemented in program code executable by a computing device, such that they may be stored in a storage device for execution by the computing device, and in some cases may be executed in a sequence different from that shown herein. Or the described steps can be implemented by making them into individual integrated circuit modules respectively, or by making multiple modules or steps among them into a single integrated circuit module. As such, the present disclosure is not limited to any specific hardware and software basis.
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Manufacturing & Machinery (AREA)
- Quality & Reliability (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
一种意图指令的确定方法及装置、存储介质及电子装置,涉及智慧家庭技术领域,该意图指令的确定方法包括:根据家电设备的状态转移图生成家电设备的多个第一意图节点,其中,状态转移图用于指示家电设备的状态之间的转移关系,多个第一意图节点分别用于存储家电设备的多个第一状态,以及用于指示家电设备将当前时刻的第二状态改变为第一状态的第一意图指令(S202);使用关联算法对状态转移图进行计算,得到多个第一状态之间的关联关系,其中,关联关系包括:多个第一状态之间的先后关系和多个第一状态的权重(S204);在接收到目标对象发送的用于控制家电设备的语音指令的情况下,根据多个第一意图节点和关联关系确定与语音指令对应的第一意图指令(S206)。
Description
本公开要求于2022年8月18日提交中国专利局、申请号为202210994006.X、发明名称“意图指令的确定方法及装置、存储介质及电子装置”的中国专利申请的优先权,其全部内容通过引用结合在本公开中。
本公开涉及智慧家庭技术领域,具体而言,涉及一种意图指令的确定方法及装置、存储介质及电子装置。
相关技术中,随着智能家电语音控制***的普及,越来越多的智能家电支持语音控制,目前,随着智能家居普及率的提高,各大厂商都在智能家居的研发上投入了大量的力量。智能家居的一项基本任务就是意图识别,但由于智能家电通常有多个设备与用户进行绑定,在语音控制***的使用过程中会面临说同样一句话,不知道选择哪个具体设备的问题。
针对相关技术中,现有的意图识别方案识别速度慢,识别准确率低等问题,尚未提出有效的解决方案。
因此,有必要对相关技术予以改良以克服相关技术中的所述缺陷。
发明内容
本公开实施例提供了一种意图指令的确定方法及装置、存储介质及电子装置,以至少解决相关技术现有的意图识别方案识别速度慢,识别准确率低的问题。
根据本公开实施例的一方面,提供一种意图指令的确定方法,包括:根据家电设备的状态转移图生成所述家电设备的多个第一意图节点,其中,所述状态转移图用于指示家电设备的状态之间的转移关系,所述多个第一意图节点分别用于 存储所述家电设备的多个第一状态,以及用于指示所述家电设备将当前时刻的第二状态改变为第一状态的第一意图指令;使用关联算法对所述状态转移图进行计算,得到所述多个第一状态之间的关联关系,其中,所述关联关系包括:所述多个第一状态之间的先后关系和多个第一状态的权重;在接收到目标对象发送的用于控制所述家电设备的语音指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令。
根据本公开实施例的另一方面,还提供了一种意图指令的确定装置,包括:生成模块,设置为根据家电设备的状态转移图生成所述家电设备的多个第一意图节点,其中,所述状态转移图用于指示家电设备的状态之间的转移关系,所述多个第一意图节点分别用于存储所述家电设备的多个第一状态,以及用于指示所述家电设备将当前时刻的第二状态改变为第一状态的第一意图指令;计算模块,设置为使用关联算法对所述状态转移图进行计算,得到所述多个第一状态之间的关联关系,其中,所述关联关系包括:所述多个第一状态之间的先后关系和多个第一状态的权重;确定模块,设置为在接收到目标对象发送的用于控制所述家电设备的语音指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令。
根据本公开实施例的又一方面,还提供了一种计算机可读的存储介质,该计算机可读的存储介质中存储有计算机程序,其中,该计算机程序被设置为运行时执行上述意图指令的确定方法。
根据本公开实施例的又一方面,还提供了一种电子装置,包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序,其中,上述处理器通过计算机程序执行上述意图指令的确定方法。
通过本公开,根据家电设备的用于指示家电设备的状态之间的转移关系的状态转移图生成所述家电设备的分别用于存储所述家电设备的多个第一状态以及用于指示所述家电设备将当前时刻的第二状态改变为第一状态的第一意图指令的多个第一意图节点;使用关联算法对所述状态转移图进行计算,得到所述多个第一状态之间的关联关系,其中,所述关联关系包括:所述多个第一状态之间的 先后关系和多个第一状态的权重;在接收到目标对象发送的用于控制所述家电设备的语音指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令;通过生成上述多个第一意图节点和关联关系,生成意图知识图谱,以存储意图,从而在意图来临时,在意图知识图谱中匹配意图,命中意图知识图谱中的边的情况下确定对应的意图。
此处的附图被并入说明书中并构成本说明书的一部分,示出了符合本公开的实施例,并与说明书一起用于解释本公开的原理。
为了更清楚地说明本公开实施例或现有技术中的技术方案,下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍,显而易见地,对于本领域普通技术人员而言,在不付出创造性劳动性的前提下,还可以根据这些附图获得其他的附图。
图1是本申请实施例的一种意图指令的确定方法的计算机终端的硬件结构框图;
图2是根据本公开实施例的一种可选的意图指令的确定方法的硬件环境示意图;
图3是根据本公开实施例的一种可选的意图指令的确定方法的流程图;
图4是根据本公开实施例的一种可选的意图指令的确定方法的流程示意图;
图5是根据本公开实施例的一种可选的家电设备的状态转移图;
