Text intention recognition model training and text intention recognition

Abstract

Disclosed in the present description are a text intention recognition model training method and apparatus, and a text intention recognition method and apparatus. The method comprises: using a product description sample text and a reference product selection condition corresponding to the product description sample text to train an initial text intention recognition model, so that the initial text intention recognition model adjusts model parameters on the basis of a learned text intention corresponding to the product description sample text, and a sample product selection condition output by the initial text intention recognition model can thus progressively align with the reference product selection condition, thereby obtaining a target text intention recognition model. Thus, by means of the target text intention recognition model, a target product selection condition for a product is quickly determined on the basis of the description of a user for the product.

Description

Text intent recognition model training, text intent recognition Technical Field

[0001] This specification relates to the field of data processing technology, and in particular to the training of text intent recognition models and the recognition of text intent. Background Art

[0002] In recent years, with the rapid development of information technology, various organizations have launched online products to quickly promote their products. However, when faced with so many online products, consumers find it difficult to quickly determine the results they want based on their own intentions. Summary of the Invention

[0003] This specification provides a text intent recognition model training method, a text intent recognition method and a device, and the technical solution is as follows.

[0004] In the first aspect, the present specification provides a text intent recognition model training method, the method comprising: creating an initial text intent recognition model, obtaining a product description sample text and a reference product selection condition corresponding to the product description sample text; inputting the product description sample text into the initial text intent recognition model for model training, determining a sample product selection intention vector corresponding to the product description sample text through the initial text intent recognition model, and determining a sample product selection condition based on the sample product selection intention vector; during the model training process, adjusting the model parameters of the initial text intent recognition model based on the sample product selection condition and the reference product selection condition to obtain a target text intent recognition model.

[0005] In the second aspect, this specification provides a text intent recognition method, which includes: obtaining a product description target text input by a user; inputting the product description target text into a target text intent recognition model, determining a target product selection intention vector corresponding to the product description target text through the target text intent recognition model, and determining a target product selection condition based on the target product selection intention vector, and outputting the target product selection condition for the product description target text; screening a product set based on the target product selection condition to obtain a target product for the product description target text.

[0006] In the third aspect, this specification provides a text intent recognition model training device, which includes: a creation module, suitable for creating an initial text intent recognition model, obtaining a product description sample text and a reference product selection condition corresponding to the product description sample text; a processing module, suitable for inputting the product description sample text into the initial text intent recognition model for model training, determining the sample product selection intention vector corresponding to the product description sample text through the initial text intent recognition model, and determining the sample product selection condition based on the sample product selection intention vector; a training module, suitable for adjusting the model parameters of the initial text intent recognition model based on the sample product selection condition and the reference product selection condition during the model training process to obtain a target text intent recognition model.

[0007] In the fourth aspect, this specification provides a text intent recognition device, which includes: an acquisition module, suitable for acquiring a product description target text input by a user; an input module, suitable for inputting the product description target text into a target text intent recognition model, determining a target product selection intention vector corresponding to the product description target text through the target text intent recognition model, and determining a target product selection condition based on the target product selection intention vector, and outputting the target product selection condition for the product description target text; an output module, suitable for screening a product set based on the target product selection condition to obtain a target product for the product description target text.

[0008] In a fifth aspect, this specification provides a computer storage medium, wherein the computer storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the above-mentioned method steps.

[0009] In a sixth aspect, this specification provides an electronic device, which may include: a processor and a memory; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the above-mentioned method steps.

[0010] In a seventh aspect, this specification provides a computer program product, which stores at least one instruction, and the at least one instruction is loaded by a processor to execute any one of the above method steps.

[0011] The beneficial effects brought about by the technical solutions provided in some embodiments of this specification include at least: using product description sample text and reference product selection conditions corresponding to the product description sample text to train the initial text intent recognition model, so that the initial text intent recognition model continuously adjusts the model parameters based on the text intent corresponding to the learned product description sample text, so that the sample product selection conditions output by the initial text intent recognition model can continue to approach the reference product selection conditions, thereby obtaining a target text intent recognition model, so as to realize the use of the target text intent recognition model to quickly determine the target product selection conditions for a product based on the user's expected description of a product, thereby solving the technical problem that consumers find it difficult to quickly determine the screening results they want based on their own intentions when facing many online products. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] FIG1 is a scenario diagram of a text intent recognition model training system provided in this specification.

[0013] FIG2 is a flow chart of a text intent recognition model training method provided in an embodiment of this specification.

[0014] FIG3 is a schematic diagram of a process for determining sample selection conditions corresponding to a sample association index vector according to an embodiment of this specification.

[0015] FIG4 is a schematic diagram of a process for determining a sample selection index vector library according to an embodiment of this specification.

[0016] FIG5 is a schematic diagram of a process for obtaining a sample association index vector according to an embodiment of this specification.

[0017] FIG6 is a flow chart of determining a sample selection intention vector according to an embodiment of this specification.

[0018] FIG7 is a schematic diagram of a process for determining at least one target keyword corresponding to a product description sample text according to an embodiment of this specification.

[0019] FIG8 is a flow chart of a text intent recognition method provided in an embodiment of this specification.

[0020] FIG9 is a schematic structural diagram of a text intent recognition model training device provided in an embodiment of this specification.

[0021] FIG10 is a schematic structural diagram of a text intent recognition device provided in an embodiment of this specification.

[0022] FIG11 is a schematic structural diagram of an electronic device provided in this specification.

[0023] FIG12 is a schematic diagram of the structure of the operating system and user space provided in this specification.

[0024] FIG13 is an architectural diagram of the Android operating system in FIG12 .

[0025] FIG14 is an architectural diagram of the IOS operating system in FIG12 . DETAILED DESCRIPTION

[0026] In the description of this specification, it should be understood that the terms "first", "second", etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance. In the description of this specification, it should be noted that, unless otherwise expressly specified and limited, "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices. For those of ordinary skill in the art, the specific meanings of the above terms in this specification can be understood according to the specific circumstances. In addition, in the description of this specification, unless otherwise specified, "multiple" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0027] The present specification is described in detail below with reference to specific embodiments.

[0028] Please refer to Figure 1, which is a scenario diagram of a text intent recognition model training system provided in this specification. As shown in Figure 1, the text intent recognition model training system may include at least a client cluster and a service platform 100.

[0029] The client cluster may include at least one client, as shown in FIG1 , specifically including client 1 corresponding to user 1, client 2 corresponding to user 2, ..., client n corresponding to user n, where n is an integer greater than 0.

[0030] Each client in the client cluster can be an electronic device with communication capabilities, including but not limited to wearable devices, handheld devices, personal computers, tablet computers, in-vehicle devices, smartphones, computing devices, or other processing devices connected to a wireless modem. Electronic devices may be called different names in different networks, such as user equipment, access terminal, subscriber unit, subscriber station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent or user device, cellular phone, cordless phone, personal digital assistant (PDA), electronic devices in 5G network or future evolution network, etc.

[0031] The service platform 100 can be a separate server device, such as a rack-mounted, blade, tower, or cabinet-mounted server device, or a workstation, mainframe computer, or other hardware device with strong computing capabilities; it can also be a server cluster composed of multiple servers. The servers in the service cluster can be symmetrically composed, wherein each server has equivalent functions and status in the transaction link, and each server can provide services to the outside world independently. The independent service can be understood as not requiring the assistance of other servers.

[0032] In one or more embodiments of the present specification, the service platform 100 may establish a communication connection with at least one client in the client cluster, and complete data interaction during the text intent recognition model training process based on the communication connection, such as online transaction data interaction. For example, the service platform 100 may obtain a target text intent recognition model based on the text intent recognition model training method of the present specification to realize text intent recognition for the client; for example, the service platform 100 may obtain training data from the client.

[0033] It should be noted that the service platform 100 and at least one client in the client cluster establish a communication connection through a network for interactive communication, wherein the network can be a wireless network or a wired network, the wireless network includes but is not limited to a cellular network, a wireless local area network, an infrared network or a Bluetooth network, and the wired network includes but is not limited to Ethernet, a universal serial bus (USB) or a controller area network. In one or more embodiments of the specification, technologies and / or formats including Hypertext Markup Language (HTML) and Extensible Markup Language (XML) are used to represent data exchanged over the network (such as a target compressed package). In addition, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), and Internet Protocol Security (IPsec) can also be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can also be used to replace or supplement the above-mentioned data communication technologies.

