Business opportunity-based enterprise information recommendation method and device, equipment and medium

Abstract

The invention discloses an enterprise information recommendation method and device based on business opportunities, equipment and a medium. The method comprises the following steps: calling a large model to process unstructured commodity text information input by a user, and outputting a target product term; querying at least one candidate customs code matched with the target product term and historical matching data associated with the target product term from a historical business database; obtaining an interpretation description text of each candidate customs code; determining an association weight between the target product term and each candidate customs code according to historical matching data associated with the target product term and the explanation description text of each candidate customs code; and according to the association weight between the target product term and each candidate customs code, selecting an optimal customs code and at least one candidate customs code from the candidate customs codes, and recommending the optimal customs code and the at least one candidate customs code to the user. The method can effectively solve the problem that artificial rules are poor in adaptation to network new words, industry abbreviations and semantic ambiguity.

Description

Technical Field

[0001] This invention relates to the fields of information recommendation and artificial intelligence technology, and in particular to a method, apparatus, electronic device and storage medium for recommending enterprise information based on business opportunities. Background Technology

[0002] Customs codes are a standardized coding system used to associate product terms with commodity classifications. They contain thousands of sub-codes and are updated annually according to policy changes. Their core function is to establish a mapping relationship between products and specific classification identifiers, providing a unified basis for customs supervision, classification statistics, and other aspects of international trade.

[0003] The current mainstream method for recommending customs codes is a manually pre-set keyword-code mapping rule scheme. Its core principle relies on manually setting correspondence rules between product terms and customs codes, and then matching codes by determining whether the product terms contain specific keywords. This method has certain shortcomings: it relies solely on the rigid judgment of keyword presence and cannot handle situations where product terminology is inaccurate, resulting in low accuracy in the recommended customs codes. Therefore, a new information recommendation method for recommending customs codes is urgently needed. Summary of the Invention

[0004] This invention provides a method, apparatus, electronic device, storage medium, and computer program product for recommending enterprise information based on business opportunities.

[0005] According to one aspect of the present invention, a method for recommending enterprise information based on business opportunities is provided, comprising:

[0006] The large model is invoked to process the unstructured product text information input by the user and output the target product terminology;

[0007] Retrieve from the historical business database at least one candidate customs code that matches the target product term, as well as historical matching data associated with the target product term;

[0008] Obtain the explanatory description text for each candidate customs code;

[0009] Based on historical matching data associated with the target product term and the explanatory description text of each candidate customs code, determine the association weight between the target product term and each candidate customs code;

[0010] Based on the association weight between the target product terminology and each candidate customs code, an optimal customs code and at least one alternative customs code are selected from the candidate customs codes as recommendation information and recommended to the user.

[0011] According to another aspect of the present invention, a business opportunity-based enterprise information recommendation device is provided, comprising:

[0012] The model processing module is used to call a large model to process the unstructured product text information input by the user and output the target product terms.

[0013] The query module is used to query the historical business database for at least one candidate customs code that matches the target product term, as well as historical matching data associated with the target product term;

[0014] The acquisition module is used to acquire the explanatory description text for each candidate customs code;

[0015] The weight calculation module is used to determine the association weight between the target product term and each candidate customs code based on historical matching data associated with the target product term and the explanatory description text of each candidate customs code.

[0016] The recommendation module is used to select an optimal customs code and at least one alternative customs code from the candidate customs codes based on the association weight between the target product term and each candidate customs code, and recommend them to the user.

[0017] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0018] At least one processor; and

[0019] A memory that is communicatively connected to at least one processor; wherein,

[0020] The memory stores a computer program that can be executed by at least one processor, such that the at least one processor is able to perform the business opportunity-based enterprise information recommendation method according to embodiments of the present invention.

[0021] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the business opportunity-based enterprise information recommendation method of the present invention.

[0022] According to another aspect of the present invention, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the steps in the above-described method.

[0023] The technical solution of this invention can identify new words and abbreviations through multi-round parsing of a large model and output standardized terms. The two-factor model does not rely on rigid keyword matching, but makes a comprehensive judgment through historical statistics and semantic association, which improves the fault tolerance and makes the adaptation rate significantly high. This solves the problem that traditional manual rules have poor adaptation ability to new online words, industry abbreviations and semantic ambiguity.