图6是根据本公开实施例的一种可选的意图指令的确定装置的结构框图。
为了使本技术领域的人员更好地理解本公开方案,下面将根据本公开实施例中的附图,对本公开实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例仅仅是本公开一部分的实施例,而不是全部的实施例。基于本公开中的 实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都应当属于本公开保护的范围。
需要说明的是,本公开的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象,而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换,以便这里描述的本公开的实施例能够以除了在这里图示或描述的那些以外的顺序实施。此外,术语“包括”和“具有”以及他们的任何变形,意图在于覆盖不排他的包含,例如,包含了一系列步骤或单元的过程、方法、***、产品或设备不必限于清楚地列出的那些步骤或单元,而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
本申请实施例所提供的方法实施例可以在计算机终端或者类似的运算装置中执行。以运行在计算机终端上为例,图1是本申请实施例的一种意图指令的确定方法的计算机终端的硬件结构框图。如图1所示,计算机终端可以包括一个或多个(图1中仅示出一个)处理器202(处理器202可以包括但不限于微处理器MCU或可编程逻辑器件FPGA等的处理装置)和用于存储数据的存储器204,在一个示例性实施例中,上述计算机终端还可以包括用于通信功能的传输设备106以及输入输出设备108。本领域普通技术人员可以理解,图1所示的结构仅为示意,其并不对上述计算机终端的结构造成限定。例如,计算机终端还可包括比图1中所示更多或者更少的组件,或者具有与图1所示等同功能或比图1所示功能更多的不同的配置。
存储器204可用于存储计算机程序,例如,应用软件的软件程序以及模块,如本申请实施例中的意图指令的确定方法对应的计算机程序,处理器202通过运行存储在存储器204内的计算机程序,从而执行各种功能应用以及数据处理,即实现上述的方法。存储器204可包括高速随机存储器,还可包括非易失性存储器,如一个或者多个磁性存储装置、闪存、或者其他非易失性固态存储器。在一些实例中,存储器204可进一步包括相对于处理器202远程设置的存储器,这些远程存储器可以通过网络连接至计算机终端。上述网络的实例包括但不限于互联网、 企业内部网、局域网、移动通信网及其组合。
传输设备106用于经由一个网络接收或者发送数据。上述的网络具体实例可包括计算机终端的通信供应商提供的无线网络。在一个实例中,传输设备106包括一个网络适配器(Network Interface Controller,简称为NIC),其可通过基站与其他网络设备相连从而可与互联网进行通讯。在一个实例中,传输设备106可以为射频(Radio Frequency,简称为RF)模块,其用于通过无线方式与互联网进行通讯。
根据本公开实施例的一个方面,提供了一种意图指令的确定方法。该意图指令的确定方法广泛应用于智慧家庭(Smart Home)、智能家居、智能家用设备生态、智慧住宅(IntelligenceHouse)生态等全屋智能数字化控制应用场景。可选地,在本实施例中,上述意图指令的确定方法可以应用于如图2所示的由多个终端设备102和服务器104所构成的硬件环境中。如图2所示,服务器104通过网络与多个终端设备102进行连接,可设置为为终端或终端上安装的客户端提供服务(如应用服务等),可在服务器上或独立于服务器设置数据库,设置为为服务器104提供数据存储服务,可在服务器上或独立于服务器配置云计算和/或边缘计算服务,设置为为服务器104提供数据运算服务。
上述网络可以包括但不限于以下至少之一:有线网络,无线网络。上述有线网络可以包括但不限于以下至少之一:广域网,城域网,局域网,上述无线网络可以包括但不限于以下至少之一:WIFI(Wireless Fidelity,无线保真),蓝牙。终端设备202可以并不限定于为PC、手机、平板电脑、智能空调、智能烟机、智能冰箱、智能烤箱、智能炉灶、智能洗衣机、智能热水器、智能洗涤设备、智能洗碗机、智能投影设备、智能电视、智能晾衣架、智能窗帘、智能影音、智能插座、智能音响、智能音箱、智能新风设备、智能厨卫设备、智能卫浴设备、智能扫地机器人、智能擦窗机器人、智能拖地机器人、智能空气净化设备、智能蒸箱、智能微波炉、智能厨宝、智能净化器、智能饮水机、智能门锁等。
在本实施例中提供了一种意图指令的确定方法,包括但不限于应用于计算机***,图3是根据本公开实施例的意图指令的确定方法的流程图,该流程包括如 下步骤:
步骤S202:根据家电设备的状态转移图生成所述家电设备的多个第一意图节点,其中,所述状态转移图用于指示家电设备的状态之间的转移关系,所述多个第一意图节点分别用于存储所述家电设备的多个第一状态,以及用于指示所述家电设备将当前时刻的第二状态改变为第一状态的第一意图指令;
步骤S204:使用关联算法对所述状态转移图进行计算,得到所述多个第一状态之间的关联关系,其中,所述关联关系包括:所述多个第一状态之间的先后关系和多个第一状态的权重;
步骤S206:在接收到目标对象发送的用于控制所述家电设备的语音指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令。
通过上述步骤,根据家电设备的用于指示家电设备的状态之间的转移关系的状态转移图生成所述家电设备的分别用于存储所述家电设备的多个第一状态以及用于指示所述家电设备将当前时刻的第二状态改变为第一状态的第一意图指令的多个第一意图节点;使用关联算法对所述状态转移图进行计算,得到所述多个第一状态之间的关联关系,其中,所述关联关系包括:所述多个第一状态之间的先后关系和多个第一状态的权重;在接收到目标对象发送的用于控制所述家电设备的语音指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令;通过生成上述多个第一意图节点和关联关系,生成意图知识图谱,以存储意图,从而在意图来临时,在意图知识图谱中匹配意图,命中意图知识图谱中的边的情况下确定对应的意图;采用上述技术方案,解决了相关技术中现有的意图识别方案识别速度慢,识别准确率低的问题;实现了提升意图识别速度和识别准确率的技术效果。
可选的,上述步骤S206:在接收到目标对象的语音指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令,可以通过以下方案来实现,包括:根据所述多个第一意图节点和所述关联关系生成意图知识图谱,其中,所述意图知识图谱包括多个第一意图查询向量,所述第一 意图查询向量用于指示两个第一意图节点之间的关联关系;获取所述家电设备在当前时刻的环境参数,并根据所述语音指令和所述环境参数生成第二意图查询向量;在所述意图知识图谱中匹配与所述第二意图查询向量对应的第一意图查询向量,以确定与所述语音指令对应的第一意图指令。
