Business management method, system and equipment based on AI keyword annotation and medium

Abstract

The invention discloses a business management method, system and device based on AI keyword annotation and a medium, and relates to the technical field of transaction platform business management, and the method comprises the following steps: obtaining text content output by a client in a business platform; performing keyword annotation and semantic analysis on the text content, and extracting demand information of the customer; on the basis of keyword annotation and semantic analysis, the transaction intentions of the clients are extracted, meanwhile, on the basis of demand information and the transaction intentions, intention classification is carried out on the clients, and the clients are divided into different transaction probability groups; the method and the device are used for solving the problems of resource waste and no help for business completion caused by unreasonable resource allocation of business personnel for different clients in the existing transaction platform business management technology.

Description

Technical Field

[0001] The present invention relates to the technical field of transaction platform business management, and specifically to a business management method, system, device and medium based on AI keyword labeling. Background Art

[0002] Trading platform business management technology refers to the various technologies and methods used to manage, optimize and support the operation of online trading platforms. These technologies cover multiple aspects from transaction matching, order management, payment settlement to risk control, data analysis and user management, ensuring that the trading platform can operate efficiently, safely and stably.

[0003] The existing transaction platform business management technology usually manages customers in a unified manner, matches business personnel for customers based on their needs, and different customers have different needs for products, which further leads to different transaction rates after connecting with customers. If business personnel invest too many resources in a customer with a low transaction rate, it will lead to a waste of resources and will not help the completion of the business. For example, in the patent application with publication number CN106779480A, a "sales management system and sales management implementation method" is disclosed. The purpose of this solution is to make the sales main line clear and make the work of sales personnel more organized, so as to facilitate the timely follow-up of multiple business main lines. However, in the management process, the classification of customers is not accurate. Classification based on the image information of the customer's business card is not accurate. It should focus on the classification of the customer's demand for the transaction. For customers with high demand, sales personnel invest more resources to promote transactions. However, this solution does not make a clear distinction between different customers, resulting in excessive waste of resources invested by sales personnel. The existing transaction platform business management technology also has the problem that business personnel allocate resources to different customers unreasonably, resulting in waste of resources and no help for the completion of the business. Summary of the invention

[0004] The present invention aims to solve one of the technical problems in the prior art to at least a certain extent, by obtaining the text content output by the customer in the business platform, then performing keyword annotation and semantic analysis on the text content, extracting the customer's demand information, and then extracting the customer's transaction intention based on the keyword annotation and semantic analysis, and at the same time classifying the customer's intention based on the demand information and transaction intention, and dividing the customer into different transaction probability groups, so as to solve the problem that the existing transaction platform business management technology still has unreasonable resource allocation for different customers by business personnel, resulting in waste of resources and no help for the completion of business.

[0005] To achieve the above objectives, in a first aspect, the present application provides a business management method based on AI keyword annotation, comprising the following steps:

[0006] Obtain the text content output by the customer in the business platform;

[0007] Perform keyword annotation and semantic analysis on the text content to extract the customer's demand information;

[0008] Based on keyword annotation and semantic analysis, extract the customer's transaction intention. At the same time, based on the demand information and transaction intention, classify the customers, and divide the customers into different transaction probability groups.

[0009] Further, obtaining the text content output by the customer in the business platform includes the following sub-steps:

[0010] Obtain the voice data output by the customer in the business platform;

[0011] Perform text conversion on the voice data, and convert the language data into text content through speech-to-text conversion;

[0012] Extract the text data output by the customer in the business platform and mark it as text content.

