Greeting generation method and device, electronic equipment and storage medium

Abstract

The invention relates to a greeting generation method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining personal information of a first user, personal information of a second user and behavior information of the second user based on a greeting generation instruction of the first user; respectively extracting a first key information fragment of the first user and a second key information fragment of the second user from the personal information of the first user and the personal information of the second user; acquiring first preference information of the second user based on the behavior information of the second user; integrating the first key information fragment, the second key information fragment and the first preference information into an input corpus part in a prompt template, inputting the input corpus part into a large language model, and generating greetings of the first user to the second user through the large language model. According to the embodiment of the invention, personalized greetings suitable for both parties are generated for the user in a scene needing to be greetted.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet data processing, and in particular to a greeting generation method, device, electronic equipment and storage medium. Background Art

[0002] In many scenarios of work and life, in two-way communication scenarios through online social tools, the two parties usually send greetings to each other before entering the formal topic. In addition, for some one-way communication scenarios, such as sending emails, messages, etc., greetings are also entered first. Traditional greetings are usually simple "Hello", "Hi", "Good morning", "Dear", etc., which are used as the start of formal topics. In other scenarios, when communicating with another person through social tools, emails, etc. and want to get a response from the other party, the greeting sent needs to have certain characteristics to arouse the other party's interest. For example, some recruitment platforms provide chat tools for job seekers and recruiters to communicate. When a job seeker has determined a desired position, he can send a message to the recruiter of the position through the chat tool. Similarly, when the recruiter has determined a job seeker, he can also send a message to the job seeker through the chat tool. In this case, in order to get a response from the other party, the greeting sent for the first time will play a particularly important role. In addition, the recruitment platform also provides a resume delivery window. When job seekers deliver their resumes to a certain position, they can send a greeting at the same time. In order to provide better services to users, the recruitment platform provides some fixed greeting templates. When applying, users can check different greeting templates and combine them to generate a greeting.

[0003] However, this method has many shortcomings: first, lack of personalization and adaptability. Greetings generated by combining preset templates lack pertinence and personalization, cannot reflect the characteristics of individual job seekers, and are not adaptable to positions. Secondly, the content of the template is limited and the amount of information is insufficient. In addition, the form is single, and most of the current greetings are presented in text form, lacking support from other multimedia forms. Therefore, the greetings generated by the current method cannot effectively improve the success rate of job hunting or recruitment. Summary of the invention

[0004] In view of the technical problems existing in the prior art, the present invention proposes a greeting generation method, device, electronic device and storage medium, which are used to generate personalized greetings suitable for both parties for users in scenarios where greetings are required.

[0005] In order to solve the above technical problem, according to one aspect of the present invention, the present invention provides a greeting generation method, comprising the following steps:

[0006] Based on the greeting generation instruction of the first user, the personal information of the first user, the personal information of the second user, and the behavior information of the second user are obtained, wherein when the first user is a job seeker / recruiter, the personal information of the first user is the resume information of the job seeker / the job description information published by the recruiter, and when the second user is a recruiter / job seeker, the personal information of the second user is the job description information published by the recruiter / the resume information of the job seeker;

[0007] Extracting a first key information fragment of the first user from the personal information of the first user;

[0008] extracting a second key information fragment of the second user from the personal information of the second user;

[0009] acquiring first preference information of the second user based on the behavior information of the second user;

[0010] Integrating the first key information segment, the second key information segment and the first preference information into the input corpus portion of the prompt template; and

[0011] The prompt template is input into the large language model, and the large language model generates a greeting from the first user to the second user, wherein the content of the greeting includes at least one or more of the personal profile information of the first user, the advantage information of the first user, the matching information of the first user to the second user, and the matching information of the first user's preference to the second user.

[0012] Optionally, when extracting a first key information segment of the first user from the personal information of the first user, a first industry category is extracted from the personal information of the first user; and / or when extracting a second key information segment of the second user from the personal information of the second user, a second industry category is extracted from the personal information of the target second user; correspondingly, the first industry category and / or the second industry category is integrated into the task description part of the prompt template; correspondingly, the large language model generates a greeting from the first user to the second user suitable for the first industry and / or the second industry.

[0013] Optionally, the greeting generation method further comprises the following steps:

[0014] Obtaining behavior information of the first user; and

[0015] Acquire second preference information of the first user based on the behavior information of the first user;

[0016] Correspondingly, the first key information segment, the second key information segment, the first preference information and the second preference information are integrated into the input corpus part of the prompt template and input into the trained large language model, and the large language model generates a greeting from the first user to the second user, wherein the greeting also corresponds to the matching information of the second user's preference for the first user.

[0017] Optionally, when the first user is a job seeker, the behavior of the first user includes one or more of viewing, submitting and online chatting behaviors for positions, and the behavior information includes corresponding behavior categories and a position sequence recorded in chronological order of the behaviors; when the first user is a recruiter, the behavior of the first user includes one or more of viewing, downloading and online chatting behaviors for resumes, and the behavior information includes corresponding behavior categories and a resume identification sequence recorded in chronological order of the behaviors.

[0018] Optionally, the greeting generated by the large language model is output to the first user in the form of text, audio, or video / animation.

[0019] Optionally, when the greeting generated by the large language model is output to the first user, the output parameters set by the first user are queried, and the corresponding greeting is output based on the output parameters, wherein the output parameters include an output format and sub-parameters corresponding to the output format.

