WO2022134421A1 - Procédé et appareil de réponse intelligente basée sur un graphe multi-connaissances, dispositif informatique et support de stockage - Google Patents

Procédé et appareil de réponse intelligente basée sur un graphe multi-connaissances, dispositif informatique et support de stockage Download PDF

Info

Publication number
WO2022134421A1
WO2022134421A1 PCT/CN2021/091291 CN2021091291W WO2022134421A1 WO 2022134421 A1 WO2022134421 A1 WO 2022134421A1 CN 2021091291 W CN2021091291 W CN 2021091291W WO 2022134421 A1 WO2022134421 A1 WO 2022134421A1
Authority
WO
WIPO (PCT)
Prior art keywords
current
entity name
question
entity
knowledge graph
Prior art date
Application number
PCT/CN2021/091291
Other languages
English (en)
Chinese (zh)
Inventor
张松
Original Assignee
平安科技（深圳）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 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2022134421A1 publication Critical patent/WO2022134421A1/fr

Links

Images

Classifications

    • 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
    • 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
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3344—Query execution using natural language analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/237—Lexical tools
    • G06F40/242—Dictionaries
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/237—Lexical tools
    • G06F40/247—Thesauruses; Synonyms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/279—Recognition of textual entities
    • G06F40/284—Lexical analysis, e.g. tokenisation or collocates
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/279—Recognition of textual entities
    • G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
    • G06F40/295—Named entity recognition

