CN117633189A - Question-answering session processing method, device, equipment and storage medium - Google Patents
Question-answering session processing method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN117633189A CN117633189A CN202311688184.0A CN202311688184A CN117633189A CN 117633189 A CN117633189 A CN 117633189A CN 202311688184 A CN202311688184 A CN 202311688184A CN 117633189 A CN117633189 A CN 117633189A
- Authority
- CN
- China
- Prior art keywords
- target
- question
- message
- interaction
- job seeker
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/022—Knowledge engineering; Knowledge acquisition
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Human Computer Interaction (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention provides a question-answering session processing method, a device, equipment and a storage medium, which are applied to a server, wherein the method comprises the following steps: acquiring an interaction message aiming at a target post and sent by a job hunting terminal, wherein the interaction message comprises a first interaction message belonging to an answer type and/or a second interaction message belonging to a question type; responding to the first interactive message in the interactive message, forwarding the first interactive message to a first module of the target large language model, so that the first module determines whether a job seeker of the job seeker end is matched with a target post according to the first interactive message; and in response to the second interactive message existing in the interactive message, forwarding the second interactive message to a second module of the target large language model, so that the second module extracts target content from the question-answer knowledge base according to the key fields extracted from the second interactive message to reply to the second interactive message. The scheme realizes accurate reply and smooth question and answer of the job hunting terminal through the target large language model.
Description
Technical Field
The present invention relates to the field of artificial intelligence technologies, and in particular, to a method, an apparatus, a device, and a storage medium for processing a question-answer session.
Background
The large language model (Large Language Model, LLM) refers to a deep learning model trained using large amounts of text data. The large language model may handle a variety of natural language tasks, such as: understanding the meaning of language text, generating natural language text, realizing question-answering, dialogue, etc., is one of the key ways to realize artificial intelligence.
In the recruitment scenario, in order to improve the conversation efficiency between the job seeker and the recruiter, the recruiter generally configures a large language model to quickly respond to conversation information sent by the job seeker through the large language model. However, in practical application, although the large language model improves the response speed during interaction, the accuracy of reply cannot be guaranteed, and the situation of answering questions easily occurs, so that the user experience is affected.
Disclosure of Invention
The embodiment of the invention provides a method, a device, equipment and a storage medium for processing a question-answer session, which are used for improving the accuracy and fluency of the question-answer session when a server side and a job-seeking side are in man-machine interaction.
In a first aspect, an embodiment of the present invention provides a method for processing a question-answer session, which is applied to a server, where the method includes:
acquiring an interaction message aiming at a target post and sent by a job hunting terminal, wherein the interaction message comprises a first interaction message belonging to an answer type and/or a second interaction message belonging to a question type;
Responding to the first interactive message in the interactive message, forwarding the first interactive message to a first module of a target large language model, so that the first module determines whether a job seeker of the job seeker end is matched with the target post according to the first interactive message;
and in response to the existence of the second interactive message in the interactive message, forwarding the second interactive message to a second module of a target large language model, so that the second module extracts target content from a question-answer knowledge base according to key fields extracted from the second interactive message to reply to the second interactive message.
In a second aspect, an embodiment of the present invention provides a question-answering session processing apparatus, applied to a server, where the apparatus includes:
the system comprises an acquisition module, a query module and a query module, wherein the acquisition module is used for acquiring an interaction message aiming at a target post and sent by a job hunting terminal, and the interaction message comprises a first interaction message belonging to an answer type and/or a second interaction message belonging to a question type;
the first processing module is used for responding to the first interaction message in the interaction message, forwarding the first interaction message to a first module of a target large language model, so that the first module determines whether a job seeker of the job seeker is matched with the target post according to the first interaction message;
And the second processing module is used for forwarding the second interactive message to a second module of a target large language model in response to the existence of the second interactive message in the interactive message, so that the second module extracts target content from a question-answer knowledge base according to the key field extracted from the second interactive message to reply to the second interactive message.
In a third aspect, an embodiment of the present invention provides an electronic device, including: a memory, a processor, a communication interface; wherein the memory has executable code stored thereon, which when executed by the processor, causes the processor to at least implement the question-answering session processing method according to the first aspect.
In a fourth aspect, embodiments of the present invention provide a non-transitory machine-readable storage medium having executable code stored thereon, which when executed by a processor of an electronic device, causes the processor to at least implement a method for processing a question-answer session as described in the first aspect.
In the scheme provided by the embodiment of the invention, firstly, the interactive message aiming at the target post and sent by the job seeker is obtained. Wherein the interactive message may comprise a first interactive message belonging to the answer type and/or a second interactive message belonging to the question type, depending on the information type. Then, in response to the first interaction message existing in the interaction message, forwarding the first interaction message to a first module of the target large language model, so as to determine whether a job seeker of a job seeker end is matched with a target post or not according to the first interaction message through the first module; and in response to the existence of the second interactive message in the interactive message, forwarding the second interactive message to a second module of the target large language model, so that target content is extracted from a question-answer knowledge base through the second module according to key fields extracted from the second interactive message to reply to the second interactive message. According to the scheme, the different functional modules (the first module and the second module) for processing the different types of interaction messages are configured in the target large language model, so that accurate reply and smooth question-answering of the job hunting terminal are realized.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic diagram of a hardware execution environment of a question-answering session processing method according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a cloud computing environment of a question-answer session processing method according to an embodiment of the present invention;
fig. 3 is an application schematic diagram of a question-answer session processing method according to an embodiment of the present invention;
fig. 4 is a flowchart of a question-answering session processing method provided by an embodiment of the present invention;
FIG. 5 is a flowchart of another method for processing a question-answer session according to an embodiment of the present invention;
FIG. 6 is a flowchart of another method for processing a question-answer session according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of a question-answering session processing device according to an embodiment of the present invention;
fig. 8 is a schematic structural diagram of an electronic device corresponding to the question-answer session processing apparatus provided in the embodiment shown in fig. 7.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that, the user information (including but not limited to user equipment information, user personal information, etc.) and the data (including but not limited to data for analysis, stored data, presented data, etc.) related to the embodiments of the present invention are information and data authorized by the user or fully authorized by each party, and the collection, use and processing of the related data need to comply with the related laws and regulations and standards of the related country and region, and provide corresponding operation entries for the user to select authorization or rejection.