[0020] Optionally, the greeting generation method further comprises the following steps:

[0021] Constructing a data set for a greeting generation task, wherein each instance sample in the data set includes at least a task description and a corpus, wherein the corpus includes an input corpus and a corresponding output corpus;

[0022] Dividing the dataset into a training set, a test set, and an evaluation set;

[0023] Training a large language model based on the training set of the greeting generation task, testing the large language model based on the test set, evaluating the large language model based on the evaluation set, and ending the training of the large language model after the evaluation requirements are met;

[0024] The task description is used to describe the task of the large language model as generating a greeting from a first user to a second user based on an input corpus, or generating a greeting from a first user to a second user in a specific industry based on an input corpus, or generating a greeting with a specific emotional color from a first user to a second user based on an input corpus;

[0025] The input corpus correspondingly includes: personal key information of the first user, personal key information of the second user, and first preference information of the second user; or the input corpus includes personal key information of the first user, personal key information of the second user, first preference information of the second user, and second preference information of the first user;

[0026] The output corpus corresponds to: the specific content of the greeting from the first user to the second user; or the specific content of the greeting from the first user to the second user suitable for a specific industry; or the specific content of the greeting from the first user to the second user with a specific emotional color.

[0027] According to another aspect of the present invention, the present invention further provides a greeting generation device, including a data acquisition module, a personal information extraction module, a preference information acquisition module, a prompt template integration module and a large language model module, wherein the data acquisition module is configured to acquire the personal information of the first user, the personal information of the second user and the second user behavior information based on the greeting generation instruction of the first user, wherein when the first user is a job seeker / recruiter, the personal information of the first user is the resume information of the job seeker / the job description information published by the recruiter, and when the second user is a recruiter / job seeker, the personal information of the first user is the job description information published by the recruiter / the resume information of the job seeker; the personal information extraction module is configured to obtain the personal information of the first user, the personal information of the second user and the behavior information of the second user from the first user. A first key information segment of a first user is extracted from the personal information of a first user; a second key information segment of a second user is extracted from the personal information of a second user; the preference information acquisition module is configured to acquire the first preference information of the second user based on the behavior information of the second user; the prompt template integration module is configured to integrate the first key information segment, the second key information segment and the first preference information into the prompt template; the large language model module is configured to input the prompt template into the large language model, and the large language model generates a greeting from the first user to the second user, wherein the content of the greeting includes at least the personal information of the first user, the advantage information of the first user, the matching information of the first user to the second user and the matching information method preferred by the second user.

[0028] According to another aspect of the present invention, the present invention also provides an electronic device, which includes a processor and a memory, wherein a computer program instruction set is stored in the memory, and when the processor executes the computer program instruction set in the memory, the aforementioned greeting generation method is implemented.

[0029] According to another aspect of the present invention, the present invention further provides a computer-readable storage medium, characterized in that a computer program instruction set is stored on the computer-readable storage medium, and the computer program instruction set implements the aforementioned greeting generation method when executed by a processor.

[0030] According to another aspect of the present invention, the present invention further provides a computer program product, characterized in that it includes a computer program instruction set, and when the computer program instruction set is executed by a processor, the aforementioned greeting generation method is implemented.

[0031] The greeting message generated for the user in the embodiment of the present invention in the scenario where greeting is required is highly personalized, contains a large amount of information, is suitable for the specific circumstances of both parties, and can effectively improve the success rate of the other party's response. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The preferred embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, wherein:

[0033] Figure 1 is a flow chart of a method for generating a greeting according to an embodiment of the present invention;

[0034] Figure 2 is a principle block diagram of a greeting generation device according to an embodiment of the present invention;

[0035] Figure 3 is a principle block diagram of a recruitment platform according to an embodiment of the present invention; and

[0036] Figure 4 It is a schematic diagram of the hardware structure principle of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] In the following detailed description, reference may be made to the various specification drawings that are part of the present application and are used to illustrate specific embodiments of the present application. In the accompanying drawings, similar reference numerals describe substantially similar components in different figures. The various specific embodiments of the present application are described below in sufficient detail so that a person of ordinary skill in the art with relevant knowledge and skills in the art can implement the technical solutions of the present application. It should be understood that other embodiments may also be used or structural, logical or electrical changes may be made to the embodiments of the present application.

[0039] The present invention provides a greeting generation method, device, electronic device and storage medium for scenarios requiring greetings, which can generate personalized and targeted greetings for users who actively initiate communication requests based on the actual conditions of both users. The scenario requiring greetings is, for example, a scenario in which a first-end user initiates a chat conversation with a second-end user through a chat tool provided by a recruitment platform, where the first-end user and the second-end user are job seekers and recruiters, respectively. Another scenario requiring greetings is, for example, a scenario in which a job seeker submits a resume to a certain position through a resume submission window provided by a recruitment platform, and so on.

[0040] See also Figure 1 , Figure 1 1 is a flow chart of a method for generating a greeting according to an embodiment of the present invention, the method comprising the following steps:

[0041] Step S11, based on the greeting generation instruction of the first user, obtain the personal information of the first user, the personal information of the second user and the behavior information of the second user.

[0042] Step S12: extracting a first key information segment of the first user from the personal information of the first user.