Definitions

  • the present application relates to the field of knowledge graphs of big data, and in particular, to an intelligent reply method, apparatus, computer equipment and storage medium based on multi-knowledge graphs.
  • chat robots carried in servers are generally classified into chat robots, task (vertical) robots, and question answering (QA) robots according to application scenarios.
  • QA question answering
  • the embodiments of the present application provide an intelligent answering method, device, computer equipment and storage medium based on multi-knowledge graphs, aiming to solve the problem of accurate keyword matching in the process of intelligent questioning and answering with users in the prior art.
  • an embodiment of the present application provides an intelligent reply method based on a multi-knowledge graph, which includes:
  • Receive the dialogue text sent by the user terminal perform intention recognition on the dialogue text to obtain an intention recognition keyword set, and obtain the current entity name corresponding to each intention recognition keyword in the intention recognition keyword set, so as to form the current entity name set;
  • the corresponding first current associated question set is obtained in the local first knowledge graph library, and the first current associated question set is sent to the client; wherein the first current associated question set is The number of association question subsets included in the current association question set is the same as the number of current entity names included in the current entity name set, and each current entity name corresponds to one of the association questions in the first current association question set Subset;
  • the first question click instruction corresponding to the first current associated question set sent by the client is detected, the first target entity name corresponding to the first question click instruction is obtained, and the first target entity name corresponding to the first target entity name is obtained.
  • a set of associated entity names obtaining a corresponding second current associated problem set in the local first knowledge graph library according to the names of each first associated entity in the first associated entity name set, and sending the second current associated problem set to The user terminal; wherein, the number of association question subsets included in the second current association question set is the same as the number of first association entity names included in the first association entity name set, and each first association entity a name corresponding to one of the subset of contextual questions in the second current set of contextual questions; and
  • the second target entity name corresponding to the second question click instruction is acquired and the completion flag is added as the first output result, and the The first output result and the current user portrait are mapped one-to-one for binding.
  • an embodiment of the present application provides an intelligent replying device based on a multi-knowledge graph, which includes:
  • connection establishment unit configured to establish a connection with the user terminal and obtain a current user portrait corresponding to the user terminal if an intelligent session connection instruction sent by the user terminal is detected
  • the current entity recognition unit is used to receive the dialogue text sent by the user terminal, perform intention recognition on the dialogue text to obtain an intention recognition keyword set, and obtain the current entity corresponding to each intention recognition keyword in the intention recognition keyword set. name, to make up the current set of entity names;
  • the first question set sending unit is configured to obtain the corresponding first current associated question set in the local first knowledge graph library according to each current entity name in the current entity name set, and send the first current associated question set to User terminal; wherein, the number of association question subsets included in the first current association question set is the same as the number of current entity names included in the current entity name set, and each current entity name corresponds to the first one of the subset of related questions in the current set of related questions;
  • the second question set sending unit is configured to acquire the name of the first target entity corresponding to the first question click instruction and obtain the The first associated entity name set corresponding to the first target entity name, according to each first associated entity name in the first associated entity name set, obtain the corresponding second current associated problem set in the local first knowledge graph library, The second current associated question set is sent to the client; wherein, the number of associated question subsets included in the second current associated question set and the number of first associated entity names included in the first associated entity name set are the same, and each first associated entity name corresponds to one of the associated problem subsets in the second current associated problem set; and
  • the result binding unit is configured to acquire the second target entity name addition completion identifier corresponding to the second question click instruction corresponding to the second question click instruction as the first an output result, the first output result and the current user portrait are mapped one-to-one to bind.
  • an embodiment of the present application further provides a computer device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer
  • the program implements the following steps:
  • Receive the dialogue text sent by the user terminal perform intention recognition on the dialogue text to obtain an intention recognition keyword set, and obtain the current entity name corresponding to each intention recognition keyword in the intention recognition keyword set, so as to form the current entity name set;
  • the corresponding first current associated question set is obtained in the local first knowledge graph library, and the first current associated question set is sent to the client; wherein the first current associated question set is The number of association question subsets included in the current association question set is the same as the number of current entity names included in the current entity name set, and each current entity name corresponds to one of the association questions in the first current association question set Subset;
  • the first question click instruction corresponding to the first current associated question set sent by the client is detected, the first target entity name corresponding to the first question click instruction is obtained, and the first target entity name corresponding to the first target entity name is obtained.
  • a set of associated entity names obtaining a corresponding second current associated problem set in the local first knowledge graph library according to the names of each first associated entity in the first associated entity name set, and sending the second current associated problem set to The user terminal; wherein, the number of association question subsets included in the second current association question set is the same as the number of first association entity names included in the first association entity name set, and each first association entity a name corresponding to one of the subset of contextual questions in the second current set of contextual questions; and
  • the second target entity name corresponding to the second question click instruction is acquired and the completion flag is added as the first output result, and the The first output result and the current user portrait are mapped one-to-one for binding.
  • embodiments of the present application further provide a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when executed by a processor, the computer program causes the processor to perform the following operations :
  • Receive the dialogue text sent by the user terminal perform intention recognition on the dialogue text to obtain an intention recognition keyword set, and obtain the current entity name corresponding to each intention recognition keyword in the intention recognition keyword set, so as to form the current entity name set;
  • the corresponding first current associated question set is obtained in the local first knowledge graph library, and the first current associated question set is sent to the client; wherein the first current associated question set is The number of association question subsets included in the current association question set is the same as the number of current entity names included in the current entity name set, and each current entity name corresponds to one of the association questions in the first current association question set Subset;
  • the first question click instruction corresponding to the first current associated question set sent by the client is detected, the first target entity name corresponding to the first question click instruction is obtained, and the first target entity name corresponding to the first target entity name is obtained.
  • a set of associated entity names obtaining a corresponding second current associated problem set in the local first knowledge graph library according to the names of each first associated entity in the first associated entity name set, and sending the second current associated problem set to The user terminal; wherein, the number of association question subsets included in the second current association question set is the same as the number of first association entity names included in the first association entity name set, and each first association entity a name corresponding to one of the subset of contextual questions in the second current set of contextual questions; and
  • the second target entity name corresponding to the second question click instruction is acquired and the completion flag is added as the first output result, and the The first output result and the current user portrait are mapped one-to-one for binding.
  • the embodiments of the present application provide an intelligent answering method, device, computer equipment and storage medium based on multiple knowledge graphs, to determine whether a user establishes a session with an intelligent dialogue robot for the first time, so as to call associated question sets from different knowledge graph libraries. Users with different levels of knowledge can recommend corresponding information accordingly, which can effectively improve the efficiency and accuracy of information acquisition by users.
  • FIG. 1 is a schematic diagram of an application scenario of an intelligent reply method based on a multi-knowledge graph provided by an embodiment of the present application;
  • FIG. 2 is a schematic flowchart of an intelligent reply method based on a multi-knowledge graph provided by an embodiment of the present application
  • FIG. 3 is a schematic block diagram of an intelligent replying device based on a multi-knowledge graph provided by an embodiment of the present application
  • FIG. 4 is a schematic block diagram of a computer device provided by an embodiment of the present application.
  • FIG. 1 is a schematic diagram of an application scenario of the multi-knowledge graph-based intelligent reply method provided by the embodiment of the application
  • FIG. 2 is a schematic flowchart of the multi-knowledge graph-based intelligent reply method provided by the embodiment of the application. , the multi-knowledge graph-based intelligent reply method is applied in the server, and the method is executed by the application software installed in the server.
  • the method includes steps S101-S109.
  • S101 Receive and save a first knowledge graph library and a second knowledge graph library uploaded by a business server; wherein, the first knowledge graph library includes a plurality of first sub-knowledge graphs, and each first sub-knowledge graph corresponds to several entity name, and there is an association relationship between entities corresponding to the entity name; the second knowledge graph library includes a plurality of second sub-knowledge graphs, each second sub-knowledge graph corresponds to several entity names, and the entity name corresponds to There are associations between entities.
  • the R&D personnel can upload the pre-built first knowledge graph library and the second knowledge graph library to the server.
  • the knowledge graph library Once the above-mentioned knowledge graph library is updated, it can be directly uploaded from the business server to the server for updating.
  • product knowledge or information from the server can be insurance products, wealth management products, electronic products, sports equipment products, etc.
  • the multiple first sub-knowledge maps included in the first knowledge map database correspond to A collection of answer texts for questions concerned by users in the first understanding stage of some products (which can be understood as the demand stimulation stage, where users pay more attention to the acquisition of product popularization knowledge), and multiple first subsections included in the first knowledge graph library.