[0034] The text intent recognition model training system embodiment provided in this specification and the text intent recognition model training method in one or more embodiments are of the same concept. The execution subject corresponding to the text intent recognition model training method involved in one or more embodiments of the specification can be the above-mentioned service platform 100; the execution subject corresponding to the text intent recognition model training method involved in one or more embodiments of the specification can also be the electronic device corresponding to the client, which is specifically determined based on the actual application environment. The embodiment of the text intent recognition model training system and its implementation process can be detailed in the following method embodiment, which will not be repeated here.

[0035] Based on the scenario diagram shown in FIG1 , the text intent recognition model training method provided by one or more embodiments of this specification is introduced in detail below.

[0036] Please refer to Figure 2, which is a flowchart illustrating a method for training a text intent recognition model according to an embodiment of this specification. This method can be implemented using a computer program and can be run on a text intent recognition model training device based on the von Neumann architecture. The computer program can be integrated into an application or run as a standalone tool application. The text intent recognition model training device can be a service platform.

[0037] Specifically, the text intent recognition model training method includes the following steps.

[0038] S202: Create an initial text intent recognition model, obtain product description sample text and reference product selection conditions corresponding to the product description sample text.

[0039] The initial text intent recognition model may be an untrained or incompletely trained initial model for recognizing text intent. The initial text intent recognition model may include multiple modules.

[0040] Optionally, the initial text intent recognition model may include at least an input module, a conditional screening module, and an output module. Specifically, the input module is used to receive sample product description text, the conditional screening module is used to determine sample product selection conditions based on the sample product description text, and the output module is used to output the sample product selection conditions.

[0041] Optionally, the initial text intent recognition model may include at least an input module, a vectorization module, a conditional screening module, and an output module. Specifically, the input module and the output module may refer to the description above. The vectorization module is used to vectorize the product description sample text received by the input module to obtain a sample product selection intent vector corresponding to the product description sample text. The conditional screening module is used to determine the sample product selection conditions based on the sample product selection intent vector.

[0042] The sample product description text can be obtained based on the user's historical input text, or can be obtained by converting the user's historical voice input into text. Furthermore, the sample product description text can be a description text for a product, where the product can be a virtual digital product or a real physical product.

[0043] The reference selection criteria for the sample product description text can be understood as the standard selection criteria for the sample product description text. Specifically, when the sample product description text reads, "Recommend a principal-guaranteed fund product for me" or "What fund products can make money every year?", the standard selection criteria for the sample product description text can be, "Year-round positive return; natural annual return rate is a long-term high profit rate; maximum drawdown is a low drawdown rate." When the sample product description text reads, "Are there fund products that have decent returns but are unlikely to lose money?", the standard selection criteria for the sample product description text can be, "High cost-effective return rate; Sharpe ratio is preferred for long-term profits; maximum drawdown is able to rise and resist decline." The Sharpe ratio is an indicator used to measure the excess return of an investment portfolio, which takes into account the return and risk of the investment portfolio.

[0044] S204: Input the product description sample text into the initial text intent recognition model for model training, determine the sample product selection intention vector corresponding to the product description sample text through the initial text intent recognition model, and determine the sample product selection conditions based on the sample product selection intention vector.

[0045] In the process of inputting the product description sample text into the initial text intent recognition model for model training, the initial text intent recognition model continuously learns the text intent corresponding to the product description sample text.

[0046] After the initial text intent recognition model receives the product description sample text, it determines the sample product selection intention vector corresponding to the product description sample text. The keywords in the product description sample text can be vectorized to obtain the sample product selection intention vector; or the keywords in the product description sample text can be filtered, and the target keywords obtained after filtering can be vectorized to obtain the sample product selection intention vector. Here, the keywords in the product description sample text can be filtered based on the application scenario corresponding to the initial text intent recognition model, that is, the keywords with low relevance to the application scenario corresponding to the initial text intent recognition model can be eliminated.

[0047] After obtaining the sample product selection intention vector corresponding to the product description sample text, query at least one sample product selection indicator corresponding to the sample product selection intention vector, and determine the sample product selection condition based on the at least one sample product selection indicator.

[0048] S206: During the model training process, the model parameters of the initial text intent recognition model are adjusted based on the sample selection conditions and the reference selection conditions to obtain the target text intent recognition model.

[0049] Among them, after obtaining the sample product selection conditions corresponding to the product description sample text based on the initial text intent recognition model, a loss function can be constructed based on the parameters corresponding to the sample product selection conditions and the parameters corresponding to the reference product selection conditions, and the parameter values ​​corresponding to the sample product selection conditions and the reference product selection conditions are substituted into the loss function to obtain the loss value.

[0050] Afterwards, the model parameters of the initial text intent recognition model can be adjusted based on the loss value to continuously reduce the loss value, allowing the initial text intent recognition model to continuously learn the text intent corresponding to the product description sample text. When the loss value corresponding to the loss function converges, it indicates that the initial text intent recognition model has fully learned the text intent corresponding to the product description sample text. At this point, the initial text intent recognition model corresponding to the convergence loss value can be used as the target text intent recognition model.

[0051] Specifically, in the process of adjusting the model parameters of the initial text intent recognition model, when the initial text intent recognition model includes at least an input module, a conditional screening module and an output module, at least the module parameters corresponding to the conditional screening module can be adjusted; when the initial text intent recognition model includes at least an input module, a vectorization module, a conditional screening module and an output module, at least the module parameters corresponding to the vectorization module can be adjusted, or the module parameters corresponding to the vectorization module and the conditional screening module can be adjusted at the same time.

[0052] In the embodiment provided in this specification, an initial text intent recognition model is first created, and the initial text intent recognition model is trained using product description sample text and reference product selection conditions corresponding to the product description sample text, so that the initial text intent recognition model continuously adjusts model parameters based on the text intent corresponding to the learned product description sample text, so that the sample product selection conditions output by the initial text intent recognition model can continue to approach the reference product selection conditions, thereby obtaining a target text intent recognition model, so that the target text intent recognition model can quickly determine the target product selection conditions for a product based on the user's expected description of a product, thereby solving the technical problem that consumers find it difficult to quickly determine the screening results they want based on their own intentions when facing many online products.

[0053] This solution uses the target text intent recognition model to accurately identify the user's product selection intention based on the user's ambiguous product description text, and associate it with the corresponding product selection conditions, thereby quickly helping the user select products that meet their real needs.

[0054] In an embodiment provided in this specification, S204 determines the sample product selection conditions based on the sample product selection intention vector, including: performing vector matching on the sample product selection intention vector based on the sample product selection indicator vector library to obtain a sample association indicator vector, and determining the sample product selection conditions corresponding to the sample association indicator vector based on the initial text intention recognition model; wherein the sample product selection indicator vector library includes a product selection indicator vector generated based on at least one product screening indicator and a product selection intention text corresponding to the product screening indicator.

[0055] After obtaining the sample product selection intent vector, vector matching can be used to associate it with the product selection indicator vectors in the sample product selection indicator vector library for the scenario corresponding to the product description sample text. The sample product selection indicator vector library includes one or more product selection indicator vectors, which are generated based on product screening indicators and the product selection intent text corresponding to the product screening indicators. The product selection intent text can be determined based on the product screening indicators.

[0056] For example, when the product screening indicator is positive rate of return, the product selection intention text can be "buy fund products with a high probability of positive returns". For example, when the product screening indicator is high cost-effectiveness, the product selection intention text can be "buy fund products with a high cost-effectiveness of return and risk".