[0013] Further, performing keyword annotation and semantic analysis on the text content to extract the customer's demand information and transaction intention includes the following sub-steps:

[0014] Perform word segmentation on the text content through jieba word segmentation to obtain text word segments. Compare with the stop word database, screen out and remove the stop words in the text word segments, and mark the remaining text word segments as text keywords;

[0015] Perform semantic analysis on the text keywords, and identify the demand keywords in the text keywords through NLP technology. The demand keywords are defined as keywords that can represent or imply the customer's demand for a certain commodity;

[0016] Obtain the first punctuation mark after the demand keyword (excluding quotation marks), and mark the sentence formed by combining the text keywords between the demand keyword and the punctuation mark as the sentence to be confirmed;

[0017] Perform semantic analysis on the sentence to be confirmed, and analyze whether the sentence to be confirmed is smooth through NLP technology. If so, mark the sentence to be confirmed as the demand name, otherwise output a signal indicating that the sentence is not smooth;

[0018] If a signal indicating that the sentence is not smooth is output, then remove the text keyword at the end of the sentence to be confirmed and repeat the analysis until the demand name is obtained;

[0019] Query whether there is a time keyword in the text keyword. The time keyword is defined as a keyword representing time or date. If there is a time keyword in the text keyword, mark the time keyword as the required time; otherwise, output a signal indicating no time requirement.

[0020] Mark the required name and the required time together as the required information.

[0021] Further, based on keyword annotation and semantic analysis, extract the customer's intention to make a deal. At the same time, based on the required information and the intention to make a deal, classify the customers' intentions and divide the customers into different groups with different deal probabilities, including the following sub-steps:

[0022] Based on keyword annotation and semantic analysis, extract the customer's intention to make a deal.

[0023] Based on the required information and the intention to make a deal, classify the customers' intentions and divide the customers into different groups with different deal probabilities.

[0024] Further, extracting the customer's intention to make a deal based on keyword annotation and semantic analysis includes the following sub-steps:

[0025] Name the number of times the customer asks for product details regarding the required name as the product inquiry times. When the customer asks about the performance, appearance, and functions of the product, it is regarded as asking for product details; name the number of times the customer asks about the price of the product regarding the required name as the price inquiry times; name the number of times the customer asks about the time regarding the required name as the time inquiry times.

[0026] Analyze the text keyword and the text content through NLP technology, and count the product inquiry times, price inquiry times, and time inquiry times.

[0027] Obtain the historical inquiry records, which store the historical data of the product inquiry times, price inquiry times, and time inquiry times, named as the historical product inquiry times, historical price inquiry times, and historical time inquiry times respectively, and also store whether the customer finally makes a deal.

[0028] Obtain the historical product inquiry times, historical price inquiry times, and historical time inquiry times of the customers who finally make a deal, and mark them as HT1, HT2, and HT3 respectively; obtain the historical product inquiry times, historical price inquiry times, and historical time inquiry times of the customers who finally do not make a deal, and mark them as HF1, HF2, and HF3 respectively.

[0029] Obtain the median of HT1 and the median of HF1, mark them as ZA1 and ZA2 respectively, then calculate the median of ZA1 and ZA2, mark it as the commodity inquiry threshold, represented by the symbol A; obtain the median of HT2 and the median of HF2, mark them as ZB1 and ZB2 respectively, then calculate the median of ZB1 and ZB2, mark it as the price inquiry threshold, represented by the symbol B; obtain the median of HT3 and the median of HF3, mark them as ZC1 and ZC2 respectively, then calculate the median of ZC1 and ZC2, mark it as the time inquiry threshold, represented by the symbol C;

[0030] Calculate A / (A + B + C), mark the calculation result as the commodity inquiry weight, represented by the symbol K1, calculate B / (A + B + C), mark the calculation result as the price inquiry weight, represented by the symbol K2, calculate C / (A + B + C), mark the calculation result as the time inquiry weight, represented by the symbol K3;

[0031] Obtain max(HF1), remove the HT1 that is less than or equal to max(HF1), then obtain min(HT1), mark it as U1, where max() is the maximum operator and min() is the minimum operator; obtain max(HF2), remove the HT2 that is less than or equal to max(HF2), then obtain min(HT2), mark it as U2; obtain max(HF3), remove the HT3 that is less than or equal to max(HF3), then obtain min(HT3), mark it as U3;

[0032] Calculate U1×K1 + U2×K2 + U3×K3, mark the calculation result as the transaction index, represented by the symbol R;

[0033] Based on the calculated commodity inquiry weight, price inquiry weight, time inquiry weight and transaction index, calculate the transaction intention of each customer in real time.