在获取到第一意图节点和关联关系之后,先根据第一意图节点和关联关系生成意图知识图谱,意图知识图谱存储了各个第一意图节点之间的关联关系,意图指示图谱的每一个第一意图查询向量即一条边都指示了一个意图;然后获取家电设备在当前时刻的环境参数,从而根据语音指令和环境参数生成第二意图查询向量;在意图知识图谱中匹配与第二意图查询向量对应的第一意图查询向量,以查询出语音指令对应的意图。
在生成意图知识图谱之后,家电设备还会将个人的意图知识图谱存储在本地,以供家电设备进行边缘计算,为用户提供更快地服务;同时,家电设备还会定期将意图知识图谱推送到云端,云端将所有人的习惯汇总为一个超级知识,从而可以对其他的新用户进行冷启动推荐,即对于新用户而言,没有历史数据,***无法为其进行推荐,因此可以采集所有人的习惯,从而预测新用户的习惯,并对其进行推荐,为用户带来更好的使用体验。
可选的,上述步骤:在所述意图知识图谱中匹配与所述第二意图查询向量对应的第一意图查询向量,以确定与所述语音指令对应的第一意图指令,包括以下步骤:确定所述家电设备是否存在前序动作,其中,所述前序动作用于指示在接收到所述语音指令之前所述家电设备执行的上一个动作;在所述家电设备存在前序动作的情况下,根据所述前序动作和所述第二意图查询向量在所述意图知识图谱中确定与所述第二意图查询向量对应的第一意图查询向量;获取与所述第一意图查询向量对应的第一意图节点存储的第一意图指令。
在进行查询之前,先确定该家电设备是否存在前序动作,即在识别该语音指令之前,该家电设备是否执行过其他动作,或家电设备是否处于其他状态,若存在前序动作,则第二意图查询向量融合前序动作在意图知识图谱中进行查询与第二意图查询向量对应的第一意图查询向量,并获取与该第一意图查询向量对应的 第一意图节点存储的第一意图指令,以实现更快速地意图指令的查询。
基于上述步骤,根据所述语音指令和所述环境参数生成第二意图查询向量之后,所述方法还包括:在未匹配到与所述第二意图查询向量对应的第一意图查询向量的情况下,将所述第二意图查询向量输入到预设网络模型中进行泛化处理;在所述泛化处理后的第二意图查询向量命中所述意图知识图谱中的第一意图查询向量的情况下,获取与所述第一意图查询向量对应的第一意图节点存储的第一意图指令;在所述泛化处理后的第二意图查询向量未命中所述意图知识图谱中的第一意图查询向量的情况下,继续通过所述预设网络模型处理所述第二意图查询向量,以得到与所述第二意图查询向量对应的第二意图指令。
若在意图知识图谱中未匹配到和第二意图查询向量对应的第一意图查询向量,则说明当前的意图知识图谱中尚未存储对应的意图,则将该第二意图查询向量输入到预设网络模型中进行泛化处理,预设网络模型可以是GCN深度学习网络模型,将泛化处理后的第二意图查询向量在意图知识图谱中进行匹配,若泛化处理后的第二意图查询向量命中了意图知识图谱中的第一意图查询向量,则获取与该第一意图查询向量对应的第一意图节点存储的第一意图指令;若未命中,则继续通过该预设网络模型处理该第二意图查询向量,以得到与该第二意图查询向量对应的第二意图指令,即识别出意图。
需要说明的是,使用GCN图深度卷积网络能够更好地对图数据进行建模,例如使用关联算法计算出的序列变成三元组可以直接放入GCN中进行训练,不需要再进行额外的数据处理,节省时间。且GCN计算出的直接就是图数据结构的预测,生成的节点既可以变成意图知识图谱中的节点,也可以结合两个部件(模型部件+知识图谱部件)进行泛化意图识别,效果好于CNN等传统的深度学习网络。
可选的,执行上述步骤:继续通过所述预设网络模型处理所述第二意图查询向量,以得到与所述第二意图查询向量对应的第二意图指令之后,所述方法还包括:根据所述第二意图指令生成第二意图节点;根据所述第二意图节点、所述多个第一意图节点和所述关联关系再次生成意图知识图谱。
在通过预设网络模型处理得到第二意图指令之后,为了更新意图知识图谱, 令意图知识图谱越来越完善,家电设备还需要根据该第二意图指令生成第二意图节点,并根据该第二意图节点,以及多个第一意图节点、关联关系再次生成意图知识图谱,以实现在不用重新调试代码的情况下,在线添加新的意图,更快地迭代更新意图知识图谱,满足用户需求。
进一步的,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令之后,所述方法还包括:获取所述家电设备执行所述第一意图指令之前的第三状态,以及根据所述第一意图指令确定所述家电设备执行所述第一意图指令之后的第四状态;根据所述家电设备的状态转移图、所述第三状态和所述第四状态确定所述第一意图指令是否可执行;在确定所述第一意图指令可执行的情况下,控制所述家电设备执行所述第一意图指令。
在得到第一意图指令之后,还需要获取家电设备执行该第一意图指令之前的第三状态,即此时的状态,以及根据该第一意图指令确定该家电设备执行该第一意图指令之后的第四状态,即预测执行第一意图指令之后的状态,根据该家电设备的状态转移图、第三状态和第四状态确定该第一意图指令是否可执行,若可执行,则控制该家电设备执行该第一意图指令。
可选的,上述步骤:根据所述家电设备的状态转移图、所述第三状态和所述第四状态确定所述第一意图指令是否可执行,包括以下方案:在所述状态转移图中获取与所述第三状态存在转移关系的多个第五状态,其中,所述转移关系用于指示所述家电设备的状态从所述第三状态转移至所述第五状态;在所述多个第五状态中包括所述第四状态的情况下,确定所述第一意图指令可执行;在所述多个第五状态中不包括所述第四状态的情况下,确定所述第一意图指令不可执行。
状态转移图指示了该家电设备的状态转移关系,即指示该家电设备在当前状态下能转换成什么状态,因此先获取该状态转移图中与该第三状态存在转移关系的,且是能从第三状态转换到的多个第五状态,若确定该多个第五状态中包括第四状态的情况下,则认定该第一意图指令可执行,否则认为该第一意图指令不可执行,并禁止家电设备执行该第一意图指令,防止对家电设备造成损害。
显然,上述所描述的实施例仅仅是本公开一部分的实施例,而不是全部的实 施例。为了更好的理解上述意图指令的确定方法,以下根据实施例对上述过程进行说明,但不用于限定本公开实施例的技术方案,具体地:
实施例1
以下根据图4对实施例1的方案进行描述,图4是根据本公开实施例的一种可选的意图指令的确定方法的流程示意图,如图4所示,具有以下步骤:
步骤S302:开始在意图知识图谱中进行意图查询;
步骤S304:判断生成的意图查询向量是否命中意图知识图谱中的边,若命中,则执行步骤S306,若未命中,执行步骤S312;
步骤S306:获取意图查询向量命中的边;
步骤S308:确定边存储的意图指令;
步骤S310:执行意图指令;
步骤S312:将意图查询向量输入到GCN深度学习模型进行泛化处理,并在意图知识图谱中对泛化处理后的意图查询向量进行查询,若命中则执行步骤S306,若未命中则执行步骤S314;
步骤S314:通过GCN深度学习模型进行意图识别,得到新的意图指令,将新的意图指令存入意图知识图谱,以更新意图知识图谱;
步骤S316:判断新的意图指令是否能执行,若可执行则执行步骤S310,若不可执行则执行步骤S318;