Some embodiments of the present invention are described in detail below with reference to the accompanying drawings. In the case where there is no conflict between the embodiments, the following embodiments and features in the embodiments may be combined with each other. In addition, the sequence of steps in the method embodiments described below is only an example and is not strictly limited.
The large language model (Large Language Model, LLM) refers to a deep learning model trained using large amounts of text data. The large language model may handle a variety of natural language tasks, such as: understanding meaning of language text, generating natural language text, implementing questions and answers, dialogue, etc., is one of the key ways to implement artificial intelligence.
In application scenes such as shopping and recruitment, in order to improve conversation efficiency, a merchant or recruiter is generally configured with a large language model, namely a commonly-called customer service robot, and the customer end input requirement is understood through the large language model, so that the problem of the customer end can be quickly answered. However, in the actual use process, the situations that the large language model replies inaccurately to the problem and the problem is not asked by the user exist, and the user experience is affected.
In this embodiment, in order to facilitate understanding, taking a recruitment scenario as an example, a question-answer session processing method is provided, where the method is applied to a server, and includes: firstly, an interaction message aiming at a target post and sent by a job hunting terminal is obtained. Wherein the interactive message may comprise a first interactive message belonging to the answer type and/or a second interactive message belonging to the question type, depending on the information type. Then, in response to the first interaction message existing in the interaction message, forwarding the first interaction message to a first module of the target large language model, so as to determine whether a job seeker of a job seeker end is matched with a target post or not according to the first interaction message through the first module; and in response to the existence of the second interactive message in the interactive message, forwarding the second interactive message to a second module of the target large language model, so that target content is extracted from a question-answer knowledge base through the second module according to key fields extracted from the second interactive message to reply to the second interactive message. In the scheme, the capacity of the target large language model is modularized, and the accurate reply and smooth question and answer to the job hunting terminal are realized by respectively configuring a first module for processing the interactive message of the answer type and a second module for processing the interactive message of the question type in the target large language model.
The method for processing the question-answer session provided by the embodiment of the invention is described below.
Fig. 1 is a schematic diagram of a hardware execution environment of a question-answer session processing method according to an embodiment of the present invention, and as shown in fig. 1, the hardware execution environment of the question-answer session processing method may be formed by job seeker device 101 and server device 102, where job seeker device 101 is communicatively connected to server device 102. The job hunting terminal device 101 may be a certain terminal device, such as a smart phone, a tablet computer, a PC, and the like. The server device 102 may be a server of a content provider or a cloud server of a cloud service provider.
In an actual recruitment scenario, the server device 102 is also communicatively coupled to a recruitment device (not shown in fig. 1). The service side serves as a recruitment service provider, recruitment related services are deployed on the service side device 102, and the job seeker device 101 and the recruitment side device implement use of the recruitment related services by deploying application programs corresponding to the recruitment related services deployed on the service side device 102. For example: the job seeker device 101 delivers job seekers to the recruiter through the application, and the recruiter device sends job invitations and the like to the job seeker through the application.
In this embodiment, a large language model is deployed in the server to replace the recruitment end to perform a question-answer session with the job seeker.
In practical applications, the server device 102 may be a separate physical server or a physical server cluster maintained by a content provider, or may be a cloud server, called a computing node, maintained by a cloud server. Fig. 2 is a schematic diagram of a cloud computing environment of a question-answer session processing method according to an embodiment of the present invention, where in the cloud computing environment shown in fig. 2, a plurality of computing nodes (cloud servers) may be distributed (201-1, 201-2, … shown in fig. 2) and each computing node has processing resources such as computation and storage. In a cloud computing environment, a service may be provided by multiple computing nodes, although one computing node may also provide one or more services, such as service a, service B, service C, and service D illustrated in fig. 2. The service may be provided in the cloud computing environment by providing a service interface to the outside, and the job seeker device invokes the service interface to use the corresponding service. The service interface includes a software development kit (Software Development Kit, abbreviated as SDK), an application program interface (Application Programming Interface, abbreviated as API), and the like.
The services described above are deployed in accordance with various virtualization technologies supported by the cloud computing environment, such as virtual machine, container-based virtualization technologies. Taking container-based virtualization techniques as an example, several containers corresponding to one service may be assembled into one container group (pod). A service B such as that illustrated in fig. 2 may be configured with one or more pods, each of which may include an agent and one or more containers. One or more containers in the pod are used to process requests related to one or more corresponding functions of the service, and agents in the pod are used to control network functions related to the service, such as routing, load balancing, etc.
