Service post recommendation method and device, storage medium and program product

Abstract

The embodiment of the invention provides a service post recommendation method and device, a storage medium and a program product. In the embodiment of the invention, dialogue information and a model cue word are input into an AI question and answer service model based on artificial intelligence, and service demand information of a user is identified from the dialogue information; under the condition that the service demand information is identified, recalling a plurality of candidate service posts for the user according to the service demand information; and selecting a target service post from the plurality of candidate service posts and recommending the target service post to the user. The AI question and answer service model based on artificial intelligence can play a role of an artificial customer service to dialog with the user, and can also identify the service demand information of the user from the dialogue information and accurately recommend the service post meeting the user demand to the user according to the service demand information. And the service posts meeting the requirements of the user are recommended to the user in the dialogue process of the dialogue page, so that the conversion rate of the service posts can be improved.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to a method, device, storage medium and program product for recommending service posts. Background Art

[0002] In the field of e-commerce, service posts are a form of text used to display merchant services on e-commerce platforms. They are similar to advertisements and convey the unique service content and advantages provided by merchants through carefully designed titles and content. These service posts can not only promote the sales of services or goods, but also enhance the brand awareness of merchants, increase user engagement, attract user attention and convey the service content of merchants. It can be seen that posts play an important role in the field of e-commerce.

[0003] However, how to accurately recommend service posts that meet user needs to users and make the posts effective is a technical problem that needs to be solved urgently. Summary of the invention

[0004] Multiple aspects of the present application provide a service post recommendation method, device, storage medium and program product for accurately recommending service posts that meet user needs to users.

[0005] An embodiment of the present application provides a method for recommending service posts, including: displaying a conversation interface, and conducting at least one round of conversation with a user based on the conversation interface, wherein at least the user's conversation information is displayed on the conversation interface; inputting the conversation information and model prompt words into an AI question-and-answer service model based on artificial intelligence, wherein the model prompt words include model role prompt words, which are used to prompt the AI ​​question-and-answer service model to at least provide a post recommendation service for the user, wherein the AI ​​question-and-answer service model is obtained by fine-tuning a pre-trained language model using sample data in a target field to which the service posts belong; identifying the user's service demand information from the conversation information under the prompt of the model role prompt words; in the case where the service demand information is identified, recalling multiple candidate service posts for the user according to the service demand information; selecting a target service post from the multiple candidate service posts and recommending the target service post to the user.

[0006] The application embodiment also provides an electronic device, including: a memory and a processor; the memory is used to store a computer program; the processor is coupled to the memory and is used to execute the computer program to implement the steps in the above method.

[0007] An embodiment of the present application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor implements the steps in the above method.

[0008] The embodiment of the present application also provides a computer program product, which includes a computer program / instruction. When the computer program / instruction is executed by a processor, the processor is enabled to implement the steps in the above-mentioned digital human live broadcast method.

[0009] In an embodiment of the present application, the conversation information and model prompt words are input into an AI question-and-answer service model based on artificial intelligence, and the user's service demand information is identified from the conversation information; when the service demand information is identified, multiple candidate service posts are recalled for the user based on the service demand information; a target service post is selected from multiple candidate service posts and the target service post is recommended to the user. The AI ​​question-and-answer service model based on artificial intelligence can not only play the role of a human customer service to talk to the user, but also can identify the user's service demand information from the conversation information and accurately recommend service posts that meet the user's needs to the user based on the service demand information, and recommend service posts that meet the user's needs to the user during the conversation process on the conversation page, which can improve the conversion rate of service posts. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0011] Figure 1a A flowchart of a method for recommending service posts provided by an exemplary embodiment of the present application;

[0012] Figure 1b A flowchart of another method for recommending service posts provided for another exemplary embodiment of the present application;

[0013] Figure 1c A schematic diagram of a business district provided for an exemplary embodiment of the present application;

[0014] Figure 2 A schematic structural diagram of an electronic device provided as an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0016] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0017] In response to the existing technical problem of not being able to accurately recommend service posts that meet user needs to users, in an embodiment of the present application, the conversation information and model prompt words are input into an AI question-and-answer service model based on artificial intelligence to identify the user's service demand information from the conversation information; when the service demand information is identified, multiple candidate service posts are recalled for the user based on the service demand information; a target service post is selected from multiple candidate service posts and the target service post is recommended to the user. The AI ​​question-and-answer service model based on artificial intelligence can not only play the role of a human customer service to talk to users, but also can identify the user's service demand information from the conversation information and accurately recommend service posts that meet user needs to users based on the service demand information, and recommend service posts that meet user needs to users during the conversation process on the conversation page, which can improve the conversion rate of service posts.

[0018] A solution provided by an embodiment of the present application is described in detail below in conjunction with the accompanying drawings.

[0019] Figure 1a A flowchart of a method for recommending service posts provided by an exemplary embodiment of the present application is provided. Figure 1b Another flowchart of the recommended method for posting services. Figure 1a and 1b As shown, including:

[0020] 101. Displaying a conversation interface, and conducting at least one round of conversation with the user based on the conversation interface, wherein at least conversation information of the user is displayed on the conversation interface;

[0021] 102. Input the dialogue information and model prompt words into an AI question-answering service model based on artificial intelligence, wherein the model prompt words include model role prompt words, which are used to prompt the AI ​​question-answering service model to at least provide post recommendation services to users, and the AI ​​question-answering service model is obtained by fine-tuning a pre-trained language model using sample data in the target domain to which the service posts belong;

[0022] 103. Under the prompt of the model role prompt word, identify the user's service demand information from the conversation information; when the service demand information is identified, recall multiple candidate service posts for the user according to the service demand information; select a target service post from the multiple candidate service posts and recommend the target service post to the user.

[0023] In the embodiment of the present application, a service application is installed on the user's terminal device, and the service application provides a conversation interface. Among them, the service application can be an independently running APP or a small program that depends on the APP to run. In addition, this embodiment does not limit the implementation form of the terminal device. For example, the terminal device can be a smart handheld device, such as a smart phone, a tablet computer, a laptop or a desktop computer, etc.; for another example, the terminal device can also be a smart wearable device, such as a smart watch, a smart bracelet, etc.; for another example, the terminal device can also be various smart home appliances with display screens, such as smart TVs, smart large screens or smart robots, etc.

[0024] In an embodiment of the present application, in response to the user's conversation operation, a conversation interface is displayed, and the user can have at least one round of conversation with the customer service of the application based on the conversation interface. During the conversation, the conversation information between the customer service and the user will be displayed in the conversation interface. Among them, the customer service can be a manual customer service or an AI robot. In the case where the customer service is an AI robot, the AI ​​question and answer service model based on artificial intelligence can play the role of the AI ​​robot to have a conversation with the user and answer questions for the user. The embodiment of the present application focuses on using the AI ​​question and answer service model based on artificial intelligence to play the role of the AI ​​robot to have a conversation with the user. The AI ​​question and answer service model is obtained by fine-tuning the pre-trained language model using sample data in the target field to which the service post belongs. It should be noted that the field involved in the sample data is the same field as the field involved in the post recommendation service provided.

