Product text display method and device

Abstract

The invention provides a product text display method and device, relates to the field of large model and machine learning, can be applied to the financial field and other fields, and comprises the following steps: collecting user side information and product side information through preset burying point information; identifying the product side information by using a large model to obtain product data, and generating corresponding weights according to a plurality of content types in the product data; analyzing through a preset machine learning model according to the user side information to obtain predicted browsing data, and assigning values to the weights according to the predicted browsing data to obtain user interests and preferences; and constructing a product display text according to the user interest preference and the product data. Customized presentation of the product specification is realized by utilizing machine learning and a large model technology, so that the time for investors to look up information is saved, the investment decision-making efficiency is improved, the customer satisfaction is increased, and the user experience is improved.

Description

Technical Field

[0001] This application relates to the fields of large models and machine learning, and can be applied to the financial field and other fields. In particular, it refers to a method and device for product text display. Background Art

[0002] Existing product manuals are presented exactly the same for all investors. However, different customers are interested in different information. The current presentation method makes it difficult for a large number of customers to quickly obtain the information they want, which in turn affects the decision-making of product transactions. In addition, when the same user browses an investment product and then browses the next product, they often need to repeatedly read the product manual, resulting in poor user experience and low efficiency. Summary of the Invention

[0003] The purpose of this application is to provide a method and device for product text display, so as to solve the problem that users cannot efficiently and targeted obtain product information through product manuals during the investment decision-making process.

[0004] To achieve the above purpose, the product text display method provided by this application includes: collecting user-side information and product-side information through preset buried-point information; using a large model to identify the product-side information to obtain product data, and generating corresponding weights according to multiple content types in the product data; analyzing the user-side information through a preset machine learning model to obtain predicted browsing data, and assigning values to the weights according to the predicted browsing data to obtain user interest preferences; constructing a product display text according to the user interest preferences and the product data.

[0005] In the above product text display method, optionally, constructing a product display text according to the user interest preferences and the product data further includes: collecting eye movement data of the user viewing the product display text, and generating feedback data according to the eye movement data; adjusting the machine learning model according to the feedback data.

[0006] In the above product text display method, optionally, using a large model to identify the product-side information to obtain product data further includes: comparing the product data with the historical product data in the historical data of the corresponding user to obtain difference data; constructing a product display text according to the difference data and the historical product data.

[0007] In the above product text display method, optionally, the user-side information includes: browsing data of the user in a preset scenario and user portrait data generated according to pre-stored user information; wherein, the user portrait data includes user identity attribute information and user historical transaction information; the product-side information includes the product information viewed by the user in the browsing data and the product description text data of the product corresponding to the product information.

[0008] In the above product text display method, optionally, using a large model to identify the product-side information to obtain product data includes: classifying and extracting keywords from the product description text data through the large model to obtain multiple semantic units; comparing the semantic units with a preset thesaurus to obtain valid semantic information, and converting the product description text data into multiple descriptive text segments according to the valid semantic information; obtaining product data according to the descriptive text segments.

[0009] In the above product text display method, optionally, generating corresponding weights according to multiple content types in the product data includes: extracting descriptive semantic units according to the content types of the valid semantic information, and calculating the corresponding weights through the text distance between the descriptive semantic units and other semantic units in the valid semantic information.

[0010] In the above product text display method, optionally, constructing a product display text according to the user interest preference and the product data includes: sorting the descriptive text segments according to the user interest preference to obtain a text display priority; combining the descriptive text segments according to the text display priority to generate a product display text.

[0011] This application also provides a product text display device, which includes a collection module, an identification module, an analysis module, and a generation module; the collection module is used to collect user-side information and product-side information through preset buried point information; the identification module is used to use a large model to identify the product-side information to obtain product data, and generate corresponding weights according to multiple content types in the product data; the analysis module is used to analyze and obtain predicted browsing data according to the user-side information through a preset machine learning model, and assign values to the weights according to the predicted browsing data to obtain user interest preferences; the generation module is used to construct a product display text according to the user interest preferences and the product data.

[0012] This application also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the above method is implemented.

[0013] This application also provides a computer-readable storage medium, which stores a computer program for executing the above method.

[0014] This application also provides a computer program product, including a computer program / instructions, and when the computer program / instructions are executed by a processor, the steps of the above method are implemented.