[0024] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart illustrating a business opportunity-based enterprise information recommendation method provided in an embodiment of the present invention.

[0027] Figure 2 This is a flowchart illustrating another enterprise information recommendation method based on business opportunities provided by an embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram of the structure of the enterprise information recommendation device based on business opportunities provided in an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram of the structure of an electronic device that implements the business opportunity-based enterprise information recommendation method according to an embodiment of the present invention. Detailed Implementation

[0030] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0031] Example 1

[0032] Figure 1This is a flowchart of a business opportunity-based enterprise information recommendation method provided by an embodiment of the present invention. This embodiment can be applied to the scenario of recommending customs codes. The method can be executed by an information recommendation device, which can be implemented in hardware and / or software and can be configured in an electronic device.

[0033] like Figure 1 As shown, business opportunity-based enterprise information recommendation methods include:

[0034] S101. Call the large model to process the unstructured product text information input by the user and output the target product terminology.

[0035] Among them, unstructured product text information refers to text that has not been standardized and contains messy information such as product name, brand, specifications, and language. For example, unstructured product text information is "Bujías de encendidopara motor de coche (brand XYZ) specifications 10x20cm"; large model refers to LLM (large language model) with semantic parsing capabilities, which is used to process unstructured text and extract core information; target product terminology refers to standardized product terminology, which is a unified and standardized expression of product name.

[0036] In some embodiments, after receiving unstructured product text information input by the user, the unstructured product text information is parsed step by step through a multi-round chain-like thinking prompting engineering-driven large model. For example, the language of the text is first identified, irrelevant information (such as brand and specifications) is cleaned, core product features are extracted, and finally a unique standardized target product terminology is output. If necessary, the core attributes of the product can be output simultaneously.

[0037] Understandably, by using a large model to transform unstructured commodity text information into structured target product data, the differences in the expression of unstructured text (such as foreign languages, abbreviations, and redundant information) are eliminated, providing a unified data foundation for subsequent association with customs codes, avoiding code matching failures caused by non-standard expressions, and replacing manual text processing operations, thus improving information processing efficiency.

[0038] S102. Query the historical business database for at least one candidate customs code that matches the target product term, as well as the historical matching data associated with the target product term.

[0039] The historical business database refers to a database that stores past records of matching product terms with customs codes, including data such as the total number of occurrences of the product term and the historical co-occurrence counts of the product term with each customs code. The candidate customs code refers to the set of all customs codes in the historical data that have matching records with the target product term. The candidate customs code set C = {c_1, c_2, ..., c_n}, where c_1, c_2, ..., c_n represent different candidate customs codes. For example, the candidate customs code set corresponding to the target product term "spark plug" is as follows: {851110, 854800, 850790, 851180, 854710}. The historical matching data includes two core types of data: the total number of occurrences of the target product term Freq(p) and the historical co-occurrence counts of the target product term with each candidate customs code Freq(p, c_i), where p is the target product term and c_i is any candidate customs code that co-occurs with the target product term p.

[0040] In practice, the target product term output by S101 is used as the query condition to access the historical business database. Specifically, firstly, all customs codes that have matching records with the target product term are queried to form a set of candidate customs codes. Then, historical matching data is retrieved, which includes Freq(p,c_i) and Freq(p) corresponding to each candidate code in the set, and an association mapping between the target product term, candidate customs code, and historical matching data is established.

[0041] S103. Obtain the explanatory description text for each candidate customs code.

[0042] The explanatory description text refers to the standardized Chinese description text of the commodity category corresponding to each customs code in the official customs code database. It is used to clarify the commodity reference of the customs code. For example, the explanatory description text of candidate customs code 851110 is "spark plug", and that of 851180 is "other devices for electric ignition and starting of engines".

[0043] In practice, based on the set of candidate customs codes determined in S102, the official database of customs codes is accessed, and the official Chinese explanatory text corresponding to each candidate customs code is retrieved one by one to ensure that the explanatory text completely matches the commodity category of the customs code, thus providing a unified reference benchmark for subsequent semantic analysis.

