Live broadcast interaction method and device based on artificial intelligence, equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, and discloses a live broadcast interaction method, device and equipment based on artificial intelligence, and a storage medium, and the method comprises the steps: obtaining the historical transaction information of a live broadcast product in a live broadcast room; performing audience classification processing on the live product according to the historical transaction information to obtain audience attributes of the live product; receiving a user question for the live broadcast product in the live broadcast room; and through a pre-trained large language model, according to the user questions, the audience group attributes and the preset knowledge base of the live broadcast products, interaction prompting is carried out on the anchor in the live broadcast room. The method can be applied to an automobile financial service live broadcast application scene, and the response accuracy, timeliness and individuation level of the live broadcast scene can be ensured.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based live interactive method, device, equipment and storage medium. Background Art

[0002] In the financial sector, text classification models, such as intelligent customer service Q&A models and intelligent outbound Q&A models, are often used to automatically answer user questions about financial products by matching answers from a pre-set answer library. For example, in the auto finance business, users can automatically be provided with answers to frequently asked questions about auto finance products such as auto loans, leases, and financial leases.

[0003] However, since financial products often involve many complex and ever-changing details, such as the amount, term, repayment method, and interest rate discounts of auto loans, the above-mentioned text classification model, which relies on a preset answer library, is difficult to effectively respond to users' personalized questions about financial products in live broadcast scenarios that require real-time interaction with users, and cannot ensure the accuracy and timeliness of the answers. Summary of the Invention

[0004] The present invention provides a live interactive method, apparatus, computer equipment and storage medium based on artificial intelligence to solve the technical problem that a text classification model that relies on a preset answer library is difficult to ensure the personalization level, accuracy and timeliness of answers in a live broadcast scenario.

[0005] In a first aspect, a live interactive method based on artificial intelligence is provided, comprising:

[0006] Get the historical transaction information of live broadcast products in the live broadcast room;

[0007] Classify the audience group of the live broadcast product based on the historical transaction information to obtain the audience group attributes of the live broadcast product;

[0008] Receiving questions from users in the live broadcast room regarding the live broadcast product;

[0009] Through a pre-trained large language model, interactive prompts are given to the host of the live broadcast room based on the user's questions, the audience group attributes and the preset knowledge base of the live broadcast product.

[0010] In a second aspect, a live interactive device based on artificial intelligence is provided, comprising:

[0011] The information acquisition module is used to obtain the historical transaction information of the live broadcast products in the live broadcast room;

[0012] A group classification module is used to classify the audience groups of the live broadcast product based on the historical transaction information to obtain the audience group attributes of the live broadcast product;

[0013] A question receiving module, configured to receive questions from users in the live broadcast room regarding the live broadcast product;

[0014] The interactive prompt module is used to provide interactive prompts to the host of the live broadcast room based on the user's questions, the audience group attributes and the preset knowledge base of the live broadcast product through a pre-trained large language model.

[0015] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned artificial intelligence-based live interactive method are implemented.

[0016] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned artificial intelligence-based live interactive method are implemented.

[0017] In the solution implemented by the above-mentioned artificial intelligence-based live interactive method, device, computer equipment and storage medium, historical transaction information of live broadcast products in the live broadcast room is obtained; based on the historical transaction information, the live broadcast products are classified by audience group to obtain the audience group attributes of the live broadcast products; user questions about the live broadcast products in the live broadcast room are received; and interactive prompts are provided to the host of the live broadcast room based on the user questions, audience group attributes and the preset knowledge base of the live broadcast products through a pre-trained large language model. In the present invention, the audience group attributes of the live broadcast products are identified based on the historical transaction information, so that the large language model can be used to quickly, flexibly and intelligently provide the host of the live broadcast room with high-quality real-time interactive prompts that fit the user's interests and needs based on the user questions, audience group attributes and the preset knowledge base of the live broadcast products. This can improve the professionalism and pertinence of the live broadcast interaction, optimize the live broadcast interaction effect, and make the live broadcast interaction more accurate, smooth, natural and vivid, thereby ensuring the accuracy, timeliness and personalization of the response in the live broadcast scene, thereby enhancing the user's sense of participation and stickiness, and further improving user satisfaction and live broadcast conversion rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0019] Figure 1 1 is a schematic diagram of an application environment of a live interactive method based on artificial intelligence in one embodiment of the present invention;