[0043] Step S13: extracting a second key information segment of the second user from the personal information of the second user.

[0044] Step S14: acquiring first preference information of the second user based on the behavior information of the second user.

[0045] Step S15: Integrate the first key information segment, the second key information segment and the first preference information into the input corpus part of the prompt template.

[0046] Step S16, inputting the prompt template into the large language model, and generating a greeting from the first user to the second user through the large language model, wherein the content of the greeting includes at least the personal information of the first user, the matching information of the first user to the target position, and the matching information to the second user's preference.

[0047] Step S17: output the greeting to the first user.

[0048] The first user and the second user in the present invention are respectively a job seeker and a recruiter. When the first user is a job seeker and the second user is a recruiter, the personal information of the first user is the resume information of the job seeker, and the personal information of the second user is the job description information posted by the recruiter; when the first user is a recruiter and the second user is a job seeker, the personal information of the first user is the job description information posted by the recruiter, and the personal information of the second user is the resume information of the job seeker.

[0049] The recruitment platform provides a greeting generation instruction operation tool, such as a button, in the interface corresponding to the various scenarios in which greetings need to be generated. The first user clicks the button when he wants to generate a greeting for him. When the button information is received, it is determined that the greeting generation instruction of the first user is received. At this time, in step S11, after receiving the greeting generation instruction of the first user, the first user identity identifier and the corresponding second user identity identifier are parsed therefrom. Then, based on the first user identity identifier, the personal information of the second user, the personal information of the second user, and the behavior information of the second user are obtained from the database.

[0050] In step S12 and step S13, the first key information fragment of the first user is extracted from the personal information of the first user, and the second key information fragment of the second user is extracted from the personal information of the second user. The key information may include one or more types of content. When there are multiple types of content, multiple types of content may be included in one key information fragment in the form of structured data, or each key information fragment may correspond to one type of content and be arranged in a preset order. The key information extracted from the resume information includes, but is not limited to, types such as "educational background", "work experience", and "skills and specialties", and the key information extracted from the job description information includes, but is not limited to, types such as "description of duties", "education requirements", "experience requirements", and "salary and benefits".

[0051] When multiple types of content are included in a key information fragment in the form of structured data, these key information are combined in the form of fields to form a key information fragment. For example, a field is represented by square brackets, and the brackets are separated by semicolons. The field is separated into two parts by colons. The first part is the field name, and the second part is the field content, that is, the extracted key information. A key information fragment data extracted from resume information is, for example, "[Educational background: undergraduate; computer major]; [Work experience: 1 year internship at XX company; 3 years of front-end R&D at YY company]; [Skills and expertise: front-end development]". A key information fragment data extracted from job description information is, for example, "[Description of responsibilities: front-end development]; [Education requirements: undergraduate]; [Experience requirements: 2 years of development experience]", etc.

[0052] In order to better understand and serve users, the recruitment platform records the user's behavior on the recruitment platform and the corresponding information involved, and stores them in the database. Each record data in the database includes the user ID, the time when the user behavior occurs, the behavior category, the peer user ID corresponding to the behavior, and the behavior data. For job seekers, the job search behavior category includes one or more of the viewing behavior, delivery behavior, and online chat behavior of positions. For recruitment users, the behavior category includes one or more of the viewing behavior, downloading behavior, and online chat behavior of resumes. Each behavior category corresponds to a kind of behavior data. For job seekers, the behavior data is, for example, the job title or job ID. For recruitment users, the behavior data is, for example, the resume ID or the name of the job seeker corresponding to the resume. Therefore, all the behavior record data of the user within a period of time can be queried in the database through the user ID, and the behavior category, occurrence time, and corresponding behavior data can be extracted from each behavior record data. Thus, a position sequence recorded in the order of the time behavior occurrence corresponding to the behavior category or a resume sequence recorded in the order of the time behavior occurrence corresponding to the behavior category is obtained. When the first user is a job seeker and the second user is a recruiter, their behavior information is a sequence of positions recorded in chronological order according to the corresponding behavior categories. When the first user is a recruiter and the second user is a job seeker, their behavior information is a sequence of resume identifications recorded in chronological order according to the corresponding behavior categories.

[0053] In step S14, the preferences of the second user are obtained according to the behavior information of the second user in step S13, which are referred to as first preference information for the sake of distinction. In one embodiment, the behavior information of the second user is input into a trained model, and the first preference information of the second user is output through the model. When the second user is a recruiting user, the first preference information may be, for example, the specific education level of the job seeker, the specific years of work experience of the job seeker, the location of the job seeker, etc. When the second user is a job seeker, the first preference information may be, for example, the specific work location, the specific job level, the specific salary, etc.

[0054] The present invention adopts a large language model (LLM) to generate greetings. Specifically, in step S15, the first key information segment, the second key information segment and the first preference information obtained above are integrated into the input corpus part of the prompt template, and in step S16, the prompt template is input to the LLM, and the greeting is output after being processed by the LLM.

[0055] The LLM is a large language model that has been fine-tuned and trained, such as the Generative Pre-Trained Transformer (GPT) series models developed by Open AI, LaMDA and PaLM developed by Google, LLaMA developed by Meta, Wenxin Yiyan developed by Baidu, Tongyi Qianwen of Alibaba, Hunyuan of Tencent, and the large language model named "AliceMind" developed by Alibaba Damo Academy, etc. Developers in this field can apply any LLM according to actual conditions.