[0044] Understandably, this step provides a crucial basis for resolving semantic ambiguity issues. By clarifying the official description of candidate codes, the semantic relevance between target product terms and customs codes can be accurately determined, avoiding mismatches caused by ambiguous code meanings.

[0045] S104. Based on the historical matching data associated with the target product term and the explanatory description text of each candidate customs code, determine the association weight between the target product term and each candidate customs code.

[0046] Among them, the association weight is a quantitative indicator that measures the confidence of the match between the target product term and the candidate customs code. The value range is [0,1]. The higher the value, the higher the matching accuracy. It can be represented by W(p,c_i). Historical matching data is used to calculate the historical frequency factor, which reflects the historical probability of the target product term being classified into a certain candidate customs code. The explanatory description text is used to calculate the semantic similarity factor, which reflects the degree of semantic association between the target product term and the official meaning of the candidate customs code.

[0047] In some embodiments, the association weight between the target product term and each candidate customs code is determined based on historical matching data associated with the target product term and the explanatory description text of each candidate customs code, including:

[0048] S1041. For any candidate customs code c_i, determine the historical frequency factor based on the historical co-occurrence count Freq(p,c_i) of the target product term p and the candidate customs code c_i in the historical business database, and the total occurrence count Freq(p) of the target product term in the historical database.

[0049] For example, the historical frequency factor F_i of candidate customs code c_i can be calculated using the following formula:

[0050] F_i = Freq(p,c_i) / Freq(p); where the historical frequency factor F_i can reflect the historical proportion of the target product term p being assigned to candidate customs code c_i.

[0051] S1042. Determine the semantic similarity factor based on the similarity between the explanatory description text of the candidate customs code and the target product terminology.

[0052] A pre-trained semantic similarity calculation model can be loaded first. Then, the similarity between the explanatory description text of the candidate customs code and the target product term can be calculated using the pre-trained semantic similarity calculation model, and the similarity calculation result can be used as the semantic similarity factor S_i. Specifically, the description texts of the target product term p and the candidate customs code c_i are respectively input into the semantic similarity calculation model and transformed into 768-dimensional semantic vectors V_p and V_{c_i}. The cosine similarity between vectors V_p and V_{c_i} is calculated to obtain the semantic similarity factor S_i=SemanticSim(p,c_i).

[0053] S1043. Based on historical frequency factors and semantic similarity factors, determine the association weight between the target product term and the candidate customs code according to the preset weight calculation formula.

[0054] The association weight W(p,c_i) is calculated using the following formula:

[0055] W(p,c_i)=λ×[Freq(p,c_i) / Freq(p)]+(1-λ)×SemanticSim(p,c_i);

[0056] Wherein, F_i=[Freq(p,c_i) / Freq(p)] is the historical frequency factor, which solves the statistical confidence problem; SemanticSim}(p,c_i) is the semantic similarity between the product term p and the official Chinese description of the customs code c_i; λ is a balancing parameter with a default value of 0.6 (adjusting the weight ratio of historical frequency and semantic similarity), which can be configured in the range of [0.3,0.8] according to the business scenario.

[0057] Understandably, the calculated association weights take into account both historical statistical patterns (to address confidence issues) and semantic associations (to address new terms and ambiguities), avoiding biases caused by relying solely on historical data or semantics. This makes the association weights more aligned with actual business scenarios, significantly improving the accuracy of coding matching, and far surpassing the adaptability of traditional manual rules.

[0058] S105. Based on the association weight between the target product terminology and each candidate customs code, select an optimal customs code and at least one alternative customs code from the candidate customs codes as recommendation information and recommend them to the user.

[0059] Among them, the optimal customs code refers to the candidate code with the highest association weight, that is, the candidate customs code with the highest confidence; the alternative customs code refers to the candidate code with the second highest association weight, which is used to deal with multi-code adaptation scenarios; the recommendation information is a comprehensive information including the optimal customs code, the alternative customs codes and their corresponding weight values ​​(confidence).

[0060] In practice, candidate customs codes are sorted in descending order according to their association weights; the candidate customs code with the highest association weight is selected as the optimal customs code, and the N candidate customs codes following the optimal customs code are selected as alternative customs codes; the optimal customs code and its association weight, as well as the alternative customs codes and their association weights, are used as recommendation information and recommended to users in a visual manner.