In operation, executing a request from a job hunting end device may require invoking one or more services in the cloud computing environment, and executing one or more functions of one service may require invoking one or more functions of another service. As shown in fig. 2, after receiving a request sent by the job seeker device, the service a may call the service B, and the service B may request the service D to perform one or more functions.
In the cloud computing environment, the embodiment of the invention provides an application schematic diagram of a question-answer session processing method shown in fig. 3.
In fig. 3, a question-answer session processing service and a corresponding service interface are provided in a cloud computing environment, a job-asking terminal device invokes the service interface to trigger a request for the question-answer session processing service, the request includes an interaction message for a target post sent by the job-asking terminal, the cloud computing environment responds to the request, determines a computing node for responding to the request and providing the question-answer session processing service, determines a target large language model for responding to the interaction message by using processing resources in the computing node, and processes the interaction message through a corresponding module in the target large language model so as to feed back a processing result to the job-asking terminal device for display.
The following describes in detail the execution procedure of the question-answer session processing method provided by the embodiment of the present invention with reference to fig. 4. The question-answer session processing method is executed by the server device, for example: is executed by a computing node in the cloud computing environment.
Fig. 4 is a flowchart of a question-answering session processing method provided by an embodiment of the present invention, where the method is applied to a server, and as shown in fig. 4, the method may include the following steps:
401. and acquiring an interaction message aiming at a target post and sent by the job seeker, wherein the interaction message comprises a first interaction message belonging to an answer type and/or a second interaction message belonging to a question type.
402. And in response to the first interactive message in the interactive message, forwarding the first interactive message to a first module of the target large language model, so that the first module determines whether the job seeker of the job seeker end is matched with the target post according to the first interactive message.
403. And in response to the second interactive message existing in the interactive message, forwarding the second interactive message to a second module of the target large language model, so that the second module extracts target content from the question-answer knowledge base according to the key fields extracted from the second interactive message to reply to the second interactive message.
As shown in fig. 3, in the question-answer session processing method in this embodiment, the interactive messages sent by the job seeker for the target posts are divided into two types according to the message types, namely an answer type and a question type. For convenience of distinction, the interactive message belonging to the answer type is referred to as a first interactive message, and the interactive message belonging to the question type is referred to as a second interactive message.
A first module and a second module are arranged in the target large language model corresponding to the server side, and the first module is configured to process a first interaction message, namely, whether a job seeker of the job seeker is matched with a target post or not is determined according to the first interaction message; the second module is configured to process the second interactive message by extracting the target content from the question-answer knowledge base based on the key fields extracted from the second interactive message to reply to the second interactive message.
In this embodiment, optionally, the target large language model may be deployed directly on the server, or may be deployed on a location independent of the server, where the server can communicate with the target large language model.
After the interactive message for the target post sent by the job hunting terminal is obtained, determining the message type contained in the interactive message. Specifically, if the interactive message is an answer message, only the first interactive message belonging to the answer type is included; if the interactive message is a question message, only a second interactive message belonging to the question type is included; if the interactive message is a question-answer message, the answer part in the question-answer message belongs to an answer type, the question part in the question-answer message belongs to a question type, and the interactive message comprises a first interactive message belonging to the answer type and a second interactive message belonging to the question type.
If the first interactive message exists in the interactive messages, the first interactive message is forwarded to a first module of the target large language model, so that whether the job seeker of the job seeker is matched with the target post or not is determined through the first module according to the first interactive message.
As a way of optionally determining whether the job seeker of the job seeker end matches the target position, first, extracting, by the first module, job seeker information related to at least one position recruitment condition of the target position from the first interaction information; and then, determining whether the job seeker of the job seeker end is matched with the target post according to the matching condition of the job seeker information and at least one post recruitment condition.
Wherein, at least one post recruitment condition of the target post, namely the recruitment requirement of the target post. For example, recruitment requirements for a target post include: age 45 years or less, at least 5 years of practise experience, having a premium mechanic certificate, etc. Job seeker information, i.e., job seeker personal information corresponding to recruitment requirements for the target post, such as: job seekers are 35 years old, 6 years old from practise, have advanced mechanic certificates, and so on.
In the specific implementation process, if the job seeker information is matched with at least one post recruitment condition, determining that the job seeker of the job seeker end is matched with the target post. For example, assume that at least one post recruitment condition for a post is: the age of the job seeker is less than 45 years and the working age is 5 years, if the job seeker information corresponds to 38 years, and the working age is 6 years, the job seeker information is considered to be matched with at least one post recruitment condition, and the job seeker at the job seeker end is determined to be matched with the post. And in response to determining that the job seeker of the job seeker is matched with the target post, sending an about-face message aiming at the target post to the job seeker. The job seeker is used for checking whether the job seeker participates in the interview of the target post, and the interview is completed by the job seeker at the job seeker end and the recruiter at the recruiter end in an on-line or off-line mode. And responding to the confirmation operation of the job seeker on the appointment message, and sending an interview notification to the recruiter corresponding to the target post so that the recruiter of the recruiter interviews the job seeker at the appointment time and the appointment place.
If the job seeker information is not matched with at least one post recruitment condition, determining that the job seeker of the job seeker end is not matched with the target post. And sending an ending message aiming at the target post to the job seeker in response to the fact that the job seeker of the job seeker is not matched with the target post. The end message is used for informing the job seeker that the conditions of the job seeker are not matched with recruitment requirements of the target post, and further interviewing cannot be performed.