[0057] It should be noted that the association between a separate product screening indicator and a product description sample text is difficult to judge directly, so the association between the product screening indicator and the product description sample text can be determined by adding a product selection intention text to the product screening indicator. In this way, the association between the product selection indicator vector generated based on the product screening indicator and the product selection intention text corresponding to the product screening indicator and the sample product selection intention vector can be judged more directly and quickly. That is, by performing vector matching on each product selection indicator vector in the sample product selection indicator vector library and the sample product selection intention vector, the product selection indicator vector with a vector matching degree higher than the preset matching value is used as the sample association indicator vector. It is easy to understand that the higher the vector matching degree, the higher the matching degree between the sample product selection intention vector and the product screening indicator corresponding to each of the product selection indicator vectors.

[0058] After obtaining the sample association index vector, the product screening index corresponding to the sample association index vector is determined, and the sample product selection condition is generated based on the product screening index corresponding to each sample association index vector.

[0059] In the embodiment provided in this specification, the sample product selection intention vector is vector-matched by each product selection indicator vector in the sample product selection indicator vector library, so as to obtain an associated indicator vector that matches the sample product selection intention vector, that is, an associated indicator vector with a high similarity to the sample product selection intention vector. Since the higher the degree of vector matching, the higher the degree of matching between the sample product selection intention vector and the product selection indicator vector, the higher the degree of matching between the product screening indicators corresponding to each of the sample product selection intention vector and the product selection indicator vector. Therefore, the sample product selection conditions corresponding to the product description sample text can be determined based on the sample associated indicator vector.

[0060] Please refer to Figure 3, which is a flowchart of an embodiment of this specification that provides a method for determining the sample selection conditions corresponding to the sample association index vector. Specifically: when creating an initial text intent recognition model, it includes: using a basic large language generation model to form a conditional screening module, and creating an initial text intent recognition model that at least includes the conditional screening module. That is, when the initial text intent recognition model includes at least the conditional screening module, the above embodiment determines the sample selection conditions corresponding to the sample association index vector based on the initial text intent recognition model, including the following steps.

[0061] S302: Generate a sample indicator screening prompt template for the sample-related indicator vector.

[0062] Among them, in order to enable the conditional screening module to output the expected result based on the sample association indicator vector. Therefore, it is necessary to construct a sample indicator screening prompt template for the sample association indicator vector, that is, to construct a sample indicator screening prompt template for the sample association indicator vector that is easy to be understood by the basic large language generation model corresponding to the conditional screening module. The basic large language generation model can refer to a language model containing hundreds of billions (or more) parameters, which are trained on a large amount of text data, such as the model GPT (Generative Pre-Trained Transformer, generative pre-training model)-3.5, the large language model trained based on the Pathway distributed training architecture, the Galactica large language model and the LLaMA large language model.

[0063] In a feasible implementation, a sample indicator screening prompt template can be constructed for the product screening indicator corresponding to the sample association indicator vector, that is, a standard modified template statement is constructed for the product screening indicator so that the conditional screening module can more easily understand the product screening indicator.

[0064] In another feasible implementation, a sample indicator screening prompt template may be directly constructed for the sample association indicator vector, that is, a standard modified template vector may be constructed for the sample association indicator vector so that the condition screening module can more easily understand the sample association indicator vector.

[0065] S304: Using the sample indicator screening prompt template to control the condition screening module to perform indicator screening processing on the sample associated indicator vector based on the sample product selection intention vector to obtain a sample product selection condition containing at least one product screening indicator.

[0066] Among them, the sample indicator screening prompt template is first filled based on the sample associated indicator vector, and then the filled sample indicator screening prompt template is input into the condition screening module. The condition screening module filters the filled sample indicator screening prompt template based on the sample product selection intention vector to obtain the filtered and filled sample indicator screening prompt template. Based on the filtered and filled sample indicator screening prompt template, it is determined that it contains at least one product screening indicator. Then, the sample product selection condition is generated based on the at least one product screening indicator.

[0067] In the process of the conditional screening module screening the filled sample indicator screening prompt template based on the sample product selection intention vector, the conditional screening module can determine the degree of matching between the sample product selection intention vector and the filled sample indicator screening prompt template, and eliminate the filled sample indicator screening prompt template with a lower degree of matching.

[0068] In the embodiment provided in this specification, the sample indicator screening prompt template makes it easier for the conditional screening module to understand the sample-related indicator vector, so as to realize the indicator screening processing of the sample-related indicator vector through the sample selection intention vector, and at the same time, the expected results can be output for the sample-related indicator vector according to the preset format.

[0069] Please refer to Figure 4, which is a schematic diagram of a process for determining a sample selection index vector library according to an embodiment of this specification. Exemplarily, the method includes the following steps.

[0070] S402: Obtain a product screening indicator library and product selection intention text for a product description sample text, wherein the product screening indicator library includes at least one product screening indicator.

[0071] Among them, the product screening index library and product selection intention text can be obtained based on the application scenario by determining the application scenario of the product description sample text. Specifically, when the application scenario of the product description sample text is a fund product selection platform, the product screening index library can be constructed for the label type pre-set for the fund product on the fund product selection platform. Similarly, the product selection intention text can be generated based on the corresponding product screening indicators in the product screening index library. Of course, when the application scenario of the product description sample text is a shopping platform, the product screening index library can be constructed for the label type pre-set for the product on the shopping platform. Similarly, the product selection intention text can be generated based on the corresponding product screening indicators in the product screening index library.

[0072] S404: Construct a set of associated product screening indicators corresponding to the product selection intention text from a product screening indicator library.

[0073] The same product selection intent text corresponds to at least one product screening metric. Therefore, the product screening metrics corresponding to the same product selection intent text can be put into the same set to obtain a set of associated product screening metrics. For example, when the product selection intent text is "profitable," the corresponding product screening metrics include "positive yield" and "rising resistance to decline."

[0074] S406: Perform vectorization processing on the associated product screening indicator set and the product selection intention text to obtain a product selection indicator vector corresponding to the product screening indicator in the associated product screening indicator set.

[0075] Among them, the product selection intention text corresponding to the same associated product screening indicator set is the same. The characteristic vector of each product screening indicator in the associated product screening indicator set and the characteristic vector of the product selection intention text can be determined. Then, the characteristic vector of each product screening indicator and the characteristic vector of the product selection intention text are spliced ​​respectively to obtain the product selection indicator vector corresponding to the product screening indicator in the associated product screening indicator set.

[0076] Specifically, vectorization refers to the process of mapping data (such as text and audio) into a low-dimensional vector space, making it more usable by machine learning models. This mapping process can be learned during model training or processed using a pre-trained model.

[0077] In a feasible implementation, the sample product selection intention vector can be used as a query vector to perform a vector query on the product selection index vector. For example, the product selection index vector can be vectorized and stored, that is, the product selection index vector can be stored based on a vector structure, because the data correlation search of the vector during the vector query process is actually an operation between vectors.

[0078] S408: Generate a sample product selection index vector library including all product selection index vectors.

[0079] Among them, the product selection index vectors corresponding to the product screening index in each associated product screening index set are obtained respectively, and then all the obtained product selection index vectors are put into the same set to obtain a sample product selection index vector library including all product selection index vectors.

[0080] In the embodiment provided in this specification, a set of related product screening indicators is determined from a product screening indicator library based on the product selection intention text, thereby realizing batch generation of product selection indicator vectors corresponding to the product screening indicators based on the product selection intention text and the related product screening indicator set, thereby quickly obtaining a sample product selection indicator vector library including all the said product selection indicator vectors.

[0081] Please refer to Figure 5, which is a schematic diagram of a process for obtaining a sample association index vector according to an embodiment of this specification. For example, in the above embodiment, vector matching of the sample selection intention vector based on the sample selection index vector library is performed to obtain the sample association index vector, including the following steps.

[0082] S502: Determine the vector similarity between the product selection index vector in the sample product selection index vector library and the sample product selection intention vector.

[0083] Among them, the vector similarity between each product selection index vector in the sample product selection index vector library and the sample product selection intention vector can be calculated based on the vector similarity calculation formula.

[0084] Specifically, vector similarity calculation formulas include, but are not limited to, cosine similarity and Jaccard similarity. Cosine similarity: Cosine similarity is a widely used similarity calculation method in vector space. It calculates the cosine of the angle between two vectors. Cosine similarity ranges from -1 to 1, with values ​​closer to 1 indicating greater similarity between the two vectors. Jaccard similarity ranges from 0 to 1, with values ​​closer to 1 indicating greater similarity between the two vectors.