[0015] The beneficial technical effects of this application are as follows: By using machine learning and large model technologies, it realizes the customized presentation of product manuals, saves the time for investors to consult information, improves the efficiency of investment decisions, increases customer satisfaction, and enhances the user experience. Description of the Drawings

[0016] The drawings described herein are used to provide a further understanding of this application, form a part of this application, and do not limit this application. In the drawings:

[0017] Figure 1 It is a flowchart showing the method for presenting product text provided by an embodiment of this application;

[0018] Figure 2 It is a flowchart showing the process for obtaining product data provided by an embodiment of this application;

[0019] Figure 3 It is a flowchart showing the process for generating product display text provided by an embodiment of this application;

[0020] Figure 4 It is a flowchart showing the process for adjusting a machine learning model provided by an embodiment of this application;

[0021] Figure 5 It is a flowchart showing the process for generating product display text provided by an embodiment of this application;

[0022] Figure 6 It is a schematic structural diagram of a product text display device provided by an embodiment of this application;

[0023] Figure 7 It is a schematic structural diagram of an electronic device provided by an embodiment of this application. Detailed Embodiments

[0024] The following will describe in detail the embodiments of this application in conjunction with the drawings, so as to fully understand how this application uses technical means to solve technical problems and the implementation process of achieving technical effects and implement accordingly. It should be noted that as long as there is no conflict, the various embodiments in this application and the various features in each embodiment can be combined with each other, and the formed technical solutions are all within the protection scope of this application.

[0025] In addition, the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0026] The information collected in the technical solution of this application, namely user-side information and product-side information, is information and data authorized by the user or fully authorized by all parties. Moreover, for the processing of relevant data such as collection, storage, use, processing, transmission, provision, disclosure, and application, all comply with the relevant laws, regulations, and standards of relevant countries and regions, necessary confidentiality measures are taken, it does not violate public order and good customs, and corresponding operation entrances are provided for users to choose to authorize or reject.

[0027] The product text display method provided by this application provides corresponding operation entrances for users to choose to agree or reject the automated decision-making results; if the user chooses to reject, the expert decision-making process will be entered.

[0028] Please refer to Figure 1 As shown, for the product text display method provided by this application, the method includes:

[0029] S101 Collect user-side information and product-side information through preset buried-point information;

[0030] S102 Use a large model to identify the product-side information to obtain product data, and generate corresponding weights according to multiple content types in the product data;

[0031] S103 Analyze through a preset machine learning model according to the user-side information to obtain predicted browsing data, and assign values to the weights according to the predicted browsing data to obtain user interest preferences;

[0032] S104 Construct product display text according to the user interest preferences and the product data.

[0033] In this way, the customization presentation of product manuals is realized by using machine learning and large model technologies, saving the time for investors to consult information, improving the efficiency of investment decisions, increasing customer satisfaction, and enhancing the user experience. The specific implementation processes of each step will be described one by one in the subsequent steps and will not be elaborated here one by one.

[0034] In the above embodiment, the user-side information includes: the browsing data of the user in a preset scenario and the user portrait data generated according to the pre-stored user information; among them, the user portrait data includes the identity attribute information and the user's historical transaction information of the user; the product-side information includes the product information viewed by the user in the browsing data and the product description text data of the product corresponding to the product information.

[0035] Specifically, in actual work, in this embodiment, it mainly involves obtaining user-side information and product-side information; the system obtains buried-point information. It obtains the operations (such as clicks, swipes, returns, no operations, etc.) and page stay duration, page stay range, etc. when the customer usually browses investment products such as fund wealth management. At the same time, it obtains the customer portrait in the system. This data includes customer number, age, gender, occupation, location (whether it is a developed area), whether there is a loan or credit card repayment due, deposit balance, purchase history of investment products (such as whether the product is a stable type, whether it is a long-term holding product, purchase amount / share), whether it is a private banking customer, customer risk rating, customer questionnaire filling, etc. The system obtains the product information clicked by the customer and the original text information of the product manual; through large model technology, it identifies the obtained product manual information.

[0036] Please refer to Figure 2 As shown, in an embodiment of the present application, using a large model to identify the product-side information to obtain product data includes:

[0037] S201 Classify and extract keywords from the product description text data through a large model to obtain multiple semantic units;

[0038] S202 Compare the semantic units with a preset thesaurus to obtain effective semantic information, and convert the product description text data into multiple descriptive text blocks according to the effective semantic information;

[0039] S203 Obtain product data according to the descriptive text blocks.