If the second interactive message exists in the interactive message, forwarding the second interactive message to a second module of the target large language model, so that target content is extracted from the question-answer knowledge base through the second module according to the key fields extracted from the second interactive message to reply to the second interactive message.
The question-answer knowledge base comprises a plurality of question-answer pairs corresponding to the target large language model in the job-seeking and recruitment scene. The plurality of question-answer pairs includes: question-answer pairs corresponding to training samples used in the training process of the target large language model, and question-answer pairs corresponding to newly generated question-answer sessions in the using process of the target large language model. The question-answer pairs corresponding to the training samples further comprise: a first question-answer pair corresponding to a first training sample used when performing question-answer training on the target large language model that is independent of the target post, and a second question-answer pair corresponding to a second training sample used when performing question-answer training on the target large language model that is dependent on the target post.
As a way of optionally generating a reply message corresponding to the second interaction message, the key field extracted from the second interaction message by the second module may be first extracted; then, at least one target question-answer pair is extracted from a question-answer knowledge base according to the extracted key fields; each target question-answer pair comprises target questions and target answers, and the similarity between the target questions and the key fields in at least one target question-answer pair is larger than a preset similarity threshold; and then, generating a reply message corresponding to the second interaction message according to the target answer in the at least one target question-answer pair and the post description information corresponding to the target post.
Because the number of the second question-answer pairs is smaller than that of the first question-answer pairs, and the question-answer corresponding to the question-answer session newly generated in the using process of the target large language model corresponds to the recent application scene of the target large language model, optionally, extracting at least one target question-answer pair from the question-answer knowledge base comprises: querying at least one target question-answer pair matched with the extracted key field from a second question-answer pair of a question-answer knowledge base or a question-answer pair corresponding to a newly generated question-answer session of the target large language model in the use process; if the second question-answer pair and the question-answer pair corresponding to the question-answer session newly generated in the using process do not have at least one target question-answer pair matched with the extracted key field, querying at least one target question-answer pair matched with the extracted key field from the first question-answer pair of the question-answer knowledge base.
The post description information corresponding to the target post can be understood as a priori knowledge corresponding to the target post, such as: target posts relate to interpretation information of concepts, etc. Optionally, post description information corresponding to different posts respectively can be preconfigured in the target large language model so as to be used in a question-answering session. The post description information is used for assisting a second module of the target large language model to better understand the second interaction information and generate a reply message corresponding to the second interaction message more accurately.
In this embodiment, when the server side performs a question-answer session on the job seeker, first, the server side obtains an interaction message for the target post sent by the job seeker. Wherein the interactive message may comprise a first interactive message belonging to the answer type and/or a second interactive message belonging to the question type, depending on the information type. Then, in response to the first interaction message existing in the interaction message, forwarding the first interaction message to a first module of the target large language model, so as to determine whether a job seeker of a job seeker end is matched with a target post or not according to the first interaction message through the first module; and in response to the existence of the second interactive message in the interactive message, forwarding the second interactive message to a second module of the target large language model, so that target content is extracted from a question-answer knowledge base through the second module according to key fields extracted from the second interactive message to reply to the second interactive message. The method comprises the steps of modularization of the capability of the target large language model, namely, respectively configuring a first module for processing the interactive message of the answering type and a second module for processing the interactive message of the question type in the target large language model, so that the interactive information sent by the job hunting terminal is differentiated according to the message type, and the accurate answer and fluent question and answer of a question and answer session are improved.
In another embodiment, the server further has a questioning capability, and the questioning capability may be implemented by a third module of the target large language model. The following description is made with reference to specific embodiments.
Fig. 5 is a flowchart of another question-answering session processing method provided by the embodiment of the present invention, where the method is applied to a server, and as shown in fig. 5, the method may include the following steps:
501. and acquiring an interaction message aiming at a target post and sent by the job seeker, wherein the interaction message comprises a first interaction message belonging to an answer type and/or a second interaction message belonging to a question type.
502. And in response to the first interactive message existing in the interactive message, forwarding the first interactive message to a first module of the target large language model, so that the first module extracts job seeker information related to at least one position recruitment condition of the target position from the first interactive message.
503. And determining the target post recruitment condition in which the corresponding job seeker information is not acquired in the post recruitment conditions of the target post.
504. And generating a target interaction message belonging to the questioning type according to the target post recruitment condition through a third module of the target large language model so as to prompt the job seeker to feed back job seeker information related to the target post recruitment condition, so that the first module determines whether the job seeker of the job seeker is matched with the target post according to the job seeker information and at least one post recruitment condition.
505. And in response to the second interactive message existing in the interactive message, forwarding the second interactive message to a second module of the target large language model, so that the second module extracts target content from the question-answer knowledge base according to the key fields extracted from the second interactive message to reply to the second interactive message.
The specific implementation process of step 501, step 502 and step 503 may refer to the foregoing embodiments, and the details of this embodiment are not repeated.
It can be appreciated that when interacting with the job seeker, the first module of the target large language model needs to obtain job seeker information related to at least one job recruitment condition of the target job from the first interaction information to further determine whether the job seeker of the job seeker matches the target job. If the method can actively generate the corresponding question related to at least one post recruitment condition of the target post and feed the question back to the job seeker, the conversation efficiency is improved.
Therefore, in this embodiment, the server may continuously determine, during the session, whether the target post recruitment condition exists in the post recruitment conditions of the target post, where the corresponding job seeker information is not acquired, for example: and confirming once at intervals, or confirming once after each interactive message for the target post sent by the job seeker is obtained.