[0085] S504: Screening the vector similarities to obtain target vector similarities.

[0086] Among them, after obtaining the vector similarity between each product selection index vector and the sample product selection intention vector, the vector similarity is screened based on the preset similarity, and the vector similarity less than the preset similarity is eliminated, and then the target vector similarity is obtained; or, the obtained similarity is sorted in descending order, and the vector similarity with the first preset position is taken as the target vector similarity.

[0087] S506: Determine a product selection index vector corresponding to the target vector similarity, and use the product selection index vector corresponding to the target vector similarity as a sample association index vector.

[0088] After obtaining the target vector similarity, we search for the product selection indicator vector corresponding to the target vector similarity. If the product selection indicator vector corresponding to the target vector similarity matches the vector similarity of the sample product selection intent vector, this indicates that the textual intents corresponding to the product selection indicator vector corresponding to the target vector similarity and the sample product selection intent vector are similar. Therefore, the product selection indicator vector corresponding to the target vector similarity can be used as the sample association indicator vector.

[0089] In the embodiment provided in this specification, by calculating the vector similarity between each product selection index vector in the sample product selection index vector library and the sample product selection intention vector, the product selection index vectors are screened by vector similarity to obtain a sample association index vector that is close to the text intention corresponding to the sample product selection intention vector.

[0090] Please refer to Figure 6, which provides a schematic diagram of a process for determining a sample product selection intent vector according to an embodiment of this specification. Exemplarily, the initial text intent recognition model includes a vectorization module, and determining the sample product selection intent vector corresponding to the sample text of the product description through the initial text intent recognition model includes the following steps.

[0091] S602: Determine at least one target keyword corresponding to the product description sample text.

[0092] The product description sample text is segmented to obtain multiple keywords, and then at least one target keyword is determined from the multiple keywords.

[0093] Specifically, the application scenario information corresponding to the product description sample text can be determined. Then, based on the application scenario information, application scenario attributes such as financial attributes, shopping attributes, etc. can be determined. At least one target keyword associated with the application scenario attribute is determined from multiple keywords.

[0094] S604: Determine the semantic features corresponding to the target keyword based on the vectorization module, and generate a sample product selection intention vector corresponding to the product description sample text based on the semantic features.

[0095] Among them, the neural network model in the vectorization module is used to learn the semantic features corresponding to the target keyword, and the semantic features are transformed into a sample product selection intention vector corresponding to the sample text.

[0096] It should be noted that a product description sample text can correspond to one or more target keywords, and each target keyword corresponds to a sample product selection intention vector. Therefore, a product description sample text can correspond to one or more sample product selection intention vectors.

[0097] In the embodiment provided in this specification, by determining at least one target keyword corresponding to the product description sample text, the semantic features of the target keyword are extracted, so as to generate a sample product selection intention vector corresponding to the product description sample text, so that the sample product selection intention vector can more comprehensively represent the text intention of the product description sample text.

[0098] Please refer to FIG. 7. FIG. 7 is a schematic flowchart of a method for determining at least one target keyword corresponding to a product description sample text provided by an embodiment of this specification. Exemplarily, determining at least one target keyword corresponding to a product description sample text includes the following steps.

[0099] S702: Extract reference keywords from the product description sample text based on the vectorization module to obtain reference keywords.

[0100] Among them, the reference keyword can be a keyword with specific semantics in the product description sample text, that is, the reference keyword can not include auxiliary words without specific semantics such as "de", "le", etc. Therefore, the neural network learning model in the vectorization module can be used to extract reference keywords from the product description sample text.

[0101] S704: Obtain the application scenario attribute corresponding to the product description sample text, and determine the association matching value between the reference keyword and the application scenario attribute.

[0102] The application scenario information corresponding to the sample product description text is determined, and then application scenario attributes, such as financial attributes, shopping attributes, etc., are determined based on the application scenario information. After obtaining the application scenario attributes corresponding to the sample product description text, the association matching value between the application scenario attributes corresponding to the sample product description text and the corresponding reference keyword can be searched based on a pre-established mapping table of association matching values ​​between application scenario attributes and keywords.

[0103] A higher correlation match value indicates a stronger correlation between the reference keyword and the application scenario attribute; a lower correlation match value indicates a weaker correlation between the reference keyword and the application scenario attribute. A weaker correlation between a reference keyword and the application scenario attribute indicates that the reference keyword is not suitable as a target keyword to characterize the textual intent of the product description sample text. A stronger correlation between a reference keyword and the application scenario attribute indicates that the reference keyword is suitable as a target keyword to characterize the textual intent of the product description sample text.

[0104] S706: Sort the associated matching values ​​to obtain a sorting result.

[0105] The obtained associated matching values ​​are arranged in descending order to obtain a sorting result arranged in descending order; or the obtained associated matching values ​​are arranged in ascending order to obtain a sorting result arranged in ascending order.

[0106] S708: Determine at least one target-related matching value based on the ranking result, determine a reference keyword corresponding to the target-related matching value, and use the reference keyword corresponding to the target-related matching value as the target keyword.

[0107] After obtaining the sorting results, the ones with higher correlation matching values ​​are used as target correlation matching values ​​based on the sorting results. For example, the first three or the first correlation matching values ​​with higher correlation matching values ​​can be used as target correlation matching values.

[0108] After obtaining the target relevance match value, the corresponding reference keyword is searched for and used as the target keyword. As can be seen, the target keyword is more closely associated with the application scenario attributes, making it more targeted and more suitable for the model to understand the textual intent of the product description sample text.

[0109] In the embodiments provided in this specification, by determining the associated matching values ​​between the reference keywords and the application scenario attributes in the product description sample text, a target keyword with a strong correlation with the application scenario attributes is obtained, so that the screened target keyword is more targeted to the application scenario corresponding to the product description sample text, thereby making it easier for the model to understand the textual intent of the product description sample text based on the target keyword.

[0110] Please refer to Figure 8, which is a flowchart illustrating a method for identifying text intent according to an embodiment of this specification. This method can be implemented using a computer program and run on a text intent recognition device based on the von Neumann architecture. The computer program can be integrated into an application or run as a standalone tool application. The text intent recognition device can be a service platform.

[0111] Specifically, the text intent recognition method includes the following steps.

[0112] S802: Acquire the product description target text input by the user.

[0113] Among them, when the current user faces the online product display interface, since the existing online product display interface is rich in functions and relatively complex, even if a label filtering service is provided, the user still needs to make manual selections. In the embodiment provided in this specification, the user can input the product description target text for the online product display interface based on their own needs to quickly determine the user's target product selection conditions. For example, for fund products, the user can input "Recommend me a fund that guarantees principal" to quickly determine the user's target product selection conditions. Of course, the user can also directly input the voice corresponding to the product description target text, and then convert the voice into text to obtain the product description target text.

[0114] S804: Input the product description target text into the target text intention recognition model, determine the target product selection intention vector corresponding to the product description target text through the target text intention recognition model, determine the target product selection conditions based on the target product selection intention vector, and output the target product selection conditions for the product description target text.

[0115] Among them, after the target text intent recognition model receives the product description target text, it determines the target product selection intention vector corresponding to the product description target text. Among them, the keywords in the product description target text can be vectorized to obtain the target product selection intention vector; or the keywords in the product description target text can be filtered, and the target keywords obtained after filtering are vectorized to obtain the target product selection intention vector. Here, the keywords in the product description target text can be filtered based on the application scenario corresponding to the target text intent recognition model, that is, the keywords with low relevance to the application scenario corresponding to the target text intent recognition model are eliminated.

[0116] After obtaining the target product selection intention vector corresponding to the target text of the product description, query at least one target product selection indicator corresponding to the target product selection intention vector, and determine the target product selection condition based on the at least one target product selection indicator.

[0117] S806: Filter the product set based on the target product selection condition to obtain target products for the target product description text.

[0118] Among them, after obtaining the target product selection conditions, the product set can be filtered according to the target product selection conditions. Specifically, the product set can be filtered based on the matching degree between the product labels of each product in the product set and the target product selection conditions, and the products that meet the target product selection conditions can be filtered out from the product set to obtain the target product for the product description target text.