[0020] Figure 2 This is a flow chart of a live interactive method based on artificial intelligence in one embodiment of the present invention;

[0021] Figure 3 yes Figure 2 A schematic flow chart of a specific implementation of step S40;

[0022] Figure 4 1 is a structural diagram of a live interactive device based on artificial intelligence in one embodiment of the present invention;

[0023] Figure 5 is a structural diagram of a computer device in one embodiment of the present invention;

[0024] Figure 6 FIG. 2 is another structural diagram of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0026] The live interactive method based on artificial intelligence provided by the embodiment of the present invention can be applied in Figure 1In an application environment, a client communicates with a server via a network. The server can obtain historical transaction information for live broadcast products in the live broadcast room through the client; based on this historical transaction information, it classifies the live broadcast products by audience group to obtain the audience attributes of the live broadcast products; receive user questions about the live broadcast products in the live broadcast room; and, using a pre-trained large language model, provide interactive prompts to the host in the live broadcast room based on user questions, audience attributes, and a pre-set knowledge base of the live broadcast products. In this way, the audience attributes of the live broadcast products are identified based on historical transaction information. The large language model, based on user questions, audience attributes, and a pre-set knowledge base of the live broadcast products, can quickly, flexibly, and intelligently provide the host with high-quality real-time interactive prompts tailored to user interests and needs. This can enhance the professionalism and pertinence of live broadcast interactions, optimize the interactive effects of live broadcasts, and make them more precise, smooth, natural, and vivid. This ensures the accuracy, timeliness, and personalization of responses in live broadcast scenarios, thereby enhancing user engagement and stickiness, and thereby improving user satisfaction and live broadcast conversion rates. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented as a standalone server or a server cluster consisting of multiple servers. The present invention is described in detail below through specific examples.

[0027] It should be noted that in various specific embodiments of the present invention, when it comes to the need to perform relevant processing based on data related to the user's identity or characteristics, such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of such data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present invention needs to obtain the user's sensitive personal information, it will obtain the user's separate permission or consent through a pop-up window or jump to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present invention will be obtained.

[0028] See also Figure 2 As shown, Figure 2 A flowchart of a live interactive method based on artificial intelligence provided by an embodiment of the present invention includes the following steps:

[0029] S10: Obtain historical transaction information of live broadcast products in the live broadcast room.

[0030] The artificial intelligence-based live interactive method provided by the present invention can be applied to live broadcast scenarios of financial services. Based on artificial intelligence and natural language processing technologies, it can provide high-quality interactive prompts to the host in response to user questions about financial products in the live broadcast room of financial services, ensuring the accuracy, timeliness, and personalization of responses in the live broadcast scene, thereby promoting the promotion and sales of financial products. For example, it can be applied to live broadcast scenarios of automobile finance services. In the live broadcast room of automobile finance services, it can generate high-quality automobile marketing response scripts in response to user questions about automobile financial products such as auto loans, leases, or financial leases, provide interactive prompts to the host, ensuring the accuracy, timeliness, and personalization of responses in the live broadcast scene of automobile finance services, increasing user participation, and promoting the sales conversion of automobile financial products.

[0031] For ease of understanding, several terms involved in the present invention are first explained:

[0032] Artificial intelligence (AI) is a new technical discipline that studies and develops theories, methods, technologies, and application systems for simulating, extending, and expanding human intelligence. A branch of computer science, AI seeks to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. AI can simulate the information processes of human consciousness and thinking. It also encompasses the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results.

[0033] Natural language processing (NLP): NLP uses computers to process, understand, and apply human languages ​​(such as Chinese and English). A branch of artificial intelligence, NLP is an interdisciplinary field between computer science and linguistics, often referred to as computational linguistics. Natural language processing encompasses grammatical analysis, semantic analysis, and discourse comprehension. Natural language processing is commonly used in technical fields such as machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, information intent recognition, information extraction and filtering, text classification and clustering, public opinion analysis, and opinion mining. It encompasses data mining, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research related to language processing, and linguistics research related to language computing.