[0034] Furthermore, calculating the transaction intention of each customer in real time based on the calculated commodity inquiry weight, price inquiry weight, time inquiry weight and transaction index includes the following sub - steps:

[0035] Obtain the number of commodity inquiries, the number of price inquiries and the number of time inquiries of the customer, mark them as Y1, Y2 and Y3 respectively;

[0036] Set the first increase threshold, represented by the symbol S, calculate Y1 / A, mark the integer part of the calculation result as X1, increase K1 by S×X1 to get P1; calculate Y2 / B, mark the integer part of the calculation result as X2, increase K2 by S×X2 to get P2; calculate Y3 / C, mark the integer part of the calculation result as X3, increase K3 by S×X3 to get P3;

[0037] Through the formula Calculate the customer's intention to close a deal, where the intention to close a deal represents the probability of closing a deal. Here, L is the intention to close a deal.

[0038] Furthermore, based on the demand information and the intention to close a deal, classify the customers' intentions, and divide the customers into different groups of deal - closing probabilities, including the following sub - steps:

[0039] Set up a high - probability deal - closing group, a medium - probability deal - closing group, and a low - probability deal - closing group;

[0040] Set a first probability threshold and a second probability threshold, where the first probability threshold is greater than the second probability threshold;

[0041] Compare the customer's intention to close a deal with the first probability threshold and the second probability threshold. If the intention to close a deal is greater than or equal to the first probability threshold, output a high - probability signal; if the intention to close a deal is less than the first probability threshold and greater than or equal to the second probability threshold, output a medium - probability signal; if the intention to close a deal is less than the second probability threshold, output a low - probability signal;

[0042] If a high - probability signal is output, classify the customer into the high - probability deal - closing group; if a medium - probability signal is output, classify the customer into the medium - probability deal - closing group; if a low - probability signal is output, classify the customer into the low - probability deal - closing group.

[0043] In a second aspect, the present application provides a business management system based on AI keyword annotation, including a text acquisition module, a semantic analysis module, and a deal - closing probability classification module; the text acquisition module and the semantic analysis module are respectively connected to the deal - closing probability classification module for data connection;

[0044] The text acquisition module is used to acquire the text content output by the customer on the business platform;

[0045] The semantic analysis module is used to perform keyword annotation and semantic analysis on the text content to extract the customer's demand information;

[0046] The deal - closing probability classification module is used to extract the customer's intention to close a deal based on keyword annotation and semantic analysis, and at the same time, based on the demand information and the intention to close a deal, classify the customers' intentions and divide the customers into different groups of deal - closing probabilities.

[0047] In a third aspect, the present application provides an electronic device, including a processor and a memory, where the memory stores computer - readable instructions, and when the computer - readable instructions are executed by the processor, the steps in the above - mentioned method are run.

[0048] In a fourth aspect, the present application provides a storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps in the above - mentioned method are run.

[0049] Advantages of the present invention: By obtaining the text content output by the customer on the business platform, then performing keyword annotation and semantic analysis on the text content to extract the customer's demand information, and based on the keyword annotation and semantic analysis, extracting the customer's intention to close a deal. The advantage lies in that the intention to close a deal reveals the degree of importance the customer attaches to the transaction, that is, the probability of the transaction being completed. By analyzing the content of the customer's inquiry from three dimensions: the details of the product, the price of the product, and the time, the inquiry about the product details reveals the degree of interest the customer has in the product, the inquiry about the product price reveals the customer's sensitivity to price, and the inquiry about time reveals the urgency of the customer's demand for the product. Through comprehensive analysis, the probability of the customer closing a deal, that is, the intention to close a deal, can be obtained, improving the effectiveness and accuracy of business management;