步骤S318:退出程序。
通过以上步骤,通过生成意图知识图谱来存储意图,从而能够快速地查询意图,若生成的意图查询向量未能在意图知识图谱中匹配到,则说明意图知识图谱中没有存储,通过GCN深度学习模型先对意图查询向量进行泛化处理,再进行一次匹配,若匹配成功则获取意图,若匹配失败则通过GCN深度学习模型识别意图,生成新的意图指令,实时将新的意图指令存入意图知识图谱,以对意图知 识图谱进行更新,最后判断新的意图指令是否能够执行,若不可执行则禁止执行该指令,避免对家电设备造成损害;采用上述技术方案,解决了相关技术中现有的意图识别方案识别速度慢,识别准确率低的问题;实现了提升意图识别速度和识别准确率的技术效果。
实施例2
以下根据图5对实施例2的方案进行描述,图5是根据本公开实施例的一种可选的家电设备的状态转移图,如图5所示:
图5给出了一种家电设备的状态转移图的示例,用于体现家电设备的各个状态之间的转移关系,譬如,空调设备需要先进行开机,然后才能进行升温降温、模式调节等操作,无论处于什么状态,均可以进行关机操作。
需要说明的是,状态转移图是根据设备本身的属性绘制而成的,因此每个型号的机器都有独一无二的状态转移图。
根据上述状态转移图,家电设备会自动生成一套完整的关系,即意图知识图谱,意图知识图谱中的每一个节点为一种明确的意图,或是一个明确的可以执行的电器操作,通过关联算法将这些动作的先后关系以及频繁程度计算出来,然后使用一些工具如pyspark将这些关联关系和节点转变为意图知识图谱,并将意图知识图谱存入一个容量很大性能很好地知识图谱数据库中,如nebula数据库。
在该意图知识图谱中,每个节点存储有与其他节点之间的关系,节点本身存储有意图发生时本身对应的属性如发生时刻的时间、空间、用户性别、年龄等,节点之间的关系表明了意图之间发生的顺序,当意图发生时,家电设备根据当时的环境变量和接收到的指令生成一个意图查询向量,并在意图知识图谱中进行查询,若查询命中则获取对应的节点存储的意图指令,并执行该指令对应的动作;若未命中则将意图查询向量输入到GCN深度学习网络中进行判断。
若GCN深度学习网络能够对意图查询向量进行泛化,并且泛化后的意图查询向量能够匹配意图知识图谱中已有的意图节点,则执行相应的意图指令,若不能命中,则由GCN深度学习网络中进行判断,识别出新的意图,并将该意图也 存入意图知识图谱中,并判断该意图是否可执行,若可执行,即该动作是无风险的,则直接控制动作执行模块进行动作执行,否则动作将不会执行,避免给家电设备造成损害。
需要说明的是,每个查询即需要判断的意图包含环境属性,可以为任意的可以获得的用户属性值,如性别、年龄、查询时所处的房间、当时的天气、温度、湿度等。
通过上述方案,根据家电设备的用于指示家电设备的状态之间的转移关系的状态转移图生成所述家电设备的分别用于存储所述家电设备的多个第一状态以及用于指示所述家电设备将当前时刻的第二状态改变为第一状态的第一意图指令的多个第一意图节点;使用关联算法对所述状态转移图进行计算,得到所述多个第一状态之间的关联关系,其中,所述关联关系包括:所述多个第一状态之间的先后关系和多个第一状态的权重;在接收到目标对象发送的用于控制所述家电设备的语音指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令;通过生成上述多个第一意图节点和关联关系,生成意图知识图谱,以存储意图,从而在意图来临时,在意图知识图谱中匹配意图,命中意图知识图谱中的边的情况下确定对应的意图;采用上述技术方案,解决了相关技术中现有的意图识别方案识别速度慢,识别准确率低的问题;实现了提升意图识别速度和识别准确率的技术效果。
通过以上的实施方式的描述,本领域的技术人员可以清楚地了解到根据上述实施例的方法可借助软件加必需的通用硬件平台的方式来实现,当然也可以通过硬件,但很多情况下前者是更佳的实施方式。基于这样的理解,本公开的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质(如ROM/RAM、磁碟、光盘)中,包括若干指令用以使得一台终端设备(可以是手机,计算机,服务器,或者网络设备等)执行本公开各个实施例的方法。
在本实施例中还提供了一种意图指令的确定装置,该装置设置为实现上述实施例及优选实施方式,已经进行过说明的不再赘述。如以下所使用的,术语“模 块”可以实现预定功能的软件和/或硬件的组合。尽管以下实施例所描述的设备较佳地以软件来实现,但是硬件,或者软件和硬件的组合的实现也是可能并被构想的。
图6是根据本公开实施例的一种可选的意图指令的确定装置的结构框图,该装置包括:
生成模块52,设置为根据家电设备的状态转移图生成所述家电设备的多个第一意图节点,其中,所述状态转移图用于指示家电设备的状态之间的转移关系,所述多个第一意图节点分别用于存储所述家电设备的多个第一状态,以及用于指示所述家电设备将当前时刻的第二状态改变为第一状态的第一意图指令;
计算模块54,设置为使用关联算法对所述状态转移图进行计算,得到所述多个第一状态之间的关联关系,其中,所述关联关系包括:所述多个第一状态之间的先后关系和多个第一状态的权重;
确定模块56,设置为在接收到目标对象发送的用于控制所述家电设备的语音指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令。
通过上述装置,根据家电设备的用于指示家电设备的状态之间的转移关系的状态转移图生成所述家电设备的分别用于存储所述家电设备的多个第一状态以及用于指示所述家电设备将当前时刻的第二状态改变为第一状态的第一意图指令的多个第一意图节点;使用关联算法对所述状态转移图进行计算,得到所述多个第一状态之间的关联关系,其中,所述关联关系包括:所述多个第一状态之间的先后关系和多个第一状态的权重;在接收到目标对象发送的用于控制所述家电设备的语音指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令;通过生成上述多个第一意图节点和关联关系,生成意图知识图谱,以存储意图,从而在意图来临时,在意图知识图谱中匹配意图,命中意图知识图谱中的边的情况下确定对应的意图;采用上述技术方案,解决了相关技术中现有的意图识别方案识别速度慢,识别准确率低的问题;实现了提升意图识别速度和识别准确率的技术效果。
可选的,执行上述步骤S206的过程中,确定模块56,还设置为根据所述多个第一意图节点和所述关联关系生成意图知识图谱,其中,所述意图知识图谱包括多个第一意图查询向量,所述第一意图查询向量用于指示两个第一意图节点之间的关联关系;获取所述家电设备在当前时刻的环境参数,并根据所述语音指令和所述环境参数生成第二意图查询向量;在所述意图知识图谱中匹配与所述第二意图查询向量对应的第一意图查询向量,以确定与所述语音指令对应的第一意图指令。
在获取到第一意图节点和关联关系之后,先根据第一意图节点和关联关系生成意图知识图谱,意图知识图谱存储了各个第一意图节点之间的关联关系,意图指示图谱的每一个第一意图查询向量即一条边都指示了一个意图;然后获取家电设备在当前时刻的环境参数,从而根据语音指令和环境参数生成第二意图查询向量;在意图知识图谱中匹配与第二意图查询向量对应的第一意图查询向量,以查询出语音指令对应的意图。