[0034] Large language models are natural language processing models based on deep learning technology, designed to understand and generate natural language. By training on massive amounts of text data, they can identify patterns, structures, and semantics in language, enabling them to generate coherent and meaningful language output in a variety of tasks, such as text generation, automatic translation, sentiment analysis, and question answering. Large language models typically use the Transformer architecture and have strong contextual understanding capabilities. For example, the GPT series (such as GPT-4) and the DeepSeek series (such as DeepSeek-V3) are typical large language models that perform well in various natural language processing tasks and can handle complex language understanding and generation tasks.

[0035] Pre-set knowledge base: Contains knowledge documents related to live streaming products. These documents cover detailed information about live streaming products, such as brand positioning information (e.g., selling points, brand introduction, brand tone, and brand characteristics of auto finance products), market positioning information (e.g., the primary consumer groups targeted by auto finance products), terms of service, ordering processes, promotional information, marketing content (e.g., advertising copy and marketing materials that may be of interest to different audience groups), and answers to frequently asked questions. It is understood that the pre-set knowledge base is highly scalable and maintainable, and can adapt to changes in live streaming products and update their detailed information.

[0036] Based on this, the artificial intelligence-based live interactive method provided by the present invention is described in detail below.

[0037] The server can obtain historical transaction information of live broadcast products in the live broadcast room.

[0038] Among them, live broadcast products include auto finance products such as auto loans, leases or financial leases.

[0039] Historical transaction information includes but is not limited to: product type (for example, specific auto finance products, such as auto loans, auto leases, financial leases, etc.), number of transactions (for example, the number of auto finance products successfully ordered during the live broadcast, such as the number of loan contracts and lease contracts concluded through the live broadcast, etc.), transaction amount (for example, the total amount related to auto finance products, such as the total loan amount, the total lease amount, the total financial lease amount, etc.), basic information of the transaction user (such as the transaction user's age, gender, occupation, income level, region, etc.), transaction time, interactive behavior information of the transaction user (such as the length of live broadcast viewing, the frequency of likes or comments, the time of entering the live broadcast room), or feedback information of the transaction user (such as the transaction user's satisfaction with the ordered auto finance products), etc.

[0040] S20: Classify the audience groups of the live broadcast products based on the historical transaction information to obtain the audience group attributes of the live broadcast products.

[0041] Based on historical transaction information, the audience of live broadcast products can be classified and processed, and the audience attributes of live broadcast products can be quickly and accurately classified, thereby providing targeted strategic support for the subsequent real-time response script generation.

[0042] In step S20 of some embodiments, feature extraction processing may be performed on historical transaction information to obtain key transaction features; based on the key transaction features, audience group attribute identification processing may be performed on the live broadcast product to obtain audience group attributes.

[0043] In step S20, the historical transaction information is first subjected to feature extraction to obtain key transaction features. These key transaction features are key features that effectively reflect the transaction decision of the transaction user, facilitating identification of the audience attributes of the live broadcast product.

[0044] For example, feature extraction is performed on the basic information of the transaction user to obtain the basic features of the transaction user; feature extraction is performed on the interactive behavior information of the transaction user to obtain the behavioral characteristics of the transaction user; feature extraction is performed on the product type, transaction quantity, transaction amount, transaction time or feedback information of the transaction user to obtain the demand characteristics of the transaction user; the basic characteristics of the transaction user, the behavioral characteristics of the transaction user and the demand characteristics of the transaction user are fused (such as splicing) to obtain the key features of the transaction.

[0045] This allows for quick and accurate identification of audience attributes based on key transaction characteristics, such as business owners, young car buyers, families, or new energy and environmentally friendly vehicle owners.

[0046] For example, a clustering algorithm (such as K-means) or a decision tree algorithm can be used to identify and process transaction users based on key transaction features to obtain audience group attributes.

[0047] Specifically, clustering algorithms (such as K-means) can be used to classify and label the transaction users based on the key transaction features to obtain the labels of the transaction users, or a decision tree algorithm can be used to predict the labels of the transaction users of the live broadcast product based on the key transaction features, thereby using the labels of the transaction users as audience group attributes.

[0048] Therefore, by identifying the audience attributes of live broadcast products, we can accurately locate which group prefers and is more interested in live broadcast products, which helps to clarify which documents in the preset knowledge base fit the needs and interests of the audience, thereby providing a targeted reference basis for the subsequent real-time response script generation.

[0049] S30: Receive questions from users regarding the live broadcast product in the live broadcast room.