[0025] In an embodiment of the present application, the AI ​​question-and-answer service model based on artificial intelligence corresponds to a prompt word, which is used to guide the model to generate a specific type of answer or perform a specific task. For example, the prompt word can be: "Role name: ***-your exclusive service steward", "Task: When the user just enters the corresponding page, you must output the opening words. Your main purpose is to help users find the service merchants they need. You need to answer all questions raised by the user. If the service is not related, generate answers within your ability and guide the user to express their service demands at the end", "Requirements: (1) When the user describes more than one service, ask the user about their actual needs and use the get_user_mes_list function to obtain a list of merchant information; (2) Do not ask the user about their location or area; (3) When the user mentions "recommendation", "recommended service", "recommended merchant", "introducing merchant", "help me find", "I want to find", "is there XX service", "give me a merchant number", "give me a merchant (4) Cleaning and sweeping of any items in the house is considered “cleaning and cleaning”; (5) Famous wines, famous watches, etc. are considered luxury gifts; (6) If the customer asks for a phone number or WeChat, do not provide any contact information, but guide the user to answer the required service; (7) You can expose that you are an AI robot. If you cannot answer, reply “This question is very profound. I’m sorry that I can’t answer it yet, but I will keep learning; (8) Please use a friendly language style and avoid using overly professional or complex expressions; (9) Please ensure that the language is natural, simple, and clear; (10) Please pay attention to the context when replying and do not repeat yourself; (11) Do not reply to questions that violate laws and regulations or public order and good morals.

[0026] This embodiment does not limit the conversation information between the AI ​​question-and-answer service model based on artificial intelligence and the user. For example, when the user just enters the conversation interface, under the guidance of the opening prompt, the AI ​​question-and-answer service model based on artificial intelligence can output the opening words, and the opening words can be the words preset in the AI ​​question-and-answer service model based on artificial intelligence or the words pre-learned by the AI ​​question-and-answer service model based on artificial intelligence. The opening words can be, for example: "Hi, I am ***, your little assistant in life. What service do you want to find today? Tell me your needs and I will help you!", "Good morning / afternoon / evening! I am ***, do you need to find any special service today? Please tell me at any time and I will help you find it." (The time period will be automatically filled according to the time of the day), "Good morning / afternoon / evening! I am ***, what kind of service do you want to find? Tell me your needs and I will recommend it to you!" (The time period will be automatically filled according to the time of the day), "Hello, I am ***! Do you need help finding a service merchant? Let me know your requirements and I will provide suitable options. ", "Hi, I am ***! Do you have any life needs? Tell me what you have in mind, and I will find the right service for you.", "Good evening! I am ***. What kind of service are you looking for nearby? Tell me your needs, and I will recommend it to you!", "Hello, I am ***! Do you have any special needs today? Let me help you find the right service provider.", "Hi, I am ***! What service are you looking for? Tell me your requirements, and I will help you find it quickly.", "Hello! I am ***. What kind of service are you looking for? Tell me your needs, and I will help you find nearby businesses.", "Hi, I am ***. What special service are you looking for? Tell me your needs, and I will recommend the most suitable option to you.".

[0027] Correspondingly, the service-side of the service application may also have a corresponding conversation interface, and the AI ​​question-and-answer service model of artificial intelligence may conduct at least one round of dialogue with the user based on the conversation interface, and at least the user's dialogue information is displayed on the conversation interface. In the process of corresponding with the user, the AI ​​question-and-answer service model of artificial intelligence may also provide the user with a post recommendation service, so that the user can understand the content details of the post and promote the conversion rate of the post.

[0028] In an embodiment of the present application, after obtaining the user's conversation information, the user's conversation information and model prompt words are input into an AI question-and-answer service model based on artificial intelligence. The prompt words include model role prompt words, which are used to prompt the AI ​​question-and-answer service model to at least provide post recommendation services to the user.

[0029] In an embodiment of the present application, after the user's conversation information and model prompt words are input into the AI ​​question-and-answer service model based on artificial intelligence, the user's service demand information can be identified from the conversation information under the prompt of the model role prompt words, so as to determine the matching service categories and service objects based on the service demand information, and recommend suitable posts to the user based on the matching service categories and service objects.

[0030] In an embodiment of the present application, the service demand information includes: a target service category and a target service object under the target service category, wherein the target service category is the service category to which the service demand information belongs, and the service category can be used to distinguish the fields of different services. Through this level of classification, it is convenient to provide accurate post recommendation services in specific fields. The target service object is a service or physical commodity that is included in the target service category and is adapted to the user's demand service information. Through this level of classification, it is convenient to provide more accurate post recommendation services in specific fields. Service categories can be, for example, nanny, confinement nanny, moving, second-hand recycling, maintenance, etc. The service objects of nanny and confinement nanny can be, for example, hourly workers, day shift nanny / confinement nanny, live-in nanny / confinement nanny, etc. The service objects included in moving can be: personal moving, family moving, company moving, etc. The service objects included in second-hand recycling can be: mobile phone recycling, computer recycling, air conditioner recycling, washing machine recycling, etc. The service objects included in maintenance can be, for example, mobile phone repair, computer repair, air conditioner repair, washing machine repair, etc.

[0031] Based on the above service demand information, the user's service demand information is identified from the conversation information. The optional implementation method includes: using the first AI tool to match the conversation information with the known service category, and if there is a match, the matching service category is used as the target service category; using the second AI tool to match the conversation information with each service object under the target service category, and if there is a match, the matching service object is used as the target service object. Among them, the first AI tool and the second AI tool refer to software tools that use artificial intelligence technology to complete the above functional services, or the first AI tool and the second AI tool are interfaces for calling the above functional services of the AI ​​question-and-answer service model based on artificial intelligence. The combination of the first AI tool and the second AI tool can improve the efficiency of matching the adapted target service object based on the user's service demand information. In addition, the first AI tool is used to first determine the target service category that is adapted to the user's service demand information, and then the second AI tool is used to determine the adapted target service object, that is, first determine the field that is adapted to the user's service demand information, and then determine the adapted target service object from the field, which can improve the accuracy of matching the target service object.

[0032] Wherein, when the first AI tool and the second AI tool are software tools, the first AI tool and the second AI tool can be implemented by the following function (the function includes the first AI tool and the second AI tool, the first part of the function is the first AI tool, and the tail of the function is the second AI tool):