[0061] To facilitate understanding of this scheme, we will use the recommended customs code for "spark plugs" as an example:

[0062] The user-inputted unstructured product text information: “Bujías de encendido para motor decoche”;

[0063] After parsing the unstructured product text information, the large model outputs the standardized target product term p as "spark plug";

[0064] Retrieve the candidate customs code set C={851110,854800,850790,851180,854710} associated with the target product term p from the historical database; obtain historical matching data Freq(p)=501730, Freq(p,851110)=385700, Freq(p,854800)=115300, etc.

[0065] Historical frequency factors for each candidate customs code:

[0066] F_{851110}=385700 / 501730=0.7687

[0067] F_{851180}=220 / 501730=0.0004;

[0068] F_{854710}=210 / 501730=0.0004;

[0069] F_{850790}=300 / 501730=0.0006;

[0070] F_{854800}=115300 / 501730=0.2298;

[0071] The semantic similarity factor between the explanatory description text of each candidate customs code and the target product term p:

[0072] S_{851110}=1.0000; The Chinese description of 851110 is "spark plug";

[0073] S_{851180}=0.2920; The Chinese description of 851180 is "other devices for electric ignition and starting of the engine";

[0074] S_{854710}=0.289; the Chinese description of 854710 is "ceramic insulating parts";

[0075] S_{850790}=0.131; The Chinese description of 850790 is "lead-acid battery parts";

[0076] S_{854800}=0.05; The Chinese description of 854800 is "electrical parts of a machine or equipment not listed in other headings of this chapter".

[0077] Association weight calculation:

[0078] W(p,851110)=0.6×0.7687+0.4×1.0000=0.8612;

[0079] W(p,851180)=0.6×0.0004+0.4×0.2920=0.1170;

[0080] Output: The optimal customs code is 851110 (confidence level 86.12%), and the alternative customs codes are 854800 (0.1580), 851180 (0.1170), 854710 (0.1158), and 850790 (0.0528).

[0081] In this embodiment of the invention, new words and abbreviations can be identified through multi-round parsing of a large model, and standardized terms can be output. The two-factor model of S104 does not rely on rigid keyword matching, but makes a comprehensive judgment through historical statistics and semantic association, which improves the fault tolerance and makes the adaptation rate significantly high. This solves the problem that traditional manual rules have poor adaptation ability to new words on the Internet, industry abbreviations and semantic ambiguity.

[0082] Example 2

[0083] Figure 2 A flowchart illustrating a business opportunity-based enterprise information recommendation method is provided as an embodiment of the present invention. See also... Figure 2 The method includes the following steps:

[0084] S201. Call the large model to process the unstructured product text information input by the user and output the target product terminology.

[0085] S202. Query the historical business database for at least one candidate customs code that matches the target product term, as well as historical matching data associated with the target product term.

[0086] The candidate code set C = {c_1, c_2, ..., c_n}, where c_1, c_2, ..., c_n represent different candidate customs codes. The historical matching data contains two core types of data: the total occurrence count of the target product term Freq(p), and the historical co-occurrence count of the target product term with each candidate customs code Freq(p, c_i), where p is the target product term, and c_i is the historical co-occurrence count of the target product term p and the candidate customs code c_i.

[0087] S203. Construct a knowledge graph based on the target product term, at least one candidate customs code that matches the target product term, and historical matching data associated with the target product term.

[0088] The knowledge graph refers to a dynamic weighted knowledge graph built around the target product terminology, candidate customs codes, and historical matching data. It is a structured knowledge carrier that carries the relationship between the three and quantifies the strength of the relationship. The nodes include the target product terminology (p) and candidate customs codes (c_1, c_2, ..., c_n). The edges and weights are initially defined by the historical matching data (Freq(p), Freq(p, c_i)) and can be dynamically updated later. The specific update process can be found in the following description.

[0089] In specific implementation, the target product term p and candidate customs codes (c_1, c_2, ..., c_n) are treated as independent nodes, and the node attributes are defined (e.g., the attribute of p is standardized product term, and the attribute of c_i is customs code). Connection edges and initial weights are established: connection edges are established between p and each c_i. Based on the historical matching data in S202, Freq(p, c_i) / Freq(p) (initial value of historical frequency factor) is used as the initial weight of the connection edge. Simultaneously, the original historical data (Freq(p), Freq(p, c_i)) is stored in the node. The overall structure of the above node-connection edge-weight-historical data is stored in the dynamic knowledge graph module to form an initial knowledge graph that can be iteratively optimized later.