[0050] Receive real-time user questions about the live broadcast product in the live broadcast room. For example, a user might ask, "I have some questions about the financing lease plan. Can you tell me if I can choose to purchase this car after the lease expires?"

[0051] S40: Through a pre-trained large language model, interactive prompts are provided to the host in the live broadcast room based on user questions, audience attributes, and the preset knowledge base of the live broadcast product.

[0052] User questions and audience attributes are input into a pre-trained large language model. Based on the user questions and audience attributes, the pre-trained large language model, with the help of the preset knowledge base of the live broadcast product, intelligently provides interactive prompts to the anchor in the live broadcast room, helping the anchor to respond to user questions quickly and accurately, thereby improving the interactive experience and conversion rate of the live broadcast room.

[0053] See also Figure 2 In some embodiments, step S40 may include but is not limited to the following steps:

[0054] S41: Classify the user's question using a large language model to obtain the question type of the user's question.

[0055] S42: Using a large language model, knowledge retrieval is performed from a preset knowledge base based on audience attributes to obtain knowledge documents that match the audience attributes.

[0056] S43: Using a large language model, a response script is generated based on the question type and knowledge document to obtain a response script.

[0057] S44: Provide interactive prompts to the host based on the reply words.

[0058] For steps S41-S44, the user's questions can first be classified through a large language model to obtain the question type of the user's question. For example, the user's question "I have some questions about the financial leasing plan. Can you tell me if I can choose to buy this car after the lease expires?" is classified, and the question types obtained are, for example, questions about the financial leasing plan or questions about options after the lease expires.

[0059] Then, using the large language model, based on audience attributes, the pre-set knowledge base quickly retrieves knowledge documents that best match the audience's needs or interests. For example, if the audience attributes are business owners purchasing cars, the knowledge documents that match this audience attribute are those related to low capital utilization. This pre-set knowledge base allows the large language model to be provided with appropriate reference information in a timely manner, enhancing the professionalism of subsequent interactions.

[0060] The large language model is then used to generate and process responses based on the question type and knowledge documents, resulting in personalized responses, such as responses ① or ②:

[0061] Response 1: "With our financing lease program, you can spread the cost of the vehicle over a longer period of time, paying a fixed monthly lease payment. This helps you better plan your finances and ensure your company's liquidity. Furthermore, a portion of the lease fee can be deducted as an operating expense, further reducing your tax burden. This not only meets your business needs but also optimizes your financial structure. Why not?"

[0062] Response Strategies ②: "Everyone knows how crucial cash flow is to our business. By choosing our financing lease plan, you can not only use your new car immediately, but also spread the purchase price over the next few months, paying a fixed monthly rent, giving you more flexibility. Furthermore, these rent payments can be counted as operating expenses, allowing for tax deductions upfront, saving you considerable money. This way, you can meet your car needs while also improving your financial statements."

[0063] Finally, the reply script will be provided to the anchor, providing interactive suggestions for the anchor.

[0064] In this way, with the support of a large language model and a preset knowledge base, it is possible to respond to user questions in real time and provide real-time interactive prompts, which can help the anchor respond to user questions more efficiently, improve the professionalism and relevance of the answers, and enhance the anchor's expressiveness, thereby improving the live broadcast response effect in the live broadcast room, avoiding awkward silences or inaccurate answers, and helping to enhance the interaction between the anchor and the audience, and improving user participation and stickiness.

[0065] It can be understood that, taking the large language model GPT-4 as an example, the process of pre-training an untrained large language model to obtain a trained large language model includes: obtaining historical transaction information samples and historical user question samples of historical live broadcast products in historical live broadcast rooms, as well as the actual reply words of the anchors in historical live broadcast rooms; identifying audience attribute samples based on historical transaction information samples; inputting audience attribute samples, historical user question samples and actual reply words into GPT-4 for iterative training. During the training process, GPT-4 takes the actual reply words as the target, uses the preset knowledge base to generate predicted reply words, and calculates the similarity between the actual reply words and the predicted reply words, and adjusts the parameters of GPT-4 according to the similarity to optimize the performance of GPT-4 until the similarity is greater than or equal to the preset similarity threshold, stops optimizing GPT-4, and obtains a trained large language model that meets the requirements.