  • the knowledge graph is mainly based on product-related popular science introduction; the multiple second sub-knowledge graphs included in the second knowledge graph base correspond to the second understanding stage of some products by the user (it can be understood as the concept introduction stage, users pay more attention to The acquisition of detailed parameters of the product) is a collection of answer texts for the concerned question, and the plurality of second sub-knowledge graphs included in the second knowledge graph base are the detailed parameters related to the product.
  • the second is the server.
  • the server can store the first knowledge graph library and the second knowledge graph library uploaded by the business server, and also store the user portraits corresponding to multiple users. These data are used as the background of the intelligent dialogue robot deployed in the server. data.
  • the relationship between the tag keywords in the user portrait and the entity names in the above two knowledge graph bases includes completely different, similar, and identical relationships.
  • a corresponding set of associated questions can be obtained from the first knowledge graph library or the second knowledge graph library through the tag keywords corresponding to the user and sent to the user terminal.
  • the third is the client.
  • Each client stores a corresponding user portrait in the server. If a certain client establishes a dialogue with the intelligent dialogue robot in the server, the user portrait of the user of the client can be obtained in the server.
  • the server After the server receives the first knowledge graph library and the second knowledge graph library uploaded by the business server, and stores the user portraits corresponding to multiple users, these data can be used in the dialog established between the user end and the intelligent dialog robot to conduct Recommendations for questions or knowledge points.
  • the server After a certain client (for example, denoted as client A) establishes a communication connection with the server and opens an intelligent session, the server establishes a connection with the client, and obtains the current user portrait corresponding to the client .
  • client A for example, denoted as client A
  • the server establishes a connection with the client, and obtains the current user portrait corresponding to the client .
  • the current user portrait can be understood as the basic portrait feature of the user.
  • the process of obtaining the user portrait corresponding to the user terminal is as follows:
  • the user unique identification code corresponding to the user terminal is obtained, and the current user portrait corresponding to the user terminal is obtained from the local user portrait library according to the user unique identification code.
  • the server in order to obtain the user portrait of the client more accurately, after the client establishes a connection with the server, the server obtains the user's unique identification code (such as a phone number, ID number, etc.) corresponding to the client. At the same time, according to the user's unique identification code in the local user portrait library of the server, the current user portrait uniquely corresponding to the user terminal can be obtained.
  • the user's unique identification code such as a phone number, ID number, etc.
  • S103 Receive the dialogue text sent by the user terminal, perform intention recognition on the dialogue text to obtain an intention recognition keyword set, and acquire the current entity names corresponding to each intention recognition keyword in the intention recognition keyword set, so as to form a current entity name. Entity name set.
  • the user terminal when the user terminal establishes a connection with the server and starts a dialogue with the intelligent dialogue robot, it is necessary to preliminarily analyze the user's intention according to the received dialogue text. Based on the conversation topics that the current user may be generating, extract the entities of the knowledge graph associated with this conversation process (such as the two entities of child + health). Since it is the first time for the client to establish a session with the intelligent dialogue robot, the sub-knowledge in the demand-stimulation stage can be called in the first knowledge map library based on the core entities extracted from this dialogue flow (such as the child + health in the above example). Atlas.
  • the step of performing intent recognition on the dialogue text in step S103 to obtain an intent recognition keyword set includes:
  • the dialogue text is subjected to word segmentation based on a probabilistic and statistical word segmentation model, and a word segmentation result corresponding to the dialogue text is obtained;
  • keywords that do not exceed the preset ranking threshold after the word segmentation results are sorted in descending order by the frequency-inverse text frequency index are extracted to form an intent identification keyword set.
  • the dialogue text is segmented based on the probability and statistical word segmentation model
  • C C1C2...Cm
  • C is the Chinese character string to be segmented
  • W W1W2...Wn
  • W is the cut
  • Wa, Wb, ..., Wk are all possible segmentation schemes of C.
  • the word segmentation model based on probability statistics can find the target word string W, so that W satisfies: P(W
  • C) MAX(P(Wa
  • the word string W obtained by the above word segmentation model is the word string with the largest estimated probability.
  • the dialogue text can be segmented based on the probability and statistics word segmentation model, so as to obtain a word segmentation result corresponding to the dialogue text.
  • the word frequency-inverse text frequency index model is used to extract the keywords in the word segmentation results.
  • the word frequency-inverse text frequency index model is the TF-IDF model, and TF-IDF is the abbreviation of Term Frequency-Inverse Document Frequency. Extracting the keywords that are located before the preset ranking value after the word segmentation results are sorted in descending order of frequency-inverse text frequency index, to form an intent-identifying keyword set.
  • the word frequency-inverse text frequency index model can accurately extract the core keywords in the text without manual reading.
  • step S103 the step of obtaining the current entity names corresponding to each intent identification keyword in the intent identification keyword set, so as to form the current entity name set, includes:
  • the current entity names corresponding to each intent identification keyword are obtained from the entity name database, so as to form a current entity name set.
  • the server locally pre-stores an entity name library composed of entity names in the first knowledge graph library and the sub-knowledge graphs included in the second knowledge graph library, and also stores a database for judging words and A thesaurus dictionary of similarity between words.
  • all the words included in the synonym words are organized in one or several tree structures (this tree or trees are recorded as the word forest), and the two words that need to be judged for the similarity of the words are found in the word forest. node, the path length of the two nodes can be used as the semantic distance between the two words (which can also be understood as the word similarity).
  • the intent recognition keyword set includes two intent keywords of child and health, and the word with the closest path length to the child in the word forest is parenting.
  • the intent keyword of child can be converted is the current entity name of parenting; in the same way, the word with the closest path length to health in the word forest is health big data.
  • the intention keyword of health can be converted into health big data.
  • S104 Acquire a corresponding first current associated question set in a local first knowledge graph library according to each current entity name in the current entity name set, and send the first current associated question set to the client; wherein, the The number of association question subsets included in the first current association question set is the same as the number of current entity names included in the current entity name set, and each current entity name corresponds to one of the first current association question sets. Associated problem subsets.
  • the local first knowledge graph library of each current entity name in the current entity name set can be used at this time. Obtain the corresponding first current related question set from , so as to guide users to consult health big data + parenting related questions.
  • the intelligent dialogue robot mainly calls the sub-knowledge map corresponding to the entity of parenting in the first knowledge map database, and also calls the sub-knowledge map corresponding to the entity of health big data in the first knowledge map database, and obtains the above-mentioned sub-knowledge After the graph, the respective associated questions corresponding to the above sub-knowledge graphs are called to form a first first current associated question set, and the first current associated question set is sent to the user terminal.
  • the first question click instruction corresponding to the first current set of related questions sent by the user terminal obtain the first target entity name corresponding to the first question click instruction, and obtain the first target entity name corresponding to the first question click instruction.
  • the first set of associated entity names, according to the name of each first associated entity in the first associated entity name set, the corresponding second current associated problem set is obtained in the local first knowledge graph library, and the second current associated problem set is Send to the user terminal; wherein, the number of association question subsets included in the second current association question set is the same as the number of first association entity names included in the first association entity name set, and each first association The associated entity name corresponds to one of the associated problem subsets in the second current associated problem set.
  • the user terminal can manually click to select one of the most interesting questions to generate the first question Click Command. For example, if a user clicks on a breast-related question under the entity name of health big data (specifically, breast health big data in 2019), the key entity name (mammary gland) in this question is extracted, and then according to the entity name of breast, and The name of the entity (health big data) to which the currently viewing (2019 breast health big data) is associated is related to another entity (breast hyperplasia + female cancer) according to the association relationship between the entities in the first knowledge atlas ), and at the same time show the related questions below these two related entities (how to avoid breast hyperplasia in post-80s women, and what are the diseases with high incidence of female cancer?).
  • entity name of health big data specifically, breast health big data in 2019
  • the name of the entity (health big data) to which the currently viewing (2019 breast health big data) is associated is related to another entity (breast hyperplasia + female cancer) according to the association relationship between
  • the two associated entities in the above example are composed of the first associated entity name set corresponding to the first target entity name. After the first associated entity name set is obtained, the corresponding entity can be obtained in the local first knowledge graph library of the server.
  • the second current associated question set is sent to the user terminal, so as to guide the user to inquire about breast hyperplasia + female cancer-related questions.
  • the user terminal can manually click to select one of the most interesting questions to generate the second question Click Command. For example, if the user finally clicks on the diseases with high incidence of female cancer, according to this related question, it is extracted that the user A has completed the demand stimulation stage. At this time, a "completed scientific knowledge of women's cancer" is output as the first output result.
  • This first data result can be added as a user label to the current user portrait corresponding to the user, or the first data result can be added to the user's current user portrait.
  • the output result is mapped one-to-one with the current user portrait for binding.
  • the server when the server detects another intelligent session connection instruction sent by the client, it may be that the previous user A and the intelligent dialogue robot establish a session again to learn more about the relevant details of the product (that is, the user It has already gone through the demand stimulation stage and entered the concept introduction stage), it is also possible that another user B other than user A has established a conversation with the intelligent dialogue robot for the first time to understand the popular science knowledge of the product (that is, the user has not yet experienced the demand stimulation stage), At this time, the corresponding target user portrait can be obtained according to the connection instruction with another intelligent session. Then, it is determined whether the target user portrait corresponds to a first output result, so as to determine whether the user establishes a conversation with the intelligent dialogue robot for the first time.