[0090] Understandably, integrating scattered product terms, customs codes, and historical data into a structured graph avoids data fragmentation and provides a clear data correlation logic for subsequent quantitative calculations.

[0091] S204. Obtain the explanatory description text for each candidate customs code.

[0092] S205. Based on historical matching data associated with the target product term and the explanatory description text of each candidate customs code, determine the association weight between the target product term and each candidate customs code.

[0093] The association weight W(p,c_i) is calculated using the following formula:

[0094] W(p,c_i)=λ×[Freq(p,c_i) / Freq(p)]+(1-λ)×SemanticSim(p,c_i);

[0095] Wherein, F_i=[Freq(p,c_i) / Freq(p)] is the historical frequency factor calculated based on historical matching data, which solves the statistical confidence problem; SemanticSim}(p,c_i) is the semantic similarity factor determined based on the semantic similarity between the target product term p and the official Chinese explanation description text of the customs code c_i; λ is a balancing parameter with a default value of 0.6 (adjusting the weight ratio of historical frequency and semantic similarity), which can be configured in the range of [0.3,0.8] according to the business scenario.

[0096] It should be noted that the semantic similarity calculation process is as follows: using a pre-trained semantic similarity calculation model, the similarity between the explanatory text of the candidate customs code and the target product terminology is calculated, and the similarity calculation result is used as the semantic similarity factor.

[0097] S206. Based on the association weight between the target product term and each candidate customs code, select an optimal customs code and at least one alternative customs code from the candidate customs codes as recommendation information and recommend them to the user.

[0098] After the optimal and alternative customs codes are recommended to the user through the above steps, the user can view the recommendation results and provide feedback on them. For details, please refer to S207-S209.

[0099] S207. In response to the user's confirmation that the recommended optimal customs code is correct, increment the total occurrence count of the target product term by one, and increment the historical co-occurrence count of the target product term and the optimal candidate customs code by one.

[0100] Optionally, upon receiving feedback from the user confirming the correctness of the optimal customs code, locate the total occurrence count field corresponding to the target product term p and increment its value by 1 (e.g., the original Freq(p) = 501730, updated to 501731); locate the historical co-occurrence count field between the target product term p and the optimal customs code c_optimal and increment its value by 1 (e.g., the original Freq(p,c_optimal) = 385700, updated to 385701).

[0101] S208. In response to the user's feedback instruction that the optimal customs code is incorrect, based on the new optimal customs code specified by the user, increment the total occurrence count of the target product term by one, and increment the historical co-occurrence count of the target product term and the new optimal candidate customs code by one.

[0102] Among them, the new optimal customs code (c_corrected) is the customs code that the user specifies based on actual business judgment and determines that the recommended optimal customs code is incorrect. It is a customs code that truly matches the target product term p. It can be a code outside the original candidate code set.

[0103] After the user specifies a new optimal customs code, the total occurrence count Freq(p) of the target product term p is incremented by one, and the historical co-occurrence count Freq(p,c_corrected) of the target product term p and the new optimal customs code c_corrected is incremented by one. Furthermore, the updated Freq(p), Freq(p,c_corrected), and the newly added mapping relationships can be synchronized to the knowledge graph to supplement or update the corresponding nodes and associated edges.

[0104] Understandably, this step can promptly correct system recommendation biases by adding user-confirmed correct codes and matching relationships to the historical database, thus preventing similar erroneous recommendations from recurring.

[0105] S209. In response to the user's feedback instruction that any alternative customs code is also the correct optimal customs code, add K to the total number of occurrences of the target product term and add K to the historical co-occurrence counts of the target product term and the alternative customs code; where K is an integer greater than 1.

[0106] If, based on the feedback instructions, the user confirms that a certain alternative customs code c_alternative is also correct (meaning there are multiple correct customs codes corresponding to the target product term p), then Freq(p, c_alternative) is increased by a large value K (K>1), and Freq(p) is also increased by K to significantly increase the historical statistical weight of the alternative customs code.

