WO2024021407A1 - 知识图谱的更新方法和装置、存储介质及电子装置 - Google Patents

知识图谱的更新方法和装置、存储介质及电子装置 Download PDF

Info

Publication number
WO2024021407A1
WO2024021407A1 PCT/CN2022/135124 CN2022135124W WO2024021407A1 WO 2024021407 A1 WO2024021407 A1 WO 2024021407A1 CN 2022135124 W CN2022135124 W CN 2022135124W WO 2024021407 A1 WO2024021407 A1 WO 2024021407A1
Authority
WO
WIPO (PCT)
Prior art keywords
question
answer text
knowledge graph
knowledge
text
Prior art date
Application number
PCT/CN2022/135124
Other languages
English (en)
French (fr)
Inventor
邓邱伟
窦方正
刘朝振
张旭
区波
翟建光
Original Assignee
青岛海尔科技有限公司
青岛海尔智能家电科技有限公司
海尔智家股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 青岛海尔科技有限公司, 青岛海尔智能家电科技有限公司, 海尔智家股份有限公司 filed Critical 青岛海尔科技有限公司
Publication of WO2024021407A1 publication Critical patent/WO2024021407A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/02—Knowledge representation; Symbolic representation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G06F16/3329—Natural language query formulation or dialogue systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Definitions