  • S108 Determine whether the target user portrait corresponds to the binding of the first output result.
  • judging whether the target user portrait corresponds to the first output result is to judge whether the user establishes a session with the intelligent dialogue robot for the first time, so that the first output result can be called more accurately according to the user's situation in the subsequent steps.
  • the target user portrait is bound with the first output result, which means that the user who has completed at least one communication before and the intelligent dialogue robot establish a conversation again to learn more about the relevant details of the product. Then obtain the set of associated questions from the first knowledge graph library as in the above steps, but obtain the corresponding third current associated question set from the second local knowledge graph library according to the entity name corresponding to the first output result, and set the The third current associated question set is sent to the client.
  • users who have a preliminary understanding of product popularization knowledge can be directed to obtain more professional knowledge from the second knowledge graph library, without the need to understand product popularization knowledge again, thereby improving the efficiency and accuracy of information acquisition by users.
  • step S108 it further includes:
  • the intent recognition keyword set is a step of acquiring the current entity name corresponding to each intent recognition keyword in the intent recognition keyword set, so as to form the current entity name set.
  • the target user portrait is not bound with the first output result, indicating that the user is establishing a session with the intelligent dialogue robot for the first time.
  • the target user portrait can be updated as the current user portrait, and the execution is returned to Step S103. In this way, corresponding information is recommended for users with different knowledge mastery degrees, which can effectively improve the efficiency and accuracy of information acquisition by users.
  • step S109 it further includes:
  • corresponding summary information is obtained based on the first current related question set, the second current related question set, and the third current related question set.
  • the current association problem set, the second current association problem set, and the third current association problem set are obtained by hashing, for example, by using the sha256s algorithm. Uploading summary information to the blockchain ensures its security and fairness and transparency to users.
  • the user equipment can download the summary information from the blockchain, so as to verify whether the first current set of related questions, the second set of current related questions and the third current set of related questions have been tampered with.
  • the blockchain referred to in this example is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information to verify its Validity of information (anti-counterfeiting) and generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.
  • the method realizes that users with different knowledge mastery levels recommend corresponding information according to the corresponding knowledge graph, which can effectively improve the efficiency and accuracy of information acquisition by users.
  • the embodiments of the present application further provide an intelligent replying device based on a multi-knowledge graph, and the multi-knowledge graph-based intelligent replying device is used to execute any of the foregoing embodiments of the multi-knowledge graph-based intelligent replying method.
  • FIG. 3 is a schematic block diagram of an intelligent replying device based on a multi-knowledge graph provided by an embodiment of the present application.
  • the multi-knowledge graph-based intelligent answering device 100 may be configured in a server.
  • the intelligent replying device 100 based on multiple knowledge graphs includes: a knowledge graph database establishment unit 101 , a connection establishment unit 102 , a current entity identification unit 103 , a first question set sending unit 104 , and a second question set sending unit 105 , a result binding unit 106 , a target user portrait acquisition unit 107 , a target user portrait judgment unit 108 , and a third question set sending unit 109 .
  • the knowledge graph library establishing unit 101 is configured to receive and save the first knowledge graph library and the second knowledge graph library uploaded by the business server; wherein, the first knowledge graph library includes a plurality of first sub-knowledge graphs, each A sub-knowledge graph corresponds to several entity names, and there is an association relationship between the entities corresponding to the entity names; the second knowledge graph library includes a plurality of second sub-knowledge graphs, and each second sub-knowledge graph corresponds to several entities name, and there is an association relationship between the entities corresponding to the entity name.
  • the server receives the first knowledge graph library and the second knowledge graph library uploaded by the service server, and stores the user portraits corresponding to multiple users, these data can be used for the client and the intelligent dialogue robot In the established dialogue, recommend questions or knowledge points.
  • the connection establishing unit 102 is configured to establish a connection with the user terminal and acquire the current user portrait corresponding to the user terminal if the intelligent session connection instruction sent by the user terminal is detected.
  • the server After a certain client (for example, denoted as client A) establishes a communication connection with the server and opens an intelligent session, the server establishes a connection with the client, and obtains the current user portrait corresponding to the client .
  • client A for example, denoted as client A
  • the server establishes a connection with the client, and obtains the current user portrait corresponding to the client .
  • the current user portrait can be understood as the basic portrait feature of the user.
  • the process of obtaining the user portrait corresponding to the user terminal is as follows:
  • the user unique identification code corresponding to the user terminal is obtained, and the current user portrait corresponding to the user terminal is obtained from the local user portrait library according to the user unique identification code.
  • the server in order to obtain the user portrait of the client more accurately, after the client establishes a connection with the server, the server obtains the user's unique identification code (such as a phone number, ID number, etc.) corresponding to the client. At the same time, according to the user's unique identification code in the local user portrait library of the server, the current user portrait uniquely corresponding to the user terminal can be obtained.
  • the user's unique identification code such as a phone number, ID number, etc.
  • the current entity recognition unit 103 is configured to receive the dialogue text sent by the user terminal, perform intention recognition on the dialogue text to obtain an intention recognition keyword set, and obtain the current entity corresponding to each intention recognition keyword in the intention recognition keyword set. Entity names to make up the current set of entity names.
  • the user terminal when the user terminal establishes a connection with the server and starts a dialogue with the intelligent dialogue robot, it is necessary to preliminarily analyze the user's intention according to the received dialogue text. Based on the conversation topics that the current user may be generating, extract the entities of the knowledge graph associated in this conversation process (for example, the two entities of child + health). Since it is the first time for the client to establish a session with the intelligent dialogue robot, the sub-knowledge in the demand-stimulation stage can be called in the first knowledge map library based on the core entities extracted from this dialogue flow (such as the child + health in the above example). Atlas.
  • the current entity identification unit 103 includes:
  • a word segmentation unit used to segment the dialogue text through a word segmentation model based on probability statistics to obtain a word segmentation result corresponding to the dialogue text;
  • the keyword extraction unit is used for extracting keywords that do not exceed a preset ranking threshold after the word segmentation results are sorted in descending order of frequency-inverse text frequency index through a word frequency-inverse text frequency index model to form an intent-identifying keyword set.
  • the dialogue text is segmented based on the probability and statistical word segmentation model
  • C C1C2...Cm
  • C is the Chinese character string to be segmented
  • W W1W2...Wn
  • W is the cut
  • Wa, Wb, ..., Wk are all possible segmentation schemes of C.
  • the word segmentation model based on probability statistics can find the target word string W, so that W satisfies: P(W
  • C) MAX(P(Wa
  • the word string W obtained by the above word segmentation model is the word string with the largest estimated probability.
  • the dialogue text can be segmented based on the probability and statistics word segmentation model, so as to obtain a word segmentation result corresponding to the dialogue text.
  • the word frequency-inverse text frequency index model is used to extract the keywords in the word segmentation results.
  • the word frequency-inverse text frequency index model is the TF-IDF model, and TF-IDF is the abbreviation of Term Frequency-Inverse Document Frequency. Extracting the keywords that are located before the preset ranking value after the word segmentation results are sorted in descending order of frequency-inverse text frequency index, to form an intent-identifying keyword set. Through the word frequency-inverse text frequency index model, the core keywords in the text can be accurately extracted without manual reading.
  • the current entity identification unit 103 further includes:
  • a word forest obtaining unit used to call the pre-stored entity name library and thesaurus dictionary, and obtain the word forest corresponding to the thesaurus dictionary
  • the current entity name set obtaining unit is configured to obtain the current entity names corresponding to each intent identification keyword in the entity name database according to the word forest, so as to form a current entity name set.
  • the server locally pre-stores an entity name library composed of entity names in the first knowledge graph library and the sub-knowledge graphs included in the second knowledge graph library, and also stores a database for judging words and A thesaurus dictionary of similarity between words.
  • all the words included in the synonym words are organized in one or several tree structures (this tree or trees are recorded as the word forest), and the two words that need to be judged for the similarity of the words are found in the word forest. node, the path length of the two nodes can be used as the semantic distance between the two words (which can also be understood as the word similarity).
  • the intent recognition keyword set includes two intent keywords of child and health, and the word with the closest path length to the child in the word forest is parenting.
  • the intent keyword of child can be converted is the current entity name of parenting; in the same way, the word with the closest path length to health in the word forest is health big data.
  • the intention keyword of health can be converted into health big data.
  • the first question set sending unit 104 is configured to obtain the corresponding first current associated question set in the local first knowledge graph library according to each current entity name in the current entity name set, and send the first current associated question set to the user terminal; wherein the number of association question subsets included in the first current association question set is the same as the number of current entity names included in the current entity name set, and each current entity name corresponds to the first A subset of contextual questions in a current set of contextual questions.
  • the local first knowledge graph library of each current entity name in the current entity name set can be used at this time. Obtain the corresponding first current related question set from , so as to guide users to consult health big data + parenting related questions.