[0033] [{

[0034] "type":"function",

[0035] "function":{

[0036] "name":"get_user_mes_list",

[0037] "description":"Get the service content that users want",

[0038] "parameters":{

[0039] "type":"object",

[0040] "properties":{

[0041] "workContent":{

[0042] "type":"string",

[0043] "description":"Business service content that users want",

[0044] "enum":["Nanny / monthly nanny, hourly worker","Nanny / monthly nanny, live-in nanny","Nanny / monthly nanny, day shift nanny","Nanny / monthly nanny, day shift childcare nanny","Nanny / monthly nanny, live-in childcare nanny","Nanny / monthly nanny, month-long nanny","Nanny / monthly nanny, aunt recruitment","Nanny / monthly nanny, care worker / accompanying worker","Nanny / monthly nanny, storage and sorting","Nanny / monthly nanny, lactation","Moving, resident moving","Moving, manual labor","Moving, cross-city moving","Moving, company moving","Moving, factory moving and cargo moving","Moving, air conditioner moving","Moving, equipment lifting","Moving, piano moving","Moving, Japanese moving","Moving, van moving","Business registration, company registration"," Industrial and commercial registration, accounting agent", "Industrial and commercial registration, tax verification", "Industrial and commercial registration, company transfer", "Industrial and commercial registration, company cancellation / change", "Industrial and commercial registration, license / qualification processing", "Industrial and commercial registration, industrial and commercial annual inspection", "Industrial and commercial registration, trademarks and patents", "Industrial and commercial registration, general taxpayer application", "Cleaning and cleaning, daily cleaning", "Cleaning and cleaning, deep cleaning", "Cleaning and cleaning, land reclamation cleaning", "Cleaning and cleaning, formaldehyde removal", "Cleaning and cleaning, pest control", "Cleaning and cleaning, home appliance cleaning", "Cleaning and cleaning, fabric cleaning", "Cleaning and cleaning, stone maintenance", "Cleaning and cleaning, storage and organization", "Cleaning and cleaning, rental cleaning", "Cleaning and cleaning, glass cleaning", "Spray-painted signs, light boxes / signs", "Spray Painted signs, background / image wall", "Spray-painted signs, outdoor advertising", "Spray-painted signs, display rack production", "Spray-painted signs, banners / pennants / medals", "Spray-painted signs, LED display screens", "Spray-painted signs, plaque production", "Spray-painted signs, doorhead production and installation", "Spray-painted signs, advertising design and production", "Spray-painted signs, luminous characters", "Spray-painted signs, spray painting", "Home appliance repair, air conditioning repair", "Home appliance repair, washing machine repair", "Home appliance repair, water heater repair", "Home appliance repair, gas stove repair", "Home appliance repair, refrigerator repair", "Home appliance repair, TV / projector repair", "Home appliance repair, wall-mounted boiler repair", "Home appliance repair, range hood repair", "Home appliance repair, air conditioning fluorination", "Home appliance repair, Air conditioning disassembly and assembly", "Ceremony and celebration, venue layout", "Ceremony and celebration, celebration company", "Ceremony and celebration, performance", "Ceremony and celebration, ceremony model", "Ceremony and celebration, event planning", "Ceremony and celebration, event host", "Rental, building materials / construction", "Rental, air conditioning", "Rental, forklift", "Rental, generator", "Rental, electronic equipment", "Rental, crane", "Rental, excavator", "Rental, loader", "Rental, truck", "Rental, articulated arm truck", "Rental, scaffolding", "Rental, road roller", "Rental, forklift", "Rental, hanging basket", "Rental, lift", "Rental, ladder truck", "Rental, bulldozer", "Rental, grader", "Rental, spider truck", "Rental,Wrecker", "Rental, Sprinkler", "Rental, Paving Machine", "Rental, Collision Buffer Vehicle", "Rental, Pile Driver", "Rental, Ramming Machine", "Rental, Pump Truck", "Rental, Concrete Mixer", "Rental, Tile Press", "Rental, Printer", "Rental, Copier", "Rental, Display", "Rental, Projector", "Rental, All-in-One", "Rental, Desktop Computer", "Rental, Tablet", "Rental, Laptop", "Rental, Television", "Rental, Dehumidifier", "Rental, Oven", "Rental, Sweeper", "Rental, Coffee Machine", "Rental, Dishwasher", "Rental, Ice Maker", "Rental, Aromatherapy Diffuser", "Rental, Tables and Chairs", "Rental, Sofa", "Rental, Screen Wind","Rental, bar tables and chairs","Rental, beds","Rental, dresses","Rental, wedding dresses","Rental, ancient costumes","Rental, Hanfu","Rental, children's clothes","Rental, performance clothes","Rental, chorus clothes","Rental, dance clothes","Rental, puppet clothes","Rental, national costumes","Rental, host clothes","Rental, multimedia classrooms","Rental, tiered classrooms","Rental, conference rooms","Rental, training classrooms","Rental, shops","Rental, celebration venues","Rental, factory warehouses","Rental, tents","Rental, container rooms","Rental, mobile board rooms","Rental, tents","Rental, fences","Rental, one-meter line","Rental, parasols","Rental Rental, iron fence", "Rental, mobile toilet", "Rental, LED screen rental", "Rental, lighting and sound", "Rental, stage equipment", "Rental, stage truss", "Rental, voting machine", "Rental, buzzer", "Rental, timer", "Rental, answering machine", "Rental, walkie-talkie", "Rental, VR equipment", "Rental, amusement equipment", "Rental, electronic entertainment equipment", "Rental, display cabinet, display board, display wall", "Unlock opening / lock change / lock repair, unlocking / lock change / lock repair", "Unlock opening / lock change / lock repair, security door unlocking", "Unlock opening / lock change / lock repair, safe unlocking", "Unlock opening / lock change / lock repair, key duplication", "Unlock opening / lock change / lock repair, car unlocking", "Home improvement service, whole house decoration", "Home Home decoration services, partial decoration", "Home decoration services, wall painting", "Home decoration services, installation / removal", "Home decoration services, coating / renovation", "Home decoration services, looking for casual workers", "Home decoration services, wall painting", "Home decoration services, house inspection / supervision", "Home decoration services, furniture customization", "Home decoration services, decoration design", "Public decoration services, partial decoration", "Public decoration services, overall decoration", "Public decoration services, installation / removal", "Public decoration services, looking for casual workers", "Public decoration services", "Public decoration services, wall painting", "Public decoration services, wall painting", "Public decoration services, garbage removal", "Public decoration services, flooring", "House repair / waterproofing, waterproofing and leak repair", "House repair / waterproofing, heating repair / floor heating", "House repair / waterproofing,Wall and floor repair", "House repair / waterproofing, door and window repair", "House repair / waterproofing, circuit and lamp repair", "House repair / waterproofing, kitchen and bathroom sanitary ware repair", "House repair / waterproofing, punching", "House repair / waterproofing, furniture repair and maintenance", "House repair / waterproofing, repair of water pipes and faucets", "House repair / waterproofing, pipe unblocking", "Pipe unblocking / cleaning, toilet unblocking", "Pipe unblocking / cleaning, floor drain unblocking", "Pipe unblocking / cleaning, water pipe unblocking", "Pipe unblocking / cleaning, floor heating cleaning", "Pipe unblocking / cleaning, sewer unblocking", "Pipe unblocking / cleaning, septic tank cleaning", "Pipe unblocking / cleaning, sink unblocking", "Pipe unblocking / cleaning, salvage", "Pipe unblocking / cleaning, Commercial pipe unblocking", "Pipe unblocking / cleaning, municipal silt removal", "Financial accounting / assessment, bookkeeping agency", "Financial accounting / assessment, financial accounting", "Financial accounting / assessment, tax declaration", "Financial accounting / assessment, tax consultation", "Financial accounting / assessment, financial and tax problems", "Financial accounting / assessment, capital verification", "Financial accounting / assessment, financial audit", "Financial accounting / assessment, asset assessment", "Financial accounting / assessment, project cost", "Financial accounting / assessment, tax planning", "House construction / renovation, construction waste removal", "House construction / renovation, house demolition / wall demolition", "House construction / renovation, container / mobile house", "House construction / renovation, wall drilling", "House construction / renovation, old house Reconstruction","House building / reconstruction and renovation, self-built house / loft","House building / reconstruction and renovation, metal appliance recycling","House building / reconstruction and renovation, cast-in-place concrete","House building / reconstruction and renovation, steel structure addition","House building / reconstruction and renovation, wall to beam / partition","Mobile phone / computer / digital / office recycling, computer / tablet","Mobile phone / computer / digital / office recycling, mobile phone","Mobile phone / computer / digital / office recycling, digital products","Mobile phone / computer / digital / office recycling, electronic hardware","Mobile phone / computer / digital / office recycling, office supplies","Mobile phone / computer / digital / office recycling","Furniture / electrical appliances / hotel equipment recycling, office furniture","Furniture / electrical appliances / hotel equipment recycling, civilian furniture", "Furniture / Electrical Appliances / Hotel Equipment Recycling, Hotel Equipment", "Furniture / Electrical Appliances / Hotel Equipment Recycling, Kitchen Equipment", "Furniture / Electrical Appliances / Hotel Equipment Recycling, Stationery", "Furniture / Electrical Appliances / Hotel Equipment Recycling, Beauty Salon", "Furniture / Electrical Appliances / Hotel Equipment Recycling, Air Conditioner", "Furniture / Electrical Appliances / Hotel Equipment Recycling, Home Appliances", "Gifts / Luxury Goods / Gold and Silver Recycling, Gold and Silver Jewelry", "Gifts / Luxury Goods / Gold and Silver Recycling, Shopping Cards", "Gifts / Luxury Goods / Gold and Silver Recycling, Luxury Goods", "Gifts / Luxury Goods / Gold and Silver Recycling, Artworks", "Gifts / Luxury Goods / Gold and Silver Recycling, Famous Wine and Old Wine Recycling", "Gifts / Luxury Goods / Gold and Silver Recycling, Supplement Recycling", "Gifts / Luxury Goods / Gold and Silver Recycling, Famous Tea Recycling","Gifts / Luxury Goods / Gold and Silver Recycling, Jade Products", "Industrial Materials / Metal Waste Recycling, Scrap Metal", "Industrial Materials / Metal Waste Recycling, Industrial Waste", "Industrial Materials / Metal Waste Recycling, Industrial Equipment", "Industrial Materials / Metal Waste Recycling, Clothing and Leather", "Industrial Materials / Metal Waste Recycling, Scrap Recycling", "Industrial Materials / Metal Waste Recycling, Rare Metal Recycling"],