[0056] The fine-tuning and training process of LLM is described as follows:

[0057] Collect data. The collected data includes resume data, position data, behavioral data of job seekers, behavioral data of recruiters, and greetings used when the two types of users communicate in various scenarios. Resume data includes, for example, basic information (such as name, education, work experience, etc.), skills, expertise, project experience, etc. Position data includes, for example, position title, position description, position requirements, salary and benefits, etc. The behavioral data of job seekers includes a position sequence corresponding to the behavior category and recorded in the order of the behavior occurrence time; the behavioral data of recruiters includes a resume sequence corresponding to the behavior category and recorded in the order of the behavior occurrence time.

[0058] Data processing. Specifically includes: extracting key information fragments from resume data; extracting key information fragments from job data; determining the preference information of the recruiting user based on the behavioral data of the recruiting user, such as education, years of work, professional skills, etc.; determining the preference information of the job-seeking user based on the behavioral data of the job-seeking user, such as industry, region, job type, salary, etc. Wherein, the preference information can be generated by a model. The collected greetings are sorted and improved based on the key information fragments extracted from the resume data and the key information fragments extracted from the job data, so that the greetings include the personal information of the first user, the advantage information of the first user, the matching information with the second user, and the matching information of the second user's preference. For example, when the first user is a job-seeking user and the second user is a recruiting user, the greetings include the name of the job-seeking user, his own advantages (such as rich experience or relevant certificates that can prove his ability), closely related information to the recruitment position (such as information that matches the recruitment position, such as position, work experience), and information matching the preferences of the recruiting user (such as education, projects participated in, etc.).

[0059] Construct a data set for a greeting generation task. Each instance sample in the data set includes at least a task description and a corpus, wherein the corpus includes an input corpus and a corresponding output corpus. The task description is used to express that the task of the LLM is to generate a greeting from a first user to a second user based on an input corpus. The input corpus includes a first key information segment extracted from the personal information of the first user, a second key information segment extracted from the personal information of the second user, and the preference information of the second user. When the LLM is used to generate a greeting from a job seeker to a recruiter, the first user is a job seeker, the second user is a recruiter, the first key information segment is a key information segment extracted from the resume data of the job seeker, and the second key information segment is a key information segment extracted from the position data.

[0060] The data set is divided into a training set, a test set and an evaluation set. The large language model is trained based on the training set of the greeting generation task, that is, the instance samples in the data set are input into the large language model, and the LLM learns the corpus according to the corresponding task described in the task description. After a certain amount of sample learning, the large language model is tested with the test set, and the large language model is evaluated with the evaluation set. After meeting the evaluation requirements, the training of the large language model is terminated.

[0061] In one embodiment, the LLM can be deployed on a private server to protect the privacy of the data, or the LLM interface on the open source platform can be called to send the above data set to the open source platform for fine-tuning and training of the LLM. The specific application method of LLM can be selected according to the actual scenario.

[0062] In the present invention, the generated greeting includes at least the first user's personal profile information, advantage information, the first user's matching information for the target position, and the matching information for the second user's preference. Not only personalization is achieved, but the greeting also includes information matching the second user, thereby effectively increasing the attractiveness to the second user, thereby achieving the purpose of successfully establishing a dialogue.

[0063] In another embodiment, see Figure 1 ,exist Figure 1On the basis of, when acquiring data in step S11, the behavior information of the first user is also acquired. After acquiring the first preference information of the second user in step S14, the second preference information of the first user is also acquired according to the behavior information of the first user, and in step S15, the second preference information is also added to the input corpus, so that the greeting generated by LLM can include the preference matching information of the second user and the first user. For example, when the job seeker is the first user, and his preferred work location is XX city, and the job location of the recruitment user's recruitment position is also in XX city, then "the work location of this position in your company is exactly the place I like" can be added to the greeting, so that more matching information can be displayed to the recruitment user, and the purpose of two-way suitability can be expressed to the recruitment user, thereby improving the success rate of establishing a dialogue.

[0064] Furthermore, in another embodiment, in order to improve the consistency of the greeting expression generated by the large language model with the industry, in another embodiment, when extracting the first key information segment of the first user from the personal information of the first user, the first industry category is extracted from the personal information of the first user; and / or when extracting the second key information segment of the second user from the personal information of the second user, the second industry category is extracted from the personal information of the target second user; correspondingly, the first industry category and / or the second industry category is integrated into the task description part of the prompt template; therefore, when the large language model generates a greeting according to the prompt template, it can generate a greeting from the first user to the second user suitable for the first industry and / or the second industry. For example, in the Internet industry, examples of generated greetings include "Hello! I am a job seeker who is passionate about the Internet industry. I am proficient in technology, good at Java development, and have a solid programming foundation and rich project experience. I am fascinated by the rapid iteration and innovative culture of the Internet industry, and I am eager to continue to learn and grow in this dynamic and creative field. I believe that my skills and enthusiasm can bring new vitality and innovative ideas to your company, and I look forward to the opportunity to have in-depth exchanges with you!" In the financial industry, examples of generated greetings include "Hello! I am a job seeker who is passionate about the financial industry, with solid financial knowledge and rich practical experience. I am good at data analysis, have a keen insight into market trends, and am familiar with the design and risk management of financial products. I believe that my professional skills and meticulous work attitude can bring value to your company, and I look forward to the opportunity to discuss opportunities and challenges in the financial field with you!".