If the target post recruitment condition of the target post does not acquire the corresponding job seeker information exists, generating a target interaction message belonging to the question type according to the target post recruitment condition through a third module of the target large language model, and sending the target interaction message to a job seeker so as to prompt the job seeker to feed back the job seeker information related to the target post recruitment condition.
And then, the first module of the target large language model determines new job seeker information according to the first interaction message corresponding to the target interaction message, and determines whether the job seeker of the job seeker end is matched with the target job according to all the job seeker information corresponding to the job seeker and at least one job recruitment condition. Wherein, the whole job seeker information can be extracted from different first interactive messages.
In this embodiment, by actively generating the target interaction message related to the at least one post recruitment condition of the target post by the third module of the target large language model, flexibility of the target large language model is improved, and efficiency of the question-answer session of the server in the recruitment scene is higher.
Fig. 6 is a flowchart of another method for processing a question-answer session, which is applied to a server, and as shown in fig. 6, may include the following steps:
601. And acquiring an interaction message aiming at a target post and sent by the job seeker, wherein the interaction message comprises a first interaction message belonging to an answer type and/or a second interaction message belonging to a question type.
602. And determining the current target interaction field according to the target post and/or the interaction information.
603. And determining a target large language model which is applicable to the interaction field and is the same as the target interaction field from a plurality of large language models which are configured by the server and are respectively applicable to different interaction fields.
604. And in response to the first interactive message in the interactive message, forwarding the first interactive message to a first module of the target large language model, so that the first module determines whether the job seeker of the job seeker end is matched with the target post according to the first interactive message.
605. And in response to the second interactive message existing in the interactive message, forwarding the second interactive message to a second module of the target large language model, so that the second module extracts target content from the question-answer knowledge base according to the key fields extracted from the second interactive message to reply to the second interactive message.
The specific implementation process of step 601, step 604 and step 605 may refer to the foregoing embodiments, and the details of this embodiment are not described herein.
In this embodiment, in order to further improve accuracy and fluency of the question-answer session between the server and the job-seeking end, the interaction field applicable to the target large language model used by the server in this embodiment is the same as the target interaction field corresponding to the current session.
The interaction field may be understood as a field associated with a target post in a current job hunting and recruitment scene, and the like. Different interaction domain partitioning rules, corresponding interaction domains may differ. For example, when the interaction fields are divided in terms of industries to which posts belong, the interaction fields can be divided into: home administration field, truck driver field, house renting field, moving field, etc. When the interaction domain is divided according to the domain to which the specific semantic content related to the interaction information in the session belongs, the interaction domain may be divided into: post content consultation field, post related policy consultation field, post salary treatment consultation field, etc.
It can be understood that different interaction fields have the specificity of the fields, the expressive power of the same large language model in different interaction fields is also different, and the accuracy of the session message cannot be ensured by adopting a single large language model to respond to the interaction information of all the interaction fields.
To solve this problem, in this embodiment, after the interaction message for the target post sent by the job seeker is obtained, the current target interaction field is determined according to the target post and/or the interaction information; and then, determining a target large language model which is the same as the target interaction field in the applicable interaction field from a plurality of large language models which are configured by the server and are respectively applicable to different interaction fields.
As a way of optionally determining the target interaction field, a correspondence relationship between a plurality of different interaction fields and their corresponding descriptors may be established in advance. After the interactive message aiming at the target post and sent by the job hunting terminal is obtained, extracting keywords from the target post and/or the interactive message; and then, determining a target post and/or a target interaction field corresponding to the interaction message by calculating the similarity between the extracted keywords and the descriptors corresponding to different interaction fields.
For example, the description words corresponding to various different question fields and the keywords extracted from the target post and/or the interaction message may be vectorized, then cosine similarity between the description word corresponding vector and the keyword corresponding vector is calculated, and if cosine similarity between the keyword corresponding vector and the description word corresponding vector of a certain interaction field x is greater than a set similarity threshold, the target interaction field corresponding to the target post and/or the interaction message is determined to be the interaction field x.
As another alternative to determining the target interaction field, the current corresponding target interaction field may also be determined directly according to the field to which the target post belongs. For example, if the target post is a nurse, the currently corresponding target interaction domain may be determined to be a home domain.
And then determining whether target large language models corresponding to the target interaction field exist in a plurality of large language models respectively suitable for different interaction fields configured by the server side. If the target large language model corresponding to the target interaction field does not exist, the conversation is carried out in a non-large language model mode, for example, corresponding prompt information is output to prompt the conversation to be carried out in a manual mode. And if the target large language model corresponding to the target interaction field exists, determining to conduct conversation through the target large language model.
In this embodiment, a plurality of large language models respectively applicable to different interaction fields are configured at a server, and a correspondence exists between the plurality of large language models and the interaction field. In practical application, whether the interaction fields are divided according to industries to which posts belong or the interaction fields are divided according to the fields to which specific semantic content related to a conversation belongs, conversation interaction is performed through a target large language model matched with a target interaction field corresponding to a current conversation, and therefore accuracy of a question-answering conversation performed by a server in an actual process is improved.