  • the present disclosure relates to the field of smart home technology, and specifically to a knowledge graph updating method and device, a storage medium and an electronic device.
  • Embodiments of the present disclosure provide a knowledge graph updating method and device, a storage medium and an electronic device to at least solve the problem in related technologies that the knowledge graph updating efficiency is low and the knowledge graph cannot provide timely feedback on the question and answer of the target object during the updating process. Response and other issues.
  • a method for updating a knowledge graph which includes: when the first knowledge graph cannot respond to the first question and answer text corresponding to the target object, obtaining an external domain knowledge document;
  • the first question and answer text and the external domain knowledge document are input into a machine reading comprehension model to obtain a target answer text corresponding to the first question and answer text, where the machine reading comprehension model uses multiple sets of data to pass the machine
  • each set of data in the multiple sets of data includes: a question and answer text, a preset external domain knowledge document, and an answer text corresponding to the question and answer text and the preset external domain knowledge document; according to
  • the target answer text provides question and answer feedback to the target object, and the content of the first knowledge graph is updated based on the target answer text and the first question and answer text to obtain a second knowledge graph.
  • a device for updating a knowledge graph including: an acquisition module configured to: when the first knowledge graph cannot respond to the first question and answer text corresponding to the target object, Obtain external domain knowledge documents; the answering module is configured to input the first question and answer text and the external domain knowledge document into the machine reading comprehension model to obtain the target answer text corresponding to the first question and answer text, wherein, The machine reading comprehension model is trained through machine learning using multiple sets of data.
  • Each set of data in the multiple sets of data includes: question and answer text, preset external domain knowledge documents, and the question and answer text and preset The answer text corresponding to the external domain knowledge document; an update module configured to provide question and answer feedback to the target object according to the target answer text, and to update the first knowledge graph based on the target answer text and the first question and answer text. Update the content to obtain the second knowledge graph.
  • a computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the above-mentioned knowledge graph when running. Update 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-mentioned steps through the computer program.
  • Update method of knowledge graph including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the above-mentioned steps through the computer program.
  • the external domain knowledge document when the first knowledge graph cannot respond to the first question and answer text corresponding to the target object, the external domain knowledge document is obtained; the first question and answer text and the external domain knowledge document are input into the machine reading comprehension model , to obtain the target answer text corresponding to the first question and answer text, where the machine reading comprehension model is trained through machine learning using multiple sets of data.
  • Each set of data in the multiple sets of data includes: question and answer text, preset external Domain knowledge document, as well as the answer text corresponding to the question and answer text and the preset external domain knowledge document; provide question and answer feedback to the target object based on the target answer text, and update the content of the first knowledge graph based on the target answer text and the first question and answer text,
  • the second knowledge graph is obtained; the above technical solution is adopted to solve the problems in related technologies such as the low efficiency of knowledge graph update and the inability of the knowledge graph to feed back the target object's questions and answers in time for effective response during the update process.
  • Figure 1 is a schematic diagram of the hardware environment of a knowledge graph updating method according to an embodiment of the present disclosure
  • Figure 2 is a flow chart of a method for updating a knowledge graph according to an embodiment of the present disclosure
  • FIG. 3 is a structural block diagram of the Q&A system architecture according to an embodiment of the present disclosure.
  • Figure 4 is a structural block diagram of a knowledge graph updating device according to an embodiment of the present disclosure.
  • FIG. 5 is a structural block diagram of an electronic device according to an embodiment of the present disclosure.
  • a method for updating a knowledge graph is provided.
  • This knowledge graph update method 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 knowledge graph updating method can be applied to a hardware environment composed of a terminal device 102 and a server 104 as shown in FIG. 1 .
  • the server 104 is connected to the terminal device 102 through the network and can be used to provide services (such as application services, etc.) for the terminal or the client installed on the terminal.
  • the database can be set on the server or independently of the server, and is set to 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 are configured 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 102 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.
  • FIG. 1 is a flow chart of a knowledge graph updating method according to an embodiment of the present disclosure. The flow includes the following steps:
  • Step S202 Obtain external domain knowledge documents when the first knowledge graph cannot respond to the first question and answer text corresponding to the target object;
  • the corresponding external domain knowledge document is a description document whose description field is a description of various devices. That is to say, the above-mentioned external domain knowledge documents can be understood as various types of documents in the corresponding fields of the knowledge graph.
  • the acquisition method of the documents can be active input by the target object, or actively collected from the network according to the knowledge field corresponding to the current first knowledge graph. document, this disclosure is not too limited.
  • Step S204 Input the first question and answer text and the external domain knowledge document into a machine reading comprehension model to obtain a target answer text corresponding to the first question and answer text, wherein the machine reading comprehension model uses multiple A set of data is trained through machine learning.
  • Each set of data in the multiple sets of data includes: a question and answer text, a preset external domain knowledge document, and an answer text corresponding to the question and answer text and the preset external domain knowledge document. ;
  • Step S206 Provide question and answer feedback to the target object based on the target answer text, and update the content of the first knowledge graph based on the target answer text and the first question and answer text to obtain a second knowledge graph.
  • the external domain knowledge document is obtained; the first question and answer text and the external domain knowledge document are input into the machine reading comprehension model to Obtain the target answer text corresponding to the first question and answer text, in which the machine reading comprehension model is trained through machine learning using multiple sets of data.
  • Each set of data in the multiple sets of data includes: question and answer text, preset external domain knowledge documents , as well as the answer text corresponding to the question and answer text and the preset external domain knowledge document; provide question and answer feedback to the target object according to the target answer text, and update the content of the first knowledge graph based on the target answer text and the first question and answer text to obtain the second Knowledge graph; using the above technical solution, it solves the problems in related technologies such as the low efficiency of knowledge graph update, and the inability of the knowledge graph to promptly feedback the target object's questions and answers during the update process to effectively respond.
  • the question and answer text is processed and responded, and when the first knowledge graph cannot support the response to the first question and answer text, the trained machine reading comprehension model is used to process the first question and answer text and external domain knowledge documents to obtain the first question and answer.
  • the target answer text corresponding to the text is used to provide question and answer feedback to the target object, and the first knowledge graph is updated to the second knowledge graph containing new knowledge, ensuring that the update of the knowledge graph does not conflict with the use of the target object, improving Improve the update efficiency of knowledge graph.
  • the above method before updating the content of the first knowledge graph based on the target answer text and the first question and answer text, the above method further includes: counting the number of occurrences of the first question and answer text; If the number of occurrences is greater than or equal to the first preset threshold, it is determined to update the first knowledge graph online; if the number of occurrences is less than the first preset threshold, it is determined to update the first knowledge graph online. Maps are updated offline.
  • the first knowledge graph when updating the first knowledge graph, it can also be selected, used jointly, or updated conditionally according to needs.
  • high-frequency user questions equivalent to the above-mentioned first question and answer text
  • the knowledge Equivalent to the above target answer text
  • the question and answer efficiency will be higher.
  • low-frequency user questions which may only appear once or twice
  • only offline updates can be used.
  • online updates when online updates have an impact on the performance of the question and answer system, they can be discarded or a compromise strategy can be adopted.
  • the target object's demand frequency for the answer text corresponding to the first Q&A text based on the occurrence of the first Q&A text, so that the length of the answer cycle can be determined based on the demand frequency, and the first Q&A with high demand frequency can be determined.
  • the answer text corresponding to the text is updated in the knowledge graph in real time, ensuring that the knowledge graph can achieve rapid response to the target object in a short period of time.
  • the answer text corresponding to the first question and answer text can be temporarily cached Or store it, and when the target object is not using the knowledge graph for information query, the knowledge graph is updated using the stored answer text, thereby ensuring the target object's use experience while improving the update efficiency of the corresponding knowledge graph during idle time, so that the knowledge graph can be used in The target object can cover as much information as possible when used.
  • the content of the first knowledge graph is updated based on the target answer text and the first question and answer text.
  • the above method further includes: determining the first knowledge graph for graph update. Available space resources corresponding to the storage space; when the available space resources are less than or equal to the first preset resource threshold, and the data traffic corresponding to the first question and answer text is greater than the preset traffic threshold, it is determined that the available space is The resource only stores the target answer text corresponding to the first question and answer text, and uses the target answer text to update the content of the first knowledge graph; when the available space resources are greater than the first preset resource threshold, and the third When the data traffic corresponding to a question and answer text is less than or equal to the preset traffic threshold, it is determined that the first question and answer text, the target answer text and the external domain knowledge document are correspondingly stored in the available space resource, And use the first question and answer text, the target answer text and the external domain knowledge document to update the content of the first knowledge graph.
  • the amount of data traffic that the device or application platform can support and the corresponding space resources can also be determined to determine the amount of data traffic that the device or application platform can support and the corresponding space resources. How to update and use the data generated in the update method, so that the normal operating functions of the device or application platform are not affected while ensuring the update effect, and use the assistance of data to achieve the accuracy of the question and answer function corresponding to the device or application platform. promote.
  • updating the content of the first knowledge graph based on the target answer text and the first question and answer text further includes: obtaining preset question category rules; The rules determine the question category of the first question and answer text output by the machine reading comprehension model, and the corresponding frequency of the first question and answer text in each question category; use the frequency as a weight to calculate the first question and answer text.
  • An update value corresponding to a question and answer text when there are multiple first question and answer texts at the same time, determine to use the target first question and answer text with the largest update value and the target answer text pair corresponding to the target first question and answer text.
  • the first knowledge graph performs content updating.
  • using the frequency as a weight to calculate the update value corresponding to the first question and answer text includes: based on the popularity values corresponding to different question categories in the first knowledge graph, determining the The first question and answer text corresponds to the target popularity value of the question category; the target popularity value is multiplied by the frequency, and the value corresponding to the product result is used as the update value corresponding to the first question and answer text.
  • the above method before inputting the first question and answer text and the external domain knowledge document into the machine reading comprehension model to obtain the target answer text corresponding to the first question and answer text, the above method further includes: When new device knowledge needs to be added to the first knowledge graph, a third knowledge graph corresponding to the new device knowledge is obtained; and the third knowledge graph is added to the external domain knowledge document.
  • the third knowledge graph when there is a third knowledge graph whose content is to be updated to the first knowledge graph, it means that the first knowledge graph has expanded the device support range. At this time, in order to improve the update efficiency, the third knowledge graph is added to the external domain. For knowledge documents, when the target object uses the data corresponding to the third knowledge graph, the first knowledge graph is updated, or when the target object does not use the first knowledge graph, the third knowledge graph is used to update the first knowledge graph offline. renew.
  • question and answer feedback is provided to the target object based on the target answer text, and the content of the first knowledge graph is updated based on the target answer text and the first question and answer text, and the first knowledge graph is obtained.
  • the above method also includes: using the second knowledge graph to replace the first knowledge graph that cannot respond to the first question and answer text corresponding to the target object as the current knowledge graph; and performing recognition processing on the first question and answer text.
  • a common implementation method for machine question answering is knowledge base question answering (KBQA), which is based on knowledge base (KB) or knowledge graph (KG). Its principle is to build a human knowledge system that can be understood by machines, and then use natural language understanding (NLU) and Knowledge bases are aligned to find relevant answers.
  • NLU natural language understanding
  • building a knowledge base (graph) in a complex field requires large-scale domain data and multiple subdivision steps, from knowledge acquisition, knowledge mining, knowledge storage to knowledge reasoning, and this link requires a large amount of resource costs. Including but not limited to the large labor costs of data and model production, the hardware costs of knowledge storage, and the cost of computing resources used for training various models and inference on the link.
  • the present disclosure designs a question-and-answer system architecture based on knowledge graphs assisted by machine reading comprehension technology.
  • This architecture not only uses machine reading comprehension technology to achieve question and answer questions about knowledge not covered in the knowledge graph, but also based on machine reading.
  • the result of understanding integrates new knowledge into the knowledge graph and realizes online updating of the knowledge graph.
  • This architecture solves the problem of still ensuring the Q&A effect without updating the knowledge graph offline.
  • the offline update of the knowledge graph can choose the appropriate time as needed, and can use the user's input during Q&A and machine reading comprehension answers as additional auxiliary information.