[0040] Furthermore, generating corresponding weights according to multiple content types in the product data includes: extracting descriptive semantic units according to the content type of the effective semantic information, and calculating the corresponding weights through the text distance between the descriptive semantic units and other semantic units in the effective semantic information.

[0041] Specifically, in actual work, the main purpose of the above embodiments is to process the original text of the product manual. Using large model technology, the information obtained from the product manual is processed. First, the large model identifies the information obtained, which includes but is not limited to text information and graphic information, etc. The model automatically classifies and organizes the obtained information, extracts keywords, and regenerates a concise version of the text or an organized chart, and outputs the organized content to the user. In this embodiment, the large model can split the text data in the manual into the smallest semantic units, such as words and single characters, etc. And compare the split words with the thesaurus. The thesaurus can be divided into a noun thesaurus and an adjective thesaurus. The noun thesaurus can be further divided into an income thesaurus, a risk thesaurus, a market view thesaurus, etc. according to the content often shown in the product manual. For example, the field "The product has a high investment return" can be split into several semantic units such as "the", "product", "investment", "return", "high", etc. Words such as "product", "investment", and "return" are all nouns, but the word "product" is not substantially helpful for investment and is not included in the thesaurus, and the system does not process it. The words "investment" and "return" are often mentioned in the product manual, so the system compares these words with the thesaurus. The word "return" has a 95% chance of belonging to the product income module and a 5% probability of belonging to the risk and other modules in the historical product manuals that have been launched. The system will first judge it as a word in the yield part. The word "high" follows "return", and the model will calculate the weight according to the distance between the words. The closer the distance to the modified word, the higher the weight of the modifier modifying this noun. If punctuation marks such as commas and periods are encountered, the weight will be reduced accordingly. The weight reduction of a comma is less than that of a period. Therefore, the system compares "high" with the thesaurus and finds that it is an adjective, so it is an adjective of "return". In this way, when outputting information to the user, the field "Income: High return" can be input. Regarding the probability problem of the attribution section of the above-mentioned word "return", it can be judged based on logistic regression. The logistic regression formula is:

[0042]

[0043] In the above formula, P is the probability that the predicted value belongs to this module, and x i is the number of times this noun appears in this thesaurus in historical statistics. β i is the correlation coefficient, which is the weight given to this noun by the trained system. For example, the number of times a word appears in the yield thesaurus in history is x i times. According to the correlation coefficient β i, if the calculated p-value is greater than a certain threshold, it can be directly considered that this term belongs to this part; if it is less than a certain threshold, it can be judged whether it belongs to this part according to the context. If the context is greater than a certain threshold, it can be directly considered that this word belongs to a certain module; the thresholds for different words are different, and the thresholds are obtained through training, allowing the machine to judge by itself without human factors.

[0044] For example, if this text description is "The product has a high return on investment", because the word "investment" cannot be judged whether it belongs to the yield part, according to the semantic decomposition of the context mentioned above, because the word "return" is included in the yield thesaurus many times, and the p-value is 95%, it can be directly inferred that the word "investment" also belongs to the yield part vocabulary. When the system outputs, it can present the words "Yield: High return on investment" to consumers.

[0045] In an embodiment of the present application, when processing the user-side information in the above step S103, the obtained user information and buried point information can be used as the variable x 1 , x 2 ... Using the machine learning LightGBM algorithm, the collected information is input into the algorithm to predict the browsing content of the customer. The content refers to the refined information such as text images processed by the large model from the product manual that the customer wants to see; the input information includes, for example, the customer's age is greater than N years old, whether the customer's permanent residence belongs to a developed area, whether the customer's annual income is greater than X yuan, whether the customer has a credit card, whether the customer has a loan, the customer's risk assessment rating (R1-R5), the risk rating of the investment products historically purchased by the customer, etc.