  • the intelligent dialogue robot mainly calls the sub-knowledge map corresponding to the entity of parenting in the first knowledge map database, and also calls the sub-knowledge map corresponding to the entity of health big data in the first knowledge map database, and obtains the above-mentioned sub-knowledge After the graph, the respective associated questions corresponding to the above sub-knowledge graphs are called to form a first first current associated question set, and the first current associated question set is sent to the user terminal.
  • the second question set sending unit 105 is configured to obtain the name of the first target entity corresponding to the first question click instruction and obtain the the first associated entity name set corresponding to the first target entity name, obtain the corresponding second current associated problem set in the local first knowledge graph library according to the first associated entity name in the first associated entity name set, and set the The second current association question set is sent to the user terminal; wherein the number of association question subsets included in the second current association question set is the same as the number of first association entity names included in the first association entity name set. The numbers are the same, and each first associated entity name corresponds to one of the associated problem subsets in the second current associated problem set.
  • the user terminal can manually click to select one of the most interesting questions to generate the first question Click Command. For example, if a user clicks on a breast-related question under the entity name of health big data (specifically, breast health big data in 2019), the key entity name (mammary gland) in this question is extracted, and then according to the entity name of breast, and The name of the entity (health big data) to which the currently viewing (2019 breast health big data) is associated is related to another entity (breast hyperplasia + female cancer) according to the association relationship between the entities in the first knowledge atlas ), and at the same time show the related questions below these two related entities (how to avoid breast hyperplasia in post-80s women, and what are the diseases with high incidence of female cancer?).
  • entity name of health big data specifically, breast health big data in 2019
  • the name of the entity (health big data) to which the currently viewing (2019 breast health big data) is associated is related to another entity (breast hyperplasia + female cancer) according to the association relationship between
  • the two associated entities in the above example are composed of a first associated entity name set corresponding to the name of the first target entity. After the first associated entity name set is obtained, the corresponding entity can be obtained in the local first knowledge graph library of the server.
  • the second current associated question set is sent to the user terminal, so as to guide the user to inquire about breast hyperplasia + female cancer-related questions.
  • the result binding unit 106 is configured to acquire the second target entity name corresponding to the second question click instruction and add the completion flag as For the first output result, the first output result and the current user portrait are mapped one-to-one to bind.
  • the user terminal can manually click to select one of the most interesting questions to generate the second question Click Command. For example, if the user finally clicks on the diseases with high incidence of female cancer, according to this related question, it is extracted that the user A has completed the demand stimulation stage. At this time, a "completed scientific knowledge of women's cancer" is output as the first output result.
  • This first data result can be added as a user label to the current user portrait corresponding to the user, or the first data result can be added to the user's current user portrait.
  • the output result is mapped one-to-one with the current user portrait for binding.
  • the target user portrait acquisition unit 107 is configured to acquire a target user portrait corresponding to another intelligent conversation connection instruction if another intelligent conversation connection instruction sent by the user terminal is detected.
  • the server when the server detects another intelligent session connection instruction sent by the client, it may be that the previous user A and the intelligent dialogue robot establish a session again to learn more about the relevant details of the product (that is, the user It has already gone through the demand stimulation stage and entered the concept introduction stage), it is also possible that another user B other than user A has established a conversation with the intelligent dialogue robot for the first time to understand the popular science knowledge of the product (that is, the user has not yet experienced the demand stimulation stage), At this time, the corresponding target user portrait can be acquired according to the connection instruction with another intelligent session. Then, it is determined whether the target user portrait corresponds to a first output result, so as to determine whether the user establishes a conversation with the intelligent dialogue robot for the first time.
  • the target user portrait judgment unit 108 is configured to judge whether the target user portrait corresponds to the binding of the first output result.
  • judging whether the target user portrait corresponds to the first output result is to judge whether the user establishes a session with the intelligent dialogue robot for the first time, so that the first output result can be called more accurately according to the user's situation in the subsequent steps.
  • the third question set sending unit 109 is configured to, if the target user portrait is bound with a first output result, obtain a corresponding third current associated question set in the local second knowledge graph library according to the first output result , and send the third current associated question set to the client.
  • the target user portrait is bound with the first output result, which means that the user who has completed at least one communication before and the intelligent dialogue robot establish a conversation again to learn more about the relevant details of the product. Then obtain the set of associated questions from the first knowledge graph library as in the above steps, but obtain the corresponding third current associated question set from the second local knowledge graph library according to the entity name corresponding to the first output result, and set the The third current associated question set is sent to the client.
  • users who have a preliminary understanding of product popularization knowledge can be directed to obtain more professional knowledge from the second knowledge graph library, without the need to understand product popularization knowledge again, thereby improving the efficiency and accuracy of information acquisition by users.
  • the multi-knowledge graph-based intelligent reply apparatus 100 further includes:
  • the return control unit is configured to update the target user portrait as the current user portrait if the target user portrait is not bound with the first output result, return the execution to receive the dialogue text sent by the user terminal, and respond to the dialogue A step of performing intent recognition on the text to obtain an intent recognition keyword set, and acquiring the current entity names corresponding to each intent recognition keyword in the intent recognition keyword set, so as to form a current entity name set.
  • the target user portrait is not bound with the first output result, indicating that the user is establishing a session with the intelligent dialogue robot for the first time.
  • the target user portrait can be updated as the current user portrait, and the execution is returned to Receive the dialogue text sent by the user terminal, perform intention recognition on the dialogue text to obtain an intention recognition keyword set, and obtain the current entity name corresponding to each intention recognition keyword in the intention recognition keyword set, so as to form the current entity name set of steps.
  • corresponding information is recommended for users with different knowledge mastery degrees, which can effectively improve the efficiency and accuracy of information acquisition by users.
  • the intelligent replying device 100 based on the multi-knowledge graph further includes:
  • An uploading unit configured to upload the first current related problem set, the second current related problem set and the third current related problem set to the blockchain.
  • corresponding summary information is obtained based on the first current related question set, the second current related question set, and the third current related question set.
  • the current association problem set, the second current association problem set, and the third current association problem set are obtained by hashing, for example, by using the sha256s algorithm. Uploading summary information to the blockchain ensures its security and fairness and transparency to users.
  • the user equipment can download the summary information from the blockchain, so as to verify whether the first current set of related questions, the second set of current related questions and the third current set of related questions have been tampered with.
  • the blockchain referred to in this example is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information to verify its Validity of information (anti-counterfeiting) and generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.
  • the device implements the recommendation of corresponding information according to the corresponding knowledge graph for users with different knowledge mastery degrees, which can effectively improve the efficiency and accuracy of information acquisition by the users.
  • the above-mentioned multi-knowledge graph-based intelligent replying apparatus can be implemented in the form of a computer program, and the computer program can be executed on a computer device as shown in FIG. 4 .
  • FIG. 4 is a schematic block diagram of a computer device provided by an embodiment of the present application.
  • the computer device 500 is a server, and the server may be an independent server or a server cluster composed of multiple servers.
  • the computer device 500 includes a processor 502 , a memory and a network interface 505 connected through a system bus 501 , wherein the memory may include a non-volatile storage medium 503 and an internal memory 504 .
  • the nonvolatile storage medium 503 can store an operating system 5031 and a computer program 5032 .
  • the processor 502 can execute the intelligent reply method based on the multi-knowledge graph.
  • the processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500 .
  • the internal memory 504 provides an environment for running the computer program 5032 in the non-volatile storage medium 503.
  • the processor 502 can execute the intelligent answering method based on the multi-knowledge graph.
  • the network interface 505 is used for network communication, such as providing transmission of data information.
  • the network interface 505 is used for network communication, such as providing transmission of data information.
  • FIG. 4 is only a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device 500 to which the solution of the present application is applied.
  • the specific computer device 500 may include more or fewer components than shown, or combine certain components, or have a different arrangement of components.
  • the processor 502 is configured to run the computer program 5032 stored in the memory, so as to implement the multi-knowledge graph-based intelligent reply method disclosed in the embodiment of the present application.
  • the embodiment of the computer device shown in FIG. 4 does not constitute a limitation on the specific structure of the computer device, and in other embodiments, the computer device may include more or less components than those shown in the figure, Either some components are combined, or different component arrangements.
  • the computer device may only include a memory and a processor. In such an embodiment, the structures and functions of the memory and the processor are the same as those of the embodiment shown in FIG. 4 , and details are not repeated here.
  • the processor 502 may be a central processing unit (Central Processing Unit, CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), Application Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • the general-purpose processor can be a microprocessor or the processor can also be any conventional processor or the like.
  • a computer-readable storage medium may be a non-volatile computer-readable storage medium, or a volatile computer-readable storage medium.
  • the computer-readable storage medium stores a computer program, wherein when the computer program is executed by the processor, the multi-knowledge graph-based intelligent reply method disclosed in the embodiments of the present application is implemented.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Mathematical Physics (AREA)
  • Animal Behavior & Ethology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