[0065] Similarly, when fine-tuning and training the large language model, the collected data is classified by industry category to obtain training data sets for different industries. The industry category is specified in the task description part of each instance sample, and greetings suitable for the industry category are used in the output corpus, i.e., greetings.

[0066] Therefore, when the greeting generation instruction of the first user is obtained during the application process, the industry category is extracted and provided to the LLM, so that the LLM generates a greeting suitable for the industry category.

[0067] In another embodiment, in order to make the greetings generated by LLM have certain emotional colors, such as respect, positivity, enthusiasm, calmness, etc., when fine-tuning and training the large language model, the corresponding specific emotional colors are specified in the task description part, and at the same time, the greetings as output corpus are expressed in a way that conforms to the corresponding emotional colors. Therefore, when generating greetings, LLM performs sentiment analysis on the input corpus, so that the generated greetings are closer to human emotions and conform to specific emotional colors, thereby increasing the affinity of communication. For example, a serious greeting: "Hello! I am a professional Java development engineer with a solid programming foundation and rich project experience. I am proficient in Java and Spring framework, and am good at system design and optimization. I have significantly improved system performance during my internship. I love programming and focus on teamwork. I look forward to the opportunity to contribute my professional skills to your company and jointly promote the efficient development of the project." For example, a relaxed and pleasant greeting: "Hey, hello! I am a Java development engineer who loves programming, and I am also a PPT expert (I have won the first prize of KOS)! When I was interning at Hunan Fenghui Yinjia Technology, I not only played with Java and Spring framework, but also successfully increased the system access speed by 10%. I like to solve technical problems and I also love to grow with the team. I look forward to talking to you and see if we can create more possibilities together!".

[0068] In step S17, the greeting generated by the large language model is output to the first user in the form of text, audio or video. In order to better serve the user, output parameters are further provided, and the output parameters include output format and sub-parameters corresponding to the output format. For example, the output format can be text format, audio format or video format. Sub-parameters corresponding to the text format are, for example, font, font size, language, etc. Sub-parameters corresponding to the audio format are, for example, timbre, speech speed, volume, language, etc. of different male or female voices. Sub-parameters corresponding to the video format include two categories, one is video parameters related to the video, and the other is audio parameters. Among them, the video parameters that can be set by the user are, for example, characters in the video. In one embodiment, virtual characters are provided to users as characters in the video / animation, and virtual characters with different images are provided for users to choose from.

[0069] In one embodiment, one of the output formats and the corresponding sub-parameters are used as the default output parameters, and each output format has a corresponding default sub-parameter. For example, a Chinese text expression with a certain font and font size is used as the default output parameter. When outputting a greeting to the first user, first check whether the user has set the output parameter. If the user has set the output parameter, the parameter set by the user is used for output. If not, the default output parameter is used. Wherein, when outputting in text mode, in one embodiment, the greeting is displayed in a text window. When outputting in audio mode, the text of the greeting is first converted into speech and output through the playback device of the user device. When outputting in video / animation mode, if the user has set a virtual character, the image of the virtual character set by the user is used as the character in the video, and the corresponding audio parameters are used to convert the text into speech output. Wherein, when outputting speech, a specific algorithm can be used to calculate the lip shape and expression of the virtual character that matches the speech content, so that the character in the video is more realistic. Through the output parameters provided above, the user's choices and expressions are increased, making the greeting more vivid and diverse.

[0070] In addition, when the greeting is output to the first user, a corresponding text window is provided in the interface for interaction with the user, and the text window is used to record the text data of the greeting in various output formats, so as to facilitate the first user to modify the greeting output by the large language model. In addition, further, the interface also provides operation buttons for saving, text to audio, text to animation / video, and recording. The save operation button is used to store the greeting output by the current large language model. The text to audio operation button is used to convert the current text format greeting into audio, and the text to animation / video operation button is used to convert the current text format greeting into animation / video. Further, when the user clicks the text to audio operation button or the text to animation / video operation button, a corresponding parameter configuration window is also provided based on the operation, so that the user can configure audio parameters such as timbre, volume, language, etc., and a variety of virtual characters are provided in the video parameters for the user to choose. After the first user has configured the parameters, the text can be converted into audio or video / animation according to the configured parameters. When the user determines to send an audio greeting to the second user, the user can also choose to use his or her original voice. The user can click the recording operation button to read the greeting in the text window, thereby obtaining the greeting in the audio format of the user's original voice.

[0071] Therefore, when the first user receives the greeting, he or she can modify it and choose to send the greeting to the second user in the form of text, audio, or video, which is diverse and novel.

[0072] When the first user is a job seeker and the second user is a recruiter, the job seeker not only saves the time and energy of writing greetings, but also the generated personalized greetings can better show the advantages and adaptability of the job seeker, thereby increasing the favor and interest of the recruiter in the job seeker, and further increasing the success rate of the job seeker submitting resumes, establishing chat sessions, etc. For the recruiter, the greetings generated by the method of the present invention can enable the recruiter to understand the profile of the job seeker and the matching of the job with the posted position, thereby improving the efficiency of the recruiter in screening resumes, thereby improving the user experience and efficiency of the recruitment process as a whole.