[0046] The goal of LightGBM regression is to minimize the squared loss, that is, the square of the difference between the actual value and the predicted value. The loss function is

[0047]

[0048] Among them, y i is the actual value of the i-th sample, and F(x i ) is the predicted value of the model for the i-th sample. During the training of this model, the model needs to update the parameters to reduce the loss. Gradient boosting uses the gradient descent method to solve the negative gradient of the loss function, and then updates the model parameters. For the squared loss, the update rule can be expressed as:

[0049] F (t) (x i ) = F (t-1) (x i ) + η·h t (x i )

[0050] Among them, F (t) (xi ) is the predicted value of the model for sample x after the t-th iteration, η is the learning rate, and h i ; h t (x i ) is the negative gradient of the t-th tree for sample x i .

[0051] In the prediction process mentioned in this step, F(x i ) is the predicted value of the model for the reading content that the customer i expects to obtain. In the previous step when the large model processes the product instruction manual text image, weights w corresponding to each field (semantic unit) of the disassembled product instruction manual are generated; using the above model, through the input customer information x i , the content that the customer hopes to see is predicted, and the weight w is assigned; the higher the weight w, the higher the order of this part of the content in the display process.

[0052] Please refer to Figure 3 shown. In an embodiment of the present application, constructing a product display text according to the user interest preference and the product data includes:

[0053] S301 Sort the description text segments according to the user interest preference to obtain a text display priority;

[0054] S302 Combine the description text segments according to the text display priority to generate a product display text.

[0055] In actual work, this example mainly pushes the generated product instruction manual to the target customer group. Specifically, according to the obtained predicted content, the product instruction manual information processed by the large model is pushed to the customer according to the preference. According to the weight w of the semantic unit mentioned in step S132, the system will automatically judge the points that the user is concerned about and actively display this part of the content in front of the product instruction manual. For example, the product instruction manual of a product is divided into a risk part, a return part, a market view part, an investment target logic part, etc. Taking the return part as an example, this part shows the 1-year average return rate, 3-year average return rate, and the value by which the return part exceeds the market of this product. The large model divides the return part into various return rate segments mentioned above according to semantics, and automatically attaches an initial weight w to each segment. Using the LightGBM regression algorithm mentioned above, according to the user's previous usage habits, such as the user's recent investment search records are all short-term funds, and the user's concerns or historical investment records are mainly cash products. Based on this, the system can adjust the value of the negative gradient function h t (x i ), such as recording short-term funds as 1, medium- and long-term funds as -1; cash products as 1, asset management products as -1, etc., thereby affecting the predicted value F (t) (xi ) value. (t) (x i ) directly affects the value of w. In this case, because customers are more concerned about short-term investment, F (t) (x i ) has 2 more points (short-term +1, cash +1), so the weight w of the 1-year average return rate section is significantly higher than other sections, and it is displayed first when it is provided to users for preview. The rest of the sections are analogous to this model. The large model can classify this part according to semantic units, that is, the content in a section, such as short-term yield and long-term yield under the yield section, and can also be classified according to large sections, such as yield section, risk section, etc.

[0056] Please refer to Figure 4 As shown, in one embodiment of the present application, constructing a product display text according to the user interest preference and the product data further includes:

[0057] S401 collects eye movement data of a user viewing the product display text, and generates feedback data according to the eye movement data;

[0058] S402 adjusts the machine learning model according to the feedback data.

[0059] Specifically, in actual work, the above embodiment can be used to calibrate the push model; subsequent optimization can collect user behavior to calibrate the model; such as the customer's eye movement data when seeing the recommendation result: if the customer watches the first part of the recommended product manual for a long time, the data is recorded and a positive score feedback is output to the model, and the model will record the data; if the customer's eye movement data is statistically significantly shorter than that of other customers, a negative score feedback system is output and recorded in the negative gradient function h t (x i ); behaviors recorded as positive scores include but are not limited to eye movement duration, page dwell time, product manual segment dwell time (for example, in the product manual presentation method, the longer the user dwells on the content closer to the front, which means that the content is important to the user and the user pays more attention to it), placing an order after browsing the manual or selecting the "not interested" button; behaviors recorded as negative scores include but are not limited to clicking on the original text of the product manual to browse, the browsing time of the last part of the segmented product manual is statistically significantly longer than the browsing time of the first half, and the browsing time is significantly shorter than that of other users, etc., so as to achieve a more accurate presentation order.

[0060] Please refer to Figure 5 As shown, in one embodiment of the present application, using a large model to identify the product side information to obtain product data further includes:

[0061] S501 compares the product data with the historical product data in the historical data of the corresponding user by type to obtain difference data;

[0062] S502 constructs a product display text according to the difference data and the historical product data.