Procédé et un appareil de réponse intelligente basée sur un graphe multi-connaissances, dispositif informatique et support de stockage, associés à des mégadonnées et à des graphes de connaissances. Le procédé consiste : si une autre instruction de connexion de session intelligente envoyée par un client est détectée, à obtenir un portrait d'utilisateur cible correspondant à une autre instruction de connexion de session intelligente (S107) ; à déterminer si le portrait d'utilisateur cible est lié de manière correspondante à un premier résultat de sortie ou non (S108) ; et si un premier résultat de sortie est lié dans le portrait d'utilisateur cible, à obtenir un troisième ensemble de problèmes d'association actuels correspondant à partir d'une seconde bibliothèque de graphes de connaissances locale selon le premier résultat de sortie, et à envoyer le troisième ensemble de problèmes d'association actuels au client (S109). Selon le procédé, l'ensemble de problèmes d'association est appelé dans différentes bibliothèques de graphes de connaissances selon des noms d'entités clés correspondant aux portraits de l'utilisateur, des informations correspondantes sont recommandées de manière correspondante aux utilisateurs ayant différents degrés de maîtrise de connaissances, et l'efficacité et la précision d'acquisition d'informations des utilisateurs peuvent être efficacement améliorées.
PCT/CN2021/091291 2020-12-25 2021-04-30 Procédé et appareil de réponse intelligente basée sur un graphe multi-connaissances, dispositif informatique et support de stockage WO2022134421A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011562101.X 2020-12-25
CN202011562101.XA CN112650854B (zh) 2020-12-25 2020-12-25 基于多知识图谱的智能答复方法、装置及计算机设备