[0045] },

[0046] "object":{

[0047] "type":"string",

[0048] "description":"Subject matter or user purpose, such as air conditioners, cordyceps sinensis, mobile phones, liquor, furniture, electrical appliances, etc."

[0049] }

[0050] },

[0051] "required":["object","workContent"]

[0052] }

[0053] }

[0054] }].

[0055] In some embodiments, a first AI tool is used to match conversation information with known service categories. If a match occurs, the matching service category is used as a target service category. Optional implementation methods include: performing word segmentation processing on the conversation information to obtain multiple first word segmentations, and performing feature extraction on the multiple first word segmentations and each known service category to obtain multiple first word segmentation features and features of each known service category; calculating a first similarity between each first word segmentation feature and a feature of each known service category, and if there is a first target similarity greater than a first similarity threshold, using the service category corresponding to the first target similarity as the target service category.

[0056] Optionally, the dialogue information is segmented to obtain a plurality of first segmented words, including: preprocessing the text of the dialogue information, such as removing punctuation marks, converting to lowercase (for case-sensitive languages), removing stop words, etc., so as to reduce the complexity of subsequent processing and improve the accuracy of segmentation; further, different segmentation algorithms are selected according to different languages ​​and application scenarios. For example, for languages ​​without obvious separators such as Chinese, commonly used segmentation algorithms include but are not limited to: rule-based methods, statistical methods, and deep learning-based methods. Rule-based methods such as forward maximum matching (MM) and reverse maximum matching (RMM) use predefined dictionaries and rules for segmentation. Statistical methods such as hidden Markov models (HMM) and conditional random fields (CRF) use large-scale corpora and statistical information for segmentation. Deep learning-based methods can use neural network models such as recurrent neural networks (RNN) and long short-term memory networks (Long Short-Term Memory Networks). Memory, LSTM), Transformer, etc.; further, use word segmentation tools or libraries to implement word segmentation. For example, for Chinese, you can use jieba word segmentation library, pkuseg, THULAC, etc. These tools usually have integrated a variety of word segmentation algorithms and allow users to customize dictionaries to improve the accuracy of word segmentation; further, input the preprocessed text into the word segmentation tool, and the word segmentation tool segments the text according to the built-in algorithm and dictionary. For example, when using the jieba word segmentation library, you can directly get the word segmentation result list through the lcut function; further, on the basis of word segmentation, perform part-of-speech tagging on each word segmentation, that is, mark each word segmentation result with its part of speech (noun, verb, etc.) to facilitate further natural language processing tasks, so as to obtain multiple first word segmentations.

[0057] In some embodiments, the features of the multiple first word segmentation and the features of each known service category may be features in the form of vectors, and feature extraction is performed on the multiple first word segmentation and each known service category to obtain the multiple first word segmentation features and the features of each known service category, including: the multiple first word segmentation and each known service category are input into an encoder, and feature extraction is performed through a self-attention learning mechanism to obtain the multiple first word segmentation features and the features of each known service category. Among them, the features of the multiple first word segmentation may be user intention features.

[0058] In other embodiments, feature extraction is performed on multiple first participles and each known service category to obtain multiple first participle features and features of each known service category, including: using TF-IDF or a deep learning model to extract features of multiple first participles and each known service category. Among them, the deep learning model can be an RNN model or an LSTM model. Optionally, feature extraction is performed on multiple first participles and each known service category using a TF-IDF model, including: using TfidfVectorizer to extract features of multiple first participles and features of each known service category from a scikit-learn library, and more specifically, converting text data into TF-IDF feature vectors through a fit_transform method.

[0059] Optionally, the first similarity between each first segmentation feature and the features of each known service category is calculated, and if there is a first target similarity greater than a first similarity threshold, the service category corresponding to the first target similarity is used as the target service category, including: calculating the first similarity between each first segmentation feature and the features of each known service category, if there is a first target similarity between any first segmentation and the features of any known service category whose similarity is greater than the first similarity threshold, the service category corresponding to the first target similarity is used as the target service category; or calculating the first similarity between each first segmentation feature and the features of each known service category, if there is a first target similarity between a preset number of first segmentations and the features of the same known service category whose similarity is greater than the first similarity threshold, the service category corresponding to the first target similarity is used as the target service category. The preset number may be part or all of the first segmentations. It should be noted that, in the case where there are multiple first target similarities greater than the first similarity threshold, the service category corresponding to the maximum value of the first target similarity may be used as the target service category.

[0060] Regardless of which of the above similarity calculation methods is used, the similarity calculation between the first word segmentation feature and the known service category feature will be involved. Optionally, the first word segmentation feature and the known service category feature can be feature vectors, and the first similarity between the first word segmentation feature and the known service category feature is calculated, including: calculating the feature vector difference based on the first word segmentation feature vector and the known service category feature vector; normalizing the feature vector difference, and using the normalized calculation result as the first similarity.