[0050] According to the present invention, based on the demand information and the intention to close a deal, the customers are classified by intention, and the customers are divided into different groups with different probabilities of closing a deal. The advantage is that business personnel can, based on the grouping situation, invest more resources in the high-probability and medium-probability groups for closing a deal. For the low-probability group for closing a deal, a small amount of resources can be invested with the aim of increasing the customer's interest in the product, while for the high-probability group for closing a deal, it is necessary to aim at promoting the completion of the transaction. Different resources can be invested in different groups with different probabilities of closing a deal, and at the same time, the content during the operation of business personnel can also be more targeted, improving the rationality and effectiveness of business management. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is the principle block diagram of the system of the present invention;

[0052] Figure 2 is the step flow chart for calculating the weights of product inquiries, price inquiries, and time inquiries of the present invention;

[0053] Figure 3 is the step flow chart of the method of the present invention;

[0054] Figure 4 is the structural schematic diagram of the electronic device of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] Example 1, please refer to Figure 1As shown in the figure, the present application provides a business management system based on AI keyword annotation, including a text acquisition module, a semantic analysis module, and a transaction probability division module; the text acquisition module and the semantic analysis module are respectively connected to the transaction probability division module for data connection;

[0057] The text acquisition module is used to acquire the text content output by the customer on the business platform;

[0058] The text acquisition module is configured with a text acquisition strategy, and the text acquisition strategy includes:

[0059] Acquire the voice data output by the customer on the business platform;

[0060] Perform text conversion on the voice data, and convert the language data into text content through speech-to-text conversion;

[0061] Extract the text data output by the customer on the business platform and mark it as text content;

[0062] In actual application, extract the text data and voice data of the customer on the business platform, and convert them all into text content to obtain the customer's needs. The customer will ask questions about the products they are interested in on the business platform. The questions include three aspects: product details, product price, and purchase time. By analyzing the multiple inquiry records of the customer for a certain product through the subsequent semantic analysis module, the demand degree of the customer for the product can be analyzed and judged, which can provide help for business management. In this embodiment, the speech-to-text technology is completed by using an existing conversion tool, and no specific description will be given in this embodiment.

[0063] The semantic analysis module is used to perform keyword annotation and semantic analysis on the text content to extract the customer's demand information;

[0064] The text acquisition module is configured with a text acquisition strategy, and the text acquisition strategy includes:

[0065] Perform word segmentation on the text content through jieba word segmentation to obtain text word segments. Compare with the stop word database, screen out and remove the stop words in the text word segments, and mark the remaining text word segments as text keywords;

[0066] Perform semantic analysis on the text keywords, and identify the demand keywords in the text keywords through NLP technology. The demand keywords are defined as the keywords that can represent or imply the customer's demand for a certain product;

[0067] Obtain the first punctuation mark after the demand keyword, excluding quotation marks, and mark the statement formed by combining the text keywords between the demand keyword and the punctuation mark as the statement to be confirmed;

[0068] In practical applications, jieba word segmentation uses the existing jieba word segmentation technology. For the text content "Do you have product A for sale here?", the text segmentation obtained by word segmentation includes "you", "here", "do you have", "product", "A", and "for sale". Among them, "you", "here", "do you have", and "for sale" are all stop words. Therefore, the text keywords obtained are "product" and "A". By retrieving, it is found that the text keywords do not include the required keywords. At this time, the removed stop words are retrieved, and it is found that "do you have" belongs to the required keywords. The required keywords are stored in the dataset obtained by big data deep learning. Through the method of manual annotation learning, the keywords or sentence patterns that represent or imply the customer's demand for a certain product in the sentence are found and learned. For example, keywords such as "sell or not" and "look for" belong to the required keywords, or it is judged by the sentence pattern of the sentence. For example, "Do you have...?", and the ellipsis is the name of the demand. All such sentence patterns can judge the customer's demand information. In this embodiment, no more display and description will be made; in the text content "Do you have product A for sale here?", the sentence to be confirmed obtained by analysis is "product A";