[0066] In some embodiments, the large language model can be optimized and trained based on user questions, audience attributes, and response scripts.

[0067] It is also possible to continuously optimize the trained large language model based on audience attributes, user questions, and response language to ensure that the model can adjust and optimize its response strategy based on actual feedback, continuously enhance the model's adaptability and learning ability, and improve the model's long-term performance.

[0068] In step S41 of some embodiments, the user's question can be processed for intent recognition through a large language model to obtain the intention of the user's question; and the user's question can be classified according to the intention through the large language model to obtain the question type.

[0069] In step S41, the user's question is input into the large language model, which performs intent recognition processing on the user's question to obtain the intention of the user's question and accurately understand the user's question. In order to quickly identify the type of question asked by the user based on the intention of the user's question, the interactive prompt can be customized according to the substantive content of the user's question. The question types include inquiries about high interest rates, confirmation of mortgage methods, inquiries about the car financing and leasing process, expensive vehicle pricing, questions about the city where the car is to be picked up, direct rejection or lack of interest, etc.

[0070] In step S42 of some embodiments, the audience group attributes can be encoded using a large language model to obtain a query vector describing the audience group attributes; and similarity retrieval processing is performed in a preset knowledge base based on the query vector using the large language model to obtain knowledge documents.

[0071] In step S42, the audience group attributes are input into the large language model, and the large language model encodes the audience group attributes, thereby converting the audience group attributes into a query vector that describes the audience group attributes. The large language model also embeds the knowledge documents in the preset knowledge base, thereby converting the knowledge documents into knowledge document vectors. In this way, the large language model can use the ElasticSearch method or the approximate nearest neighbor (ANN) search method to perform similarity retrieval in the preset knowledge base. Specifically, the query vector and the knowledge document vector can be cosine similarity calculated or Euclidean distance calculated. The knowledge document corresponding to the knowledge document vector with a higher cosine similarity or a smaller Euclidean distance with the query vector is more in line with the requirements of the audience group attributes. Therefore, the knowledge document vector whose cosine similarity with the query vector is greater than the cosine similarity threshold or the Euclidean distance is less than the Euclidean distance threshold can be retrieved, thereby obtaining the knowledge document that meets the audience group attributes.

[0072] Through the above method, it can be ensured that the retrieved knowledge documents can provide reference information that is more in line with the needs of the audience for interactive prompts.

[0073] In step S43 of some embodiments, the question type and the knowledge document can be fused through a large language model to obtain context information; the context information can be encoded through a large language model to obtain a context vector; and the context vector can be autoregressively processed through the large language model according to the attention mechanism to generate a reply script to obtain a reply script.

[0074] In step S43, the question type and the knowledge document are fused through a large language model to obtain context information, and then the context information is encoded to obtain a context vector. Finally, an attention mechanism is used to capture the global dependency relationship between different word vectors in the context vector to obtain a global dependency feature. Based on the global dependency feature, a reply is generated through autoregressive generation.

[0075] In this way, the large language model not only relies on its own training knowledge, but can also effectively utilize the information in the knowledge documents to effectively integrate the semantics of the substantive content of the user's question and the information in the knowledge documents, thereby generating more accurate response words.

[0076] The AI-based live interaction method provided in the above embodiment obtains historical transaction information of live broadcast products in the live broadcast room; classifies the live broadcast products according to the historical transaction information to obtain the audience attributes of the live broadcast products; receives user questions about the live broadcast products in the live broadcast room; and uses a pre-trained large language model to provide interactive prompts to the host of the live broadcast room based on the user questions, audience attributes, and a preset knowledge base of the live broadcast products. In this way, the audience attributes of the live broadcast products are identified based on the historical transaction information, so that the large language model can be used to quickly, flexibly, and intelligently provide the host of the live broadcast room with high-quality real-time interactive prompts that are tailored to the user's interests and needs based on the user questions, audience attributes, and the preset knowledge base of the live broadcast products. This can improve the professionalism and pertinence of live broadcast interactions, optimize the live broadcast interaction effect, and make the live broadcast interactions more accurate, smooth, natural, and vivid. This ensures the accuracy, timeliness, and personalization of responses in the live broadcast scene, thereby enhancing user participation and stickiness, and further improving user satisfaction and live broadcast conversion rate.