  • Figure 3 is a structural block diagram of a question and answer system architecture according to an embodiment of the present disclosure.
  • the above question and answer system architecture includes: an input module 32, a machine reading comprehension module 34, a knowledge graph module 36, an update module 38, and a storage feedback module 40;
  • the input module 32 includes user input and external domain knowledge, where the user input is the text entered by the user for question and answer, and the external domain knowledge is various documents in the domain, such as the manual domain is the manual documents of various devices.
  • the knowledge graph module 36 for question and answer query or reasoning it is necessary to use intention and entity attribute recognition technology to conduct refined analysis and extraction of the user's intention and attention information.
  • Input and input processing will be recorded through the storage module as system feedback information, which can assist in updating and upgrading the knowledge graph in the future.
  • machine reading comprehension module 34 this module may introduce various machine reading comprehension technologies based on deep learning, reinforcement learning, etc., such as neural networks, interactive machine reading comprehension, etc.
  • the input of the module is user question text and external domain knowledge documents, and the output is the answer A corresponding to the user input question.
  • MRC represents a type of machine reading comprehension model that can input user questions I and external domain knowledge documents Set D gives the answer text that the user needs.
  • the update module 38 includes two parts, namely online update and offline update.
  • This functional module can be selected according to needs, used jointly or conditionally. For example, if online updates have an impact on the system performance, they can be discarded, or a compromise strategy can be adopted. For high-frequency user questions, online updates are performed and the knowledge is updated to the knowledge graph. The question and answer efficiency will be higher, while for low-frequency questions , may only appear once or twice, then you can just use offline updates.
  • Knowledge graph online updates can use various existing methods, but the principle that needs to be ensured is that it does not affect system applications. Therefore, subgraph update or distributed update are two feasible directions.
  • a storage feedback module 40 which can be designed in detail according to the traffic of the question and answer system, such as storage solutions, storage device selection, etc.
  • storage strategy there are three methods. The first one is the full storage strategy. This strategy is suitable for situations where the traffic is small and the storage resources are sufficient. All user inputs are stored and the information is fully utilized when updating the knowledge graph offline. The second one is only storing the required machines. The questions answered by reading comprehension, that is, the knowledge that cannot be covered by the knowledge graph, only consider these unknown or unmatched knowledge when updating offline.
  • This method is suitable for situations with large traffic and insufficient storage resources; the third method not only stores Machine reading comprehension answers questions, and also stores the frequency of all categories of questions.
  • the former is used, and the latter is used as auxiliary information, such as it can be regarded as a representation of importance, as a weight, for focused updates to detect the knowledge graph.
  • the optional examples of this disclosure by using MRC technology, can introduce external knowledge or unsupervised knowledge in this field to the system without updating the knowledge graph, improving the user Q&A experience and satisfying scenarios where the knowledge graph cannot cover the situation.
  • User Q&A User Q&A.
  • this type of external knowledge can be updated online into the knowledge graph through machine reading comprehension and related technologies, and through storage strategies, it can also be used as auxiliary information when the knowledge graph is updated offline.
  • This embodiment can use machine reading comprehension (MRC) technology to obtain external knowledge to meet the user's Q&A needs when the knowledge graph cannot obtain the user's Q&A.
  • MRC machine reading comprehension
  • the acquired external knowledge is used to update the knowledge graph online.
  • new knowledge It can also be recorded and used as auxiliary information when the knowledge graph is updated offline. Therefore, the domain knowledge graph no longer needs to spend a lot of resources to build a super complete domain knowledge graph at the beginning, which is relatively complete and can even support basic services. Instead, it needs to continuously iterate through the above process to finally achieve a more complete domain knowledge graph.
  • 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.
  • Figure 4 is a structural block diagram of a knowledge graph updating device according to an embodiment of the present disclosure. As shown in Figure 4, it includes:
  • the acquisition module 42 is configured to acquire external domain knowledge documents when the first knowledge graph cannot respond to the first question and answer text corresponding to the target object;
  • the answer module 44 is configured to input the first question and answer text and the external domain knowledge document into a machine reading comprehension model to obtain a target answer text corresponding to the first question and answer text, wherein the machine reading comprehension model It is trained through machine learning using multiple sets of data.
  • Each set of data in the multiple sets of data includes: question and answer text, a preset external domain knowledge document, and the correspondence between the question and answer text and the preset external domain knowledge document. the answer text;
  • the update module 46 is configured to provide question and answer feedback to the target object based on the target answer text, and update the content of the first knowledge graph based on the target answer text and the first question and answer text to obtain second knowledge. Map.
  • the external domain knowledge document is obtained; the first question and answer text and the external domain knowledge document are input into the machine reading comprehension model to Obtain the target answer text corresponding to the first question and answer text, in which the machine reading comprehension model is trained through machine learning using multiple sets of data.
  • Each set of data in the multiple sets of data includes: question and answer text, preset external domain knowledge documents , as well as the answer text corresponding to the question and answer text and the preset external domain knowledge document; provide question and answer feedback to the target object according to the target answer text, and update the content of the first knowledge graph based on the target answer text and the first question and answer text to obtain the second Knowledge graph; using the above technical solution, it solves the problems in related technologies such as the low efficiency of knowledge graph update, and the inability of the knowledge graph to promptly feedback the target object's questions and answers during the update process to effectively respond.
  • the question and answer text is processed and responded, and when the first knowledge graph cannot support the response to the first question and answer text, the trained machine reading comprehension model is used to process the first question and answer text and external domain knowledge documents to obtain the first question and answer.
  • the target answer text corresponding to the text is used to provide question and answer feedback to the target object, and the first knowledge graph is updated to the second knowledge graph containing new knowledge, ensuring that the update of the knowledge graph does not conflict with the use of the target object, improving Improve the update efficiency of knowledge graph.
  • the above device further includes: a statistical module configured to count the number of occurrences of the first question and answer text; when the number of occurrences is greater than or equal to a first preset threshold, determine whether the number of occurrences of the first question and answer text is The first knowledge graph is updated online; when the number of occurrences is less than the first preset threshold, it is determined to update the first knowledge graph offline.
  • a statistical module configured to count the number of occurrences of the first question and answer text; when the number of occurrences is greater than or equal to a first preset threshold, determine whether the number of occurrences of the first question and answer text is The first knowledge graph is updated online; when the number of occurrences is less than the first preset threshold, it is determined to update the first knowledge graph offline.
  • the above device further includes: a storage module configured to determine available space resources corresponding to the storage space used by the first knowledge graph for graph update; when the available space resources are less than or equal to the first When the resource threshold is preset and the data traffic corresponding to the first question and answer text is greater than the preset traffic threshold, it is determined that only the target answer text corresponding to the first question and answer text is stored in the available space resources, and the target answer text is used.
  • a storage module configured to determine available space resources corresponding to the storage space used by the first knowledge graph for graph update; when the available space resources are less than or equal to the first
  • the resource threshold is preset and the data traffic corresponding to the first question and answer text is greater than the preset traffic threshold, it is determined that only the target answer text corresponding to the first question and answer text is stored in the available space resources, and the target answer text is used.
  • the answer text updates the content of the first knowledge graph; when the available space resources are greater than the first preset resource threshold, and the data traffic corresponding to the first question and answer text is less than or equal to the preset traffic threshold, determine Store the first question and answer text, the target answer text and the external domain knowledge document correspondingly in the available space resource, and use the first question and answer text, the target answer text and the external domain knowledge document
  • the knowledge document updates the content of the first knowledge graph.
  • the above device further includes: a category module configured to obtain preset question category rules; and determine the question of the first question and answer text output by the machine reading comprehension model through the question category rule. category, and the corresponding frequency of the first question and answer text in each question category; using the frequency as a weight, calculate the update value corresponding to the first question and answer text; when there are multiple first question and answer texts at the same time.
  • a category module configured to obtain preset question category rules; and determine the question of the first question and answer text output by the machine reading comprehension model through the question category rule.
  • category and the corresponding frequency of the first question and answer text in each question category
  • using the frequency as a weight
  • the above category module is further configured to determine the target popularity value of the question category corresponding to the first question and answer text based on the popularity values corresponding to different question categories in the first knowledge graph; The target popularity value is multiplied by the frequency, and the value corresponding to the product result is used as the updated value corresponding to the first question and answer text.
  • the above device further includes: a graph module configured to obtain a third knowledge graph corresponding to the new device knowledge when new device knowledge needs to be added to the first knowledge graph; Add the third knowledge graph to the external domain knowledge document.
  • the above device further includes: a determination module configured to use the second knowledge graph to replace the first knowledge graph that cannot respond to the first question and answer text corresponding to the target object as the current knowledge graph; Perform recognition processing on the first question and answer text to obtain query text for querying in the graph, where the recognition processing is used to indicate the determination of the intended entity of the first question and answer text and the attributes corresponding to the first question and answer text.
  • Information input the query text into the current knowledge graph and determine the query result, wherein the query result is used to indicate whether the updated graph effectively supports the response to the first question and answer text.
  • An embodiment of the present disclosure also provides a storage medium that includes a stored program, wherein the method of any of the above items is executed when the program is run.
  • the above-mentioned storage medium may be configured to store program codes for performing the following steps:
  • S3 Provide question and answer feedback to the target object according to the target answer text, and update the content of the first knowledge graph based on the target answer text and the first question and answer text to obtain a second knowledge graph.
  • 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 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.
  • the electronic device includes a memory 702 and a processor 704.
  • the memory 702 stores a computer program
  • the processor 704 is configured to execute any of the above method embodiments through the computer program. step.
  • the above-mentioned electronic device may be located in at least one network device among multiple network devices of the computer network.
  • the above-mentioned processor may be configured to perform the following steps through a computer program:
  • S3 Provide question and answer feedback to the target object according to the target answer text, and update the content of the first knowledge graph based on the target answer text and the first question and answer text to obtain a second knowledge graph.
  • the structure shown in Figure 5 is only illustrative, and the electronic device can also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a handheld computer, and a mobile Internet device (Mobile Internet Devices, MID), PAD and other terminal equipment.
  • FIG. 5 does not limit the structure of the above-mentioned electronic device.
  • the electronic device may also include more or fewer components (such as network interfaces, etc.) than shown in FIG. 5 , or have a different configuration than that shown in FIG. 5 .
  • the memory 702 can be used to store software programs and modules, such as program instructions/modules corresponding to the communication connection methods and devices in the embodiments of the present disclosure.
  • the processor 704 executes various software programs and modules by running the software programs and modules stored in the memory 702. Function application and data processing, that is, realizing the above communication connection method.
  • Memory 702 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 702 may further include memory located remotely relative to the processor 704, and these remote memories may be connected to the terminal through a network.
  • the above-mentioned networks include but are not limited to the Internet, intranets, local area networks, mobile communication networks and combinations thereof.
  • the memory 702 may include, but is not limited to, the acquisition module 42 , the response module 44 , and the update module 46 in the communication connection device. In addition, it may also include but is not limited to other modular units in the above-mentioned communication connection device, which will not be described again in this example.
  • the above-mentioned transmission device 706 is used to receive or send data via a network.
  • Specific examples of the above-mentioned network may include wired networks and wireless networks.
  • the transmission device 706 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices and routers through network cables to communicate with the Internet or a local area network.
  • the transmission device 1106 is a radio frequency (Radio Frequency, RF) module, which is used to communicate with the Internet wirelessly.
  • RF Radio Frequency
  • the above-mentioned electronic device also includes: a display 708 configured to display the above-mentioned knowledge graph; and a connection bus 710 configured to connect various module components in the above-mentioned electronic device.
  • the above 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), Various media that can store program code, such as mobile hard drives, magnetic disks, or optical disks.
  • ROM read-only memory
  • RAM random access memory
  • program code such as mobile hard drives, magnetic disks, or optical disks.
  • 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. , optionally, 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 in a sequence different from that herein.
  • the steps shown or described are performed either individually as individual integrated circuit modules, or as multiple modules or steps among them as a single integrated circuit module. As such, the present disclosure is not limited to any specific combination of hardware and software.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Human Computer Interaction (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