[0063] Specifically, when the same customer browses different products, the differences between two or more products can be compared. In addition to the above functions, the use of large models and machine learning can also be used to compare the differences between the product manuals of different products. In the foregoing embodiment, the large model has split the semantic units of the product manual of Product A. Store this part of the information. When browsing Product B, repeat the above steps. The large model can compare the information in the manuals of Product A and Product B in the thesaurus and organize and generate the differences for output to the customer.

[0064] In actual work, the product manual of Product A states that: a certain product mainly invests in short-term bonds with low risks, has a short duration and small fluctuations, taking into account both returns and liquidity. The product manual of Product B is: This product mainly invests in medium-term and short-term bonds, and creates a continuous and stable investment return under the premise of strictly controlling fluctuations. It does not participate in stock investment, bringing a stable investment experience to investors. When processing the above text, the large model splits the investment style paragraph into semantic units such as "investment", "risk", "low", "duration", "short", "fluctuation", "small", etc. When comparing Product A and Product B, the large model will go to the semantic library to match synonyms or the same words. For example, both have the same words such as "risk", "fluctuation", etc. and synonymous descriptions such as "return", "income", etc., and there are adjectives such as "small", "low", "stable", "stable" etc. after these nouns. Match the noun thesaurus with the adjective thesaurus, and then compare the two products to obtain the conclusion that both Product A and Product B have relatively low risks and stable returns. Output this conclusion to the investor for product comparison. Reduce the reading time of investors and improve efficiency.

[0065] Please refer to Figure 6 As shown, the present application also provides a product text display device, which includes a collection module, an identification module, an analysis module and a generation module; the collection module is used to collect user-side information and product-side information through preset buried point information; the identification module is used to use a large model to identify the product-side information to obtain product data, and generate corresponding weights according to multiple content types in the product data; the analysis module is used to analyze the user-side information through a preset machine learning model to obtain predicted browsing data, and assign values to the weights according to the predicted browsing data to obtain user interest preferences; the generation module is used to construct a product display text according to the user interest preferences and the product data. Since the principle of this device to solve problems is similar to that of the product text display method, the implementation of this device can refer to the implementation of the product text display method, and the repeated parts will not be elaborated.

[0066] It should be noted that the product text display method provided by the embodiments of the present invention can be used in the financial field, and can also be used in any technical field other than the financial field. The embodiments of the present invention do not limit the application field of the product text display method.

[0067] The beneficial technical effects of this application are as follows: Using machine learning and large model technologies to achieve customized presentation of product manuals, saving investors' time in consulting information, improving the efficiency of investment decisions, increasing customer satisfaction, and enhancing user experience.

[0068] This application also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the above method is implemented.

[0069] This application also provides a computer-readable storage medium, which stores a computer program for executing the above method.

[0070] This application also provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the steps of the above method are implemented.

[0071] As Figure 7 shown, the electronic device 600 may further include: a communication module 110, an input unit 120, an audio processor 130, a display 160, and a power supply 170. It should be noted that the electronic device 600 does not necessarily have to include Figure 7 all the components shown in Figure 7 ; in addition, the electronic device 600 may further include

[0072] As Figure 7 shown, the central processing unit 100 is sometimes also referred to as a controller or an operation control unit, and may include a microprocessor or other processor devices and / or logic devices. The central processing unit 100 receives inputs and controls the operations of the various components of the electronic device 600.

[0073] Among them, the memory 140, for example, may be one or more of a buffer, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory, or other suitable devices. The above information related to failures can be stored, and in addition, programs for executing relevant information can also be stored. And the central processing unit 100 can execute the program stored in the memory 140 to achieve information storage or processing, etc.

[0074] The input unit 120 provides an input to the central processing unit 100. The input unit 120 is, for example, a button or a touch input device. The power supply 170 is used to supply power to the electronic device 600. The display 160 is used to display display objects such as images and texts. The display may be, for example, an LCD display, but is not limited thereto.

[0075] The memory 140 may be a solid-state memory, for example, a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It may also be a memory that stores information even when power is off, can be selectively erased and has more data. Examples of such a memory are sometimes referred to as EPROMs, etc. The memory 140 may also be some other type of device. The memory 140 includes a buffer memory 141 (sometimes referred to as a buffer). The memory 140 may include an application / function storage unit 142 that is used to store application programs and function programs or the processes for operating the electronic device 600 through the central processing unit 100.

[0076] The memory 140 may also include a data storage unit (data 143) that is used to store data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit (driver 144) of the memory 140 may include various drivers for the communication functions of the electronic device and / or for performing other functions of the electronic device (such as a messaging application, an address book application, etc.).

[0077] The communication module 110 is a transmitter / receiver 110 that transmits and receives signals via the antenna 111. The communication module (transmitter / receiver) 110 is coupled to the central processing unit 100 to provide an input signal and receive an output signal, which may be the same as in the case of a conventional mobile communication terminal.

[0078] Based on different communication technologies, multiple communication modules 110 may be provided in the same electronic device, such as a cellular network module, a Bluetooth module, and / or a wireless local area network module, etc. The communication module (transmitter / receiver) 110 is also coupled to the speaker 131 and the microphone 132 via the audio processor 130 to provide an audio output via the speaker 131 and receive an audio input from the microphone 132, thereby implementing normal telecommunication functions. The audio processor 130 may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 130 is also coupled to the central processing unit 100, so that recording can be performed on the local machine through the microphone 132, and the sound stored on the local machine can be played through the speaker 131.

[0079] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0080] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or the combination of blocks.

[0081] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing devices to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or the combination of blocks.

[0082] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, such that a series of operation steps are executed on the computer or other programmable devices to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable devices provide steps for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or the combination of blocks.

[0083] The specific embodiments described above further elaborate on the purpose, technical solutions, and beneficial effects of the present application. It should be understood that the above are only specific embodiments of the present application and are not used to limit the protection scope of the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A product text display method, characterized in that: The method comprises: Collect user-side information and product-side information through preset tracking point information; Using a large model to identify the product side information to obtain product data, and generating corresponding weights according to multiple content types in the product data; According to the user side information, predicted browsing data is obtained by analyzing the preset machine learning model, and the weight is assigned according to the predicted browsing data to obtain the user interest preference; Construct product display text according to the user interest preference and the product data.

2. The product text display method according to claim 1, characterized in that: Constructing product display text according to the user's interest preference and the product data further includes: Collecting eye movement data of a user viewing the product display text, and generating feedback data according to the eye movement data; The machine learning model is adjusted according to the feedback data.

3. The product text display method according to claim 1, characterized in that: Using the big model to identify the product side information to obtain product data also includes: Comparing the product data with historical product data in the historical data of the corresponding user by type to obtain difference data; A product display text is constructed according to the difference data and the historical product data.

4. The product text display method according to claim 1, characterized in that: The user-side information includes: the user's browsing data in preset scenarios and user portrait data generated based on pre-stored user information; wherein, the user portrait data includes the user's identity attribute information and the user's historical transaction information; the product-side information includes the product information viewed by the user in the browsing data and the product description text data of the product corresponding to the product information.

5. The product text display method according to claim 4, characterized in that: Using the big model to identify the product side information to obtain product data includes: Classify the product description text data through a large model to extract keywords to obtain multiple semantic units; Comparing the semantic unit with a preset word library to obtain valid semantic information, and converting the product description text data into a plurality of description text sections according to the valid semantic information; Product data is obtained according to the description text panel.

6. The product text display method according to claim 5, characterized in that: Generating corresponding weights according to multiple content types in the product data includes: A description semantic unit is extracted according to the content type of the effective semantic information, and a corresponding weight is obtained by calculating the text distance between the description semantic unit and other semantic units in the effective semantic information.

7. The product text display method according to claim 5, characterized in that: Constructing product display text according to the user's interest preference and the product data includes: Sorting the description text sections according to the user's interest preference to obtain a text display priority; The product display text is generated by combining the description text sections according to the text display priority.

8. A product text display device, characterized in that: The device comprises a collection module, an identification module, an analysis module and a generation module; The collection module is used to collect user-side information and product-side information through preset embedding point information; The recognition module is used to use the large model to recognize the product side information to obtain product data, and generate corresponding weights according to multiple content types in the product data; The analysis module is used to obtain predicted browsing data through a preset machine learning model analysis based on the user-side information, and to assign weights based on the predicted browsing data to obtain user interest preferences; The generating module is used to construct product display text according to the user interest preference and the product data.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

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

Images