在生成意图知识图谱之后,家电设备还会将个人的意图知识图谱存储在本地,以供家电设备进行边缘计算,为用户提供更快地服务;同时,家电设备还会定期将意图知识图谱推送到云端,云端将所有人的习惯汇总为一个超级知识,从而可以对其他的新用户进行冷启动推荐,即对于新用户而言,没有历史数据,***无法为其进行推荐,因此可以采集所有人的习惯,从而预测新用户的习惯,并对其进行推荐,为用户带来更好的使用体验。
可选的,确定模块56,还设置为确定所述家电设备是否存在前序动作,其中,所述前序动作用于指示在接收到所述语音指令之前所述家电设备执行的上一个动作;在所述家电设备存在前序动作的情况下,根据所述前序动作和所述第二意图查询向量在所述意图知识图谱中确定与所述第二意图查询向量对应的第一意图查询向量;获取与所述第一意图查询向量对应的第一意图节点存储的第一意图指令。
在进行查询之前,先确定该家电设备是否存在前序动作,即在识别该语音指令之前,该家电设备是否执行过其他动作,或家电设备是否处于其他状态,若存 在前序动作,则第二意图查询向量融合前序动作在意图知识图谱中进行查询与第二意图查询向量对应的第一意图查询向量,并获取与该第一意图查询向量对应的第一意图节点存储的第一意图指令,以实现更快速地意图指令的查询。
另一方面,基于上述步骤,生成模块52,还设置为在根据所述语音指令和所述环境参数生成第二意图查询向量之后,在未匹配到与所述第二意图查询向量对应的第一意图查询向量的情况下,将所述第二意图查询向量输入到预设网络模型中进行泛化处理;在所述泛化处理后的第二意图查询向量命中所述意图知识图谱中的第一意图查询向量的情况下,获取与所述第一意图查询向量对应的第一意图节点存储的第一意图指令;在所述泛化处理后的第二意图查询向量未命中所述意图知识图谱中的第一意图查询向量的情况下,继续通过所述预设网络模型处理所述第二意图查询向量,以得到与所述第二意图查询向量对应的第二意图指令。
若在意图知识图谱中未匹配到和第二意图查询向量对应的第一意图查询向量,则说明当前的意图知识图谱中尚未存储对应的意图,则将该第二意图查询向量输入到预设网络模型中进行泛化处理,预设网络模型可以是GCN深度学习网络模型,将泛化处理后的第二意图查询向量在意图知识图谱中进行匹配,若泛化处理后的第二意图查询向量命中了意图知识图谱中的第一意图查询向量,则获取与该第一意图查询向量对应的第一意图节点存储的第一意图指令;若未命中,则继续通过该预设网络模型处理该第二意图查询向量,以得到与该第二意图查询向量对应的第二意图指令,即识别出意图。
可选的,生成模块52,还设置为继续通过所述预设网络模型处理所述第二意图查询向量,以得到与所述第二意图查询向量对应的第二意图指令之后,根据所述第二意图指令生成第二意图节点;根据所述第二意图节点、所述多个第一意图节点和所述关联关系再次生成意图知识图谱。
在通过预设网络模型处理得到第二意图指令之后,为了更新意图知识图谱,令意图知识图谱越来越完善,家电设备还需要根据该第二意图指令生成第二意图节点,并根据该第二意图节点,以及多个第一意图节点、关联关系再次生成意图知识图谱,以实现在不用重新调试代码的情况下,在线添加新的意图,更快地迭 代更新意图知识图谱,满足用户需求。
进一步的,确定模块56,还设置为根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令之后,获取所述家电设备执行所述第一意图指令之前的第三状态,以及根据所述第一意图指令确定所述家电设备执行所述第一意图指令之后的第四状态;根据所述家电设备的状态转移图、所述第三状态和所述第四状态确定所述第一意图指令是否可执行;在确定所述第一意图指令可执行的情况下,控制所述家电设备执行所述第一意图指令。
在得到第一意图指令之后,还需要获取家电设备执行该第一意图指令之前的第三状态,即此时的状态,以及根据该第一意图指令确定该家电设备执行该第一意图指令之后的第四状态,即预测执行第一意图指令之后的状态,根据该家电设备的状态转移图、第三状态和第四状态确定该第一意图指令是否可执行,若可执行,则控制该家电设备执行该第一意图指令。
可选的,确定模块56,还设置为在所述状态转移图中获取与所述第三状态存在转移关系的多个第五状态,其中,所述转移关系用于指示所述家电设备的状态从所述第三状态转移至所述第五状态;在所述多个第五状态中包括所述第四状态的情况下,确定所述第一意图指令可执行;在所述多个第五状态中不包括所述第四状态的情况下,确定所述第一意图指令不可执行。
状态转移图指示了该家电设备的状态转移关系,即指示该家电设备在当前状态下能转换成什么状态,因此先获取该状态转移图中与该第三状态存在转移关系的,且是能从第三状态转换到的多个第五状态,若确定该多个第五状态中包括第四状态的情况下,则认定该第一意图指令可执行,否则认为该第一意图指令不可执行,并禁止家电设备执行该第一意图指令,防止对家电设备造成损害。
本公开的实施例还提供了一种计算机可读存储介质,该计算机可读存储介质中存储有计算机程序,其中,该计算机程序被设置为运行时执行上述任一项方法实施例中的步骤。
可选地,在本实施例中,上述存储介质可以被设置为存储用于执行以下步骤 的计算机程序:
S1,根据家电设备的状态转移图生成所述家电设备的多个第一意图节点,其中,所述状态转移图用于指示家电设备的状态之间的转移关系,所述多个第一意图节点分别用于存储所述家电设备的多个第一状态,以及用于指示所述家电设备将当前时刻的第二状态改变为第一状态的第一意图指令;
S2,使用关联算法对所述状态转移图进行计算,得到所述多个第一状态之间的关联关系,其中,所述关联关系包括:所述多个第一状态之间的先后关系和多个第一状态的权重;
S3,在接收到目标对象发送的用于控制所述家电设备的语音指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令。
在一个示例性实施例中,上述计算机可读存储介质可以包括但不限于:U盘、只读存储器(Read-Only Memory,简称为ROM)、随机存取存储器(Random Access Memory,简称为RAM)、移动硬盘、磁碟或者光盘等各种可以存储计算机程序的介质。
本实施例中的具体示例可以参考上述实施例及示例性实施方式中所描述的示例,本实施例在此不再赘述。
本公开的实施例还提供了一种电子装置,包括存储器和处理器,该存储器中存储有计算机程序,该处理器被设置为运行计算机程序以执行上述任一项方法实施例中的步骤。
可选地,在本实施例中,上述处理器可以被设置为通过计算机程序执行以下步骤:
S1,根据家电设备的状态转移图生成所述家电设备的多个第一意图节点,其中,所述状态转移图用于指示家电设备的状态之间的转移关系,所述多个第一意图节点分别用于存储所述家电设备的多个第一状态,以及用于指示所述家电设备将当前时刻的第二状态改变为第一状态的第一意图指令;