Publications (1)

Publication Number Publication Date
WO2022134421A1 true WO2022134421A1 (fr) 2022-06-30

Family

ID=75362972

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/091291 WO2022134421A1 (fr) 2020-12-25 2021-04-30 Procédé et appareil de réponse intelligente basée sur un graphe multi-connaissances, dispositif informatique et support de stockage

Country Status (2)

Country Link
CN (1) CN112650854B (fr)
WO (1) WO2022134421A1 (fr)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115412640A (zh) * 2022-11-02 2022-11-29 北京北投智慧城市科技有限公司 一种基于知识图谱的呼叫中心信息处理方法及装置
CN115810422A (zh) * 2022-12-07 2023-03-17 深圳市宁远科技股份有限公司 基于ai技术的互联网智能自动诊疗应答系统
CN116821287A (zh) * 2023-08-28 2023-09-29 湖南创星科技股份有限公司 基于知识图谱和大语言模型的用户心理画像系统及方法
CN117633254A (zh) * 2024-01-26 2024-03-01 武汉大学 一种基于知识图谱的地图检索用户画像构建方法和系统

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112650854B (zh) * 2020-12-25 2022-09-27 平安科技（深圳）有限公司 基于多知识图谱的智能答复方法、装置及计算机设备
CN113314236A (zh) * 2021-06-01 2021-08-27 南华大学 一种面向高血压的智能问答系统
CN115146075B (zh) * 2022-07-11 2023-03-10 中科雨辰科技有限公司 一种获取知识图谱的数据处理系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108519998A (zh) * 2018-03-07 2018-09-11 北京云知声信息技术有限公司 基于知识图谱的问题引导方法及装置
US20190205726A1 (en) * 2018-01-02 2019-07-04 International Business Machines Corporation Cognitive conversational agent for providing personalized insights on-the-fly
US20200004875A1 (en) * 2018-06-29 2020-01-02 International Business Machines Corporation Query expansion using a graph of question and answer vocabulary
CN110929016A (zh) * 2019-12-10 2020-03-27 北京爱医生智慧医疗科技有限公司 一种基于知识图谱的智能问答方法及装置
CN112650854A (zh) * 2020-12-25 2021-04-13 平安科技（深圳）有限公司 基于多知识图谱的智能答复方法、装置及计算机设备