[0119] Specifically, after obtaining the target product selection conditions, the target text intention recognition model can automatically enter the target product selection conditions into the label filtering service corresponding to the online product display interface based on the application scenario corresponding to the product description target text, thereby obtaining a product list of target products for the product description target text after filtering.

[0120] The embodiments provided in this specification use a target text intention recognition model to accurately identify the user's product selection intention based on the user's ambiguous product description target text, and associate it with the corresponding target product selection conditions, thereby quickly helping the user select target products that meet their real needs. The user does not need to check all the filtering conditions one by one, and can quickly select a list of target products that meet the user's conditions, saving the user's time and greatly improving the user's product selection efficiency.

[0121] In an embodiment provided in this specification, S804 determines the target product selection conditions based on the target product selection intention vector, including: performing vector matching on the target product selection intention vector based on the target product selection indicator vector library to obtain a target associated indicator vector, and determining the target product selection conditions corresponding to the target associated indicator vector based on the target text intention recognition model; wherein the target product selection indicator vector library includes a product selection indicator vector generated based on at least one product screening indicator and a product selection intention text corresponding to the product screening indicator.

[0122] After obtaining the target product selection intent vector, vector matching can be used to associate it with the product selection indicator vectors in the target product selection indicator vector library for the scenario corresponding to the target product description text. The target product selection indicator vector library includes one or more product selection indicator vectors, which are generated based on product screening indicators and the product selection intent text corresponding to the product screening indicators. The product selection intent text can be determined based on the product screening indicators.

[0123] It should be noted that the association between a separate product screening indicator and the target text of the product description is difficult to judge directly, so the association between the product screening indicator and the target text of the product description can be determined by adding the product selection intention text to the product screening indicator. In this way, the association between the product selection indicator vector generated based on the product screening indicator and the product selection intention text corresponding to the product screening indicator and the target product selection intention vector can be judged more directly and quickly. That is, by performing vector matching on each product selection indicator vector in the target product selection indicator vector library and the target product selection intention vector, the product selection indicator vector with a vector matching degree higher than the preset matching value is used as the target association indicator vector. It is easy to understand that the higher the vector matching degree, the higher the matching degree of the product screening indicators corresponding to the target product selection intention vector and the product selection indicator vector.

[0124] After obtaining the target-related indicator vector, the product screening indicator corresponding to the target-related indicator vector is determined, and the target product selection condition is generated based on the product screening indicator corresponding to each target-related indicator vector.

[0125] In the embodiment provided in this specification, the target product selection intention vector is vector matched respectively by each product selection indicator vector in the target product selection indicator vector library, so as to obtain an associated indicator vector that matches the target product selection intention vector, that is, an associated indicator vector with a high similarity to the target product selection intention vector. Since the higher the degree of vector matching, the higher the degree of matching between the target product selection intention vector and the product selection indicator vector, the target product selection conditions corresponding to the target text of the product description can be determined based on the target associated indicator vector.

[0126] In an embodiment provided in the present specification, the target text intention recognition model includes a conditional screening module based on a basic large language generation model. In the above embodiment, the target product selection condition corresponding to the target associated indicator vector is determined based on the target text intention recognition model, including: generating an indicator screening prompt template for the target associated indicator vector; using the indicator screening prompt template to control the conditional screening module to perform indicator screening processing on the target associated indicator vector based on the target product selection intention vector to obtain a target product selection condition containing at least one product screening indicator.

[0127] In order for the conditional screening module to output a desired result based on the target correlation indicator vector, it is necessary to construct a target indicator screening prompt template for the target correlation indicator vector. That is, a target indicator screening prompt template is constructed for the target correlation indicator vector that is easily understood by the basic large language generation model corresponding to the conditional screening module. The basic large language generation model can refer to a language model containing hundreds of billions (or more) parameters, which are trained on a large amount of text data.

[0128] In a feasible implementation, a target indicator screening prompt template can be constructed for the product screening indicator corresponding to the target association indicator vector, that is, a standard modified template statement is constructed for the product screening indicator so that the conditional screening module can more easily understand the product screening indicator.

[0129] In another feasible implementation, a target indicator screening prompt template may be directly constructed for the target association indicator vector, that is, a standard modified template vector may be constructed for the target association indicator vector so that the condition screening module can more easily understand the target association indicator vector.

[0130] After obtaining the indicator screening prompt template, the target indicator screening prompt template is filled based on the target associated indicator vector, and then the filled target indicator screening prompt template is input into the condition screening module. The condition screening module filters the filled target indicator screening prompt template based on the target product selection intention vector to obtain the filtered and filled target indicator screening prompt template. Based on the filtered and filled target indicator screening prompt template, it is determined that it contains at least one product screening indicator. Then, the target product selection condition is generated based on the at least one product screening indicator.

[0131] In the process of the conditional screening module screening the filled target indicator screening prompt template based on the target product selection intention vector, the conditional screening module can determine the degree of matching between the target product selection intention vector and the filled target indicator screening prompt template, and eliminate the filled target indicator screening prompt template with a lower degree of matching.

[0132] In the embodiment provided in this specification, the target indicator screening prompt template makes it easier for the conditional screening module to understand the target-related indicator vector, so as to realize the indicator screening processing of the target-related indicator vector through the target product selection intention vector, and at the same time, the target-related indicator vector can be output according to the preset format to meet the expected results.

[0133] The following will be combined with Figure 9, which is a structural schematic diagram of a text intent recognition model training device provided in an embodiment of this specification, to provide a detailed introduction to the text intent recognition model training device provided in an embodiment of this specification. It should be noted that the text intent recognition model training device shown in Figure 9 is used to execute the method of the embodiment shown in Figures 1 to 7 of this specification. For ease of explanation, only the parts related to this specification are shown. For specific technical details not disclosed, please refer to the embodiment shown in Figures 1 to 7 of this specification.

[0134] Please refer to Figure 9, which shows a structural diagram of the text intention recognition model training device of the present specification. The text intention recognition model training device 1 can be implemented as all or part of the user terminal through software, hardware or a combination of both. According to some embodiments, the text intention recognition model training device 1 includes a creation module 11, a processing module 12 and a training module 13, which are specifically used for: the creation module 11 is suitable for creating an initial text intention recognition model, obtaining a product description sample text and a reference product selection condition corresponding to the product description sample text; the processing module 12 is suitable for inputting the product description sample text into the initial text intention recognition model for model training, determining the sample product selection intention vector corresponding to the product description sample text through the initial text intention recognition model, and determining the sample product selection condition based on the sample product selection intention vector; the training module 13 is suitable for adjusting the model parameters of the initial text intention recognition model based on the sample product selection condition and the reference product selection condition during the model training process to obtain the target text intention recognition model.

[0135] Optionally, the processing module 12 is suitable for performing vector matching on the sample product selection intention vector based on the sample product selection index vector library to obtain a sample association index vector, and determining the sample product selection condition corresponding to the sample association index vector based on the initial text intention recognition model; wherein, the sample product selection index vector library includes a product selection index vector generated based on at least one product screening indicator and a product selection intention text corresponding to the product screening indicator.

[0136] Optionally, the creation module 11 includes: a creation unit, suitable for using a basic large language generation model to form a conditional screening module, and creating an initial text intention recognition model that at least includes the conditional screening module; the processing module 12 includes: a generation unit, suitable for generating a sample indicator screening prompt template for a sample associated indicator vector; an indicator screening processing unit, suitable for using a sample indicator screening prompt template to control the conditional screening module to perform indicator screening processing on the sample associated indicator vector based on the sample product selection intention vector, and obtain a sample product selection condition containing at least one product screening indicator.

[0137] Optionally, the text intention recognition model training device 1 also includes: a product description sample text association acquisition module, suitable for obtaining a product screening index library and product selection intention text for the product description sample text, the product screening index library including at least one product screening index; a construction module, suitable for constructing an associated product screening index set corresponding to the product selection intention text from the product screening index library; a vectorization module, suitable for vectorizing the associated product screening index set and the product selection intention text to obtain a product selection index vector corresponding to the product screening index in the associated product screening index set; a generation module, suitable for generating a sample product selection index vector library including all product selection indicator vectors.