[0077] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0078] In one embodiment, a live interactive device based on artificial intelligence is provided, and the live interactive device based on artificial intelligence corresponds to the live interactive method based on artificial intelligence in the above embodiment. Figure 4 As shown, the live interactive device based on artificial intelligence includes an information acquisition module 101, a group classification module 102, a question receiving module 103 and an interactive prompt module 104. The functional modules are described in detail as follows:

[0079] Information acquisition module 101, used to obtain historical transaction information of live broadcast products in the live broadcast room;

[0080] The group classification module 102 is used to classify the audience groups of the live broadcast product based on the historical transaction information to obtain the audience group attributes of the live broadcast product;

[0081] A question receiving module 103 is used to receive questions from users in the live broadcast room regarding the live broadcast product;

[0082] The interactive prompt module 104 is used to provide interactive prompts to the host of the live broadcast room based on the user's questions, the audience group attributes and the preset knowledge base of the live broadcast product through a pre-trained large language model.

[0083] In one embodiment, the interactive prompt module 104 is specifically configured to:

[0084] Classifying the user's question using the large language model to obtain the question type of the user's question;

[0085] Using the large language model, performing knowledge retrieval processing from the preset knowledge base according to the attributes of the audience group, and obtaining knowledge documents that meet the attributes of the audience group;

[0086] Using the large language model, generating a reply script based on the question type and the knowledge document to obtain a reply script;

[0087] Provide interactive prompts to the host based on the reply words.

[0088] In one embodiment, the interactive prompt module 104 is further configured to:

[0089] Performing intent recognition processing on the user's question using the large language model to obtain the user's intent in asking the question;

[0090] The user's question is classified and processed according to the intention by using the large language model to obtain the question type.

[0091] In one embodiment, the interactive prompt module 104 is further configured to:

[0092] encoding the audience group attributes using the large language model to obtain a query vector describing the audience group attributes;

[0093] By using the large language model, similarity retrieval processing is performed in the preset knowledge base according to the query vector to obtain the knowledge document.

[0094] In one embodiment, the interactive prompt module 104 is further configured to:

[0095] By using the large language model, the question type and the knowledge document are fused to obtain context information;

[0096] Encoding the context information using the large language model to obtain a context vector;

[0097] The large language model is used to perform autoregressive response speech generation processing on the context vector according to the attention mechanism to obtain the response speech.

[0098] In one embodiment, the group classification module 102 is specifically configured to:

[0099] Performing feature extraction processing on the historical transaction information to obtain key transaction features;

[0100] Based on the key transaction features, the audience group attributes of the live broadcast product are identified to obtain the audience group attributes.

[0101] In one embodiment, the artificial intelligence-based live interactive device further includes an optimization training module for:

[0102] The large language model is optimized and trained based on the user questions, the audience group attributes and the reply words.

[0103] The present invention provides a live broadcast interactive device based on artificial intelligence, which identifies the audience group attributes of live broadcast products based on historical transaction information, so that through a large language model, based on user questions, audience group attributes and a preset knowledge base of live broadcast products, high-quality real-time interactive prompts that fit the user's interests and needs can be provided to the anchor in the live broadcast room quickly, flexibly and intelligently. This can improve the professionalism and pertinence of live broadcast interaction, optimize the live broadcast interaction effect, and make the live broadcast interaction more accurate, smooth, natural and vivid, thereby ensuring the accuracy, timeliness and personalization of responses in live broadcast scenes, thereby enhancing user participation and stickiness, and further improving user satisfaction and live broadcast conversion rate.

[0104] For the specific definition of the live interactive device based on artificial intelligence, please refer to the definition of the live interactive method based on artificial intelligence above, which will not be repeated here. The various modules in the above-mentioned live interactive device based on artificial intelligence can be implemented in whole or in part by software, hardware, or a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0105] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client via a network connection. When the computer program is executed by the processor, it realizes the functions or steps on the server side of a live interactive method based on artificial intelligence.

[0106] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Figure 6 As shown. The computer device includes a processor, memory, network interface, display screen and input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps on the client side of a live interactive method based on artificial intelligence.