In practical application, in order to make the large language model have better model performance in the application scene of recruitment, prior knowledge information in the post association field is acquired before the large language model is used, for example: working content, recruitment requirements and the like of posts in different fields; and then, carrying out fine tuning training on the large language model according to the priori knowledge information, so that the large language model has the capability of carrying out conversations in a plurality of different fields associated with posts. However, in practical applications, since a large language model has the ability to perform a conversation in a certain area, it does not mean that it is good at handling conversations in that area, and therefore, it is important to confirm a conversation area (i.e., an interaction area) to which the large language model is applied in order to improve the accuracy of a conversation of the large language model.
Optionally, in this embodiment, by adding the modes of "the front target large language model is applicable to the interaction field confirmation" and "the rear target large language model is applicable to the interaction field update", the accuracy and fluency of the large language model in the question-answer session process are further improved.
The preposed large language model is suitable for an interactive field confirmation stage, and optionally, a plurality of question-answer test messages corresponding to the target large language model in a model test stage can be acquired first; and then, determining the applicable interaction field of the target large language model according to a plurality of question-answer accuracy evaluation results respectively corresponding to the plurality of question-answer test messages. For example, if the question-answer accuracy evaluation results corresponding to the question-answer test messages indicate that the question-answer of the target large language model is accurate, determining the interaction field corresponding to the question-answer test messages as the interaction field applicable to the target large language model.
Alternatively, the question-answer accuracy evaluation result may be determined by means of, for example, manual labeling or the like.
In the update stage of the interaction field applicable to the postposition large language model, optionally, a plurality of question-answer messages corresponding to the target large language model in the process of carrying out the session with the job seeker can be acquired first; and then updating the applicable interaction field of the target large language model according to a plurality of question-answer accuracy evaluation results respectively corresponding to the question-answer messages.
For example, when there are evaluation results of inaccurate questions and answers among the plurality of question and answer accuracy evaluation results respectively corresponding to the plurality of question and answer messages, it is indicated that the target large language model has general performance in the current question and answer session. Optionally, the target interaction domain may be eliminated from the applicable interaction domains of the target large language model. When a plurality of question-answer accuracy evaluation results respectively corresponding to the plurality of question-answer messages indicate that the target large language model is accurate in question-answer, the method indicates that the large language model has normal performance in the question-answer session, and the target interaction field is still used as the applicable interaction field of the large language model. Therefore, the accuracy and fluency of the target large language model in the session interaction process can be ensured by updating the mode of the target large language model in time, wherein the mode is suitable for the interaction field.
A question-answer session processing device of one or more embodiments of the present invention will be described in detail below. Those skilled in the art will appreciate that these means may be configured by the steps taught by the present solution using commercially available hardware components.
Fig. 7 is a schematic structural diagram of a question-answering session processing device provided by an embodiment of the present invention, which is applied to a server, as shown in fig. 7, and the device includes: an acquisition module 11, a first processing module 12, a second processing module 13.
The obtaining module 11 is configured to obtain an interaction message for a target post sent by a job seeker, where the interaction message includes a first interaction message belonging to an answer type and/or a second interaction message belonging to a question type.
And the first processing module 12 is configured to forward the first interaction message to a first module of a target large language model in response to the first interaction message existing in the interaction message, so that the first module determines whether a job seeker of the job seeker matches the target post according to the first interaction message.
And the second processing module 13 is configured to forward the second interaction message to a second module of the target large language model in response to the second interaction message existing in the interaction message, so that the second module extracts target content from a question-answer knowledge base according to the key field extracted from the second interaction message to reply to the second interaction message.
Optionally, the first processing module 12 is specifically configured to extract, from the first interaction information, job seeker information related to at least one job recruitment condition of the target job; and determining whether the job seeker of the job seeker end is matched with the target post according to the matching condition of the job seeker information and the at least one post recruitment condition.
Optionally, the first processing module 12 is specifically configured to determine that the job applicant at the job applicant end matches the target job if the job applicant information matches the at least one job recruitment condition; responsive to determining that a job seeker of the job seeker matches the target post, sending an about-face message for the target post to the job seeker, the about-face message being used to determine whether the job seeker is engaged in interviewing of the target post; and responding to the confirmation operation of the job hunting terminal to the about surface message, and sending a interview notification to the recruitment terminal corresponding to the target post.
Optionally, the second processing module 13 is specifically configured to extract at least one target question-answer pair from a question-answer knowledge base according to the key field extracted from the second interaction message; each target question-answer pair comprises target questions and target answers, and the similarity between the target questions in at least one target question-answer pair and the key field is larger than a preset similarity threshold; and generating a reply message corresponding to the second interaction message according to the target answer in the at least one target question-answer pair and the post description information corresponding to the target post.
Optionally, the first processing module 12 is further configured to determine a target post recruitment condition in which no corresponding job seeker information is acquired from the target post recruitment condition, where the job seeker information is acquired from the first interaction information; and generating a target interaction message belonging to a questioning type according to the target post recruitment condition through a third module of the target large language model so as to prompt the job seeker to feed back job seeker information related to the target post recruitment condition.
Optionally, the server is configured with a plurality of large language models applicable to different interaction fields, and the question-answering session processing device further includes: and a determining module. The determining module is used for determining the current target interaction field according to the target post and/or the interaction information; determining the target large language model with the same applicable interaction field as the target interaction field from the large language models
Optionally, the determining module is further configured to obtain a plurality of question-answer test messages corresponding to the target large language model in a model test stage; determining the applicable interaction field of the target large language model according to a plurality of question-answer accuracy evaluation results respectively corresponding to the plurality of question-answer test messages; and/or acquiring a plurality of question-answer messages corresponding to the target large language model in the process of carrying out the session with the job seeker; and updating the applicable interaction field of the target large language model according to a plurality of question-answer accuracy evaluation results respectively corresponding to the question-answer messages.