[0107] In this embodiment of the invention, the historical matching data of customs codes can be dynamically adjusted based on user feedback on the recommendation results, so as to adjust the knowledge graph weights in the future.

[0108] Furthermore, through steps S207-S209, the optimal customs codes corresponding to the target product terms confirmed by multiple users can be obtained. At this point, new training samples can be constructed based on the explanatory text of the correct optimal customs codes and the target product terms. Then, the semantic similarity calculation model can be fine-tuned periodically using the new training samples. This ensures the accuracy of subsequent semantic similarity calculations using the fine-tuned model.

[0109] Furthermore, based on user feedback, the frequency of occurrence of the target product term and the historical co-occurrence frequency of the target product term and some customs codes have been modified. Subsequently, the weights of the constructed knowledge graph need to be updated periodically based on the modified data, specifically as follows: For any candidate customs code associated with the target product term in the knowledge graph, a new historical frequency factor is calculated using the updated total frequency of occurrence of the target product term and the updated historical co-occurrence frequency of the target product term and the candidate customs code; using the fine-tuned semantic similarity calculation model, the similarity between the explanatory description text of the candidate customs code and the target product term is calculated to determine a new semantic similarity factor; based on the new historical frequency factor and the new semantic similarity factor, the association weight between the target product term and the candidate customs code in the knowledge graph is re-determined according to the preset weight calculation formula. This completes the iterative optimization of the dynamic knowledge graph.

[0110] Example 3

[0111] Figure 3 This is a schematic diagram of a business opportunity-based enterprise information recommendation device provided in an embodiment of the present invention. This device can execute any of the business opportunity-based enterprise information recommendation methods of the present invention. For example... Figure 3 As shown, the information recommendation device includes:

[0112] The model processing module 301 is used to call the large model to process the unstructured product text information input by the user and output the target product terms.

[0113] The query module 302 is used to query at least one candidate customs code that matches the target product term from the historical business database, as well as historical matching data associated with the target product term;

[0114] The acquisition module 303 is used to acquire the explanatory description text for each candidate customs code;

[0115] The weight calculation module 304 is used to determine the association weight between the target product term and each candidate customs code based on the historical matching data associated with the target product term and the explanatory description text of each candidate customs code.

[0116] The recommendation module 305 is used to select an optimal customs code and at least one alternative customs code from the candidate customs codes as recommendation information based on the association weight between the target product term and each candidate customs code, and recommend them to the user.

[0117] In some embodiments, the historical matching data associated with the target product term includes: the total number of occurrences of the target product term and the historical co-occurrence counts of the target product term with each candidate customs code;

[0118] In determining the association weight between the target product term and each candidate customs code based on historical matching data associated with the target product term and the explanatory description text of each candidate customs code, the weight calculation module is specifically used for:

[0119] For any candidate customs code, the historical frequency factor is determined based on the number of times the target product term and the candidate customs code co-occur in the historical business database, and the total number of times the target product term appears in the historical database.

[0120] The semantic similarity factor is determined based on the similarity between the explanatory text of the candidate customs code and the target product terminology.

[0121] Based on historical frequency factors and semantic similarity factors, the association weight between the target product term and the candidate customs code is determined according to a preset weight calculation formula.

[0122] In some embodiments, the recommendation module is specifically used to select an optimal customs code and at least one alternative customs code from the candidate customs codes as recommendation information based on the association weight between the target product term and each candidate customs code, and to recommend them to the user.

[0123] Sort the candidate customs codes in descending order according to their association weight;

[0124] The candidate customs code with the highest association weight is selected as the optimal customs code, and the N candidate customs codes ranked after the optimal customs code are selected as alternative customs codes.

[0125] The optimal customs code and its associated weight, as well as alternative customs codes and their associated weights, are used as recommendation information and presented to users in a visual manner.

[0126] In some embodiments, the device further includes a feedback processing module for:

[0127] In response to the user's confirmation that the recommended optimal customs code is correct, increment the total occurrence count of the target product term by one, and increment the historical co-occurrence count of the target product term and the optimal candidate customs code by one; or,

[0128] In response to user feedback that the optimal customs code was incorrect, based on the new optimal customs code specified by the user, increment the total occurrence count of the target product term by one, and increment the historical co-occurrence count of the target product term and the new optimal candidate customs code by one; or,

[0129] In response to the user's feedback instruction that any alternative customs code is also the correct optimal customs code, the total number of occurrences of the target product term is incremented by K, and the historical co-occurrence counts of the target product term and the alternative customs code are incremented by K; where K is an integer greater than 1.