[0061] Correspondingly, a second AI tool is used to match the conversation information with each service object under the target service category. If a match is found, the matching service object is used as the target service object. Optional implementation methods include: extracting features of each service object under the target service category to obtain features of each service object; calculating the second similarity between each first participle feature and the feature of each service object; if there is a second target similarity greater than a second similarity threshold, the service object corresponding to the second target similarity is used as the target service object.

[0062] Optionally, the second similarity between each first participle feature and the feature of each service object is calculated. If there is a second target similarity greater than the second similarity threshold, the service object corresponding to the second target similarity is used as the target service object, including: calculating the second similarity between each first participle feature and the feature of each service object, if there is a second target similarity between any first participle and the feature of any known service category whose similarity is greater than the second similarity threshold, the service object corresponding to the second target similarity is used as the target service object; or, calculating the second similarity between a preset number of first participle features and the features of the same service object, if there is a second target similarity between a preset number of first participles and the features of the same known service category whose similarity is greater than the second similarity threshold, the service object corresponding to the second target similarity is used as the target service object. Wherein, the preset number may be part or all of the first participles. For the implementation of feature extraction and similarity calculation in this embodiment, please refer to the description of the above-mentioned related embodiments, which will not be repeated here.

[0063] In other embodiments, the first AI tool is used to match the conversation information with known service categories. If a match is found, the matching service category is used as the target service category. Optional implementation methods include: extracting features from the conversation information and the known service categories to obtain features of the conversation information and each known category; calculating the fourth similarity between the features of the conversation information and each known category, and using the service category corresponding to the maximum value in the similarity as the target service category. For the implementation methods of feature extraction and similarity calculation in this embodiment, please refer to the description of the above-mentioned related embodiments, which will not be repeated here.

[0064] In an embodiment of the present application, when user service demand information is identified, multiple candidate service posts can be recalled for the user based on the user's service demand information, so as to select a target service post from the multiple candidate service posts and recommend the target service post to the user.

[0065] In some embodiments, the service demand information includes a target service category and a target service object under the target service category, and a plurality of candidate service posts are recalled for the user according to the service demand information, including at least one of the following methods:

[0066] Method 1: Based on the target service category and the target service object, from the service posts located in the business district near the user, service posts whose distance from the user's location is less than a set first distance threshold are recalled as candidate service posts. The business district near the user can be a business district with the user as the center and a preset distance as the radius. The first distance threshold is less than the preset distance, and the preset distance can be 5KM, 10KM, etc.

[0067] Method 2: Based on the target service category and target service object, service posts with a service location in the target business district where the user is located and whose user rating is greater than the set rating threshold are recalled as candidate service posts. The service location refers to a location or range where the corresponding service can be provided, and the target business district where the user is located refers to a shopping mall, a district, or a preset range.

[0068] Method 3: Based on the target service category and target service object, service posts in the target area where the user is located are recalled as candidate service posts; wherein the target area includes the business district near the user, the business district near the user does not include the business district where the user is located, and the business district near the user can be a business district within a set range, such as Figure 1c As shown, the business district near the user is outside the business district where the user is located.

[0069] It should be noted that, recalling multiple candidate service posts for the user according to the service demand information may be done by one of the above three methods or a combination of multiple methods.

[0070] In some embodiments, based on the target service category and the target service object, service posts whose service locations are located in a business district near the user and whose distance to the user's location is less than a set first distance threshold are recalled as candidate service posts, including: based on the target service category and the target service object, an initial service post belonging to the target service category and containing the target service object is obtained from service posts whose service locations are located in a business district near the user; based on the service location of the initial service post and the user's location, the distance from the initial service post to the user's location is calculated, and service posts whose distances are less than the set first distance threshold are selected from the initial service posts as candidate service posts.

[0071] Optionally, based on the target service category and the target service object, an initial service post belonging to the target service category and containing the target service object is obtained from service posts whose service locations are located in a business district near the user, including: extracting features of the target service category and the target service object respectively to obtain features of the target service category and features of the target service object; performing word segmentation on each service post whose service location is located in a business district near the user to obtain multiple second word segmentations corresponding to each post, and performing feature extraction on the multiple second word segmentations corresponding to each post to obtain multiple second word segmentation features; calculating the third similarity between each post and the target service object under the target service category according to the multiple second word segmentation features corresponding to each post, the features of the target service category, and the features of the target service object; and selecting a post whose third similarity is greater than a third similarity threshold from service posts whose service locations are located in a business district near the user as the initial service post. For the implementation of feature extraction and similarity calculation in this embodiment, please refer to the description of the above-mentioned related embodiments, which will not be repeated here.

[0072] Among them, according to the multiple second segmentation features corresponding to each post, the features of the target service category and the features of the target service object, the third similarity between each post and the target service object under the target service category is calculated. The optional implementation method includes: for the multiple second segmentation features of each post, the similarity between each segmentation feature and the features of the target service category and the features of the target service object is calculated; based on the two similarities, a weighted sum calculation is performed to obtain the third similarity between each post and the target service object under the target service category. Alternatively, for each post, all second segmentation feature segmentations are jointly and separately calculated with the features of the target service category and the features of the target service object to obtain the third similarity between each post and the target service object under the target service category. Alternatively, the comprehensive features between the features of the target service category and the features of the target service object are calculated; the similarity between each segmentation feature separately / commonly and the comprehensive features is calculated to obtain the third similarity between each post and the target service object under the target service category.

[0073] It should be noted that the first similarity, the second similarity, the third similarity and the fourth similarity in the embodiment of the present application may be the same value or different values.

[0074] In an embodiment of the present application, after obtaining multiple candidate posts, a target service post can be selected from the multiple candidate service posts and recommended to the user, so as to accurately recommend posts that are adapted to the user's service demand information, and the user does not need to spend time and effort to select from multiple posts, thereby improving the user's experience.

[0075] In some embodiments, selecting a target service post from multiple candidate service posts includes: using a ranking model to obtain historical behavior characteristics of users within a historical period, and calculating the matching degree between each candidate service post and the historical behavior characteristics as a first ranking score for each candidate service post; calculating a second ranking score for each candidate service post based on the user score of each candidate service post and the distance from the service location of each candidate service post to the user; calculating the semantic similarity between the title of each candidate service post and the target service object as a third ranking score for each candidate service post; ranking each candidate service post based on the first ranking score, the second ranking score and / or the third ranking score, and selecting a target service post from the candidate service posts based on the ranking order.

[0076] Optionally, the historical period can be any period before the current period, and the historical behavior features include but are not limited to: display, clicks, call volume, click-through rate, conversion rate and other indicators. The matching degree between each candidate service post and the historical behavior feature refers to the number of times the user executes one or all of the above historical behavior features for each candidate service post. The more times, the higher the matching degree. The matching degree between each candidate service post and the historical behavior feature is calculated as the first ranking score of each candidate service post, including: calculating the matching degree between each candidate service post and any historical behavior feature, and taking the one with the highest matching degree as the first ranking score of each candidate service post. Alternatively, the matching degree between each candidate service post and all historical behavior features is calculated as the first ranking score of each candidate service post. Among them, the sorting model can be a multi-objective model (Multi-gate Mixture-of-Experts, MMoE).