[0069] Perform semantic analysis on the sentence to be confirmed. Analyze whether the sentence to be confirmed is smooth through NLP technology. If so, mark the sentence to be confirmed as the name of the demand, otherwise output a signal indicating that the sentence is not smooth;

[0070] If a signal indicating that the sentence is not smooth is output, remove the text keyword at the end of the sentence to be confirmed and repeat the analysis until the name of the demand is obtained;

[0071] Query whether there is a time keyword in the text keywords. The time keyword is defined as a keyword representing time or date. If there is a time keyword in the text keywords, mark the time keyword as the required time, otherwise output a signal indicating no time requirement;

[0072] Mark the name of the demand and the required time together as the demand information;

[0073] In practical applications, NLP technology uses the existing natural language processing technology. By performing semantic analysis on the sentence to be confirmed, it is found that the sentence to be confirmed is smooth and can become a complete sentence or noun. Therefore, the sentence to be confirmed is marked as the name of the demand; the judgment of the time keyword is used to complete the extraction of the number of time inquiries in the subsequent analysis. In this embodiment, no specific description will be made. The existing natural language processing technology can automatically identify whether the keyword represents time information;

[0074] The deal probability division module is used to extract the customer's deal intention based on keyword annotation and semantic analysis. At the same time, based on the demand information and the deal intention, the customers are classified intention-wise and divided into different deal probability groups. The deal probability division module includes an intention analysis unit and an intention classification unit.

[0075] Please refer to Figure 2 As shown, the intention analysis unit is used to extract the customer's deal intention based on keyword annotation and semantic analysis.

[0076] The intention analysis unit is configured with an intention analysis strategy, and the intention analysis strategy includes:

[0077] The number of times a customer asks for product details regarding the demand name is named the product inquiry times. When a customer asks about the performance, appearance, and functions of a product, it is regarded as asking for product details. The number of times a customer asks about the price of a product regarding the demand name is named the price inquiry times. The number of times a customer asks about the time regarding the demand name is named the time inquiry times.

[0078] Analyze the text keywords and text content through NLP technology, and count the product inquiry times, price inquiry times, and time inquiry times.

[0079] In practical applications, identify the questions through NLP technology, then identify the text keywords in the questions, and determine whether the customer is asking about product details, product price, or time, so as to count them. In this embodiment, it is counted that the customer's product inquiry times for product A is 13, the price inquiry times is 6, and the time inquiry times is 2.

[0080] Obtain the historical inquiry records. The historical inquiry records store the historical data of the product inquiry times, price inquiry times, and time inquiry times, which are respectively named the product historical inquiry times, price historical inquiry times, and time historical inquiry times. At the same time, it also stores whether the customer finally made a deal.

[0081] Obtain the product historical inquiry times, price historical inquiry times, and time historical inquiry times of the customers who finally made a deal, and mark them as HT1, HT2, and HT3 respectively. Obtain the product historical inquiry times, price historical inquiry times, and time historical inquiry times of the customers who finally did not make a deal, and mark them as HF1, HF2, and HF3 respectively.

[0082] Get the median of HT1 and the median of HF1, mark them as ZA1 and ZA2 respectively, then calculate the median of ZA1 and ZA2, mark them as the product inquiry threshold, represented by symbol A; get the median of HT2 and the median of HF2, mark them as ZB1 and ZB2 respectively, then calculate the median of ZB1 and ZB2, mark them as the price inquiry threshold, represented by symbol B; get the median of HT3 and the median of HF3, mark them as ZC1 and ZC2 respectively, then calculate the median of ZC1 and ZC2, mark them as the time inquiry threshold, represented by symbol C;

[0083] Calculate A / (A+B+C), mark the result as the commodity inquiry weight, represented by symbol K1, calculate B / (A+B+C), mark the result as the price inquiry weight, represented by symbol K2, calculate C / (A+B+C), mark the result as the time inquiry weight, represented by symbol K3;