[0130] In some embodiments, a semantic similarity factor is determined based on the similarity between the explanatory description text of the candidate customs code and the target product terminology, including:

[0131] The similarity between the explanatory text of the candidate customs code and the target product terminology is calculated using a pre-trained semantic similarity calculation model, and the similarity calculation result is used as a semantic similarity factor.

[0132] Correspondingly, the device also includes a model fine-tuning module for:

[0133] New training samples are constructed based on the explanatory text describing the correct optimal customs code determined by the user and the target product terminology.

[0134] The semantic similarity calculation model is fine-tuned periodically using new training samples.

[0135] In some embodiments, a knowledge graph construction module is also included, for:

[0136] A knowledge graph is constructed based on the target product term, at least one candidate customs code that matches the target product term, and historical matching data associated with the target product term.

[0137] In some embodiments, a weight recalculation module is further included, for:

[0138] For any candidate customs code associated with the target product term in the knowledge graph, a new historical frequency factor is calculated using the total occurrence count of the updated target product term and the historical co-occurrence count of the updated target product term and the candidate customs code.

[0139] Using a fine-tuned semantic similarity calculation model, the similarity between the explanatory text of the candidate customs code and the target product terminology is calculated, and a new semantic similarity factor is determined.

[0140] Based on the new historical frequency factor and the new semantic similarity factor, and according to the preset weight calculation formula, the association weight between the target product term and the candidate customs code in the knowledge graph is re-determined.

[0141] The business opportunity-based enterprise information recommendation device provided in the embodiments of the present invention can execute the business opportunity-based enterprise information recommendation method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method execution.

[0142] According to embodiments of the present invention, the present invention also provides an electronic device, a readable storage medium, and a computer program product.

[0143] Example 4

[0144] Figure 4 A schematic diagram of the structure of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the invention described and / or claimed herein.

[0145] like Figure 4 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory 12 or a random access memory 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the read-only memory 12 or loaded from storage unit 18 into the random access memory 13. The random access memory 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, read-only memory 12, and random access memory 13 are interconnected via a bus 14. An input / output interface 15 is also connected to the bus 14.

[0146] Multiple components in electronic device 10 are connected to input / output interface 15, including: input unit 16; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disks, optical disks, etc.; and communication unit 19, such as network interface cards, modems, wireless transceivers, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0147] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing units, graphics processing units, various special-purpose artificial intelligence computing chips, various processors running machine learning model algorithms, digital signal processors, and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as performing business opportunity-based enterprise information recommendation methods.

[0148] In some embodiments, the business opportunity-based business information recommendation method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via read-only memory 12 and / or communication unit 19. When the computer program is loaded into random access memory 13 and executed by processor 11, one or more steps of the business opportunity-based business information recommendation method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the business opportunity-based business information recommendation method by any other suitable means (e.g., by means of firmware).

[0149] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays, application-specific integrated circuits (ASICs), application-specific standard products (ASICs), systems-on-a-chip (SoCs), complex programmable logic devices, computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0150] Computer programs used to implement the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable information recommendation device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The computer programs can be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0151] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, optical fibers, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0152] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device or liquid crystal display for displaying information to a user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0153] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet. The computing system can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having client-server relationships with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, a host product within the cloud computing service system, addressing the shortcomings of traditional physical hosts and virtual private servers, such as high management difficulty and weak business scalability.

[0154] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0155] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for recommending enterprise information based on business opportunities, characterized in that, include: The large model is invoked to process the unstructured product text information input by the user and output the target product terminology; Retrieve from the historical business database at least one candidate customs code that matches the target product term, as well as historical matching data associated with the target product term; Obtain the explanatory description text for each of the candidate customs codes; Based on historical matching data associated with the target product term and explanatory description text for each candidate customs code, the association weight between the target product term and each candidate customs code is determined; Based on the association weight between the target product term and each of the candidate customs codes, an optimal customs code and at least one alternative customs code are selected from the candidate customs codes as recommendation information and recommended to the user.