S2,使用关联算法对所述状态转移图进行计算,得到所述多个第一状态之间的关联关系,其中,所述关联关系包括:所述多个第一状态之间的先后关系和多个第一状态的权重;
S3,在接收到目标对象发送的用于控制所述家电设备的语音指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令。
在一个示例性实施例中,上述电子装置还可以包括传输设备以及输入输出设备,其中,该传输设备和上述处理器连接,该输入输出设备和上述处理器连接。
本实施例中的具体示例可以参考上述实施例及示例性实施方式中所描述的示例,本实施例在此不再赘述。
显然,本领域的技术人员应该明白,上述的本公开的各模块或各步骤可以用通用的计算装置来实现,它们可以集中在单个的计算装置上,或者分布在多个计算装置所组成的网络上,它们可以用计算装置可执行的程序代码来实现,从而,可以将它们存储在存储装置中由计算装置来执行,并且在某些情况下,可以以不同于此处的顺序执行所示出或描述的步骤,或者将它们分别制作成各个集成电路模块,或者将它们中的多个模块或步骤制作成单个集成电路模块来实现。这样,本公开不限制于任何特定的硬件和软件根据。
以上所述仅是本公开的优选实施方式,应当指出,对于本技术领域的普通技术人员来说,在不脱离本公开原理的前提下,还可以做出若干改进和润饰,这些改进和润饰也应视为本公开的保护范围。
Claims (16)
- 一种意图指令的确定方法,包括:根据家电设备的状态转移图生成所述家电设备的多个第一意图节点,其中,所述状态转移图用于指示家电设备的状态之间的转移关系,所述多个第一意图节点分别用于存储所述家电设备的多个第一状态,以及用于指示所述家电设备将当前时刻的第二状态改变为第一状态的第一意图指令;使用关联算法对所述状态转移图进行计算,得到所述多个第一状态之间的关联关系,其中,所述关联关系包括:所述多个第一状态之间的先后关系和多个第一状态的权重;在接收到目标对象发送的用于控制所述家电设备的语音指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令。
- 根据权利要求1所述的意图指令的确定方法,其中,在接收到目标对象的语音指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令,包括:根据所述多个第一意图节点和所述关联关系生成意图知识图谱,其中,所述意图知识图谱包括多个第一意图查询向量,所述第一意图查询向量用于指示两个第一意图节点之间的关联关系;获取所述家电设备在当前时刻的环境参数,并根据所述语音指令和所述环境参数生成第二意图查询向量;在所述意图知识图谱中匹配与所述第二意图查询向量对应的第一意图查询向量,以确定与所述语音指令对应的第一意图指令。
- 根据权利要求2所述的意图指令的确定方法,其中,在所述意图知识图谱中匹配与所述第二意图查询向量对应的第一意图查询向量,以确定与所述语音指令对应的第一意图指令,包括:确定所述家电设备是否存在前序动作,其中,所述前序动作用于指示在接收到所述语音指令之前所述家电设备执行的上一个动作;在所述家电设备存在前序动作的情况下,根据所述前序动作和所述第二意图查询向量在所述意图知识图谱中确定与所述第二意图查询向量对应的第一意图查询向量;获取与所述第一意图查询向量对应的第一意图节点存储的第一意图指令。
- 根据权利要求2所述的意图指令的确定方法,其中,根据所述语音指令和所述环境参数生成第二意图查询向量之后,所述方法还包括:在未匹配到与所述第二意图查询向量对应的第一意图查询向量的情况下,将所述第二意图查询向量输入到预设网络模型中进行泛化处理;在所述泛化处理后的第二意图查询向量命中所述意图知识图谱中的第一意图查询向量的情况下,获取与所述第一意图查询向量对应的第一意图节点存储的第一意图指令;在所述泛化处理后的第二意图查询向量未命中所述意图知识图谱中的第一意图查询向量的情况下,继续通过所述预设网络模型处理所述第二意图查询向量,以得到与所述第二意图查询向量对应的第二意图指令。
- 根据权利要求4所述的意图指令的确定方法,其中,继续通过所述预设网络模型处理所述第二意图查询向量,以得到与所述第二意图查询向量对应的第二意图指令之后,所述方法还包括:根据所述第二意图指令生成第二意图节点;根据所述第二意图节点、所述多个第一意图节点和所述关联关系再次生成意图知识图谱。
- 根据权利要求1所述的意图指令的确定方法,其中,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令之后,所述方法还包括:获取所述家电设备执行所述第一意图指令之前的第三状态,以及根据所述第一意图指令确定所述家电设备执行所述第一意图指令之后的第四状态;根据所述家电设备的状态转移图、所述第三状态和所述第四状态确定所述第一意图指令是否可执行;在确定所述第一意图指令可执行的情况下,控制所述家电设备执行所述第一意图指令。
- 根据权利要求6所述的意图指令的确定方法,其中,根据所述家电设备的状态转移图、所述第三状态和所述第四状态确定所述第一意图指令是否可执行,包括:在所述状态转移图中获取与所述第三状态存在转移关系的多个第五状态,其中,所述转移关系用于指示所述家电设备的状态从所述第三状态转移至所述第五状态;在所述多个第五状态中包括所述第四状态的情况下,确定所述第一意图指令可执行;在所述多个第五状态中不包括所述第四状态的情况下,确定所述第一意图指令不可执行。
- 一种意图指令的确定装置,包括:生成模块,设置为根据家电设备的状态转移图生成所述家电设备的多个第一意图节点,其中,所述状态转移图用于指示家电设备的状态之间的转移关系,所述多个第一意图节点分别用于存储所述家电设备的多个第一状态,以及用于指示所述家电设备将当前时刻的第二状态改变为第一状态的第一意图指令;计算模块,设置为使用关联算法对所述状态转移图进行计算,得到所述多个第一状态之间的关联关系,其中,所述关联关系包括:所述多个第一状态之间的先后关系和多个第一状态的权重;确定模块,设置为在接收到目标对象发送的用于控制所述家电设备的语音 指令的情况下,根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令。
- 根据权利要求8所述的意图指令的确定装置,其中,所述确定模块,还设置为根据所述多个第一意图节点和所述关联关系生成意图知识图谱,其中,所述意图知识图谱包括多个第一意图查询向量,所述第一意图查询向量用于指示两个第一意图节点之间的关联关系;获取所述家电设备在当前时刻的环境参数,并根据所述语音指令和所述环境参数生成第二意图查询向量;在所述意图知识图谱中匹配与所述第二意图查询向量对应的第一意图查询向量,以确定与所述语音指令对应的第一意图指令。
- 根据权利要求9所述的意图指令的确定装置,其中,所述确定模块,还设置为确定所述家电设备是否存在前序动作,其中,所述前序动作用于指示在接收到所述语音指令之前所述家电设备执行的上一个动作;在所述家电设备存在前序动作的情况下,根据所述前序动作和所述第二意图查询向量在所述意图知识图谱中确定与所述第二意图查询向量对应的第一意图查询向量;获取与所述第一意图查询向量对应的第一意图节点存储的第一意图指令。