本公开提供了一种知识图谱的更新方法和装置、存储介质及电子装置，涉及智慧家庭技术领域，该知识图谱的更新方法包括：在第一知识图谱无法对目标对象对应的第一问答文本进行响应的情况下，获取外部领域知识文档；将第一问答文本与外部领域知识文档输入到机器阅读理解模型中，以得到第一问答文本对应的目标回答文本，其中，机器阅读理解模型为使用多组数据通过机器学习训练出的，多组数据中的每组数据均包括：问答文本，预设的外部领域知识文档，以及问答文本和预设的外部领域知识文档对应的回答文本；根据目标回答文本向目标对象进行问答反馈，并基于目标回答文本和第一问答文本对第一知识图谱进行内容更新，得到第二知识图谱。

Description

知识图谱的更新方法和装置、存储介质及电子装置
本公开要求于2022年07月29日提交中国专利局、申请号为202210911171.4、发明名称“知识图谱的更新方法和装置、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。
技术领域
本公开涉及智慧家庭技术领域，具体而言，涉及一种知识图谱的更新方法和装置、存储介质及电子装置。
背景技术
相关技术中的，在领域知识图谱支撑的问答系统中，无法通过单一的知识图谱实现整个问答系统对现有领域全部知识的问答反馈，并且传统的基于人工设计领域知识图谱库的智能问答系统，已无法满足日益增长的用户需求，其知识图谱库的上线周期长，需实时维护，维护成本高，更新效率低下，并且在更新期间无法对用户进行问答支持，影响用户对于图谱的正常使用。
因此，针对相关技术中，知识图谱更新效率低下，知识图谱在更新过程中无法及时反馈目标对象的问答进行有效响应等问题，尚未提出有效的解决方案。
发明内容
本公开实施例提供了一种知识图谱的更新方法和装置、存储介质及电子装置，以至少解决相关技术中，知识图谱更新效率低下，知识图谱在更新过程中无法及时反馈目标对象的问答进行有效响应等问题。
根据本公开实施例的一个实施例，提供了一种知识图谱的更新方法，包括：在第一知识图谱无法对目标对象对应的第一问答文本进行响应的情况下，获取外部领域知识文档；将所述第一问答文本与所述外部领域知识文档输入到机器阅读理解模型中，以得到所述第一问答文本对应的目标回答文本，其中，所述 机器阅读理解模型为使用多组数据通过机器学习训练出的，所述多组数据中的每组数据均包括：问答文本，预设的外部领域知识文档，以及所述问答文本和预设的外部领域知识文档对应的回答文本；根据所述目标回答文本向所述目标对象进行问答反馈，并基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，得到第二知识图谱。
根据本公开实施例的另一个实施例，还提供了一种知识图谱的更新装置，包括：获取模块，设置为在第一知识图谱无法对目标对象对应的第一问答文本进行响应的情况下，获取外部领域知识文档；回答模块，设置为将所述第一问答文本与所述外部领域知识文档输入到机器阅读理解模型中，以得到所述第一问答文本对应的目标回答文本，其中，所述机器阅读理解模型为使用多组数据通过机器学习训练出的，所述多组数据中的每组数据均包括：问答文本，预设的外部领域知识文档，以及所述问答文本和预设的外部领域知识文档对应的回答文本；更新模块，设置为根据所述目标回答文本向所述目标对象进行问答反馈，并基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，得到第二知识图谱。
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知识图谱的更新方法。
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的知识图谱的更新方法。
在本公开实施例中，在第一知识图谱无法对目标对象对应的第一问答文本进行响应的情况下，获取外部领域知识文档；将第一问答文本与外部领域知识文档输入到机器阅读理解模型中，以得到第一问答文本对应的目标回答文本，其中，机器阅读理解模型为使用多组数据通过机器学习训练出的，多组数据中的每组数据均包括：问答文本，预设的外部领域知识文档，以及问答文本和预设的外部领域知识文档对应的回答文本；根据目标回答文本向目标对象进行问答反馈，并基于目标回答文本和第一问答文本对第一知识图谱进行内容更新，得到第二知识图 谱；采用上述技术方案，解决了相关技术中，知识图谱更新效率低下，知识图谱在更新过程中无法及时反馈目标对象的问答进行有效响应等问题。
附图说明
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是本公开实施例的一种知识图谱的更新方法的硬件环境示意图；
图2是根据本公开实施例的知识图谱的更新方法的流程图；
图3是根据本公开实施例的问答系统架构的结构框图；
图4是根据本公开实施例的一种知识图谱的更新装置的结构框图；
图5是根据本公开实施例的一种电子装置的结构框图。
具体实施方式
为了使本技术领域的人员更好地理解本公开方案，下面将结合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、系统、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。
根据本公开实施例的一个方面，提供了一种知识图谱的更新方法。该知识图谱的更新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，上述知识图谱的更新方法可以应用于如图1所示的由终端设备102和服务器104所构成的硬件环境中。如图1所示，服务器104通过网络与终端设备102进行连接，可用于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102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知识图谱的更新方法，应用于上述计算机终端，图2是根据本公开实施例的知识图谱的更新方法的流程图，该流程包括如下步骤：
步骤S202，在第一知识图谱无法对目标对象对应的第一问答文本进行响应的情况下，获取外部领域知识文档；
可选的，当第一知识图谱为与设备功能说明有关的图谱时，对应的外部领域 知识文档为说明书领域为各种设备的说明书文档。即上述外部领域知识文档可以理解为是知识图谱对应领域内的各类文档，该文档的获取方式可以是目标对象主动输入，或者是根据当前第一知识图谱对应的知识领域从网络中主动收集的文档，对此本公开不作过多限定。
步骤S204，将所述第一问答文本与所述外部领域知识文档输入到机器阅读理解模型中，以得到所述第一问答文本对应的目标回答文本，其中，所述机器阅读理解模型为使用多组数据通过机器学习训练出的，所述多组数据中的每组数据均包括：问答文本，预设的外部领域知识文档，以及所述问答文本和预设的外部领域知识文档对应的回答文本；
步骤S206，根据所述目标回答文本向所述目标对象进行问答反馈，并基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，得到第二知识图谱。
通过上述步骤，在第一知识图谱无法对目标对象对应的第一问答文本进行响应的情况下，获取外部领域知识文档；将第一问答文本与外部领域知识文档输入到机器阅读理解模型中，以得到第一问答文本对应的目标回答文本，其中，机器阅读理解模型为使用多组数据通过机器学习训练出的，多组数据中的每组数据均包括：问答文本，预设的外部领域知识文档，以及问答文本和预设的外部领域知识文档对应的回答文本；根据目标回答文本向目标对象进行问答反馈，并基于目标回答文本和第一问答文本对第一知识图谱进行内容更新，得到第二知识图谱；采用上述技术方案，解决了相关技术中，知识图谱更新效率低下，知识图谱在更新过程中无法及时反馈目标对象的问答进行有效响应等问题，通过第一知识图谱对目标对象的第一问答文本进行处理响应，并在第一知识图谱无法支撑响应第一问答文本的情况下，利用训练好的机器阅读理解模型对第一问答文本和外部领域知识文档进行处理，得出该第一问答文本对应的目标回答文本，使用该目标回答文本向目标对象进行问答反馈，并将第一知识图谱更新为包含新知识的第二知识图谱，保证了知识图谱更新与目标对象的使用不冲突，提升了知识图谱的更新效率。
在一个示例性实施例中，基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新之前，上述方法还包括：统计所述第一问答文本的出现次数；在所述出现次数大于或者等于第一预设阈值的情况下，确定对所述第一知识图谱进行在线更新；在所述出现次数小于第一预设阈值的情况下，确定对所述第一知识图谱进行离线更新。
换句话说，在对第一知识图谱更新时，还可以根据需求取舍、联合使用或有条件进行更新，例如，对于高频用户问题(相当于上述第一问答文本)进行在线更新，将知识(相当于上述目标回答文本)更新到知识图谱，问答效率会更高一些，而对于低频用户问题，可能仅出现一两次，那可以仅使用离线更新。此外，当在线更新对问答系统效果有影响，可以舍去，或者采取折中的策略。
通过上述实施方式可以基于第一问答文本的出现情况，识别目标对象对于该第一问答文本对应回答文本的需求频率，从而可以基于需求频率确定回答周期的时间长短，将需求频率高的第一问答文本对应回答文本实时在知识图谱中更新，保证知识图谱可以在较短的时间内实现对目标对象的快速响应，当需求频率低时，则可以暂时将对应的第一问答文本对应回答文本进行缓存或者存储，选择目标对象未使用知识图谱进行信息查询时，对知识图谱利用存储的回答文本进行更新，进而在保证目标对象使用体验的同时，提升空闲时间对应知识图谱的更新效率，使得知识图谱在目标对象使用时可以尽可能多的覆盖更大的信息范围。