[0077] Optionally, a second ranking score for each candidate service post is calculated based on the user score of each candidate service post and the distance from the service location of each candidate service post to the user, including: calculating a first initial ranking score for each candidate service post based on the user score of each candidate service post; calculating a second initial ranking score for each candidate service post based on the distance from the service location of each candidate service post to the user; calculating a second ranking score for each candidate service post based on the first initial ranking score and the second initial ranking score of each service post.

[0078] Among them, according to the user score of each candidate service post, the first initial ranking score of each candidate service post is calculated, and the optional implementation methods include: selecting the highest score and the lowest score from the user scores of each candidate service post; taking the difference between each post score and the lowest score as the numerator, and the difference between the highest score and the lowest score as the denominator, performing quotient calculation, and using the calculation result as the first initial ranking score of each candidate service post.

[0079] Among them, according to the distance from the service location of each candidate service post to the user, the second initial ranking score of each candidate service post is calculated, and the optional implementation methods include: selecting the farthest distance and the closest distance from the distance from the service location of each candidate service post to the user; using the difference between the distance from the service location of each post to the user and the closest distance as the numerator, and the difference between the farthest distance and the closest distance as the denominator, performing quotient calculation, and using the calculation result as the second initial ranking score of each candidate service post. Among them, the quotient calculation can be regarded as a normalization calculation to project the distance and the score onto the same dimension. In addition, the normalization calculation method is not limited to this, and can also be a square difference calculation, a dot product calculation, etc.

[0080] According to the first initial ranking score and the second initial ranking score of each candidate service post, the second ranking score of each candidate service post is calculated, and an optional implementation includes: performing a weighted sum calculation according to the first initial ranking score and the second initial ranking score of each service post, and using the calculation result as the second ranking score of each candidate service post. The weight ratio of the weighted sum can be 1:1, but is not limited thereto.

[0081] Optionally, the semantic similarity between the title of each candidate service post and the target service object is calculated as the calculation method of the third ranking score of each candidate service post, and the calculation method of the feature similarity in the above embodiment can be referred to. Alternatively, the BM25 value between the title of each candidate service post and the target service object can be calculated as the semantic similarity between the title of each candidate service post and the target service object, so as to serve as the third ranking score of each candidate service post.

[0082] Among them, the BM25 value between the title of each candidate service post and the target service object is calculated, and optional implementation methods include: determining the number of identical characters between the title of each candidate service post and the name of the target service object; taking the number of identical characters as the numerator and the number of characters contained in the title of each candidate service post as the denominator to obtain a first quotient value, and taking the number of identical characters as the numerator and the number of characters contained in the name of each target service object as the denominator to obtain a second quotient value, and taking the sum of the first quotient value and the second quotient value as the third ranking score of each candidate service post.

[0083] It should be noted that the method of feature extraction and similarity calculation in each embodiment of the present application can refer to the relevant description of each embodiment, and no limitation is made to this.

[0084] Further optionally, the model role prompt is also used to prompt the AI ​​question-and-answer service model to provide question-and-answer services to users, and the model prompt also includes a service guidance prompt, which is used to prompt the AI ​​question-and-answer service model to provide guidance services to users. Then, under the prompt of the model role prompt, answer information is generated for the dialogue information, and the answer information is displayed in the conversation interface for users to view. The model role prompt can be, for example, "after the user enters the conversation interface, question-and-answer services need to be provided to the user."

[0085] Further optionally, when no service demand information is identified, prompt information and / or service demand guidance information that can provide post recommendation services is generated at the end of the answer information under the prompt of the service guidance prompt word, and the prompt information and / or service demand guidance information that can provide post recommendation services is displayed in the conversation interface to guide the user to express service needs in subsequent conversations.

[0086] Among them, the prompt information and / or service demand guidance information for providing post recommendation services can be, for example: "Hello! I just recommended some businesses to you. I wonder what you think? Is there anything unsatisfactory? Please tell me and I can further adjust the recommendation for you.", "Hi, I am ***. Does the business information just provided meet your needs? If there is anything unsatisfactory, please tell me and I will help you find a more suitable option.", "Hello! Seeing that you have not replied yet, I want to confirm whether the previously recommended businesses meet your expectations? If there is anything inappropriate, you can tell me and I will make adjustments.", "Hello! I just recommended some service businesses to you. I wonder if these recommendations meet your requirements? If there is anything unsatisfactory, please tell me and I will find a more suitable option for you.", "Hi, *** is here! What do you think of the previously recommended business information? Is there anything that needs to be improved? Please tell me. Tell me, I will find a new service based on your feedback.", "Hello! I have just screened some merchants for you. Do any of them meet your needs? If you are not satisfied with anything, please tell me and I will find a more suitable option for you as soon as possible.", "Hi! Have you seen the merchants recommended before? If you are not satisfied with anything or have other needs, please tell me and I will provide you with more accurate recommendations.", "Hello! I see that you have not replied yet. I wonder if the merchants I recommended before meet your needs? If you have any questions or dissatisfaction, please tell me and I will adjust the recommendation.", "Hello! I just provided some merchant recommendations. Do any of them meet your needs? If there are any inappropriate places or other requirements, please feel free to tell me and I will make adjustments. ", "Hi, I am ***. Does the previous recommended information meet your needs? If you are not satisfied with anything, please tell me and I will optimize it based on your feedback. ", etc.

[0087] Further optionally, the model prompt words also include: style constraint prompt words, which are used to constrain the word count, format and / or style of the answer information and service demand guidance information generated by the AI ​​question-answering service model; wherein the word count of the service demand guidance information is more than the word count of the answer information. Style constraint prompt words can be, for example: "The word count should be within the range of X words", "The font format and paragraph format should be...", "The language style should meet the requirements of...".

[0088] It can be seen from the above implementation examples of this application that the AI ​​question-and-answer service model is obtained by fine-tuning the pre-trained language model using sample data in the target field to which the service post belongs. Optionally, the pre-trained language model is fine-tuned using sample data in the target field to which the service post belongs, including: sample data in the target field to which the service post belongs, the sample data containing the user's dialogue information, model role prompt words, and corresponding recommended sample service posts; inputting the dialogue information and model prompt words in the sample data into the pre-trained language model, and identifying the user's service demand information from the dialogue information under the prompt of the model role prompt words contained in the sample data; in the case of identifying the service demand information, recalling multiple candidate service posts for the user according to the service demand information; selecting a target service post from multiple candidate service posts; performing loss calculation based on the sample service post and the target service post to obtain a loss function; according to the loss function, adjusting the network parameters of each network layer of the pre-trained language model until the loss function meets the model training termination condition, so as to obtain an AI question-and-answer service model based on artificial intelligence.