[0107] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are performed:

[0108] Get the historical transaction information of live broadcast products in the live broadcast room;

[0109] Classify the audience group of the live broadcast product based on the historical transaction information to obtain the audience group attributes of the live broadcast product;

[0110] Receiving questions from users in the live broadcast room regarding the live broadcast product;

[0111] Through a pre-trained large language model, interactive prompts are given to the host of the live broadcast room based on the user's questions, the audience group attributes and the preset knowledge base of the live broadcast product.

[0112] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0113] Get the historical transaction information of live broadcast products in the live broadcast room;

[0114] Classify the audience group of the live broadcast product based on the historical transaction information to obtain the audience group attributes of the live broadcast product;

[0115] Receiving questions from users in the live broadcast room regarding the live broadcast product;

[0116] Through a pre-trained large language model, interactive prompts are given to the host of the live broadcast room based on the user's questions, the audience group attributes and the preset knowledge base of the live broadcast product.

[0117] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can be found in the relevant descriptions of the server side and the client side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0118] Those skilled in the art will appreciate that all or part of the processes in the above-described embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described embodiments. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAM bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM).

[0119] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0120] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A live interactive method based on artificial intelligence, characterized in that: include: Get the historical transaction information of live broadcast products in the live broadcast room; Classify the audience group of the live broadcast product based on the historical transaction information to obtain the audience group attributes of the live broadcast product; Receiving questions from users in the live broadcast room regarding the live broadcast product; Through a pre-trained large language model, interactive prompts are given to the host of the live broadcast room based on the user's questions, the audience group attributes and the preset knowledge base of the live broadcast product.

2. The live interactive method according to claim 1, wherein: The pre-trained large language model is used to provide interactive prompts to the host of the live broadcast room based on the user's question, the audience group attributes, and the preset knowledge base of the live broadcast product, including: Classifying the user's question using the large language model to obtain the question type of the user's question; Using the large language model, performing knowledge retrieval processing from the preset knowledge base according to the attributes of the audience group, and obtaining knowledge documents that meet the attributes of the audience group; Using the large language model, generating a reply script based on the question type and the knowledge document to obtain a reply script; Provide interactive prompts to the host based on the reply words.

3. The live interactive method according to claim 2, wherein: The large language model is used to perform question classification processing on the user's question to obtain the question type of the user's question, including: Performing intent recognition processing on the user's question using the large language model to obtain the user's intent in asking the question; The user's question is classified and processed according to the intention by using the large language model to obtain the question type.

4. The live interactive method according to claim 2, wherein: The method of performing knowledge retrieval processing from the preset knowledge base based on the attributes of the audience group using the large language model to obtain knowledge documents that meet the attributes of the audience group includes: encoding the audience group attributes using the large language model to obtain a query vector describing the audience group attributes; By using the large language model, similarity retrieval processing is performed in the preset knowledge base according to the query vector to obtain the knowledge document.

5. The live interactive method according to claim 2, wherein: The large language model is used to generate a reply speech based on the question type and the knowledge document to obtain the reply speech, including: By using the large language model, the question type and the knowledge document are fused to obtain context information; Encoding the context information using the large language model to obtain a context vector; The large language model is used to perform autoregressive response speech generation processing on the context vector according to the attention mechanism to obtain the response speech.

6. The live interactive method according to claim 1, wherein: The audience group classification processing of the live broadcast product is performed based on the historical transaction information to obtain the audience group attributes of the live broadcast product, including: Performing feature extraction processing on the historical transaction information to obtain key transaction features; Based on the key transaction features, the audience group attributes of the live broadcast product are identified to obtain the audience group attributes.

7. The live interactive method according to claim 2, wherein: After generating a reply speech based on the question type and the knowledge document using the large language model and obtaining the reply speech, the method further includes: The large language model is optimized and trained based on the user questions, the audience group attributes and the reply words.

8. A live interactive device based on artificial intelligence, characterized in that: include: The information acquisition module is used to obtain the historical transaction information of the live broadcast products in the live broadcast room; A group classification module is used to classify the audience groups of the live broadcast product based on the historical transaction information to obtain the audience group attributes of the live broadcast product; A question receiving module, configured to receive questions from users in the live broadcast room regarding the live broadcast product; The interactive prompt module is used to provide interactive prompts to the host of the live broadcast room based on the user's questions, the audience group attributes and the preset knowledge base of the live broadcast product through a pre-trained large language model.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the live interactive method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the live interactive method according to any one of claims 1 to 7 are implemented.

Images