[0138] Optionally, the processing module 12 includes: a vector similarity determination unit, suitable for determining the vector similarity between the product selection index vector in the sample product selection index vector library and the sample product selection intention vector; a target vector similarity determination unit, suitable for screening the vector similarity to obtain the target vector similarity; a sample association indicator vector determination unit, suitable for determining the product selection index vector corresponding to the target vector similarity, and using the product selection index vector corresponding to the target vector similarity as the sample association indicator vector.

[0139] Optionally, the initial text intent recognition model includes a vectorization module, and the processing module 12 includes: a target keyword determination unit, suitable for determining at least one target keyword corresponding to the product description sample text; a sample product selection intention vector determination unit, suitable for determining the semantic features corresponding to the target keyword based on the vectorization module, and generating a sample product selection intention vector corresponding to the product description sample text based on the semantic features.

[0140] Optionally, the target keyword determination unit includes: a reference keyword determination subunit, suitable for extracting reference keywords from the product description sample text based on the vectorization module to obtain reference keywords; an acquisition subunit, suitable for obtaining application scenario attributes corresponding to the product description sample text, and determining the association matching value between the reference keyword and the application scenario attribute; a sorting subunit, suitable for sorting the association matching value to obtain a sorting result; a target keyword determination subunit, suitable for determining at least one target association matching value based on the sorting result, determining the reference keyword corresponding to the target association matching value, and using the reference keyword corresponding to the target association matching value as the target keyword.

[0141] It should be noted that the text intent recognition model training device provided in the above embodiment only uses the division of the above functional modules as an example when executing the text intent recognition model training method. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the text intent recognition model training device provided in the above embodiment and the text intent recognition model training method embodiment belong to the same concept. The implementation process is detailed in the method embodiment and will not be repeated here.

[0142] The above serial numbers in this specification are for description only and do not represent the advantages or disadvantages of the embodiments.

[0143] The text intent recognition device provided in this specification will be described in detail below with reference to FIG10, which is a schematic structural diagram of a text intent recognition device provided in an embodiment of this specification. It should be noted that the text intent recognition device shown in FIG10 is used to execute the method of the embodiment shown in FIG8 of this specification. For ease of explanation, only the parts relevant to this specification are shown. For specific technical details not disclosed, please refer to the embodiment shown in FIG8 of this specification.

[0144] Please refer to Figure 10, which shows a structural diagram of the text intention recognition device of this specification. The text intention recognition device 2 can be implemented as all or part of the user terminal through software, hardware or a combination of both. According to some embodiments, the text intention recognition device 2 includes an acquisition module 21, an input module 22 and an output module 23, which are specifically used for: the acquisition module 21 is suitable for acquiring the product description target text input by the user; the input module 22 is suitable for inputting the product description target text into the target text intention recognition model, determining the target product selection intention vector corresponding to the product description target text through the target text intention recognition model and determining the target product selection conditions based on the target product selection intention vector, and outputting the target product selection conditions for the product description target text; the output module 23 is suitable for screening the product set based on the target product selection conditions to obtain the target product for the product description target text.

[0145] Optionally, the input module 22 is suitable for performing vector matching on the target product selection intention vector based on the target product selection indicator vector library to obtain a target association indicator vector, and determining the target product selection condition corresponding to the target association indicator vector based on the target text intention recognition model; wherein, the target product selection indicator vector library includes a product selection indicator vector generated based on at least one product screening indicator and a product selection intention text corresponding to the product screening indicator.

[0146] Optionally, the target text intention recognition model includes a conditional screening module based on a basic large language generation model, and the input module 22 includes: an indicator screening prompt template generation unit, suitable for generating an indicator screening prompt template for a target-related indicator vector; a target product selection condition determination unit, suitable for using the indicator screening prompt template to control the condition screening module to perform indicator screening processing on the target-related indicator vector based on the target product selection intention vector to obtain a target product selection condition containing at least one product screening indicator.

[0147] It should be noted that the text intent recognition device provided in the above embodiment only uses the division of the above functional modules as an example when executing the text intent recognition method. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the text intent recognition device provided in the above embodiment and the text intent recognition method embodiment belong to the same concept. The implementation process thereof is detailed in the method embodiment and will not be repeated here.

[0148] This specification also provides a computer storage medium, which can store multiple instructions. The instructions are suitable for being loaded by a processor and executing the text intent recognition model training method or text intent recognition method of the embodiments shown in Figures 1 to 8 above. The specific execution process can be found in the specific description of the embodiments shown in Figures 1 to 8, which will not be repeated here.

[0149] This specification also provides a computer program product, which stores at least one instruction, and the at least one instruction is loaded by the processor and executed by the text intent recognition model training method or text intent recognition method of the embodiments shown in Figures 1 to 8 above. The specific execution process can be found in the specific description of the embodiments shown in Figures 1 to 8, and will not be repeated here.

[0150] Please refer to Figure 11, which is a block diagram of the structure of an electronic device provided in an embodiment of this specification. The electronic device described in this specification may include one or more of the following components: a processor 110, a memory 120, an input device 130, an output device 140, and a bus 150. The processor 110, the memory 120, the input device 130, and the output device 140 may be connected via the bus 150.

[0151] The processor 110 may include one or more processing cores. The processor 110 utilizes various interfaces and circuits to connect various components within the electronic device. It executes instructions, programs, code sets, or instruction sets stored in the memory 120, as well as accesses data stored in the memory 120, to perform various functions of the electronic device and process data. Optionally, the processor 110 may be implemented using at least one of the following hardware forms: a digital signal processing (DSP), a field-programmable gate array (FPGA), or a programmable logic array (PLA). The processor 110 may integrate one or a combination of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. The CPU primarily processes the operating system, user interface, and application programs; the GPU is responsible for rendering and drawing display content; and the modem handles wireless communications. It is understood that the modem may not be integrated into the processor 110 and may be implemented separately via a communications chip.

[0152] The memory 120 may include a random access memory (RAM) or a read-only memory (ROM). Optionally, the memory 120 includes a non-transitory computer-readable storage medium. The memory 120 may be used to store instructions, programs, codes, code sets, or instruction sets. The memory 120 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the following various method embodiments, etc. The operating system may be an Android system, including a system deeply developed based on the Android system, an IOS system developed by Apple, including a system deeply developed based on the IOS system or other systems. The data storage area may also store data created by the electronic device during use, such as a phone book, audio and video data, chat record data, etc.

[0153] Please refer to Figure 12, which is a structural diagram of an operating system and user space provided in an embodiment of this specification. The memory 120 can be divided into an operating system space and a user space. The operating system runs in the operating system space, and native and third-party applications run in the user space. In order to ensure that different third-party applications can achieve better operating results, the operating system allocates corresponding system resources to different third-party applications. However, the requirements for system resources in different application scenarios in the same third-party application are also different. For example, in the local resource loading scenario, the third-party application has higher requirements for disk reading speed; in the animation rendering scenario, the third-party application has higher requirements for GPU performance. The operating system and the third-party application are independent of each other. The operating system often cannot perceive the current application scenario of the third-party application in a timely manner, resulting in the operating system being unable to perform targeted system resource adaptation according to the specific application scenario of the third-party application.

[0154] In order for the operating system to distinguish the specific application scenarios of third-party applications, it is necessary to open up data communication between third-party applications and the operating system so that the operating system can obtain the current scenario information of third-party applications at any time, and then perform targeted system resource adaptation based on the current scenario.