[0084] In practical applications, Figure 2 The calculation process of the product inquiry weight, price inquiry weight and time inquiry weight is shown. HT1, HT2, HT3, HF1, HF2 and HF3 are all a set, not a single value. Due to the large amount of data, this embodiment will not be specifically shown. Based on HT1 and HF1, ZA1 and ZA2 are obtained to be 8 and 3 respectively, and then the median of 8 and 3 is calculated to obtain a product inquiry threshold A of 5.5. The calculation and analysis process of the price inquiry threshold and the time inquiry threshold are the same, and this embodiment will not be specifically explained. Similarly, the price inquiry threshold B is calculated to be 5, and the time inquiry threshold C is 3. Further calculations obtain that the product inquiry weight K1 is 0.41, the price inquiry weight K2 is 0.37, and the time inquiry weight K3 is 0.22. The calculation results are rounded to two decimal places.

[0085] Get max(HF1), remove HT1 that is less than or equal to max(HF1), and then get min(HT1), marked as U1, where max() is the maximum value operator and min() is the minimum value operator; get max(HF2), remove HT2 that is less than or equal to max(HF2), and then get min(HT2), marked as U2; get max(HF3), remove HT3 that is less than or equal to max(HF3), and then get min(HT3), marked as U3;

[0086] Calculate U1×K1+U2×K2+U3×K3, and mark the result as the transaction index, represented by the symbol R;

[0087] In practical applications, due to an excessive number of data samples, HT1, HT2, HT3, HF1, HF2, and HF3 all present as a range interval. Among them, HF1 is [1, 5], HT1 is [1, 15]. Thus, max(HF1) = 5 is obtained. The data in HT1 that is less than or equal to 5 is removed. At this time, min(HT1) = 6 is obtained, that is, U1 = 6. Similarly, U2 and U3 are obtained as 6 and 2 respectively. Further calculation gives the transaction index R as 5.12;

[0088] Based on the calculated commodity inquiry weight, price inquiry weight, time inquiry weight, and transaction index, calculate the transaction intention of each customer in real time;

[0089] Obtain the number of commodity inquiries, number of price inquiries, and number of time inquiries of the customer, and mark them as Y1, Y2, and Y3 respectively;

[0090] Set the first increase threshold, represented by the symbol S. Calculate Y1 / A, mark the integer part of the calculation result as X1, and increase K1 by S×X1 to obtain P1; calculate Y2 / B, mark the integer part of the calculation result as X2, and increase K2 by S×X2 to obtain P2; calculate Y3 / C, mark the integer part of the calculation result as X3, and increase K3 by S×X3 to obtain P3;

[0091] Through the formula Calculate the transaction intention of the customer. The transaction intention represents the transaction probability, where L is the transaction intention;

[0092] In practical applications, in this embodiment, Y1 is 13, Y2 is 6, Y3 is 2, and the first increase threshold is set to 0.05. Generally, for customers with low purchase intention, their Y1, Y2, and Y3 are all less than the corresponding A, B, and C. For customers with high purchase intention, their Y1, Y2, and Y3 may exceed the corresponding A, B, and C. For example, when Y2 is greater than B, it means that the customer pays more attention to the price of the product. At this time, the proportion of price in it should be increased to evaluate the customer's true purchase intention. Each increase is based on the first increase threshold. If the increase is too small, the importance of price to the customer cannot be highlighted. If the increase is too large, the influence of price on the transaction intention will be too great. Through multiple social experiments and data calibration in this embodiment, it is verified that the best setting range of the first increase threshold is between 0.05 and 0.06. This embodiment selects 0.05 as the first increase threshold, which is more convenient for calculation. Taking Y1 as an example, where Y1 = 13, it is calculated that Y1 / A = 2.4, and the integer part is X1 = 2. Further calculation gives P1 = K1 + S×X1 = 0.41 + 0.05×2 = 0.51. Similarly, P2 and P3 are 0.42 and 0.22 respectively. The calculated transaction intention L of the customer is 1.87. The calculation result is rounded to two decimal places, that is, 187%. The transaction intention is 187%, exceeding 100%. This result is too absolute. Therefore, it is not applicable to business management. All calculation results greater than or equal to 100% need to be changed to 99%. That is, the maximum value of the transaction intention is limited to 99%, indicating that the customer is very likely to complete the transaction, but there are still certain problems with the product details, product price, or time. The business personnel need to understand and help solve the problems in time to complete the transaction;