[0073] Similarly, when the first user is a recruiter and the second user is a job seeker, the recruiter can save time and energy in writing greetings when contacting job seekers, and can highlight the matching and adaptability of the job position to job seekers, thereby attracting the job seekers' attention. When job seekers receive greetings for communication requests, they can understand the key information of the job position through the greetings, thus reducing the blindness of communication.

[0074] In another aspect, the present invention also provides a greeting generating device, see Figure 2 , Figure 2 FIG. 1 is a block diagram of a greeting generation device according to an embodiment of the present invention. Figure 3 , Figure 3 A schematic diagram of a recruitment platform according to an embodiment of the present invention. The greeting generation device 100 in this embodiment includes a data acquisition module 101, a personal information extraction module 102, a preference information acquisition module 103, a prompt template integration module 104 and a large language model module 105. The greeting generation device 100 is located in the recruitment platform 10 as a functional module of the recruitment platform 10. The recruitment platform 10 provides an interactive interface 200 and a database 300. Figure 2 and Figure 3When the first user needs to send a greeting to the second user through the interactive interface 200 in some scenarios, a greeting generation button is provided in the interactive interface. When the user clicks the button, a greeting generation instruction is issued and sent to the data acquisition module 101 in the greeting generation device 100. After receiving the greeting generation instruction of the first user, the data acquisition module 101 parses it to obtain the first user identifier and the second user identifier, and then obtains personal information and behavior information from the database 300, and sends the personal information to the personal information extraction module 102, and sends the behavior information to the preference information acquisition module 103. In one embodiment, the personal information of the first user, the personal information of the second user, and the behavior information of the second user are obtained. In another embodiment, the personal information of the first user, the personal information of the second user, the behavior information of the first user, and the behavior information of the second user are obtained. Wherein, when the first user is a job seeker / recruiter, the personal information of the first user is the resume information of the job seeker / the job description information published by the recruiter, and the second user is a recruiter / job seeker, and the personal information of the first user is the job description information published by the recruiter / the resume information of the job seeker.

[0075] After receiving the personal information, the personal information extraction module 102 extracts the first key information segment of the first user from the personal information of the first user; extracts the second key information segment of the second user from the personal information of the second user, and sends the first key information segment and the second key information segment to the prompt template integration module 104.

[0076] After receiving the behavior information, the preference information acquisition module 103 acquires the first preference information of the second user based on the behavior information of the second user, and sends it to the prompt template integration module 104. If the behavior information of the first user is also received, the second preference information of the first user is acquired from it and sent to the prompt template integration module 104.

[0077] The prompt template integration module 104 stores a prompt template (prompt) for the large language model, which includes a task description part and an input corpus part, wherein the task description part is used to express that the task of the large language model is to generate a greeting from the first user to the second user based on the input corpus. The corpus part includes a first user corpus part and a second user corpus part, and the first key information fragment received is added to the first user corpus part, and the second key information fragment and the first preference information are added to the second user corpus part, so as to obtain a complete prompt template. Then the prompt template is sent to the large language model module 105.

[0078] The large language model module 105 inputs the prompt template to the large language model, and the large language model generates a greeting from the first user to the second user. The large language model can be a large language model that has been pre-trained, fine-tuned, and task-trained and deployed to the recruitment platform 10, or a large language model that has been fine-tuned and task-trained and is located on an open source platform.

[0079] The content of the greeting includes at least the personal information of the first user, the advantage information of the first user, the matching information of the first user to the second user, and the matching information of the first user's preference to the second user.

[0080] Furthermore, when extracting the personal key information fragments, the personal information extraction module 102 also extracts the corresponding industry, for example, determines the corresponding industry according to the job description information, determines the corresponding industry according to the resume information, and sends the industry information to the prompt template integration module 104. The prompt template integration module 104 adds the industry information to the task description part in the prompt template. The industry information extracted from the first user's personal information is added to the first user input corpus part, and the industry information extracted from the second user's personal information is added to the second user input corpus part. Correspondingly, the large language model module 105 generates a greeting suitable for the industry from the first user to the second user.

[0081] For another example, when a specific emotional tone is specified in the task description part of the prompt template, LLM performs sentiment analysis on the input corpus when generating greetings, so that the generated greetings are closer to human emotions and conform to the specific emotional tone.

[0082] In the following embodiment, a job seeker generates a greeting when submitting a resume to a position recruited by a recruiter.

[0083] Among them, the resume information of the job seeker is as follows:

[0084]

Self-introduction

[0085] 1. Be able to read and communicate in English proficiently;

[0086] 2. Be familiar with the use of office software and obtain the first prize in the PPT track of Kingsoft Office Skills Certification KOS;

[0087] 3. Have a certain understanding of the software development life cycle, including requirements analysis, design, development, testing and deployment.

[0088]

Work Experience

[0089] 2023-09-01Internship as a Java development engineer at YY Company:

[0090] 1. According to product requirements, sort out the core functions of the product, complete the development and implementation of interface functions, and ensure 100% delivery according to plan;

[0091] 2. Participated in the design and development of multiple backend modules, using Java language, spring framework and corresponding middleware technologies to ensure the scalability and stability of the system. The local optimization of database design and SQL statements increased the access speed by 10%.