在一个示例性实施例中，基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，上述方法还包括：确定所述第一知识图谱用于进行图谱更新的存储空间对应的可用空间资源；在所述可用空间资源小于或者等于第一预设资源阈值，且所述第一问答文本对应的数据流量大于预设流量阈值的情况下，确定在所述可用空间资源仅存储所述第一问答文本对应的目标回答文本，使用所述目标回答文本对所述第一知识图谱进行内容更新；在所述可用空间资源大于第一预设资源阈值，且所述第一问答文本对应的数据流量小于或者等于预设流量阈值的情况下，确定在所述可用空间资源中将所述第一问答文本、所述目标回答文本以及所述外部领域知识文档进行对应存储，并使用所述第一问答文本、所述目标回答文本以及所述外部领域知识文档对所述第一知 识图谱进行内容更新。
通过上述实施例，在将上述知识图谱的更新方法应用到对应的设备或者应用平台中时，还可以通过确定设备或者应用平台所能支持的数据流量大小以及相应的空间资源来确定对知识图谱的更新方法中产生的数据如何进行更新使用，进而在保证更新效果的同时，使得设备或者应用平台对应的正常运行功能不受影响，利用数据的辅助实现对设备或者应用平台对应问答功能的准确率的提升。
在一个示例性实施例中，基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，上述方法还包括：获取预设的问题类别规则；通过所述问题类别规则确定利用所述机器阅读理解模型输出的所述第一问答文本的问题类别，以及所述第一问答文本在每一种问题类别中对应的频率；将所述频率作为权重，计算所述第一问答文本对应的更新值；在所述第一问答文本同时存在多个的情况下，确定使用所述更新值最大的目标第一问答文本以及所述目标第一问答文本对应的目标回答文本对所述第一知识图谱进行内容更新。
可以理解的是，当上述方法应用于相关问答系统时，可以根据问答系统的流量进行更新的进一步设置，可选的，从存储策略上来说，可以看作三种方法：第一种，全部存储策略，该策略适用于流量较小、存储资源充足的情况，将用户所有输入、输出均做存储，在离线更新知识图谱时，充分利用这些信息；第二种，仅存储需要机器阅读理解回答的问题，即知识图谱无法覆盖的知识，在离线更新时仅考虑这些未知或未匹配的知识，该方法适用于流量较大、存储资源不太充足的情况；第三种，不仅存储机器阅读理解回答的问题，还存储所有类别的问题的频率，前者用于，后者作为辅助信息，如：可以视为重要性的表征，作为权重，有重点的更新检测知识图谱。
在一个示例性实施例中，将所述频率作为权重，计算所述第一问答文本对应的更新值，包括：基于所述第一知识图谱中的不同问题类别分别对应的热度值，确定所述第一问答文本对应问题类别的目标热度值；将所述目标热度值与所述频率相乘，将乘积结果对应的值作为所述第一问答文本对应的更新值。
可以理解的是，由于第一知识图谱中包含的问题类别多种多样，有些问题通过知识图谱自身的数据便可以初步支撑问答，当详细问答时才需要通过对知识图 谱进行更新后才能支持，因此，为了避免频繁的更新导致知识图谱中的数据异常或者使用异常，还可以通过确定第一问答文本对应的更新值，使用更新值对待更新的问答文本对应的数据集合进行排列，从更新值最大、需求最高的第一问答文本对应数据依次更新，完成当知识图谱的全面更新。
在一个示例性实施例中，将所述第一问答文本与所述外部领域知识文档输入到机器阅读理解模型中，以得到所述第一问答文本对应的目标回答文本之前，上述方法还包括：在所述第一知识图谱中需要增加新的设备知识的情况下，获取所述新的设备知识对应的第三知识图谱；将所述第三知识图谱添加至所述外部领域知识文档。
可以理解的是，当存在待将内容更新至第一知识图谱的第三知识图谱时，说明第一知识图谱扩大的设备支撑范围，此时为了提升更新效率，将第三知识图谱添加至外部领域知识文档，当目标对象使用到第三知识图谱对应的数据时，再进行第一知识图谱的更新，或者在目标对象未使用第一知识图谱时，利用第三知识图谱对第一知识图谱进行离线更新。
在一个示例性实施例中，根据所述目标回答文本向所述目标对象进行问答反馈，并基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，得到第二知识图谱之后，上述方法还包括：使用所述第二知识图谱替换无法对目标对象对应的第一问答文本进行响应的第一知识图谱作为当前知识图谱；对所述第一问答文本进行识别处理，得到用于在图谱中进行查询的查询文本，其中，所述识别处理用于指示确定所述第一问答文本的意图实体以及所述第一问答文本对应的属性信息；将所述查询文本输入到所述当前知识图谱中，确定出查询结果，其中，所述查询结果用于指示更新后的图谱是否有效支持对所述第一问答文本的响应。
由于知识图谱在使用时，需要确定目标对象的意图以及实体属性要求，因此，在完成知识图谱的更新之后，通过问答测试确定当前更新后的知识图谱对应之前的问答能否有效的支持，进一步验证知识图谱的准确性，保证知识图谱使用后可以精准的反馈目标对象的问答。
为了更好的理解上述知识图谱的更新方法的过程，以下再结合可选实施例对 上述知识图谱的更新的实现方法流程进行说明，但不用于限定本公开实施例的技术方案。
机器问答的常见实现方法是知识库问答(KBQA)，基于知识库(KB)或知识图谱(KG)的问答，其原理是构建机器可理解的人类知识体系，然后通过自然语言理解(NLU)和知识库对齐从而找到相关答案。但要构建一个复杂领域的知识库(图谱)需要大规模的领域数据和多个细分步骤，从知识获取、知识挖掘、知识存储到知识推理，而该链路上需要耗费大量的资源成本，包括但不限于数据和模型制作的大量人工成本，知识存储的硬件成本以及链路上训练各类模型和推理使用的计算资源成本等。另外，在基于传统KBQA的智能硬件说明书场景下，每一台设备的升级，每一类设备的新品发布等，都需要将最新的设备知识融合到KB或KG中，带来了KBQA系统的频繁更新上线操作。
基于上述分析，现有两类方法，一种从知识图谱构建角度采取提前构建较完备知识图谱，按需更新的方法，一种通过一些技术从提升用户输入和知识图谱的匹配度角度。前者，存在构建和更新周期长的问题，导致拉长系统上线周期，构建与维护成本高、时效低等缺点。后者并未解决前者的问题，但是也能在一定程度上提高问答准确率，但是该类方法并未引入可获取到的大量知识图谱外的领域知识。
作为一种可选的实施例，本公开设计了通过机器阅读理解技术辅助基于知识图谱的问答系统架构，该架构不仅通过机器阅读理解技术实现知识图谱中没有覆盖的知识的问答，还基于机器阅读理解的结果将新知识融入知识图谱中，实现知识图谱的在线更新。该架构解决了对知识图谱不做离线更新的前提下，仍能保证问答效果，而知识图谱的离线更新可以按需选择合适的时机，且可以将用户问答时的输入和机器阅读理解回答作为额外的辅助信息。
可选的，图3是根据本公开实施例的问答系统架构的结构框图，上述问答系统架构包括：输入模块32、机器阅读理解模块34、知识图谱模块36、更新模块38、存储反馈模块40；
可选的，输入模块32，包括用户输入和外部领域知识，其中，用户输入为用 户问答输入的文本，外部领域知识为领域内的各类文档，如说明书领域为各种设备的说明书文档。需要说明的是，当用户输入到知识图谱模块36进行问答查询或推理时，需要通过意图和实体属性识别技术，对用户意图和关注的信息做精细化的分析与提取。输入以及对输入的处理，都会通过存储模块进行记录，作为系统反馈信息，未来能够辅助知识图谱的更新升级等。
可选的，机器阅读理解模块34，该模块可能引入各类基于深度学习、强化学习等的机器阅读理解技术，如图神经网络、交互式机器阅读理解等。模块的输入是用户问题文本和外部领域知识文档，输出为对应用户输入问题的答案A，则该模块可用公式表示为：A＝MRC(I，D)；其中，I＝w 1，w 2，...，w m，D＝{d 1，d 2，...，d n}分别为用户输入和文档集合，MRC表示一类机器阅读理解模型能够通过输入用户问题I和外部领域知识文档集合D，给出用户需要的回答文本。
可选的，更新模块38包括两部分，即在线更新和离线更新，该功能模块可以根据需求取舍、联合使用或有条件使用。比如，若在线更新对系统效果有影响，可以舍去，或者采取折中的策略，对于高频用户问题进行在线更新，将知识更新到知识图谱，问答效率会更高一些，而对于低频的问题，可能仅出现一两次，那可以仅使用离线更新。知识图谱在线更新可以采用各类现有的方法，但是需要保证的原则是，不影响系统应用，因此，子图更新或分布式更新是可行的两个方向。
可选的，存储反馈模块40，该模块可以根据问答系统的流量进行细节设计，如存储方案，存储设备选择等。从存储策略上来说，可以看三种方法。第一种，全部存储策略，该策略适用于流量较小、存储资源充足的情况，将用户所有输入均做存储，在离线更新知识图谱时，充分利用这些信息；第二种，仅存储需要机器阅读理解回答的问题，即知识图谱无法覆盖的知识，在离线更新时仅考虑这些未知或未匹配的知识，该方法适用于流量较大、存储资源不太充足的情况；第三种，不仅存储机器阅读理解回答的问题，还存储所有类别的问题的频率，前者用于，后者作为辅助信息，如可以视为重要性的表征，作为权重，有重点的更新检测知识图谱。
综上，本公开可选示例，通过使用MRC技术，能够在不更新知识图谱的情况 下，向系统引入该领域的外部知识或无监督知识，提升用户问答的体验，满足知识图谱无法覆盖场景下的用户问答。同时，这类外部知识经过机器阅读理解和相关技术能够在线更新到知识图谱中，且通过存储策略，知识图谱离线更新时，也可作为辅助信息。
作为一种可选示例，在实际应用中，当新的设备投放到市场，进入用户家庭，知识图谱就需要进行一次全链路的更新，即使是在知识图谱自动化生产平台的辅助下，如此长的构建或更新链路会导致资源、时间成本的累加，以及性能的逐级衰减。在设备说明书问答系统这种较高频次的迭代更新场景下，这种现象尤为严重。而MRC技术可以将新设备知识作为领域外部知识，无需提取结构化信息，导入知识图谱即可支撑用户问答需求。而在线和离线分支更新，能够保证知识图谱本身知识覆盖率的不断提升，满足相同用户输入下的高效率问答。