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104408639A (zh) * 2014-10-22 2015-03-11 百度在线网络技术（北京）有限公司 多轮会话交互方法和系统
CN105068661B (zh) * 2015-09-07 2018-09-07 百度在线网络技术（北京）有限公司 基于人工智能的人机交互方法和系统
US20190164064A1 (en) * 2017-11-27 2019-05-30 Shanghai Xiaoi Robot Technology Co., Ltd. Question and answer interaction method and device, and computer readable storage medium
CN110046238B (zh) * 2019-03-29 2024-03-26 华为技术有限公司 对话交互方法、图形用户界面、终端设备以及网络设备
CN110377715A (zh) * 2019-07-23 2019-10-25 天津汇智星源信息技术有限公司 基于法律知识图谱的推理式精准智能问答方法
CN110704596B (zh) * 2019-09-29 2023-03-31 北京百度网讯科技有限公司 基于话题的对话方法、装置和电子设备
CN111026886B (zh) * 2019-12-26 2023-05-02 成都航天科工大数据研究院有限公司 一种针对专业场景的多轮对话处理方法
CN111339781B (zh) * 2020-02-10 2023-05-30 科大讯飞华南人工智能研究院（广州）有限公司 意图识别方法、装置、电子设备和存储介质
CN111400479B (zh) * 2020-04-14 2023-05-23 支付宝(杭州)信息技术有限公司 针对多轮对话的问题识别方法和装置
CN111552880B (zh) * 2020-04-30 2023-06-30 杭州网易再顾科技有限公司 基于知识图谱的数据处理方法、装置、介质及电子设备

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190205726A1 (en) * 2018-01-02 2019-07-04 International Business Machines Corporation Cognitive conversational agent for providing personalized insights on-the-fly
CN108519998A (zh) * 2018-03-07 2018-09-11 北京云知声信息技术有限公司 基于知识图谱的问题引导方法及装置
US20200004875A1 (en) * 2018-06-29 2020-01-02 International Business Machines Corporation Query expansion using a graph of question and answer vocabulary
CN110929016A (zh) * 2019-12-10 2020-03-27 北京爱医生智慧医疗科技有限公司 一种基于知识图谱的智能问答方法及装置
CN112650854A (zh) * 2020-12-25 2021-04-13 平安科技（深圳）有限公司 基于多知识图谱的智能答复方法、装置及计算机设备

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115412640A (zh) * 2022-11-02 2022-11-29 北京北投智慧城市科技有限公司 一种基于知识图谱的呼叫中心信息处理方法及装置
CN115810422A (zh) * 2022-12-07 2023-03-17 深圳市宁远科技股份有限公司 基于ai技术的互联网智能自动诊疗应答系统
CN115810422B (zh) * 2022-12-07 2023-12-29 深圳市维康致远科技有限公司 基于ai技术的互联网智能自动诊疗应答系统
CN116821287A (zh) * 2023-08-28 2023-09-29 湖南创星科技股份有限公司 基于知识图谱和大语言模型的用户心理画像系统及方法
CN116821287B (zh) * 2023-08-28 2023-11-17 湖南创星科技股份有限公司 基于知识图谱和大语言模型的用户心理画像系统及方法
CN117633254A (zh) * 2024-01-26 2024-03-01 武汉大学 一种基于知识图谱的地图检索用户画像构建方法和系统
CN117633254B (zh) * 2024-01-26 2024-04-05 武汉大学 一种基于知识图谱的地图检索用户画像构建方法和系统

Also Published As

Publication number Publication date
CN112650854A (zh) 2021-04-13
CN112650854B (zh) 2022-09-27

Similar Documents

Publication Publication Date Title
WO2022134421A1 (fr) Procédé et appareil de réponse intelligente basée sur un graphe multi-connaissances, dispositif informatique et support de stockage
CN107797984B (zh) 智能交互方法、设备及存储介质
US9582757B1 (en) Scalable curation system
CN106960030B (zh) 基于人工智能的推送信息方法及装置
US10630798B2 (en) Artificial intelligence based method and apparatus for pushing news
CN110837550A (zh) 基于知识图谱的问答方法、装置、电子设备及存储介质
CN111310436B (zh) 基于人工智能的文本处理方法、装置及电子设备
CN109032381B (zh) 一种基于上下文的输入方法、装置、存储介质及终端
CN109345282A (zh) 一种业务咨询的响应方法及设备
CN109857846B (zh) 用户问句与知识点的匹配方法和装置
CN112925898B (zh) 基于人工智能的问答方法、装置、服务器及存储介质
CN108304424B (zh) 文本关键词提取方法及文本关键词提取装置
CN111694940A (zh) 一种用户报告的生成方法及终端设备
CN110147433A (zh) 一种基于字典树的文本模板提取方法
CN111552797B (zh) 名称预测模型的训练方法、装置、电子设备及存储介质
US20220329556A1 (en) Detect and alert user when sending message to incorrect recipient or sending inappropriate content to a recipient
CN112632261A (zh) 智能问答方法、装置、设备及存储介质
CN112148702A (zh) 一种文件检索的方法及设备
WO2021000400A1 (fr) Procédé et système de génération de paires de problèmes similaires de guide d'hôpital, et dispositif informatique
WO2020000752A1 (fr) Procédé et système de détermination de programme d'application mobile contrefait
CN112966076A (zh) 智能问答问题的生成方法、装置、计算机设备及存储介质
CN117112727A (zh) 适用于云计算业务的大语言模型微调指令集构建方法
WO2023179038A1 (fr) Procédé d'étiquetage de données, plateforme de développement d'ia, grappe de dispositifs informatiques et support de stockage
CN108810640B (zh) 一种电视节目的推荐方法
CN110209804B (zh) 目标语料的确定方法和装置、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: 21908432

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21908432

Country of ref document: EP

Kind code of ref document: A1