[0092] The job description information for recruiting users is as follows:

[0093]

Job Content

[0094] Participate in the design, development, testing and maintenance of the company's software products;

[0095] Analyze and solve technical problems in the software development process and write high-quality code;

[0096] Assist team members to complete project development tasks and improve the overall technical level of the team;

[0097] Cooperate with product managers to conduct demand analysis and functional design to ensure the realization of product functions.

[0098]

Job Requirements

[0099] Bachelor degree or above in computer related major;

[0100] Proficient in Java programming language and have a solid foundation in data structure and algorithm;

[0101] Have good coding habits and document writing skills, and be familiar with coding standards;

[0102] Love programming, have strong learning and problem-solving abilities;

[0103] Have good communication skills and teamwork spirit, and be able to complete work tasks independently;

[0104] Fresh graduates and experienced candidates are welcome to apply, and those with relevant industry experience will be given priority;

[0105] Deeply understand business and system modeling, flexibly use design patterns to solve defects in projects, and have good system debugging skills and corresponding problem troubleshooting capabilities through communication with multiple teams;

[0106] Optimize workflows, reduce duplication of technology, build modular, service-oriented, and platform-based systems, reduce dependencies between systems, improve system stability and scalability, and assist the team in improving development efficiency.

[0107] The preference information of the recruiting user obtained through the behavioral information of the recruiting user is "work experience".

[0108] The greeting generated according to the method of the present invention is as follows:

[0109] Dear HR, I am XX. I am very honored to have the opportunity to apply for the position of Java development engineer in your company. During my internship experience at YY Company, I not only mastered the core technologies such as Java development, Spring framework and database optimization, but also successfully improved the system performance by 10%. I am good at demand analysis, functional design and implementation, and team collaboration. I can quickly integrate into the team and contribute value. My English communication skills and office software skills will also add convenience to your international projects or document management. I look forward to displaying my talents in your company and jointly promoting technological innovation and development.

[0110] Among them, "I am XX" is personal information, "I am good at...adding convenience" is the job seeker's advantage information and matching information for the recruiter, and "In YY company...system performance 10%" is the matching information of the job seeker's preference for the recruiter.

[0111] The present invention utilizes a large language model and comprehensively considers multiple dimensions of information such as the job seeker's resume, target position description information, and the recruiter's historical viewing and downloading of resumes. The generated greeting is in line with the job seeker's background, position requirements, and recruiter preferences. It is more personalized and more closely related to the target position. Moreover, after fully considering the preferences of both parties, the greeting is more in line with the expectations and needs of both parties, which can effectively attract the attention of the HR of the target position. In the chat scenario, it can improve the success rate of establishing a chat dialogue with the HR. In the resume delivery scenario, it can improve the probability of successful delivery.

[0112] Figure 4 1 is a schematic diagram of the hardware structure principle of an electronic device according to an embodiment of the present invention. The electronic device can be implemented as a server or other various terminal devices, such as a desktop personal computer, a tablet computer, a laptop computer, a mobile phone, etc., including a processor 601 and a memory 602. The memory 602 stores a program instruction set. When the processor 601 executes the program instruction set on the memory 602, the various greeting generation methods mentioned above are implemented.

[0113] Specifically, the processor 601 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiment of the present invention.

[0114] The memory 602 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 602 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 602 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 602 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 602 is a non-volatile solid-state memory.

[0115] The memory may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Therefore, generally, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer executable instructions, and when the software is executed (e.g., by one or more processors), it can be operated to perform the greeting generation method provided by the present invention.

[0116] In one example, the electronic device may further include a communication interface 603 and a bus 604. The processor 601, the memory 602, and the communication interface 603 are connected via the bus 604 and communicate with each other.

[0117] The communication interface 603 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiment of the present invention.

[0118] Bus 604 includes hardware, software or both, and the components of online data flow billing equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 604 may include one or more buses. Although the embodiment of the present invention describes and shows a specific bus, the present invention considers any suitable bus or interconnection.

[0119] The present invention also provides a computer-readable storage medium having computer program instructions stored thereon, and the computer program instructions can be executed by a processor to implement any of the greeting generation methods in the aforementioned embodiments. The computer-readable storage medium can be any tangible medium that contains or stores computer-executable instructions for use by or in combination with an instruction execution system, device, and apparatus. The storage medium can be a transient computer-readable storage medium or a non-transient computer-readable storage medium. Non-transient computer-readable storage media may include, but are not limited to, magnetic storage devices, optical storage devices, and / or semiconductor storage devices. Examples of such storage devices include, for example, magnetic disks, optical disks based on CD, DVD, or Blu-ray technology, and persistent solid-state memories such as flash memory, solid-state drives, and the like.

[0120] The present invention also provides a computer program product, which includes a set of computer program instructions, and when the computer program instruction set is executed by a processor, any one of the greeting generation methods in the above embodiments is implemented. The computer program product includes but is not limited to application installation packages published on websites and application stores, application plug-ins, and small programs that can be run in certain applications.

[0121] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0122] The above embodiments are only used to illustrate the present invention, but not to limit the present invention. Ordinary technicians in the relevant technical field can make various changes and modifications without departing from the scope of the present invention. Therefore, all equivalent technical solutions should also fall within the scope of the present invention.