[0155] Figure 13 is an architectural diagram of the Android operating system shown in Figure 12 . Taking the Android operating system as an example, the programs and data stored in memory 120 are shown in Figure 13 . Memory 120 may store a Linux kernel layer 320, a system runtime library layer 340, an application framework layer 360, and an application layer 380. The Linux kernel layer 320, the system runtime library layer 340, and the application framework layer 360 belong to the operating system space, while the application layer 380 belongs to the user space. The Linux kernel layer 320 provides underlying drivers for various hardware components of electronic devices, such as display drivers, audio drivers, camera drivers, Bluetooth drivers, Wi-Fi drivers, and power management. The system runtime library layer 340 provides key feature support for the Android system through several C / C++ libraries. For example, the SQLite library provides database support, the OpenGL / ES library provides 3D drawing support, and the Webkit library provides browser kernel support. The system runtime library layer 340 also includes the Android runtime, which primarily provides core libraries that allow developers to write Android applications in Java. The application framework layer 360 provides various APIs that may be used when building applications. Developers can also use these APIs to build their own applications, such as activity management, window management, view management, notification management, content provider management, package management, call management, resource management, and location management. The application layer 380 runs at least one application. These applications can be native applications that come with the operating system, such as contacts, SMS, clock, and camera applications, or third-party applications developed by third-party developers, such as games, instant messaging programs, and photo enhancement programs.

[0156] Figure 14 is an architectural diagram of the iOS operating system in Figure 12. Taking the iOS operating system as an example, the programs and data stored in the memory 120 are shown in Figure 14. The iOS system includes: a core operating system layer 420 (Core OS layer), a core services layer 440 (Core Services layer), a media layer 460 (Media layer), and a touchable layer 480 (Cocoa Touch Layer). The core operating system layer 420 includes the operating system kernel, drivers, and underlying program frameworks. These underlying program frameworks provide functions closer to the hardware for use by the program frameworks located in the core services layer 440. The core services layer 440 provides system services and / or program frameworks required by applications, such as the foundation framework, account framework, advertising framework, data storage framework, network connection framework, geographic location framework, motion framework, etc. The media layer 460 provides audio-visual interfaces for applications, such as graphics and image-related interfaces, audio technology-related interfaces, video technology-related interfaces, and wireless playback (AirPlay) interfaces for audio and video transmission technology. The touchable layer 480 provides various commonly used interface-related frameworks for application development. It is responsible for user touch interaction operations on electronic devices, such as local notification services, remote push services, advertising frameworks, game tool frameworks, message user interface (UI) frameworks, UIKit frameworks, and map frameworks.

[0157] In the frameworks shown in FIG14 , those relevant to most applications include, but are not limited to, the Foundation framework in the core services layer 440 and the UIKit framework in the touchable layer 480. The Foundation framework provides many basic object classes and data types, offering fundamental system services for all applications and having nothing to do with the UI. The classes provided by the UIKit framework are the foundational UI class library for creating touch-based user interfaces. iOS applications can use the UIKit framework to provide their UIs, providing the application infrastructure for building user interfaces, drawing, handling user interaction events, responding to gestures, and so on.

[0158] Among them, the method and principle of implementing data communication between third-party applications and the operating system in the IOS system can be referred to the Android system, and this manual will not go into details here.

[0159] Among them, the input device 130 is used to receive input instructions or data, and the input device 130 includes but is not limited to a keyboard, a mouse, a camera, a microphone or a touch device. The output device 140 is used to output instructions or data, and the output device 140 includes but is not limited to a display device and a speaker. In one example, the input device 130 and the output device 140 can be combined, and the input device 130 and the output device 140 are touch screen displays, which are used to receive touch operations on or near the touch screen using any suitable object such as a finger or a touch pen, and to display the user interface of each application. The touch screen display is usually provided on the front panel of the electronic device. The touch screen display can be designed as a full screen, a curved screen or a special-shaped screen. The touch screen display can also be designed as a combination of a full screen and a curved screen, or a combination of a special-shaped screen and a curved screen, which is not limited in this specification.

[0160] In addition, those skilled in the art will understand that the structures of the electronic devices shown in the above figures do not limit the electronic devices. The electronic devices may include more or fewer components than shown, or may combine certain components, or arrange the components differently. For example, the electronic devices may also include radio frequency circuits, input units, sensors, audio circuits, wireless fidelity (WiFi) modules, power supplies, Bluetooth modules, and other components, which are not described in detail here.

[0161] In this specification, the execution entity of each step can be the electronic device described above. Optionally, the execution entity of each step is the operating system of the electronic device. The operating system can be Android, iOS, or other operating systems, which is not limited in this specification.

[0162] The electronic device of this specification may also be equipped with a display device, which may be any device capable of realizing a display function, such as a cathode ray tube display (CR), a light-emitting diode display (LED), an electronic ink screen, a liquid crystal display (LCD), a plasma display panel (PDP), etc. The user may use the display device on the electronic device 101 to view displayed text, images, videos and other information. The electronic device may be a smart phone, a tablet computer, a gaming device, an AR (Augmented Reality) device, a car, a data storage device, an audio playback device, a video playback device, a notebook, a desktop computing device, a wearable device such as an electronic watch, electronic glasses, an electronic helmet, an electronic bracelet, an electronic necklace, electronic clothing and the like.

[0163] In the electronic device shown in Figure 11, which can be a terminal, the processor 110 can be used to call the text intent recognition model training program stored in the memory 120, and specifically perform the following operations: create an initial text intent recognition model, obtain product description sample text and reference product selection conditions corresponding to the product description sample text; input the product description sample text into the initial text intent recognition model for model training, determine the sample product selection intention vector corresponding to the product description sample text through the initial text intent recognition model, and determine the sample product selection conditions based on the sample product selection intention vector; during the model training process, adjust the model parameters of the initial text intent recognition model based on the sample product selection conditions and the reference product selection conditions to obtain the target text intent recognition model.

[0164] In one embodiment, when the processor 110 determines the sample product selection conditions based on the sample product selection intention vector, it specifically performs the following operations: performing vector matching on the sample product selection intention vector based on the sample product selection indicator vector library to obtain a sample association indicator vector, and determining the sample product selection conditions corresponding to the sample association indicator vector based on the initial text intention recognition model; wherein the sample product selection indicator vector library includes a product selection indicator vector generated based on at least one product screening indicator and a product selection intention text corresponding to the product screening indicator.

[0165] In one embodiment, when executing the creation of the initial text intent recognition model, the processor 110 specifically performs the following operations: using the basic large language generation model to form a conditional screening module, and creating an initial text intent recognition model that at least includes the conditional screening module; determining the sample product selection conditions corresponding to the sample association indicator vector based on the initial text intent recognition model, including: generating a sample indicator screening prompt template for the sample association indicator vector; using the sample indicator screening prompt template to control the conditional screening module to perform indicator screening processing on the sample association indicator vector based on the sample product selection intention vector, and obtain a sample product selection condition containing at least one product screening indicator.

[0166] In one embodiment, the processor 110 further specifically performs the following operations: obtaining a product screening indicator library and a product selection intention text for a product description sample text, wherein the product screening indicator library includes at least one product screening indicator; constructing an associated product screening indicator set corresponding to the product selection intention text from the product screening indicator library; vectorizing the associated product screening indicator set and the product selection intention text to obtain a product selection indicator vector corresponding to the product screening indicator in the associated product screening indicator set; and generating a sample product selection indicator vector library including all product selection indicator vectors.

[0167] In one embodiment, when the processor 110 performs vector matching on the sample product selection intention vector based on the sample product selection index vector library to obtain the sample association index vector, it specifically performs the following operations: determining the vector similarity between the product selection index vector in the sample product selection index vector library and the sample product selection intention vector; filtering the vector similarity to obtain the target vector similarity; determining the product selection index vector corresponding to the target vector similarity, and using the product selection index vector corresponding to the target vector similarity as the sample association index vector.

[0168] In one embodiment, the initial text intent recognition model includes a vectorization module. When the processor 110 determines the sample product selection intention vector corresponding to the product description sample text through the initial text intent recognition model, it specifically performs the following operations: determining at least one target keyword corresponding to the product description sample text; determining the semantic features corresponding to the target keywords based on the vectorization module, and generating the sample product selection intention vector corresponding to the product description sample text based on the semantic features.

[0169] In one embodiment, when the processor 110 determines at least one target keyword corresponding to the product description sample text, it specifically performs the following operations: extracting reference keywords from the product description sample text based on the vectorization module to obtain reference keywords; obtaining application scenario attributes corresponding to the product description sample text, and determining the associated matching value between the reference keyword and the application scenario attribute; sorting the associated matching value to obtain a sorting result; determining at least one target associated matching value based on the sorting result, determining the reference keyword corresponding to the target associated matching value, and using the reference keyword corresponding to the target associated matching value as the target keyword.