- 根据权利要求9所述的意图指令的确定装置,其中,所述生成模块,还设置为在根据所述语音指令和所述环境参数生成第二意图查询向量之后,在未匹配到与所述第二意图查询向量对应的第一意图查询向量的情况下,将所述第二意图查询向量输入到预设网络模型中进行泛化处理;在所述泛化处理后的第二意图查询向量命中所述意图知识图谱中的第一意图查询向量的情况下,获取与所述第一意图查询向量对应的第一意图节点存储的第一意图指令;在所述泛化处理后的第二意图查询向量未命中所述意图知识图谱中的第一意图查询向量的情况下,继续通过所述预设网络模型处理所述第二意图查询向量,以得到与所述第二意图查询向量对应的第二意图指令。
- 根据权利要求11所述的意图指令的确定装置,其中,所述生成模块,还设置为继续通过所述预设网络模型处理所述第二意图查询向量,以得到与所述第二意图查询向量对应的第二意图指令之后,根据所述第二意图指令生成第二 意图节点;根据所述第二意图节点、所述多个第一意图节点和所述关联关系再次生成意图知识图谱。
- 根据权利要求8所述的意图指令的确定装置,其中,所述确定模块,还设置为根据所述多个第一意图节点和所述关联关系确定与所述语音指令对应的第一意图指令之后,获取所述家电设备执行所述第一意图指令之前的第三状态,以及根据所述第一意图指令确定所述家电设备执行所述第一意图指令之后的第四状态;根据所述家电设备的状态转移图、所述第三状态和所述第四状态确定所述第一意图指令是否可执行;在确定所述第一意图指令可执行的情况下,控制所述家电设备执行所述第一意图指令。
- 根据权利要求13所述的意图指令的确定装置,其中,确定模块,还设置为在所述状态转移图中获取与所述第三状态存在转移关系的多个第五状态,其中,所述转移关系用于指示所述家电设备的状态从所述第三状态转移至所述第五状态;在所述多个第五状态中包括所述第四状态的情况下,确定所述第一意图指令可执行;在所述多个第五状态中不包括所述第四状态的情况下,确定所述第一意图指令不可执行。
- 一种计算机可读的存储介质,所述计算机可读的存储介质包括存储的程序,其中,所述程序运行时执行权利要求1至7中任一项所述的方法。
- 一种电子装置,包括存储器和处理器,所述存储器中存储有计算机程序,所述处理器被设置为通过所述计算机程序执行权利要求1至7中任一项所述的方法。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210994006.X | 2022-08-18 | ||
CN202210994006.XA CN115453897A (zh) | 2022-08-18 | 2022-08-18 | 意图指令的确定方法及装置、存储介质及电子装置 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2024036804A1 true WO2024036804A1 (zh) | 2024-02-22 |
Family
ID=84298451
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2022/134786 WO2024036804A1 (zh) | 2022-08-18 | 2022-11-28 | 意图指令的确定方法及装置、存储介质及电子装置 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN115453897A (zh) |
WO (1) | WO2024036804A1 (zh) |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104992709A (zh) * | 2015-06-25 | 2015-10-21 | 广州视源电子科技股份有限公司 | 一种语音指令的执行方法及语音识别设备 |
CN106373566A (zh) * | 2016-08-25 | 2017-02-01 | 深圳市元征科技股份有限公司 | 数据传输控制方法及装置 |
US20190221206A1 (en) * | 2019-03-27 | 2019-07-18 | Intel Corporation | Spoken keyword detection based utterance-level wake on intent system |
CN110262273A (zh) * | 2019-07-12 | 2019-09-20 | 珠海格力电器股份有限公司 | 一种家居设备控制方法、装置、存储介质及智能家居*** |
CN111291157A (zh) * | 2020-01-21 | 2020-06-16 | 招商银行股份有限公司 | 应答响应方法、装置、终端及存储介质 |
CN111830842A (zh) * | 2020-07-10 | 2020-10-27 | 同济大学 | 一种智能家居无感化控制*** |
CN112466289A (zh) * | 2020-12-21 | 2021-03-09 | 北京百度网讯科技有限公司 | 语音指令的识别方法、装置、语音设备和存储介质 |
CN113674749A (zh) * | 2021-08-26 | 2021-11-19 | 珠海格力电器股份有限公司 | 一种控制方法、装置、电子设备和存储介质 |
CN114500139A (zh) * | 2022-01-27 | 2022-05-13 | 青岛海尔科技有限公司 | 指令组的发送方法和装置、存储介质及电子装置 |
CN114546486A (zh) * | 2022-01-28 | 2022-05-27 | 青岛海尔科技有限公司 | 向用户推荐指令的方法和装置、存储介质及电子装置 |
CN114882882A (zh) * | 2022-05-25 | 2022-08-09 | 无锡小天鹅电器有限公司 | 一种语音指令执行方法、装置、家用电器及存储介质 |
CN114915514A (zh) * | 2022-03-28 | 2022-08-16 | 青岛海尔科技有限公司 | 意图的处理方法和装置、存储介质及电子装置 |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110795532A (zh) * | 2019-10-18 | 2020-02-14 | 珠海格力电器股份有限公司 | 一种语音信息的处理方法、装置、智能终端以及存储介质 |
CN110718224B (zh) * | 2019-10-31 | 2022-01-21 | 珠海格力电器股份有限公司 | 一种语音控制方法、装置、存储介质和智能设备 |
CN113257241A (zh) * | 2020-02-11 | 2021-08-13 | 青岛海尔空调器有限总公司 | 一种智能家电及其语音控制方法 |
CN113806473A (zh) * | 2020-06-17 | 2021-12-17 | 华为技术有限公司 | 意图识别方法和电子设备 |