Claims

1. A greeting generation method, characterized in that: include: Based on the greeting generation instruction of the first user, the personal information of the first user, the personal information of the second user, and the behavior information of the second user are obtained, wherein when the first user is a job seeker / recruiter, the personal information of the first user is the resume information of the job seeker / the job description information published by the recruiter, and when the second user is a recruiter / job seeker, the personal information of the second user is the job description information published by the recruiter / the resume information of the job seeker; Extracting a first key information fragment of the first user from the personal information of the first user; extracting a second key information fragment of the second user from the personal information of the second user; acquiring first preference information of the second user based on the behavior information of the second user; Integrating the first key information segment, the second key information segment and the first preference information into the input corpus portion of the prompt template; and The prompt template is input into the large language model, and the large language model generates a greeting from the first user to the second user, wherein the content of the greeting includes at least one or more of the personal profile information of the first user, the advantage information of the first user, the matching information of the first user to the second user, and the matching information of the first user's preference to the second user.

2. The greeting generation method according to claim 1, characterized in that: When extracting a first key information segment of the first user from the personal information of the first user, a first industry category is extracted from the personal information of the first user; and / or when extracting a second key information segment of the second user from the personal information of the second user, a second industry category is extracted from the personal information of the target second user; correspondingly, the first industry category and / or the second industry category is integrated into the task description part of the prompt template; correspondingly, the large language model generates a greeting from the first user to the second user suitable for the first industry and / or the second industry.

3. The greeting generation method according to claim 1, characterized in that: Further including: Obtaining behavior information of the first user; and Acquire second preference information of the first user based on the behavior information of the first user; Correspondingly, the first key information segment, the second key information segment, the first preference information and the second preference information are integrated into the input corpus part of the prompt template and input into the trained large language model, and the large language model generates a greeting from the first user to the second user, wherein the greeting also corresponds to the matching information of the second user's preference for the first user.

4. The greeting generation method according to claim 3, characterized in that: When the first user is a job seeker, the behavior of the first user includes one or more of viewing, submitting, and online chatting on positions, and the behavior information includes corresponding behavior categories and a sequence of positions recorded in chronological order of the behaviors; when the first user is a recruiter, the behavior of the first user includes one or more of viewing, downloading, and online chatting on resumes, and the behavior information includes corresponding behavior categories and a sequence of resume identification recorded in chronological order of the behaviors.

5. The greeting generation method according to claim 1, characterized in that: The greeting generated by the large language model is output to the first user in the form of text, audio, or video / animation.

6. The greeting generation method according to claim 5, characterized in that: When the greeting generated by the large language model is output to the first user, the output parameters set by the first user are queried, and the corresponding greeting is output based on the output parameters, wherein the output parameters include an output format and sub-parameters corresponding to the output format.

7. The greeting generation method according to claim 1, characterized in that: Further including: Constructing a data set for a greeting generation task, wherein each instance sample in the data set includes at least a task description and a corpus, wherein the corpus includes an input corpus and a corresponding output corpus; Dividing the data set into a training set, a test set, and an evaluation set; Training a large language model based on the training set of the greeting generation task, testing the large language model based on the test set, evaluating the large language model based on the evaluation set, and ending the training of the large language model after the evaluation requirements are met; The task description is used to describe the task of the large language model as generating a greeting from a first user to a second user based on an input corpus, or generating a greeting from a first user to a second user in a specific industry based on an input corpus, or generating a greeting with a specific emotional color from a first user to a second user based on an input corpus; The input corpus correspondingly includes: personal key information of the first user, personal key information of the second user, and first preference information of the second user; or the input corpus includes personal key information of the first user, personal key information of the second user, first preference information of the second user, and second preference information of the first user; The output corpus corresponds to: the specific content of the greeting from the first user to the second user; or the specific content of the greeting from the first user to the second user suitable for a specific industry; or the specific content of the greeting from the first user to the second user with a specific emotional color.

8. A greeting generating device, characterized in that: include: a data acquisition module configured to generate instructions based on the greeting of the first user, and acquire personal information of the first user, personal information of the second user, and behavior information of the second user, wherein when the first user is a job seeker / recruiter, the personal information of the first user is the resume information of the job seeker / the job description information published by the recruiter, and when the second user is a recruiter / job seeker, the personal information of the first user is the job description information published by the recruiter / the resume information of the job seeker; A personal information extraction module is configured to extract a first key information segment of the first user from the personal information of the first user; and to extract a second key information segment of the second user from the personal information of the second user; a preference information acquisition module, configured to acquire first preference information of a second user based on behavior information of the second user; a prompt template integration module configured to integrate the first key information segment, the second key information segment and the first preference information into a prompt template; and The large language model module is configured to input the prompt template into the large language model, and the large language model generates a greeting from the first user to the second user, wherein the content of the greeting includes at least the personal information of the first user, the advantage information of the first user, the matching information of the first user to the second user, and the matching information method preferred by the second user.

9. An electronic device comprising a processor and a memory, characterized in that: The memory stores a computer program instruction set, and when the processor executes the computer program instruction set on the memory, the greeting generation method described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program instruction set, and when the computer program instruction set is executed by a processor, the greeting generation method according to any one of claims 1 to 7 is implemented.

11. A computer program product, characterized in that The method comprises a computer program instruction set, and when the computer program instruction set is executed by a processor, the method for generating a greeting according to any one of claims 1 to 7 is implemented.

Images