The apparatus shown in fig. 7 may perform the steps described in the foregoing embodiments, and detailed execution and technical effects are referred to in the foregoing embodiments and are not described herein.
In one possible design, the structure of the question-answering session processing apparatus shown in fig. 7 may be implemented as an electronic device, as shown in fig. 8, where the electronic device may include: memory 21, processor 22, communication interface 23. Wherein the memory 21 has stored thereon executable code which, when executed by the processor 22, causes the processor 22 to at least implement the question-answer session processing method as provided in the foregoing embodiments.
In addition, embodiments of the present invention provide a non-transitory machine-readable storage medium having executable code stored thereon, which when executed by a processor of an electronic device, causes the processor to at least implement a question-answering session processing method as provided in the foregoing embodiments.
The apparatus embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by adding necessary general purpose hardware platforms, or may be implemented by a combination of hardware and software. Based on such understanding, the foregoing aspects, in essence and portions contributing to the art, may be embodied in the form of a computer program product, which may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A question-answering session processing method is applied to a server and is characterized by comprising the following steps:
acquiring an interaction message aiming at a target post and sent by a job hunting terminal, wherein the interaction message comprises a first interaction message belonging to an answer type and/or a second interaction message belonging to a question type;
responding to the first interactive message in the interactive message, forwarding the first interactive message to a first module of a target large language model, so that the first module determines whether a job seeker of the job seeker end is matched with the target post according to the first interactive message;
and in response to the existence of the second interactive message in the interactive message, forwarding the second interactive message to a second module of a target large language model, so that the second module extracts target content from a question-answer knowledge base according to key fields extracted from the second interactive message to reply to the second interactive message.
2. The method of claim 1, wherein the first module determining whether the job seeker of the job seeker matches the target post based on the first interaction message comprises:
the first module extracts job seeker information related to at least one post recruitment condition of the target post from the first interaction information;
And determining whether the job seeker of the job seeker end is matched with the target post according to the matching condition of the job seeker information and the at least one post recruitment condition.
3. The method of claim 2, wherein the determining whether the job applicant at the job applicant end matches the target job based on the matching of the job applicant information and the at least one job recruitment condition comprises:
if the job seeker information is matched with the at least one post recruitment condition, determining that the job seeker of the job seeker end is matched with the target post;
responsive to determining that a job seeker of the job seeker matches the target post, sending an about-face message for the target post to the job seeker, the about-face message being used to determine whether the job seeker is engaged in interviewing of the target post;
and responding to the confirmation operation of the job hunting terminal to the about surface message, and sending a interview notification to the recruitment terminal corresponding to the target post.
4. The method of claim 1, wherein the second module extracting the target content from the knowledge base of questions and answers based on the key fields extracted from the second interactive message to reply to the second interactive message comprises:
The second module extracts at least one target question-answer pair from a question-answer knowledge base according to the key field extracted from the second interactive message; each target question-answer pair comprises target questions and target answers, and the similarity between the target questions in at least one target question-answer pair and the key field is larger than a preset similarity threshold;
and generating a reply message corresponding to the second interaction message according to the target answer in the at least one target question-answer pair and the post description information corresponding to the target post.
5. The method according to any one of claims 1 to 4, further comprising:
determining a target post recruitment condition in which corresponding job seeker information is not acquired in the post recruitment conditions of the target posts, wherein the job seeker information is acquired from the first interaction information;
and generating a target interaction message belonging to a questioning type according to the target post recruitment condition through a third module of the target large language model so as to prompt the job seeker to feed back job seeker information related to the target post recruitment condition.
6. The method of claim 1, wherein the server is configured with a plurality of large language models respectively applicable to different interaction fields, and after the interaction message for the target post sent by the job seeker is obtained, the method further comprises:
Determining the current target interaction field according to the target post and/or the interaction information;
and determining the target large language model with the same applicable interaction field as the target interaction field from the large language models.
7. The method of claim 6, wherein the method further comprises:
acquiring a plurality of question-answer test messages corresponding to the target large language model in a model test stage;
determining the applicable interaction field of the target large language model according to a plurality of question-answer accuracy evaluation results respectively corresponding to the plurality of question-answer test messages; and/or the number of the groups of groups,
acquiring a plurality of question-answer messages corresponding to the target large language model in the process of carrying out a session with the job seeker;
and updating the applicable interaction field of the target large language model according to a plurality of question-answer accuracy evaluation results respectively corresponding to the question-answer messages.
8. A question-answering session processing device, which is applied to a server, and comprises:
the system comprises an acquisition module, a query module and a query module, wherein the acquisition module is used for acquiring an interaction message aiming at a target post and sent by a job hunting terminal, and the interaction message comprises a first interaction message belonging to an answer type and/or a second interaction message belonging to a question type;
The first processing module is used for responding to the first interaction message in the interaction message, forwarding the first interaction message to a first module of a target large language model, so that the first module determines whether a job seeker of the job seeker is matched with the target post according to the first interaction message;
and the second processing module is used for forwarding the second interactive message to a second module of a target large language model in response to the existence of the second interactive message in the interactive message, so that the second module extracts target content from a question-answer knowledge base according to the key field extracted from the second interactive message to reply to the second interactive message.