CN112151035B (zh) * | 2020-10-14 | 2023-08-11 | 珠海格力电器股份有限公司 | 一种语音控制方法、装置、电子设备及可读存储介质 |
CN112466293A (zh) * | 2020-11-13 | 2021-03-09 | 广州视源电子科技股份有限公司 | 解码图优化方法、装置及存储介质 |
CN114093024A (zh) * | 2021-09-24 | 2022-02-25 | 张哲为 | 人体动作的识别方法、装置、设备及存储介质 |
-
2022
- 2022-08-18 CN CN202210994006.XA patent/CN115453897A/zh active Pending
- 2022-11-28 WO PCT/CN2022/134786 patent/WO2024036804A1/zh unknown
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104992709A (zh) * | 2015-06-25 | 2015-10-21 | 广州视源电子科技股份有限公司 | 一种语音指令的执行方法及语音识别设备 |
CN106373566A (zh) * | 2016-08-25 | 2017-02-01 | 深圳市元征科技股份有限公司 | 数据传输控制方法及装置 |
US20190221206A1 (en) * | 2019-03-27 | 2019-07-18 | Intel Corporation | Spoken keyword detection based utterance-level wake on intent system |
CN110262273A (zh) * | 2019-07-12 | 2019-09-20 | 珠海格力电器股份有限公司 | 一种家居设备控制方法、装置、存储介质及智能家居*** |
CN111291157A (zh) * | 2020-01-21 | 2020-06-16 | 招商银行股份有限公司 | 应答响应方法、装置、终端及存储介质 |
CN111830842A (zh) * | 2020-07-10 | 2020-10-27 | 同济大学 | 一种智能家居无感化控制*** |
CN112466289A (zh) * | 2020-12-21 | 2021-03-09 | 北京百度网讯科技有限公司 | 语音指令的识别方法、装置、语音设备和存储介质 |
CN113674749A (zh) * | 2021-08-26 | 2021-11-19 | 珠海格力电器股份有限公司 | 一种控制方法、装置、电子设备和存储介质 |
CN114500139A (zh) * | 2022-01-27 | 2022-05-13 | 青岛海尔科技有限公司 | 指令组的发送方法和装置、存储介质及电子装置 |
CN114546486A (zh) * | 2022-01-28 | 2022-05-27 | 青岛海尔科技有限公司 | 向用户推荐指令的方法和装置、存储介质及电子装置 |
CN114915514A (zh) * | 2022-03-28 | 2022-08-16 | 青岛海尔科技有限公司 | 意图的处理方法和装置、存储介质及电子装置 |
CN114882882A (zh) * | 2022-05-25 | 2022-08-09 | 无锡小天鹅电器有限公司 | 一种语音指令执行方法、装置、家用电器及存储介质 |
Also Published As
Publication number | Publication date |
---|---|
CN115453897A (zh) | 2022-12-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109257259A (zh) | 场景联动控制方法、装置及家电设备 | |
US20180305851A1 (en) | Laundry treating apparatus, control method of the laundry treating apparatus and online system including the same | |
JP2021530196A (ja) | スマートプラグの動作のシステムおよび方法 | |
WO2016065812A1 (zh) | 基于设定场景模式的智能设备控制方法和装置 | |
WO2024036694A1 (zh) | 控制指令的发送方法、控制装置、存储介质及电子装置 | |
CN109547308A (zh) | 一种智能家居的控制方法、装置、存储介质及服务器 | |
CN103970554B (zh) | 一种家电产品功能升级的控制方法 | |
CN109600285A (zh) | 一种智能家居中动态构建环境调节规则列表的方法及装置 | |
CN107888458A (zh) | 智能设备状态处理方法及装置 | |
WO2024021407A1 (zh) | 知识图谱的更新方法和装置、存储介质及电子装置 | |
WO2023207170A1 (zh) | 洗涤程序的推荐方法及装置、存储介质及电子装置 | |
CN111694280A (zh) | 一种应用场景的控制***及其控制方法 | |
CN111338227A (zh) | 基于强化学习的电子电器控制方法及控制设备、存储介质 | |
CN109167711A (zh) | 一种家电的控制方法、装置、存储介质及服务器 | |
CN115167164A (zh) | 设备场景的确定方法和装置、存储介质及电子装置 | |
CN114697150A (zh) | 命令的下发方法和装置、存储介质及电子装置 | |
CN110830334A (zh) | 一种家电设备控制方法、装置、***、服务器及家电设备 | |
WO2024001196A1 (zh) | 家电设备的控制方法、装置、存储介质及电子装置 | |
WO2024036804A1 (zh) | 意图指令的确定方法及装置、存储介质及电子装置 | |
WO2024001189A1 (zh) | 食物存储信息的确定方法及装置、存储介质及电子装置 | |
WO2024021546A1 (zh) | 行为偏好表的生成方法和装置、存储介质及电子装置 | |
WO2023168853A1 (zh) | 使用意图的预测方法和装置、存储介质及电子装置 | |
WO2024066089A1 (zh) | 烹饪计划的确定方法及装置、存储介质及电子装置 | |
CN114915514B (zh) | 意图的处理方法和装置、存储介质及电子装置 | |
CN115200146A (zh) | 关闭指令的发送方法和装置、存储介质及电子装置 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 22955573 Country of ref document: EP Kind code of ref document: A1 |