[0089] The technical solution provided by the above-mentioned embodiment of the present application, in the embodiment of the present application, inputs the dialogue information and model prompt words into the AI ​​question-and-answer service model based on artificial intelligence, identifies the user's service demand information from the dialogue information; when the service demand information is identified, recalls multiple candidate service posts for the user based on the service demand information; selects the target service post from the multiple candidate service posts and recommends the target service post to the user. The AI ​​question-and-answer service model based on artificial intelligence can not only play the role of a human customer service to talk with the user, but also can identify the user's service demand information from the dialogue information and accurately recommend service posts that meet the user's needs to the user based on the service demand information, and recommends service posts that meet the user's needs to the user during the dialogue process on the dialogue page, which can improve the conversion rate of service posts.

[0090] Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 2 As shown, it includes: a memory and a processor; the memory is used to store a computer program; the processor is coupled to the memory and is used to execute the computer program to implement the following steps:

[0091] Display a conversation interface, and conduct at least one round of conversation with the user based on the conversation interface, with at least the user's conversation information displayed on the conversation interface; input the conversation information and model prompts into an AI question-and-answer service model based on artificial intelligence, wherein the model prompts include model role prompts, which are used to prompt the AI ​​question-and-answer service model to at least provide post recommendation services to the user, and the AI ​​question-and-answer service model is obtained by fine-tuning a pre-trained language model using sample data in the target field to which the service posts belong; identify the user's service demand information from the conversation information under the prompt of the model role prompts; when the service demand information is identified, recall multiple candidate service posts for the user based on the service demand information; select a target service post from the multiple candidate service posts and recommend the target service post to the user.

[0092] In some embodiments, the service demand information includes a target service category and a target service object under the target service category. When the processor identifies the user's service demand information from the conversation information, it is specifically used to: use a first AI tool to match the conversation information with a known service category, and if there is a match, use the matched service category as the target service category; use a second AI tool to match the conversation information with each service object under the target service category, and if there is a match, use the matched service object as the target service object.

[0093] Optionally, when the processor uses the first AI tool to match the conversation information with the known service categories, and if a match occurs, takes the matching service category as the target service category, the processor is specifically used to: perform word segmentation processing on the conversation information to obtain multiple first word segmentations, and perform feature extraction on the multiple first word segmentations and each known service category respectively to obtain multiple first word segmentation features and features of each known service category; calculate a first similarity between each first word segmentation feature and a feature of each known service category, and if there is a first target similarity greater than a first similarity threshold, take the service category corresponding to the first target similarity as the target service category,

[0094] Correspondingly, when the processor uses the second AI tool to match the conversation information with each service object under the target service category, and if a match occurs, uses the matching service object as the target service object, it is specifically used to: extract features from each service object under the target service category to obtain features of each service object; calculate the second similarity between each first word segmentation feature and the feature of each service object, and if there is a second target similarity greater than a second similarity threshold, use the service object corresponding to the second target similarity as the target service object.

[0095] In some embodiments, the service demand information includes a target service category and a target service object under the target service category. When the processor recalls multiple candidate service posts for the user according to the service demand information, the processor is specifically configured to perform at least one of the following:

[0096] Based on the target service category and the target service object, service posts whose service locations are located in a business district near the user and whose distance to the user's location is less than a set first distance threshold are recalled as candidate service posts.

[0097] Based on the target service category and target service object, service posts whose service locations are located in the target business district where the user is located are recalled as candidate service posts whose user scores are greater than a set score threshold.

[0098] Based on the target service category and the target service object, service posts in the target area where the user is located are recalled as candidate service posts; wherein the target area includes the business district near the user, and the business district near the user includes the business district where the user is located.

[0099] In an embodiment of the present application, when the processor recalls service posts whose distance to the user's location is less than a set first distance threshold from service posts whose service locations are located in a business district near the user based on the target service category and the target service object, as candidate service posts, the processor is specifically used to: based on the target service category and the target service object, obtain an initial service post belonging to the target service category and containing the target service object from service posts whose service locations are located in a business district near the user; calculate the distance from the initial service post to the user's location based on the service location of the initial service post and the user's location, and select service posts whose distance is less than the set first distance threshold from the initial service posts as candidate service posts.

[0100] Optionally, when the processor obtains an initial service post belonging to the target service category and containing the target service object from service posts whose service locations are located in a business district near the user based on the target service category and the target service object, the processor is specifically used to: perform feature extraction on the target service category and the target service object respectively to obtain features of the target service category and features of the target service object; perform word segmentation on each service post whose service location is located in the business district near the user to obtain multiple second word segmentations corresponding to each post, and perform feature extraction on the multiple second word segmentations corresponding to each post to obtain multiple second word segmentation features; calculate the third similarity between each post and the target service object under the target service category based on the multiple second word segmentation features corresponding to each post, the features of the target service category, and the features of the target service object; and select, from the service posts whose service locations are located in the business district near the user, a post whose third similarity is greater than a third similarity threshold as the initial service post.

[0101] In an embodiment of the present application, when the processor selects a target service post from multiple candidate service posts, it is specifically used to: use a ranking model to obtain historical behavior characteristics of users in a historical period, and calculate the matching degree between each candidate service post and the historical behavior characteristics as the first ranking score of each candidate service post; calculate the second ranking score of each candidate service post according to the user score of each candidate service post and the distance from the service location of each candidate service post to the user; calculate the semantic similarity between the title of each candidate service post and the target service object as the third ranking score of each candidate service post; sort each candidate service post according to the first ranking score, the second ranking score and / or the third ranking score, and select the target service post from the candidate service posts according to the sorting order.

[0102] Further optionally, the model role prompt words are also used to prompt the AI ​​question and answer service model to provide question and answer services to users, and the model prompt words also include service guide prompt words, which are used to prompt the AI ​​question and answer service model to provide guidance services to users. The processor is also used to generate answer information for the dialogue information under the prompt of the model role prompt words, and display the answer information in the conversation interface for the user to view; when the service demand information is not identified, under the prompt of the service guide prompt words, prompt information and / or service demand guidance information that can provide post recommendation services is generated at the end of the answer information, and the prompt information and / or service demand guidance information that can provide post recommendation services is displayed in the conversation interface to guide users to express their service needs in subsequent conversations.

[0103] Further optionally, the model prompt words also include: style constraint prompt words, which are used to constrain the word count, format and / or style of the answer information and service demand guidance information generated by the AI ​​question and answer service model; wherein the word count of the service demand guidance information is more than the word count of the answer information.

[0104] Furthermore, if Figure 2 As shown, the server also includes: a communication component 20c, a display 20d, a power component 20e, an audio component 20f and other components. Figure 2 Only some components are shown schematically, which does not mean that the electronic device only includes Figure 2 Components shown.

[0105] The detailed implementation and beneficial effects of the electronic device provided in the embodiments of the present application have been described in detail in the aforementioned embodiments and will not be elaborated here.

[0106] The exemplary embodiments of the present application further provide a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor is caused to implement the steps in the above-mentioned method embodiments.

[0107] An exemplary embodiment of the present application further provides a computer program product, which includes a computer program / instructions. When the computer program / instructions are executed by a processor, the processor is enabled to implement the steps in the above-mentioned method embodiments.

[0108] The above-mentioned memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0109] The above-mentioned communication component is configured to facilitate wired or wireless communication between the device where the communication component is located and other devices. The device where the communication component is located can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G / LTE, 5G and other mobile communication networks, or a combination thereof. In an exemplary embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra Wide Band (UWB) technology, Bluetooth (BT) technology and other technologies.