9. An electronic device, comprising: a memory, a processor, a communication interface; wherein the memory has stored thereon executable code which, when executed by the processor, causes the processor to perform the question-answering session processing method according to any one of claims 1 to 7.
10. A non-transitory machine-readable storage medium having stored thereon executable code which, when executed by a processor of an electronic device, causes the processor to perform the question-answering session processing method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311688184.0A CN117633189A (en) | 2023-12-08 | 2023-12-08 | Question-answering session processing method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311688184.0A CN117633189A (en) | 2023-12-08 | 2023-12-08 | Question-answering session processing method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117633189A true CN117633189A (en) | 2024-03-01 |
Family
ID=90023254
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311688184.0A Pending CN117633189A (en) | 2023-12-08 | 2023-12-08 | Question-answering session processing method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117633189A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109255594A (en) * | 2018-09-30 | 2019-01-22 | 福建海峡中创网络信息技术股份有限公司 | The method and system of registration is brought, recommended to AI intelligence people hilllock information discriminating, matching based on recruitment platform together |
| CN110866096A (en) * | 2019-10-15 | 2020-03-06 | 平安科技（深圳）有限公司 | Intelligent answer control method and device, computer equipment and storage medium |
| KR20220108648A (en) * | 2021-01-27 | 2022-08-03 | 박광훈 | System for providing non face-to-face online interview platform service |
| CN115640386A (en) * | 2022-08-30 | 2023-01-24 | 胜斗士(上海)科技技术发展有限公司 | Method and apparatus for conducting dialogs based on recommended dialogs |
| CN117078220A (en) * | 2023-08-31 | 2023-11-17 | 北京五八信息技术有限公司 | Recruitment service method and device based on AI, electronic equipment and storage medium |
-
2023
- 2023-12-08 CN CN202311688184.0A patent/CN117633189A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109255594A (en) * | 2018-09-30 | 2019-01-22 | 福建海峡中创网络信息技术股份有限公司 | The method and system of registration is brought, recommended to AI intelligence people hilllock information discriminating, matching based on recruitment platform together |
| CN110866096A (en) * | 2019-10-15 | 2020-03-06 | 平安科技（深圳）有限公司 | Intelligent answer control method and device, computer equipment and storage medium |
| KR20220108648A (en) * | 2021-01-27 | 2022-08-03 | 박광훈 | System for providing non face-to-face online interview platform service |
| CN115640386A (en) * | 2022-08-30 | 2023-01-24 | 胜斗士(上海)科技技术发展有限公司 | Method and apparatus for conducting dialogs based on recommended dialogs |
| CN117078220A (en) * | 2023-08-31 | 2023-11-17 | 北京五八信息技术有限公司 | Recruitment service method and device based on AI, electronic equipment and storage medium |
Non-Patent Citations (3)
| Title |
|---|
| 唐红杰;: "基于Web的求职招聘系统分析与设计", 计算机时代, no. 06, 15 June 2013 (2013-06-15) * |
| 李忠58 技术: "人物｜李忠： AI 面试机器人打造智能 化招聘", pages 1 - 16, Retrieved from the Internet <URL:https://mp.weixin.qq.com/s/7 MgW7BiRL3u7jFuvPNnseg> * |
| 陈希;樊治平;: "考虑多种形式信息的求职者与岗位双边匹配研究", 运筹与管理, no. 06, 25 December 2009 (2009-12-25) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11956187B2 (en) | Natural language processing for information extraction | |
| CN117573834B (en) | Multi-robot dialogue method and system for software-oriented instant service platform | |
| CN110377712B (en) | Intelligent session switching method, device, equipment and storage medium | |
| CN109002515A (en) | A kind of method and apparatus of intelligent response | |
| CN119670863A (en) | Information generation method, device, electronic device and storage medium | |
| CN107103005A (en) | The collection method and device of question and answer language material | |
| CN110765253A (en) | Method, device and equipment for replying question of user in customer service scene | |
| CN109376354A (en) | Fraud recognition methods, device, electronic equipment and readable storage medium storing program for executing | |
| CN113901837A (en) | Intention understanding method, device, equipment and storage medium | |
| CN116775820A (en) | Question-answering processing method and question-answering model training method | |
| CN119513264A (en) | Intelligent question-answering method, device, equipment and medium based on job search scenario | |
| CN118069813A (en) | A method, device, terminal device and storage medium for processing dialogue interaction | |
| CN115129847A (en) | Intelligent answering method and device | |
| CN117633189A (en) | Question-answering session processing method, device, equipment and storage medium | |
| CN111858864B (en) | Method, device, electronic device and readable medium for implementing slot filling | |
| CN116932927B (en) | Data processing method and device for realizing social matching recommendation based on artificial intelligence | |
| CN112085577A (en) | On-line bidding question answering method, device, system and storage medium | |
| US20190163738A1 (en) | Development Factory | |
| CN118467700A (en) | Question-answering processing method, question-answering processing device, computing device, storage medium and program product | |
| CN117035091A (en) | Question and answer method, device, electronic equipment and storage medium | |
| CN116821463A (en) | Searching method, searching device, electronic equipment and storage medium | |
| CN117633188A (en) | Question-answering session processing method, device, equipment and storage medium | |
| CN118051782B (en) | Model training method, business processing method and related device | |
| CN111008373A (en) | Intelligent question and answer processing method and device, computer readable medium and electronic equipment | |
| CN118349630A (en) | Information processing method, system, device, storage medium and program product |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |