Product text generation method, device, computer equipment and storage medium

Abstract

The present application proposes a method, device, computer equipment and storage medium for generating a product text, the method comprising obtaining the object characteristics of a target object and initial description information related to the target object; generating an opening description text and a preferential description text related to the target object according to the initial description information; generating a content description text corresponding to the target object according to the object characteristics; generating a target product text corresponding to the target object according to the opening description text, the content description text, and the preferential description text. Through the present application, the automation of the generation of introductory product texts can be effectively realized, the batch generation of introductory product texts can be realized, the consumption of human resources can be reduced, the generation efficiency and quality of introductory product texts can be improved, and the generation effect of introductory product texts can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, device, computer equipment and storage medium for generating commodity text. Background Art

[0002] In recent years, with the rapid development of e-commerce platforms and the live video streaming industry, live streaming e-commerce business that combines the characteristics of the two has gradually become a mainstream online shopping model. The form of live streaming anchors has also changed rapidly. Not only do online celebrities and artists with their own traffic bring goods through interaction with the audience and fan effect, but there are also stores that bring goods through live streaming mainly by displaying goods. And with the rapid development of artificial intelligence, virtual idols and virtual anchors completely driven by artificial intelligence have also joined the ranks of bringing goods.

[0003] Usually, in the product introduction session, real anchors can fully introduce the selling points and preferential information of the products to promote the audience's willingness to buy. However, due to the limitations of artificial intelligence virtual anchors, the current process of introducing products mainly relies on intelligent speech synthesis technology to convert the product introduction text provided by the merchant in advance into voice reading.

[0004] In this way, the product introduction texts in virtual live broadcasts are mainly written manually, which is time-consuming and of varying quality. In addition, the product introduction text writing process cannot be carried out in batches, resulting in excessive manpower costs in the process of writing the product introduction texts, causing a waste of resources. Summary of the invention

[0005] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0006] To this end, the purpose of this application is to propose a method, device, computer equipment and storage medium for generating product texts, which can effectively realize the automation of the generation of product introduction texts, realize the batch generation of product introduction texts, reduce human resource consumption, improve the generation efficiency and quality of product introduction texts, and improve the generation effect of product introduction texts.

[0007] To achieve the above-mentioned purpose, the method for generating product text proposed in the first aspect embodiment of the present application includes: obtaining object characteristics of a target object and initial description information related to the target object; generating an opening description text and a preferential description text related to the target object based on the initial description information; generating a content description text corresponding to the target object based on the object characteristics; and generating a target product text corresponding to the target object based on the opening description text, the content description text, and the preferential description text.

[0008] The method for generating product text proposed in the first aspect of the embodiment of the present application obtains the object characteristics of the target object and the initial description information related to the target object, and generates an opening description text and a preferential description text related to the target object according to the initial description information, generates a content description text corresponding to the target object according to the object characteristics, and generates a target product text corresponding to the target object according to the opening description text, the content description text, and the preferential description text. This method can effectively realize the automation of the generation of introductory product texts, realize the batch generation of introductory product texts, reduce human resource consumption, improve the generation efficiency and quality of introductory product texts, and improve the generation effect of introductory product texts.

[0009] To achieve the above-mentioned purpose, the device for generating product text proposed in the second aspect embodiment of the present application includes: an acquisition module, which is used to acquire the object characteristics of a target object and initial description information related to the target object; a first generation module, which is used to generate an opening description text and a preferential description text related to the target object according to the initial description information; a second generation module, which is used to generate a content description text corresponding to the target object according to the object characteristics; and a third generation module, which is used to generate a target product text corresponding to the target object according to the opening description text, the content description text, and the preferential description text.

[0010] The device for generating product text proposed in the second aspect of the embodiment of the present application obtains the object characteristics of the target object and the initial description information related to the target object, and generates an opening description text and a preferential description text related to the target object according to the initial description information, generates a content description text corresponding to the target object according to the object characteristics, and generates a target product text corresponding to the target object according to the opening description text, the content description text, and the preferential description text. This can effectively realize the automation of the generation of introductory product texts, realize the batch generation of introductory product texts, reduce human resource consumption, improve the generation efficiency and quality of introductory product texts, and improve the generation effect of introductory product texts.

[0011] The third aspect embodiment of the present application proposes a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method for generating product text proposed in the first aspect embodiment of the present application is implemented.

[0012] The fourth aspect embodiment of the present application proposes a non-temporary computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for generating product text as proposed in the first aspect embodiment of the present application.

[0013] The fifth aspect embodiment of the present application proposes a computer program product. When the instruction processor in the computer program product executes, it executes the method for generating product text as proposed in the first aspect embodiment of the present application.

[0014] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0016] Figure 1 It is a flowchart of a method for generating a product text proposed in an embodiment of the present application;

[0017] Figure 2 is a flowchart of a method for generating a product text proposed in another embodiment of the present application;

[0018] Figure 3 is a flowchart of a method for generating a product text proposed in another embodiment of the present application;

[0019] Figure 4 It is a schematic diagram of the architecture of the artificial intelligence model in the embodiment of the present application;

[0020] Figure 5 It is a structural schematic diagram of a device for generating a product text proposed in an embodiment of the present application;

[0021] Figure 6 is a structural schematic diagram of a device for generating a product text proposed in another embodiment of the present application;

[0022] Figure 7 A block diagram of an exemplary computer device suitable for implementing embodiments of the present application is shown. DETAILED DESCRIPTION

[0023] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be construed as limiting the present application. On the contrary, the embodiments of the present application include all changes, modifications and equivalents that fall within the spirit and connotation of the appended claims.

[0024] Figure 1 It is a flowchart of a method for generating product text proposed in an embodiment of the present application.

[0025] It should be noted that the executor of the method for generating product text in this embodiment is a device for generating product text, which can be implemented by software and / or hardware, and can be configured in an electronic device, which may include but is not limited to a terminal, a server, etc.

[0026] like Figure 1 As shown, the method for generating the product text includes:

[0027] S101: Acquire object features of a target object and initial description information related to the target object.

[0028] The description object of the generated commodity text may be referred to as a target object. The target object may be, for example, a product to be sold. A specific example of the target object is “small household appliances”, which is not limited.

[0029] Among them, the object feature can be used to describe the attribute characteristics of the target object. Assuming that the target object is a product to be sold, the object feature can be, for example, the product type. If the target object is specifically "small household appliances", the object feature can be, for example, "household appliances category", and there is no limitation to this.

[0030] Among them, the initial description information can be used to describe some descriptive features of the target object, such as some information about the activities associated with the target object. The initial description information may include the name of the target object, the name of the associated activity, the time of the event, etc., or it may be some preferential information related to the target object. For example, if you purchase the target object, you will get A for free. The name and quantity of A for free can be called some related preferential information, and there is no restriction on this.

[0031] When obtaining the object features of the target object and the initial description information related to the target object, the above-mentioned method can specifically provide an input interface for the object features and the initial description information, receive the content input by the user, and automatically parse the content input by the user to obtain the object features and the initial description information. Alternatively, the existing image can be recognized, the product to be sold can be identified from the image, and a preset algorithm can be called to analyze its object features and the initial description information. Alternatively, the content broadcast by the host in the live broadcast room can be captured, and the target object can be determined according to its content, and its object features and the initial description information can be determined. There is no limitation to this.

[0032] S102: Generate an opening description text and a preferential description text related to the target object according to the initial description information.

[0033] In the embodiment of the present application, after the initial description information related to the target object is obtained, an opening description text and a preferential description text related to the target object can be generated according to the initial description information.

[0034] It can be understood that the embodiments of the present application can be applied in the application scenario of live broadcast, and the product text generated for the live broadcast will usually include a description text with an opening nature to carry over the subsequent specific description content. The description text with an opening nature can be called an opening description text, and the description text involving some preferential content related to the target object can be called a preferential description text.

[0035] In an embodiment of the present application, the initial description information can be input into a pre-trained text generation model to obtain an opening description text and a preferential description text related to the target object output by the text generation model.

[0036] In some other embodiments, in order to balance the generation efficiency and quality of the commodity text used for live broadcast, a template matching method can be used to generate an opening description text and a discount description text related to the target object based on the initial description information.

[0037] Alternatively, if Figure 2 As shown, Figure 2 This is a flow chart of a method for generating a product text according to another embodiment of the present application, which generates an opening description text and a preferential description text related to a target object according to initial description information, including:

[0038] S201: Generate an opening description text according to the initial description information and in combination with the first product text template.

[0039] Among them, the product text template used to assist in generating the opening description text can be called the first product text template, and correspondingly, the product text template used to assist in generating the discount description text can be called the second product text template. The product text template can, for example, include some standard words. The first product text template used to assist in generating the opening description text can include standard words related to the opening description, and the second product text template used to assist in generating the discount description text can include standard words related to the discount description. There is no limitation on this.

[0040] For example, assuming that the initial description information is the name of the target object, the name of the related activity, the time of the event, etc., the first product text template can be called to parse the standard words related to the opening description in the first product text template, and then, based on the name of the target object, the name of the related activity, the time of the event, etc., combined with the standard words related to the opening description, the opening description text is generated.

[0041] Optionally, in some embodiments, the initial description information includes: time information, activity information of the platform to which the target object belongs, and the name of the target object. The opening description text is generated based on the initial description information and combined with the first product text template. The time description sub-text corresponding to the time information, the activity description sub-text corresponding to the activity information, and the name description sub-text corresponding to the name are parsed from the first product text template respectively, and the time description sub-text, the activity description sub-text, and the name description sub-text are spliced ​​to obtain the opening description text.

[0042] The time information may be, for example, the time of an activity associated with the target object, the activity information of the platform to which the target object belongs may be, for example, the name of the activity, and the name of the target object may be, for example, "small household appliance A". Then, the time description subtext corresponding to the time information, the activity description subtext corresponding to the activity information, and the name description subtext corresponding to the name may be parsed from the first product text template respectively. The time description subtext may specifically be a segment of speech for describing the opening description of the time information, the activity description subtext may specifically be a segment of speech for describing the opening description of the activity information, and the name description subtext may specifically be a segment of speech for describing the opening description of the name. Then, the time description subtext, the activity description subtext, and the name description subtext may be spliced ​​to obtain the opening description text.

[0043] For example, the opening description text is automatically generated: the opening description text can be generated by a predefined template (a first product text template), and the first product text template is as follows:

[0044] "My friends, day b of month a is the activity day of platform c. This time platform c has various promotional activities with great discounts. Today I also got a very useful and affordable d in the live broadcast room and recommend it to everyone!"

[0045] Among them, a is the month (time information), b is the date (time information), c is the specific activity (activity information), and d is the specific recommended product (the name of the target object).

[0046] The opening description text generated based on the first product text template is, for example: "My friends, June 18th is the platform c17 618 mid-year shopping carnival day. This time the platform c has various promotional activities with great discounts. Today I also got a very useful and affordable small appliance in the live broadcast room and recommend it to everyone!".

[0047] S202: Generate a preferential description text related to the target object according to the initial description information and in combination with a second product text template, wherein the first product text template and the second product text template are different.

[0048] For example, assuming that the initial description information is some preferential information related to the target object, the second product text template can be called to parse the standard wording related to the preferential description in the second product text template, and then, based on some preferential information related to the target object, combined with the standard wording related to the preferential description, a preferential description text is generated. There is no restriction on this.

[0049] Optionally, in some embodiments, the initial description information includes: the type of offer, then based on the initial description information, in combination with the second product text template, a discount description text related to the target object is generated, and the second product text template matching the discount type is determined. If the discount type indicates that there is corresponding discount content, at least one discount description sub-text corresponding to the discount content is parsed from the second product text template, and a discount description text is generated based on the at least one discount description sub-text. If the discount type indicates that there is no discount content, the template content text in the second product text template is directly used as the discount description text.

[0050] Among them, the type of discount can be specifically used to describe the way of discount, and the types of discounts include free gifts of other goods, price discounts, and discounts for purchases over a certain amount, and there is no restriction on this.

[0051] Accordingly, when pre-configuring the second product text template, the corresponding second product text template can be configured for different discount types. In addition, if the above-mentioned discount type is any of the types of giving away other products, price discount types, and discount types for purchases above a certain amount, then it can be indicated that the discount type indicates the existence of corresponding discount content. If the discount type is not any of the above-mentioned types, then it can be indicated that the discount type indicates the absence of discount content, and there is no restriction on this.

[0052] Therefore, in the embodiment of the present application, the generated discount description text can be adapted to the specific discount type, thereby improving the flexibility and convenience of generating the discount description text, thereby effectively improving the automatic generation effect of the discount description text.

[0053] For example, the automatic generation of discount description text: the specific second product text templates designed for different discount types are as follows:

[0054] If the discount type is: give away other product types, the second product text template is: "Buy a, get b for free, buy one get one free discount", where a is the current promotional product and b is the free product.

[0055] If the discount type is: giving away other types of goods, the second product text template is: "Today you can enjoy a super low exclusive price in the live broadcast room, only xx yuan!".

[0056] If the discount type is: full-discount discount type, the second product text template is: "Get the coupon first and then buy. Get yy yuan off for purchases over xx yuan, and you only need zz yuan to buy. Be sure to get the coupon first and then buy!".

[0057] If the discount type is: none of the above three categories, or there is no discount information, it will be a general second product text template, such as: "You won't suffer a loss or be cheated, the price here is the most affordable, and it is guaranteed to be genuine, and we will compensate you ten times if it is fake!".

[0058] The above-mentioned "Today you can enjoy a super low exclusive price in the live broadcast room, only... yuan", "Get the coupon first and then buy, get... yuan off for purchases over... yuan", "Only... yuan", "Be sure to get the coupon before buying" can all be called discount description sub-texts, and then, at least one discount description sub-text can be spliced ​​to obtain a discount description text, without any restriction.

[0059] S103: Generate content description text corresponding to the target object according to the object features.

[0060] Among them, the object feature can be used to describe the attribute characteristics of the target object. Assuming that the target object is a product to be sold, the object feature can be, for example, the product type. If the target object is specifically "small household appliances", the object feature can be, for example, "household appliances category", and there is no limitation to this.

[0061] After acquiring the object features, the content description text corresponding to the target object can be generated by template matching according to the object features, or any other possible method can be used to generate the content description text corresponding to the target object, without limitation.

[0062] S104: Generate a target product text corresponding to the target object according to the opening description text, the content description text, and the discount description text.

[0063] After generating the opening description text and the preferential description text related to the target object according to the initial description information, and generating the content description text corresponding to the target object according to the object characteristics, the target product text corresponding to the target object can be generated according to the opening description text, the content description text, and the preferential description text. For example, the opening description text, the content description text, and the preferential description text can be spliced, so that the spliced ​​description text is used as the target product text corresponding to the target object.

[0064] In this embodiment, by acquiring the object characteristics of the target object and the initial description information related to the target object, and generating an opening description text and a preferential description text related to the target object based on the initial description information, generating a content description text corresponding to the target object based on the object characteristics, and generating a target product text corresponding to the target object based on the opening description text, the content description text, and the preferential description text, the automation of the generation of introductory product texts can be effectively realized, the batch generation of introductory product texts can be realized, the consumption of human resources can be reduced, the generation efficiency and quality of the introductory product texts can be improved, and the generation effect of the introductory product texts can be improved.

[0065] Figure 3 It is a flowchart of a method for generating product text proposed in another embodiment of the present application.

[0066] like Figure 3 As shown, the method for generating the product text includes:

[0067] S301: Determine the annotation content description text corresponding to the object feature.

[0068] In the embodiment of the present application, it is possible to pre-train an artificial intelligence model, which may be a neural network model or a machine learning model, so that the artificial intelligence model has the function of generating corresponding content description text according to object features.

[0069] Among them, the content description text can be a description text with a general description nature. For example, if the product text is used to assist in selling products during the live broadcast, then the description text used to generally describe the details of the product being sold can be called the content description text.

[0070] The above-mentioned annotated content description text can be a standard content description text used to train artificial intelligence models, and there is no restriction on this.

[0071] The above-mentioned determination of the annotated content description text corresponding to the object features takes into account that the target objects with different object features usually have personalized description content, so that the trained artificial intelligence model can generate content description text that is adapted to the object features, thereby improving the accuracy of the generated content description text.

[0072] S302: Determine a plurality of sample objects according to object features.

[0073] That is to say, in the embodiments of the present application, multiple sample objects whose features are compatible with the above-mentioned features can be determined. Assuming that the object feature is the product type, such as "home appliance category", multiple sample objects of multiple home appliance categories can be selected to assist in training the artificial intelligence model.

[0074] S303: Determine a plurality of actual purchase rates corresponding to the plurality of sample objects respectively.

[0075] When determining the training data for training the artificial intelligence model, multiple actual purchase rates corresponding to multiple sample objects can also be determined. The actual purchase rate can be the historical actual purchase rate of multiple products of the corresponding product type. The actual purchase rate can be obtained based on historical purchase statistics, and there is no restriction on this.

[0076] For example, for different products (sample objects), user historical purchase data can be collected to record the purchase rate corresponding to the product selling point introduction product text of each product, that is, the ratio of the number of purchasers to the number of viewers. For example, the number of viewers of mobile phone A is 100, the number of purchasers is 10, and the actual purchase rate is 10%, while the number of viewers of mobile phone B is 200, the number of purchasers is 50, and the actual purchase rate is 25%.

[0077] S304: Train an initial artificial intelligence model according to multiple actual purchase rates and object features until the target content description text and the annotated content description text output by the artificial intelligence model meet the set conditions, and then use the trained artificial intelligence model as the target artificial intelligence model.

[0078] When determining the timing of convergence of the artificial intelligence model, the loss value between the target content description text and the annotated content description text output by the artificial intelligence model can be specifically determined. If the loss value meets the set conditions, it is determined that the initial artificial intelligence model training is completed, thereby accurately assisting in determining the timing of completion of model training, ensuring the training effect of the artificial intelligence model while improving training efficiency.

[0079] Among them, determining the loss value between the target content description text and the annotated content description text output by the artificial intelligence model can be to determine the similarity between the target content description text and the annotated content description text, and determine multiple product values ​​obtained by multiplying the similarity with multiple actual purchase rates respectively, and take the average of the multiple product values ​​as the loss value, so that the actual purchase rate can be used as the reward reference content for artificial intelligence model training, and the model generation effect of the artificial intelligence model can be measured based on the dimension of purchase rate, so that the trained artificial intelligence model is adapted to the live sales scenario, thereby improving the model expression effect.

[0080] When training the initial artificial intelligence model, a reinforcement learning algorithm can be used to perform the training process, and the description of the reinforcement learning algorithm can be as follows:

[0081] Based on the reinforcement learning algorithm, the loss function may be used to calculate the similarity between the text of the high conversion rate products and the products of the same product type.

[0082] For example, if the similarity between the generated product text and the product text of mobile phone A is 0.3, and the similarity between the generated product text and the product text of mobile phone B is 0.2, the loss value may be (0.3*10%+0.2*25%) / 2=0.04.

[0083] The artificial intelligence model in the embodiment of the present application can be trained using a reinforcement learning algorithm framework based on policy gradient. In the reinforcement learning algorithm framework, the summary automatic generation model is a reinforcement learning agent, which uses an encoder-decoder framework. The environment in which the agent interacts with the agent is the words generated at the historical moment and the contextual semantic vector of the current moment; the agent's parameters are used as the policy, and then, according to the policy and the environment, the agent can take the action at the current moment, that is, generate the word at the current moment.

[0084] In the embodiment of the present application, if Figure 4 As shown, Figure 4 It is a schematic diagram of the architecture of the artificial intelligence model in the embodiment of the present application, including multiple encoders, among which a bidirectional recurrent neural network (Bi-directional Long Short-Term Memory, BiLSTM) can be used as an encoder, a detailed product introduction text a can be input, and the encoder (which can use a bidirectional recurrent neural network (Bi-directional Long Short-Term Memory, BiLSTM) as an encoder) encodes it to generate an encoded hidden layer sequence h, the formula is: h = BiLSTM (a), and then, multiple decoders (which can use a unidirectional recurrent neural network (Long Short-Term Memory, LSTM) as a decoder) use the encoded hidden layer sequence h and the attention mechanism to generate a decoded hidden layer sequence s, and then generate the current word based on this (the word can be regarded as part of the target content description text to be output). The formula is:

[0085] s t =LSTM(s t-1 ,y t-1 , c t );

[0086] where {t1, t2, ..., t m} is the decoder hidden sequence, {y1, y2, ..., y m} is the target summary, c t is the context vector at time t, generated by the attention mechanism, and the formula is as follows:

[0087]

[0088] α t =softmax(e t );

[0089] c t =∑ i α t，i h i .

[0090] Among them, u a , W a and V a is the model parameter matrix, α t，i is the source word x at time t i attention.

[0091] The decoder uses the decoder hidden state and context vector to generate the probability of the summary word w, as follows: P(w) = softmax(W b s t +V b c t );

[0092] Among them, W b and V b is the model parameter matrix.

[0093] The reward function (i.e. loss function) of the model is the similarity with the high conversion rate product text of similar products multiplied by the average purchase rate. The formula is as follows:

[0094]

[0095] Among them, x i Describes the annotation content of other sample objects of the same object type as the current object, p(x i ) is x i The corresponding actual purchase rate, N is the total number of annotated content description texts of other sample objects of the same object type. The purpose of this reward function is to make the actual purchase rate of the sales activities based on the commodity text generated by the artificial intelligence model higher. sim(x i ,x) represents the annotation content description text x i The cosine similarity with the feature vector of the target content description text x, x i The feature vector of x is defined as the mean of all hidden layer states of the decoder of the artificial intelligence model.

[0096] S305: Acquire object features of the target object and initial description information related to the target object.

[0097] S306: Generate an opening description text and a preferential description text related to the target object according to the initial description information.

[0098] The description of S305 - S306 can be specifically referred to the above embodiment, which will not be repeated here.

[0099] S307: Determine a target artificial intelligence model corresponding to the object features.

[0100] In the process of actual application, multiple artificial intelligence models can be trained according to different object features, so that there is a one-to-one correspondence between multiple artificial intelligence models and multiple object features. After the artificial intelligence model is trained and the trained artificial intelligence model is able to generate content description text that matches the object features, in the process of actually generating the content description text, it is possible to determine the target artificial intelligence model corresponding to the object features, and use the target artificial intelligence model to assist in determining the content description text.

[0101] S308: Input the object features into the target artificial intelligence model, and obtain the content description text corresponding to the target object output by the target artificial intelligence model.

[0102] After determining the target artificial intelligence model corresponding to the object features, the object features can be directly input into the target artificial intelligence model, and the content description text corresponding to the target object output by the target artificial intelligence model can be obtained. Since the target artificial intelligence model is pre-trained using a reinforcement learning algorithm and refers to the actual purchase rate of similar objects as a reward reference content, the accuracy of content description text generation can be effectively improved.

[0103] S309: Generate a target product text corresponding to the target object according to the opening description text, the content description text, and the discount description text.

[0104] The description of S309 can be specifically referred to the above embodiment, which will not be repeated here.

[0105] In this embodiment, by obtaining the object features of the target object and the initial description information related to the target object, and generating the opening description text and the preferential description text related to the target object according to the initial description information, generating the content description text corresponding to the target object according to the object features, and generating the target commodity text corresponding to the target object according to the opening description text, the content description text, and the preferential description text, it is possible to effectively realize the automation of the generation of the commodity text for introduction, realize the batch generation of the commodity text for introduction, reduce the consumption of human resources, improve the generation efficiency and generation quality of the commodity text for introduction, and improve the generation effect of the commodity text for introduction. An artificial intelligence model is pre-trained so that the artificial intelligence model has the function of generating the corresponding content description text according to the object features. By determining the loss value between the target content description text output by the artificial intelligence model and the annotated content description text, if the loss value meets the set conditions, it is determined that the initial artificial intelligence model training is completed, so that the timing of the completion of the model training can be accurately assisted, and the training effect of the artificial intelligence model is guaranteed while improving the training efficiency. By determining the similarity between the target content description text and the annotated content description text, and determining multiple product values ​​obtained by multiplying the similarity with multiple actual purchase rates respectively, and taking the average of the multiple product values ​​as the loss value, the actual purchase rate can be used as the reward reference content for the training of the artificial intelligence model, and the model generation effect of the artificial intelligence model can be measured based on the dimension of the purchase rate, so that the trained artificial intelligence model is adapted to the live sales scenario, thereby improving the model expression effect. After determining the target artificial intelligence model corresponding to the object features, the object features can be directly input into the target artificial intelligence model, and the content description text corresponding to the target object output by the target artificial intelligence model can be obtained. Since the target artificial intelligence model is pre-trained using a reinforcement learning algorithm and refers to the actual purchase rate of similar objects as the reward reference content, the accuracy of the content description text generation is effectively improved.

[0106] Figure 5 It is a structural diagram of a device for generating product text proposed in one embodiment of the present application.

[0107] like Figure 5 As shown, the product text generating device 50 includes:

[0108] An acquisition module 501 is used to acquire object features of a target object and initial description information related to the target object;

[0109] The first generating module 502 is used to generate an opening description text and a preferential description text related to the target object according to the initial description information;

[0110] The second generating module 503 is used to generate a content description text corresponding to the target object according to the object characteristics;

[0111] The third generating module 504 is used to generate a target commodity text corresponding to the target object according to the opening description text, the content description text, and the discount description text.

[0112] In some embodiments of the present application, Figure 6 As shown, the first generation module 502 includes:

[0113] The first generating submodule 5021 is used to generate an opening description text according to the initial description information and in combination with the first product text template;

[0114] The second generating submodule 5022 is used to generate a preferential description text related to the target object according to the initial description information and in combination with the second product text template, wherein the first product text template and the second product text template are different.

[0115] In some embodiments of the present application, the initial description information includes: time information, activity information of the platform to which the target object belongs, and the name of the target object, and the first generation submodule 5021 is specifically used to:

[0116] Parse the time description subtext corresponding to the time information, the activity description subtext corresponding to the activity information, and the name description subtext corresponding to the name from the first product text template;

[0117] The time description sub-text, the activity description sub-text, and the name description sub-text are concatenated to obtain the opening description text.

[0118] In some embodiments of the present application, the initial description information includes: the type of discount, and the second generation submodule 5022 is specifically used to:

[0119] Determine a second product text template that matches the offer type;

[0120] If the discount type indicates that corresponding discount content exists, parsing at least one discount description subtext corresponding to the discount content from the second product text template, and generating a discount description text according to the at least one discount description subtext;

[0121] If the discount type indicates that there is no discount content, the template content text in the second product text template is directly used as the discount description text.

[0122] In some embodiments of the present application, the second generating module 503 is specifically used to:

[0123] determining a target artificial intelligence model corresponding to the object characteristics;

[0124] The object features are input into the target artificial intelligence model, and the content description text corresponding to the target object output by the target artificial intelligence model is obtained.

[0125] In some embodiments of the present application, Figure 6 As shown, it also includes:

[0126] The training module 505 is used to determine the annotated content description text corresponding to the object features, and based on the object features, determine multiple sample objects, determine multiple actual purchase rates corresponding to the multiple sample objects, and train the initial artificial intelligence model based on the multiple actual purchase rates and object features until the target content description text output by the artificial intelligence model and the annotated content description text meet the set conditions, and then use the trained artificial intelligence model as the target artificial intelligence model.

[0127] In some embodiments of the present application, the training module 505 is specifically used to:

[0128] Determine the loss value between the target content description text output by the artificial intelligence model and the annotated content description text;

[0129] If the loss value meets the set conditions, it is determined that the initial artificial intelligence model training is completed.

[0130] In some embodiments of the present application, the training module 505 is specifically used to:

[0131] Determine the similarity between the target content description text and the annotated content description text;

[0132] A plurality of product values ​​obtained by respectively multiplying the similarity and the plurality of actual purchase rates are determined, and an average value of the plurality of product values ​​is taken as the loss value.

[0133] With the above Figures 1 to 4 Corresponding to the method for generating a product text provided in the embodiment, the present application also provides a device for generating a product text. Figures 1 to 4 The method for generating the product text provided in the embodiment corresponds to the method, so the implementation method of the method for generating the product text is also applicable to the device for generating the product text provided in the embodiment of the present application, and will not be described in detail in the embodiment of the present application.

[0134] In this embodiment, by acquiring the object characteristics of the target object and the initial description information related to the target object, and generating an opening description text and a preferential description text related to the target object based on the initial description information, generating a content description text corresponding to the target object based on the object characteristics, and generating a target product text corresponding to the target object based on the opening description text, the content description text, and the preferential description text, the automation of the generation of introductory product texts can be effectively realized, the batch generation of introductory product texts can be realized, the consumption of human resources can be reduced, the generation efficiency and quality of the introductory product texts can be improved, and the generation effect of the introductory product texts can be improved.

[0135] In order to implement the above embodiments, the present application also proposes a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method for generating product text proposed in the above embodiments of the present application is implemented.

[0136] In order to implement the above embodiments, the present application also proposes a non-temporary computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method for generating product text proposed in the above embodiments of the present application is implemented.

[0137] In order to implement the above embodiments, the present application also proposes a computer program product. When the instruction processor in the computer program product is executed, the method for generating product text proposed in the above embodiments of the present application is executed.

[0138] Figure 7 A block diagram of an exemplary computer device suitable for implementing embodiments of the present application is shown. Figure 7 The computer device 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0139] like Figure 7 As shown, the computer device 12 is in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that connects various system components (including the system memory 28 and the processing unit 16).

[0140] The bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor or a local bus using any of a variety of bus structures. For example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus and Peripheral Component Interconnection (PCI) bus.

[0141] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0142] The memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 7 Not shown, often called a "hard drive").

[0143] although Figure 7 Not shown, a disk drive for reading and writing a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing a removable non-volatile optical disk (e.g., a compact disc read only memory (CD-ROM), a digital versatile disc read only memory (DVD-ROM), or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present application.

[0144] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in the memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0145] The computer device 12 may also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), one or more devices that enable a user to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication may be performed via an input / output (I / O) interface 22. In addition, the computer device 12 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with other modules of the computer device 12 via a bus 18. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the computer device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0146] The processing unit 16 executes various functional applications and data processing by running the programs stored in the system memory 28, such as implementing the method for generating the commodity text mentioned in the above embodiment.

[0147] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0148] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

[0149] It should be noted that, in the description of this application, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" is two or more.

[0150] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0151] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0152] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0153] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0154] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0155] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0156] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for generating a product text, characterized in that: The method comprises: Acquire object features of a target object and initial description information related to the target object; Generate an opening description text and a preferential description text related to the target object according to the initial description information, wherein the opening description text refers to a description text with an opening nature, and the preferential description text refers to a description text related to the preferential content of the target object; Generating a content description text corresponding to the target object according to the object features; Generating a target product text corresponding to the target object according to the opening description text, the content description text, and the discount description text; The step of generating an opening description text and a preferential description text related to the target object according to the initial description information includes: Generate the opening description text according to the initial description information and in combination with the first product text template; generating the preferential description text related to the target object according to the initial description information and in combination with a second product text template, wherein the first product text template and the second product text template are different; The step of generating a content description text corresponding to the target object according to the object feature includes: Determining a target artificial intelligence model corresponding to the object features; Inputting the object features into the target artificial intelligence model, and obtaining a content description text corresponding to the target object output by the target artificial intelligence model; Before acquiring the object features of the target object, the method further includes: Determine the annotation content description text corresponding to the object feature; Determining a plurality of sample objects according to the object characteristics; determining a plurality of actual purchase rates respectively corresponding to the plurality of sample objects; Training an initial artificial intelligence model according to the multiple actual purchase rates and the object features until it is determined that the loss value between the target content description text output by the artificial intelligence model and the annotated content description text meets a set condition, and then using the trained artificial intelligence model as the target artificial intelligence model; The determining of the loss value between the target content description text output by the artificial intelligence model and the annotated content description text includes: Determining the similarity between the target content description text and the annotated content description text; A plurality of product values ​​obtained by respectively multiplying the similarity and the plurality of actual purchase rates are determined, and an average value of the plurality of product values ​​is used as the loss value.

2. The method according to claim 1, characterized in that The initial description information includes: time information, activity information of the platform to which the target object belongs, and the name of the target object. Then, the opening description text is generated according to the initial description information and in combination with the first commodity text template, including: Parse the first product text template to obtain a time description subtext corresponding to the time information, an activity description subtext corresponding to the activity information, and a name description subtext corresponding to the name; The time description subtext, the activity description subtext, and the name description subtext are concatenated to obtain the opening description text.

3. The method according to claim 1, characterized in that The initial description information includes: the discount type, and the generating of the discount description text related to the target object according to the initial description information and in combination with the second commodity text template includes: Determine a second product text template that matches the discount type; If the discount type indicates that there is corresponding discount content, parsing at least one discount description subtext corresponding to the discount content from the second product text template, and generating the discount description text according to the at least one discount description subtext; If the discount type indicates that the discount content does not exist, the template content text in the second commodity text template is directly used as the discount description text.

4. A device for generating product text, characterized in that: The device comprises: An acquisition module, used to acquire object features of a target object and initial description information related to the target object; A first generating module, configured to generate an opening description text and a preferential description text related to the target object according to the initial description information, wherein the opening description text refers to a description text having an opening nature, and the preferential description text refers to a description text related to the preferential content of the target object; A second generating module, used for generating a content description text corresponding to the target object according to the object features; A third generating module, configured to generate a target commodity text corresponding to the target object according to the opening description text, the content description text, and the discount description text; The first generating module comprises: A first generating submodule, configured to generate the opening description text according to the initial description information and in combination with a first product text template; A second generating submodule, configured to generate the preferential description text related to the target object according to the initial description information and in combination with a second commodity text template, wherein the first commodity text template and the second commodity text template are different; The second generating module is specifically used for: Determining a target artificial intelligence model corresponding to the object features; Inputting the object features into the target artificial intelligence model, and obtaining a content description text corresponding to the target object output by the target artificial intelligence model; The device further comprises: A training module, used to determine the annotated content description text corresponding to the object feature, and determine multiple sample objects based on the object feature, determine multiple actual purchase rates corresponding to the multiple sample objects respectively, and train an initial artificial intelligence model based on the multiple actual purchase rates and the object feature until it is determined that the loss value between the target content description text output by the artificial intelligence model and the annotated content description text meets a set condition, and then use the trained artificial intelligence model as the target artificial intelligence model; The training module is specifically used for: Determining the similarity between the target content description text and the annotated content description text; A plurality of product values ​​obtained by respectively multiplying the similarity and the plurality of actual purchase rates are determined, and an average value of the plurality of product values ​​is used as the loss value.

5. The device according to claim 4, characterized in that The initial description information includes: time information, activity information of the platform to which the target object belongs, and the name of the target object. The first generating submodule is specifically used to: Parse the first product text template to obtain a time description subtext corresponding to the time information, an activity description subtext corresponding to the activity information, and a name description subtext corresponding to the name; The time description subtext, the activity description subtext, and the name description subtext are concatenated to obtain the opening description text.

6. The device according to claim 4, characterized in that The initial description information includes: the discount type, and the second generation submodule is specifically used to: Determine a second product text template that matches the discount type; If the discount type indicates that there is corresponding discount content, parsing at least one discount description subtext corresponding to the discount content from the second product text template, and generating the discount description text according to the at least one discount description subtext; If the discount type indicates that the discount content does not exist, the template content text in the second commodity text template is directly used as the discount description text.

7. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method according to any one of claims 1 to 3 is implemented.

8. A storage medium, when instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the method according to any one of claims 1 to 3.

Images