2. The method according to claim 1, characterized in that, Historical matching data associated with the target product term includes: the total number of occurrences of the target product term and the historical co-occurrence counts of the target product term with each candidate customs code; Based on historical matching data associated with the target product term and the explanatory description text of each candidate customs code, the association weight between the target product term and each candidate customs code is determined, including: For any candidate customs code, a historical frequency factor is determined based on the number of times the target product term and the candidate customs code co-occur in the historical business database, and the total number of times the target product term appears in the historical database. The semantic similarity factor is determined based on the similarity between the explanatory text of the candidate customs code and the target product terminology. Based on the historical frequency factor and the semantic similarity factor, the association weight between the target product term and the candidate customs code is determined according to a preset weight calculation formula.

3. The method according to claim 2, characterized in that, Based on the association weight between the target product term and each of the candidate customs codes, an optimal customs code and at least one alternative customs code are selected from the candidate customs codes as recommendation information and recommended to the user, including: The candidate customs codes are sorted in descending order according to their association weight; The candidate customs code with the highest association weight is selected as the optimal customs code, and the N candidate customs codes ranked after the optimal customs code are selected as alternative customs codes. The optimal customs code and its associated weight, as well as the alternative customs codes and their associated weights, are used as recommendation information and presented to the user in a visual manner.

4. The method according to claim 3, characterized in that, Also includes: In response to the user's feedback instruction that the recommended optimal customs code is correct, the total occurrence count of the target product term is incremented by one, and the historical co-occurrence count of the target product term and the optimal candidate customs code is incremented by one; or, In response to the user's feedback instruction that the optimal customs code is incorrect, the total occurrence count of the target product term is incremented by one, and the historical co-occurrence count of the target product term and the new optimal candidate customs code is incremented by one, based on the new optimal customs code specified by the user. or, In response to the user's feedback instruction that any alternative customs code is also the correct optimal customs code, the total number of occurrences of the target product term is incremented by K, and the historical co-occurrence counts of the target product term and the alternative customs code are incremented by K; where K is an integer greater than 1.

5. The method according to claim 4, characterized in that, Based on the similarity between the explanatory description text of the candidate customs code and the target product terminology, a semantic similarity factor is determined, including: The similarity between the explanatory text of the candidate customs code and the target product terminology is calculated using a pre-trained semantic similarity calculation model, and the similarity calculation result is used as a semantic similarity factor. Accordingly, the method further includes: New training samples are constructed based on the explanatory text describing the correct optimal customs code determined by the user and the target product terminology. The semantic similarity calculation model is fine-tuned periodically using new training samples.

6. The method according to claim 5, characterized in that, Also includes: A knowledge graph is constructed based on the target product term, at least one candidate customs code that matches the target product term, and historical matching data associated with the target product term.

7. The method according to claim 6, characterized in that, Also includes: For any candidate customs code associated with the target product term in the knowledge graph, a new historical frequency factor is calculated using the updated total occurrence count of the target product term and the updated historical co-occurrence count of the target product term and the candidate customs code. Using a fine-tuned semantic similarity calculation model, the similarity between the explanatory description text of the candidate customs code and the target product terminology is calculated, and a new semantic similarity factor is determined. Based on the new historical frequency factor and the new semantic similarity factor, and according to the preset weight calculation formula, the association weight between the target product term and the candidate customs code in the knowledge graph is re-determined.

8. A business opportunity-based enterprise information recommendation device, characterized in that, include: The model processing module is used to call a large model to process the unstructured product text information input by the user and output the target product terms. The query module is used to query at least one candidate customs code that matches the target product term from the historical business database, as well as historical matching data associated with the target product term; The acquisition module is used to acquire the explanatory description text for each of the candidate customs codes; The weight calculation module is used to determine the association weight between the target product term and each candidate customs code based on historical matching data associated with the target product term and the explanatory description text of each candidate customs code. The recommendation module is used to select an optimal customs code and at least one alternative customs code from the candidate customs codes as recommendation information based on the association weight between the target product term and each candidate customs code, and recommend them to the user.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the method of any one of claims 1-7.

Images