[0170] In an embodiment provided in this specification, the processor 110 can be used to call the text intent recognition program stored in the memory 120, and specifically perform the following operations: obtain the product description target text input by the user; input the product description target text into the target text intent recognition model, determine the target product selection intention vector corresponding to the product description target text through the target text intent recognition model, and determine the target product selection conditions based on the target product selection intention vector, and output the target product selection conditions for the product description target text; screen the product set based on the target product selection conditions to obtain the target product for the product description target text.

[0171] In one embodiment, when the processor 110 determines the target product selection conditions based on the target product selection intention vector, it specifically performs the following operations: performing vector matching on the target product selection intention vector based on the target product selection indicator vector library to obtain a target association indicator vector, and determining the target product selection conditions corresponding to the target association indicator vector based on the target text intention recognition model; wherein the target product selection indicator vector library includes a product selection indicator vector generated based on at least one product screening indicator and a product selection intention text corresponding to the product screening indicator.

[0172] In one embodiment, the target text intention recognition model includes a conditional screening module based on a basic large language generation model. When the processor 110 determines the target product selection condition corresponding to the target associated indicator vector based on the target text intention recognition model, it specifically performs the following operations: generates an indicator screening prompt template for the target associated indicator vector; uses the indicator screening prompt template to control the conditional screening module to perform indicator screening processing on the target associated indicator vector based on the target product selection intention vector to obtain a target product selection condition containing at least one product screening indicator.

[0173] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware through a computer program. The program can be stored in a computer-readable storage medium, and when executed, the program can include the processes in the above-described method embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0174] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.), and signals involved in the embodiments of this specification are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. For example, the object characteristics, interactive behavior characteristics, and user information involved in this specification are all obtained with full authorization.

[0175] The above disclosure is only a preferred embodiment of this specification, and certainly cannot be used to limit the scope of rights of this specification. Therefore, equivalent changes made according to the claims of this specification are still within the scope covered by this specification.

Claims

1. A text intent recognition model training method, the method comprising: Creating an initial text intent recognition model, obtaining product description sample text and reference product selection conditions corresponding to the product description sample text; Inputting the product description sample text into the initial text intent recognition model for model training, determining the sample product selection intention vector corresponding to the product description sample text through the initial text intent recognition model, and determining the sample product selection condition based on the sample product selection intention vector; During the model training process, the model parameters of the initial text intent recognition model are adjusted based on the sample product selection conditions and the reference product selection conditions to obtain a target text intent recognition model.

2. The method according to claim 1, wherein determining the sample product selection conditions based on the sample product selection intention vector comprises: Perform vector matching on the sample product selection intention vector based on the sample product selection indicator vector library to obtain a sample association indicator vector, and determine the sample product selection condition corresponding to the sample association indicator vector based on the initial text intent recognition model; Among them, the sample product selection index vector library includes a product selection index vector generated based on at least one product screening index and the product selection intention text corresponding to the product screening index.

3. The method according to claim 2, wherein creating an initial text intent recognition model comprises: Using the basic large language generation model to build a conditional screening module, creating an initial text intent recognition model that at least includes the conditional screening module; The determining of the sample selection condition corresponding to the sample association index vector based on the initial text intent recognition model includes: generating a sample indicator screening prompt template for the sample-related indicator vector; The sample indicator screening prompt template is used to control the condition screening module to perform indicator screening processing on the sample associated indicator vector based on the sample product selection intention vector to obtain a sample product selection condition containing at least one product screening indicator.

4. The method according to claim 2, further comprising: Obtaining a product screening index library and product selection intention text for the product description sample text, wherein the product screening index library includes at least one product screening index; Constructing a set of related product screening indicators corresponding to the product selection intention text from the product screening indicator library; Performing vectorization processing on the associated product screening indicator set and the product selection intention text to obtain a product selection indicator vector corresponding to the product screening indicator in the associated product screening indicator set; Generate a sample product selection index vector library including all the product selection index vectors.

5. The method according to claim 2, wherein the step of performing vector matching on the sample product selection intention vector based on the sample product selection indicator vector library to obtain the sample association indicator vector comprises: Determining the vector similarity between the product selection indicator vector in the sample product selection indicator vector library and the sample product selection intention vector; Performing screening processing on the vector similarities to obtain target vector similarities; Determine a product selection index vector corresponding to the target vector similarity, and use the product selection index vector corresponding to the target vector similarity as a sample association index vector.

6. The method according to claim 1, wherein the initial text intent recognition model includes a vectorization module, and determining the sample product selection intent vector corresponding to the product description sample text using the initial text intent recognition model includes: Determining at least one target keyword corresponding to the product description sample text; The semantic features corresponding to the target keywords are determined based on the vectorization module, and a sample product selection intention vector corresponding to the product description sample text is generated based on the semantic features.

7. The method according to claim 6, wherein determining at least one target keyword corresponding to the product description sample text comprises: Extracting reference keywords from the product description sample text based on the vectorization module to obtain reference keywords; Obtaining application scenario attributes corresponding to the product description sample text, and determining an associated matching value between the reference keyword and the application scenario attribute; Sorting the associated matching values ​​to obtain a sorting result; At least one target association matching value is determined based on the ranking result, a reference keyword corresponding to the target association matching value is determined, and the reference keyword corresponding to the target association matching value is used as a target keyword.

8. A method for identifying text intent, comprising: Get the product description target text entered by the user; Inputting the product description target text into a target text intention recognition model, determining a target product selection intention vector corresponding to the product description target text through the target text intention recognition model, determining a target product selection condition based on the target product selection intention vector, and outputting the target product selection condition for the product description target text; The product set is screened based on the target product selection condition to obtain target products for the product description target text.

9. The method according to claim 8, wherein determining target product selection conditions based on the target product selection intention vector comprises: Perform vector matching on the target product selection intention vector based on the target product selection indicator vector library to obtain a target association indicator vector, and determine the target product selection condition corresponding to the target association indicator vector based on the target text intention recognition model; Among them, the target product selection index vector library includes a product selection index vector generated based on at least one product screening index and the product selection intention text corresponding to the product screening index.

10. The method according to claim 9, wherein the target text intention recognition model includes a conditional screening module based on a basic large language generation model. The determining of the target product selection condition corresponding to the target association index vector based on the target text intention recognition model includes: generating an indicator screening prompt template for the target-related indicator vector; The indicator screening prompt template is used to control the condition screening module to perform indicator screening processing on the target-related indicator vector based on the target product selection intention vector to obtain a target product selection condition containing at least one product screening indicator.

11. A text intent recognition model training device, the device comprising: A creation module, adapted to create an initial text intent recognition model, obtain product description sample text and reference product selection conditions corresponding to the product description sample text; a processing module adapted to input the product description sample text into the initial text intent recognition model for model training, determine a sample product selection intention vector corresponding to the product description sample text through the initial text intent recognition model, and determine a sample product selection condition based on the sample product selection intention vector; The training module is suitable for adjusting the model parameters of the initial text intent recognition model based on the sample selection conditions and the reference product selection conditions during the model training process to obtain a target text intent recognition model.

12. A text intent recognition device, comprising: An acquisition module, adapted to acquire a product description target text input by a user; an input module adapted to input the product description target text into a target text intention recognition model, determine a target product selection intention vector corresponding to the product description target text through the target text intention recognition model, determine a target product selection condition based on the target product selection intention vector, and output the target product selection condition for the product description target text; The output module is adapted to screen the product set based on the target product selection condition to obtain target products for the target product description text.

13. A computer storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the method steps according to any one of claims 1 to 7 or 8 to 10.

14. A computer program product, the computer program product storing at least one instruction, wherein the at least one instruction is loaded by a processor and executes the method steps according to any one of claims 1 to 7 or 8 to 10.

15. An electronic device comprising: A processor and a memory; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the method steps according to any one of claims 1 to 7 or 8 to 10.

Images