[0110] The above-mentioned display includes a screen, and the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation.

[0111] The power supply assembly provides power to various components of the device where the power supply assembly is located. The power supply assembly may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device where the power supply assembly is located.

[0112] The above-mentioned audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (Microphone, MIC), and when the device where the audio component is located is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in a memory or sent via a communication component. In some embodiments, the audio component also includes a speaker for outputting an audio signal.

[0113] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-readable storage media (including but not limited to disk storage, compact disc read-only memory (Compact Disc Read-Only Memory, CD-ROM), optical storage, etc.) containing computer-usable program code.

[0114] The present application is described with reference to the flowchart and / or block diagram of the method, device (system) and computer program product according to the embodiment of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, and the combination of the process and / or box in the flowchart and / or block diagram can be realized by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for realizing the function specified in one process or multiple processes in the flowchart and / or one box or multiple boxes in the block diagram.

[0115] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0116] These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0117] In a typical configuration, a computing device includes one or more processors (Central Processing Unit, CPU), input / output interface, network interface and memory.

[0118] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0119] Computer readable media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0120] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0121] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for recommending service posts, characterized in that: include: Displaying a conversation interface, and conducting at least one round of conversation with the user based on the conversation interface, wherein at least conversation information of the user is displayed on the conversation interface; Inputting the dialogue information and model prompt words into an AI question-answering service model based on artificial intelligence, wherein the model prompt words include model role prompt words, which are used to prompt the AI ​​question-answering service model to at least provide a post recommendation service for the user, wherein the AI ​​question-answering service model is obtained by fine-tuning a pre-trained language model using sample data in the target field to which the service posts belong; Under the prompt of the model role prompt word, identifying the service demand information of the user from the dialogue information; When the service demand information is identified, multiple candidate service posts are recalled for the user according to the service demand information; a target service post is selected from the multiple candidate service posts and the target service post is recommended to the user.

2. The method according to claim 1, characterized in that The service demand information includes a target service category and a target service object under the target service category, and identifying the service demand information of the user from the conversation information includes: Using the first AI tool to match the conversation information with known service categories, if a match is found, using the matching service category as a target service category; The conversation information is matched with each service object under the target service category by using a second AI tool, and if a match is found, the matching service object is used as the target service object.

3. The method according to claim 2, characterized in that Using the first AI tool to match the conversation information with known service categories, and if a match is found, using the matching service category as a target service category, including: Performing word segmentation processing on the conversation information to obtain a plurality of first word segmentations, and performing feature extraction on the plurality of first word segmentations and each known service category to obtain a plurality of first word segmentation features and a feature of each known service category; Calculate a first similarity between each first word segmentation feature and a feature of each known service category, and if there is a first target similarity greater than a first similarity threshold, use the service category corresponding to the first target similarity as a target service category; Accordingly, the conversation information is matched with each service object under the target service category by using the second AI tool, and if a match is found, the matching service object is used as the target service object, including: Extracting features of each service object under the target service category to obtain features of each service object; The second similarity between each first word segmentation feature and the feature of each service object is calculated, and if there is a second target similarity greater than a second similarity threshold, the service object corresponding to the second target similarity is used as the target service object.

4. The method according to claim 1, characterized in that: The service demand information includes a target service category and a target service object under the target service category, and a plurality of candidate service posts are recalled for the user according to the service demand information, including at least one of the following: Based on the target service category and the target service object, from the service posts whose service locations are in the business district near the user, service posts whose distances are less than a set first distance threshold from the user's location are recalled as candidate service posts; Based on the target service category and the target service object, recalling service posts whose user scores are greater than a set score threshold from service posts whose service locations are located in the target business district where the user is located as candidate service posts; Based on the target service category and the target service object, service posts in the target area where the user is located are recalled as candidate service posts; wherein the target area includes a business district near the user, and the business district near the user does not include the business district where the user is located.

5. The method according to claim 4, characterized in that Based on the target service category and the target service object, recalling service posts whose service locations are located in a business district near the user and whose distance to the user's location is less than a set first distance threshold as candidate service posts, including: Based on the target service category and the target service object, obtaining an initial service post belonging to the target service category and including the target service object from service posts whose service locations are in a business district near the user; According to the service location of the initial service post and the location of the user, the distance from the initial service post to the location of the user is calculated, and service posts whose distance is less than a set first distance threshold are selected from the initial service posts as candidate service posts.

6. The method according to claim 5, characterized in that Based on the target service category and the target service object, obtaining an initial service post belonging to the target service category and including the target service object from service posts whose service locations are in a business district near the user, including: Extracting features of the target service category and the target service object respectively to obtain features of the target service category and features of the target service object; Segmenting each service post whose service location is located in a business district near the user to obtain a plurality of second segmentations corresponding to each post, and performing feature extraction on the plurality of second segmentations corresponding to each post to obtain a plurality of second segmentation features; Calculate a third similarity between each post and the target service object under the target service category according to the plurality of second word segmentation features corresponding to each post, the feature of the target service category, and the feature of the target service object; From the service posts whose service locations are located in the business district near the user, a post whose third similarity is greater than a third similarity threshold is selected as an initial service post.

7. The method according to claim 1, characterized in that Selecting a target service post from the plurality of candidate service posts comprises: Using the ranking model to obtain the historical behavior characteristics of the user in the historical period, and calculating the matching degree between each candidate service post and the historical behavior characteristics as the first ranking score of each candidate service post; Calculate a second ranking score for each candidate service post according to the user score of each candidate service post and the distance from the service location of each candidate service post to the user; Calculating the semantic similarity between the title of each candidate service post and the target service object as the third ranking score of each candidate service post; Each candidate service post is ranked according to the first ranking score, the second ranking score and / or the third ranking score, and a target service post is selected from the candidate service posts according to the ranking order.

8. The method according to any one of claims 1 to 7, characterized in that: The model role prompt is also used to prompt the AI ​​question-and-answer service model to provide question-and-answer services to the user, and the model prompt also includes a service guidance prompt, which is used to prompt the AI ​​question-and-answer service model to provide guidance services to the user. The method also includes: Under the prompt of the model role prompt word, generate answer information for the dialogue information, and display the answer information in the conversation interface for the user to view; In the case where the service demand information is not identified, prompt information and / or service demand guidance information for providing post recommendation services is generated at the end of the answer information under the prompt of the service guidance prompt word, and the prompt information and / or service demand guidance information for providing post recommendation services is displayed in the conversation interface to guide the user to express service needs in the subsequent conversation.

9. The method according to claim 8, characterized in that The model prompt words also include: style constraint prompt words, which are used to constrain the number of words, format and / or style of the answer information and the service demand guidance information generated by the AI ​​question and answer service model; wherein the number of words in the service demand guidance information is more than the number of words in the answer information.

10. An electronic device, characterized in that: include: Memory and processor; The memory is used to store a computer program; the processor is coupled to the memory and is used to execute the computer program to implement the steps in the method according to any one of claims 1 to 9.

11. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is caused to implement the steps in the method according to any one of claims 1 to 9.

12. A computer program product, characterized in that The computer program product comprises a computer program / instruction, which, when executed by a processor, enables the processor to implement the steps of any one of the methods of claims 1 to 9.

Images