本实施例能够通过机器阅读理解(MRC)技术，在知识图谱无法获取用户问答的情况下，获取外部知识满足用户问答需求，同时，利用获取的外部知识对知识图谱进行在线更新，另外，新知识也可以记录下来，当知识图谱离线更新时作为辅助信息。因此，领域知识图谱不再需要在一开始就耗费大量资源去构建一个超级完备的，相对完备，甚至能够支持基础服务即可，而是通过上述过程不断迭代最终实现一个较为完备的领域知识图谱。
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。
图4是根据本公开实施例的一种知识图谱的更新装置的结构框图。如图4所示，包括：
获取模块42，设置为在第一知识图谱无法对目标对象对应的第一问答文本进 行响应的情况下，获取外部领域知识文档；
回答模块44，设置为将所述第一问答文本与所述外部领域知识文档输入到机器阅读理解模型中，以得到所述第一问答文本对应的目标回答文本，其中，所述机器阅读理解模型为使用多组数据通过机器学习训练出的，所述多组数据中的每组数据均包括：问答文本，预设的外部领域知识文档，以及所述问答文本和预设的外部领域知识文档对应的回答文本；
更新模块46，设置为根据所述目标回答文本向所述目标对象进行问答反馈，并基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，得到第二知识图谱。
通过上述装置，在第一知识图谱无法对目标对象对应的第一问答文本进行响应的情况下，获取外部领域知识文档；将第一问答文本与外部领域知识文档输入到机器阅读理解模型中，以得到第一问答文本对应的目标回答文本，其中，机器阅读理解模型为使用多组数据通过机器学习训练出的，多组数据中的每组数据均包括：问答文本，预设的外部领域知识文档，以及问答文本和预设的外部领域知识文档对应的回答文本；根据目标回答文本向目标对象进行问答反馈，并基于目标回答文本和第一问答文本对第一知识图谱进行内容更新，得到第二知识图谱；采用上述技术方案，解决了相关技术中，知识图谱更新效率低下，知识图谱在更新过程中无法及时反馈目标对象的问答进行有效响应等问题，通过第一知识图谱对目标对象的第一问答文本进行处理响应，并在第一知识图谱无法支撑响应第一问答文本的情况下，利用训练好的机器阅读理解模型对第一问答文本和外部领域知识文档进行处理，得出该第一问答文本对应的目标回答文本，使用该目标回答文本向目标对象进行问答反馈，并将第一知识图谱更新为包含新知识的第二知识图谱，保证了知识图谱更新与目标对象的使用不冲突，提升了知识图谱的更新效率。
在一个示例性实施例中，上述装置还包括：统计模块，设置为统计所述第一问答文本的出现次数；在所述出现次数大于或者等于第一预设阈值的情况下，确定对所述第一知识图谱进行在线更新；在所述出现次数小于第一预设阈值的情况 下，确定对所述第一知识图谱进行离线更新。
在一个示例性实施例中，上述装置还包括：存储模块，设置为确定所述第一知识图谱用于进行图谱更新的存储空间对应的可用空间资源；在所述可用空间资源小于或者等于第一预设资源阈值，且所述第一问答文本对应的数据流量大于预设流量阈值的情况下，确定在所述可用空间资源仅存储所述第一问答文本对应的目标回答文本，使用所述目标回答文本对所述第一知识图谱进行内容更新；在所述可用空间资源大于第一预设资源阈值，且所述第一问答文本对应的数据流量小于或者等于预设流量阈值的情况下，确定在所述可用空间资源中将所述第一问答文本、所述目标回答文本以及所述外部领域知识文档进行对应存储，并使用所述第一问答文本、所述目标回答文本以及所述外部领域知识文档对所述第一知识图谱进行内容更新。
在一个示例性实施例中，上述装置还包括：类别模块，设置为获取预设的问题类别规则；通过所述问题类别规则确定利用所述机器阅读理解模型输出的所述第一问答文本的问题类别，以及所述第一问答文本在每一种问题类别中对应的频率；将所述频率作为权重，计算所述第一问答文本对应的更新值；在所述第一问答文本同时存在多个的情况下，确定使用所述更新值最大的目标第一问答文本以及所述目标第一问答文本对应的目标回答文本对所述第一知识图谱进行内容更新。
在一个示例性实施例中，上述类别模块，还设置为基于所述第一知识图谱中的不同问题类别分别对应的热度值，确定所述第一问答文本对应问题类别的目标热度值；将所述目标热度值与所述频率相乘，将乘积结果对应的值作为所述第一问答文本对应的更新值。
在一个示例性实施例中，上述装置还包括：图谱模块，设置为在所述第一知识图谱中需要增加新的设备知识的情况下，获取所述新的设备知识对应的第三知识图谱；将所述第三知识图谱添加至所述外部领域知识文档。
在一个示例性实施例中，上述装置还包括：确定模块，设置为使用所述第二知识图谱替换无法对目标对象对应的第一问答文本进行响应的第一知识图谱作 为当前知识图谱；对所述第一问答文本进行识别处理，得到用于在图谱中进行查询的查询文本，其中，所述识别处理用于指示确定所述第一问答文本的意图实体以及所述第一问答文本对应的属性信息；将所述查询文本输入到所述当前知识图谱中，确定出查询结果，其中，所述查询结果用于指示更新后的图谱是否有效支持对所述第一问答文本的响应。
本公开的实施例还提供了一种存储介质，该存储介质包括存储的程序，其中，上述程序运行时执行上述任一项的方法。
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程序代码：
S1，在第一知识图谱无法对目标对象对应的第一问答文本进行响应的情况下，获取外部领域知识文档；
S2，将所述第一问答文本与所述外部领域知识文档输入到机器阅读理解模型中，以得到所述第一问答文本对应的目标回答文本，其中，所述机器阅读理解模型为使用多组数据通过机器学习训练出的，所述多组数据中的每组数据均包括：问答文本，预设的外部领域知识文档，以及所述问答文本和预设的外部领域知识文档对应的回答文本；
S3，根据所述目标回答文本向所述目标对象进行问答反馈，并基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，得到第二知识图谱。
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，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。
可选的，如图5所示，该电子装置包括存储器702和处理器704，该存储器702中存储有计算机程序，该处理器704被设置为通过计算机程序执行上述任一 项方法实施例中的步骤。
可选地，在本实施例中，上述电子装置可以位于计算机网络的多个网络设备中的至少一个网络设备。
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，在第一知识图谱无法对目标对象对应的第一问答文本进行响应的情况下，获取外部领域知识文档；
S2，将所述第一问答文本与所述外部领域知识文档输入到机器阅读理解模型中，以得到所述第一问答文本对应的目标回答文本，其中，所述机器阅读理解模型为使用多组数据通过机器学习训练出的，所述多组数据中的每组数据均包括：问答文本，预设的外部领域知识文档，以及所述问答文本和预设的外部领域知识文档对应的回答文本；
S3，根据所述目标回答文本向所述目标对象进行问答反馈，并基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，得到第二知识图谱。
可选地，本领域普通技术人员可以理解，图5所示的结构仅为示意，电子装置也可以是智能手机(如Android手机、iOS手机等)、平板电脑、掌上电脑以及移动互联网设备(Mobile Internet Devices，MID)、PAD等终端设备。图5其并不对上述电子装置的结构造成限定。例如，电子装置还可包括比图5中所示更多或者更少的组件(如网络接口等)，或者具有与图5所示不同的配置。
其中，存储器702可用于存储软件程序以及模块，如本公开实施例中的通信连接方法和装置对应的程序指令/模块，处理器704通过运行存储在存储器702内的软件程序以及模块，从而执行各种功能应用以及数据处理，即实现上述的通信连接方法。存储器702可包括高速随机存储器，还可以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702可进一步包括相对于处理器704远程设置的存储器，这些远程 存储器可以通过网络连接至终端。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。作为一种示例，如图5所示，上述存储器702中可以但不限于包括上述通信连接装置中的获取模块42、回答模块44、更新模块46。此外，还可以包括但不限于上述通信连接装置中的其他模块单元，本示例中不再赘述。
可选地，上述的传输装置706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有线网络及无线网络。在一个实例中，传输装置706包括一个网络适配器(Network Interface Controller，NIC)，其可通过网线与其他网络设备与路由器相连从而可与互联网或局域网进行通讯。在一个实例中，传输装置1106为射频(Radio Frequency，RF)模块，其用于通过无线方式与互联网进行通讯。
此外，上述电子装置还包括：显示器708，设置为显示上述知识图谱；和连接总线710，设置为连接上述电子装置中的各个模块部件。
可选地，在本实施例中，上述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程序代码的介质。
可选地，本实施例中的具体示例可以参考上述实施例及可选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，可选地，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结合。
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)

  1. 一种知识图谱的更新方法，包括：
    在第一知识图谱无法对目标对象对应的第一问答文本进行响应的情况下，获取外部领域知识文档；
    将所述第一问答文本与所述外部领域知识文档输入到机器阅读理解模型中，以得到所述第一问答文本对应的目标回答文本，其中，所述机器阅读理解模型为使用多组数据通过机器学习训练出的，所述多组数据中的每组数据均包括：问答文本，预设的外部领域知识文档，以及所述问答文本和预设的外部领域知识文档对应的回答文本；
    根据所述目标回答文本向所述目标对象进行问答反馈，并基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，得到第二知识图谱。
  2. 根据权利要求1所述的知识图谱的更新方法，其中，基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新之前，所述方法还包括：
    统计所述第一问答文本的出现次数；
    在所述出现次数大于或者等于第一预设阈值的情况下，确定对所述第一知识图谱进行在线更新；
    在所述出现次数小于第一预设阈值的情况下，确定对所述第一知识图谱进行离线更新。
  3. 根据权利要求1所述的知识图谱的更新方法，其中，基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，所述方法还包括：
    确定所述第一知识图谱用于进行图谱更新的存储空间对应的可用空间资源；
    在所述可用空间资源小于或者等于第一预设资源阈值，且所述第一问答文 本对应的数据流量大于预设流量阈值的情况下，确定在所述可用空间资源仅存储所述第一问答文本对应的目标回答文本，使用所述目标回答文本对所述第一知识图谱进行内容更新；
    在所述可用空间资源大于第一预设资源阈值，且所述第一问答文本对应的数据流量小于或者等于预设流量阈值的情况下，确定在所述可用空间资源中将所述第一问答文本、所述目标回答文本以及所述外部领域知识文档进行对应存储，并使用所述第一问答文本、所述目标回答文本以及所述外部领域知识文档对所述第一知识图谱进行内容更新。
  4. 根据权利要求1所述的知识图谱的更新方法，其中，基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，所述方法还包括：
    获取预设的问题类别规则；
    通过所述问题类别规则确定利用所述机器阅读理解模型输出的所述第一问答文本的问题类别，以及所述第一问答文本在每一种问题类别中对应的频率；
    将所述频率作为权重，计算所述第一问答文本对应的更新值；
    在所述第一问答文本同时存在多个的情况下，确定使用所述更新值最大的目标第一问答文本以及所述目标第一问答文本对应的目标回答文本对所述第一知识图谱进行内容更新。
  5. 根据权利要求4所述的知识图谱的更新方法，其中，将所述频率作为权重，计算所述第一问答文本对应的更新值，包括：
    基于所述第一知识图谱中的不同问题类别分别对应的热度值，确定所述第一问答文本对应问题类别的目标热度值；
    将所述目标热度值与所述频率相乘，将乘积结果对应的值作为所述第一问答文本对应的更新值。
  6. 根据权利要求1所述的知识图谱的更新方法，其中，将所述第一问答文本与所述外部领域知识文档输入到机器阅读理解模型中，以得到所述第一问答文本 对应的目标回答文本之前，所述方法还包括：
    在所述第一知识图谱中需要增加新的设备知识的情况下，获取所述新的设备知识对应的第三知识图谱；
    将所述第三知识图谱添加至所述外部领域知识文档。
  7. 根据权利要求1所述的知识图谱的更新方法，其中，根据所述目标回答文本向所述目标对象进行问答反馈，并基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，得到第二知识图谱之后，所述方法还包括：
    使用所述第二知识图谱替换无法对目标对象对应的第一问答文本进行响应的第一知识图谱作为当前知识图谱；
    对所述第一问答文本进行识别处理，得到用于在图谱中进行查询的查询文本，其中，所述识别处理用于指示确定所述第一问答文本的意图实体以及所述第一问答文本对应的属性信息；
    将所述查询文本输入到所述当前知识图谱中，确定出查询结果，其中，所述查询结果用于指示更新后的图谱是否有效支持对所述第一问答文本的响应。
  8. 一种知识图谱的更新装置，包括：
    获取模块，设置为在第一知识图谱无法对目标对象对应的第一问答文本进行响应的情况下，获取外部领域知识文档；
    回答模块，设置为将所述第一问答文本与所述外部领域知识文档输入到机器阅读理解模型中，以得到所述第一问答文本对应的目标回答文本，其中，所述机器阅读理解模型为使用多组数据通过机器学习训练出的，所述多组数据中的每组数据均包括：问答文本，预设的外部领域知识文档，以及所述问答文本和预设的外部领域知识文档对应的回答文本；
    更新模块，设置为根据所述目标回答文本向所述目标对象进行问答反馈，并基于所述目标回答文本和所述第一问答文本对所述第一知识图谱进行内容更新，得到第二知识图谱。
  9. 根据权利要求8所述的知识图谱的更新装置，其中，所述装置还包括：统计模块，设置为统计所述第一问答文本的出现次数；在所述出现次数大于或者等于第一预设阈值的情况下，确定对所述第一知识图谱进行在线更新；在所述出现次数小于第一预设阈值的情况下，确定对所述第一知识图谱进行离线更新。
  10. 根据权利要求8所述的知识图谱的更新装置，其中，所述装置还包括：
    存储模块，设置为确定所述第一知识图谱用于进行图谱更新的存储空间对应的可用空间资源；在所述可用空间资源小于或者等于第一预设资源阈值，且所述第一问答文本对应的数据流量大于预设流量阈值的情况下，确定在所述可用空间资源仅存储所述第一问答文本对应的目标回答文本，使用所述目标回答文本对所述第一知识图谱进行内容更新；在所述可用空间资源大于第一预设资源阈值，且所述第一问答文本对应的数据流量小于或者等于预设流量阈值的情况下，确定在所述可用空间资源中将所述第一问答文本、所述目标回答文本以及所述外部领域知识文档进行对应存储，并使用所述第一问答文本、所述目标回答文本以及所述外部领域知识文档对所述第一知识图谱进行内容更新。
  11. 根据权利要求8所述的知识图谱的更新装置，其中，所述装置还包括：
    类别模块，设置为获取预设的问题类别规则；通过所述问题类别规则确定利用所述机器阅读理解模型输出的所述第一问答文本的问题类别，以及所述第一问答文本在每一种问题类别中对应的频率；将所述频率作为权重，计算所述第一问答文本对应的更新值；在所述第一问答文本同时存在多个的情况下，确定使用所述更新值最大的目标第一问答文本以及所述目标第一问答文本对应的目标回答文本对所述第一知识图谱进行内容更新。
  12. 根据权利要求11所述的知识图谱的更新装置，其中，所述类别模块，还设置为基于所述第一知识图谱中的不同问题类别分别对应的热度值，确定所述第一问答文本对应问题类别的目标热度值；将所述目标热度值与所述频率相乘，将乘积结果对应的值作为所述第一问答文本对应的更新值。
  13. 根据权利要求8所述的知识图谱的更新装置，其中，所述装置还包括：图谱 模块，设置为在所述第一知识图谱中需要增加新的设备知识的情况下，获取所述新的设备知识对应的第三知识图谱；将所述第三知识图谱添加至所述外部领域知识文档。
  14. 根据权利要求8所述的知识图谱的更新装置，其中，所述装置还包括：
    确定模块，设置为使用所述第二知识图谱替换无法对目标对象对应的第一问答文本进行响应的第一知识图谱作为当前知识图谱；对所述第一问答文本进行识别处理，得到用于在图谱中进行查询的查询文本，其中，所述识别处理用于指示确定所述第一问答文本的意图实体以及所述第一问答文本对应的属性信息；将所述查询文本输入到所述当前知识图谱中，确定出查询结果，其中，所述查询结果用于指示更新后的图谱是否有效支持对所述第一问答文本的响应。
  15. 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上述权利要求1至7任一项中所述的方法。
  16. 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所述权利要求1至7任一项中所述的方法。
PCT/CN2022/135124 2022-07-29 2022-11-29 知识图谱的更新方法和装置、存储介质及电子装置 WO2024021407A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210911171.4A CN115358395A (zh) 2022-07-29 2022-07-29 知识图谱的更新方法和装置、存储介质及电子装置
CN202210911171.4 2022-07-29

Publications (1)

Publication Number Publication Date
WO2024021407A1 true WO2024021407A1 (zh) 2024-02-01

Family

ID=84032539

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/135124 WO2024021407A1 (zh) 2022-07-29 2022-11-29 知识图谱的更新方法和装置、存储介质及电子装置

Country Status (2)

Country Link
CN (1) CN115358395A (zh)
WO (1) WO2024021407A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115358395A (zh) * 2022-07-29 2022-11-18 青岛海尔科技有限公司 知识图谱的更新方法和装置、存储介质及电子装置
CN117725423B (zh) * 2024-02-18 2024-05-24 青岛海尔科技有限公司 基于大模型的反馈信息的生成方法及装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180107647A1 (en) * 2016-10-19 2018-04-19 International Business Machines Corporation System and method for supplementing a question answering system with mixed-language source documents
CN110390003A (zh) * 2019-06-19 2019-10-29 北京百度网讯科技有限公司 基于医疗的问答处理方法及系统、计算机设备及可读介质
CN113220895A (zh) * 2021-04-23 2021-08-06 北京大数医达科技有限公司 基于强化学习的信息处理方法、装置、终端设备
CN114780697A (zh) * 2022-04-18 2022-07-22 国网安徽省电力有限公司电力科学研究院 基于知识图谱的智能问答系统
CN115358395A (zh) * 2022-07-29 2022-11-18 青岛海尔科技有限公司 知识图谱的更新方法和装置、存储介质及电子装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180107647A1 (en) * 2016-10-19 2018-04-19 International Business Machines Corporation System and method for supplementing a question answering system with mixed-language source documents
CN110390003A (zh) * 2019-06-19 2019-10-29 北京百度网讯科技有限公司 基于医疗的问答处理方法及系统、计算机设备及可读介质
CN113220895A (zh) * 2021-04-23 2021-08-06 北京大数医达科技有限公司 基于强化学习的信息处理方法、装置、终端设备
CN114780697A (zh) * 2022-04-18 2022-07-22 国网安徽省电力有限公司电力科学研究院 基于知识图谱的智能问答系统
CN115358395A (zh) * 2022-07-29 2022-11-18 青岛海尔科技有限公司 知识图谱的更新方法和装置、存储介质及电子装置

Also Published As

Publication number Publication date
CN115358395A (zh) 2022-11-18

Similar Documents

Publication Publication Date Title
WO2024021407A1 (zh) 知识图谱的更新方法和装置、存储介质及电子装置
CN112668128B (zh) 联邦学习系统中终端设备节点的选择方法及装置
CN109885452A (zh) 性能监控方法、装置及终端设备
WO2024036694A1 (zh) 控制指令的发送方法、控制装置、存储介质及电子装置
WO2023168838A1 (zh) 语句文本的识别方法和装置、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
CN114697150A (zh) 命令的下发方法和装置、存储介质及电子装置
CN115877726A (zh) 智能家居设备的控制方法、计算机设备及存储介质
CN106910075A (zh) 移动通信客户投诉的智能处理系统和方法
WO2024001196A1 (zh) 家电设备的控制方法、装置、存储介质及电子装置
CN111368060B (zh) 对话机器人的自学习方法、装置、系统、电子设备及介质
WO2024066089A1 (zh) 烹饪计划的确定方法及装置、存储介质及电子装置
CN117573320A (zh) 任务节点的执行方法、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
CN116881752A (zh) 数据的聚类方法和装置、存储介质及电子设备
US20210042621A1 (en) Method for operation of network model and related product
CN114925158A (zh) 语句文本的意图识别方法和装置、存储介质及电子装置
CN115345225A (zh) 推荐场景的确定方法及装置、存储介质及电子装置
CN114840557A (zh) 数据的查询方法和装置、存储介质及电子装置
CN116560731A (zh) 一种数据处理方法及其相关装置
WO2024036804A1 (zh) 意图指令的确定方法及装置、存储介质及电子装置
WO2023206723A1 (zh) 语义转换方法和装置、存储介质及电子装置
CN115481317A (zh) 一种工作台场景的推荐方法、存储介质及电子装置
CN113622134A (zh) 一种洗衣控制方法、装置、设备及存储介质
CN115796185A (zh) 语义意图的确定方法及装置、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: 22952841

Country of ref document: EP

Kind code of ref document: A1