[0093] The intention classification unit is used to classify customers based on the demand information and the transaction intention, and divide the customers into different transaction probability groups;

[0094] The intention classification unit is configured with an intention classification strategy, and the intention classification strategy includes:

[0095] Set a high-probability transaction group, a medium-probability transaction group, and a low-probability transaction group;

[0096] Set a first probability threshold and a second probability threshold, and the first probability threshold is greater than the second probability threshold;

[0097] Compare the customer's transaction intention with the first probability threshold and the second probability threshold. If the transaction intention is greater than or equal to the first probability threshold, output a high-probability signal; if the transaction intention is less than the first probability threshold and greater than or equal to the second probability threshold, output a medium-probability signal; if the transaction intention is less than the second probability threshold, output a low-probability signal;

[0098] If a high - probability signal is output, the customer is classified into the high - probability transaction group; if a medium - probability signal is output, the customer is classified into the medium - probability transaction group; if a low - probability signal is output, the customer is classified into the low - probability transaction group;

[0099] In practical applications, the first probability threshold and the second probability threshold are set by the business platform itself. Since R is calculated from min(HT1), min(HT2), and min(HT3), there will be many cases where the calculation results exceed 100%. All such calculation results are changed to 99%. Therefore, there are many datasets with a transaction intention equal to 99%, and the transaction intention is strong, which can be separately classified into the high - probability transaction group. Therefore, the first probability threshold is set to 99%, and the second probability threshold is usually set to 50%. The business platform can also change it according to requirements.

[0100] Example 2, please refer to Figure 3 As shown, the present application provides a business management method based on AI keyword annotation, including the following steps:

[0101] Step S1, obtain the text content output by the customer on the business platform; Step S1 includes the following sub - steps:

[0102] Step S101, obtain the voice data output by the customer on the business platform;

[0103] Step S102, perform text conversion on the voice data, and convert the language data from voice to text to obtain the text content;

[0104] Step S103, extract the text data output by the customer on the business platform and mark it as the text content;

[0105] Step S2, perform keyword annotation and semantic analysis on the text content to extract the customer's demand information; Step S2 includes the following sub - steps:

[0106] Step S201, perform word segmentation on the text content through jieba word segmentation to obtain text word segments. By comparing with the stop - word database, screen out and remove the stop - words in the text word segments, and mark the remaining text word segments as text keywords;

[0107] Step S202, perform semantic analysis on the text keywords. Through NLP technology, identify the demand keywords in the text keywords. The demand keywords are defined as the keywords that can represent or imply the customer's demand for a certain commodity;

[0108] Step S203, obtain the first punctuation mark after the demand keyword (excluding quotation marks), and mark the sentence formed by combining the text keywords between the demand keyword and the punctuation mark as the sentence to be confirmed;

[0109] Step S204, semantic analysis is performed on the statement to be confirmed, and NLP technology is used to analyze whether the statement to be confirmed is fluent. If so, the statement to be confirmed is marked as a requirement name, otherwise a signal indicating that the statement is not fluent is output;

[0110] Step S205, if the output statement does not pass the signal, remove the text keywords at the end of the statement to be confirmed and repeat the analysis until the requirement name is obtained;

[0111] Step S206, querying whether there is a time keyword in the text keyword, where the time keyword is defined as a keyword representing time or date. If there is a time keyword in the text keyword, the time keyword is marked as the required time, otherwise a signal of no time requirement is output;

[0112] Step S207, marking the demand name and the demand time together as demand information;

[0113] Step S3, based on keyword tagging and semantic analysis, extract the customer's transaction intention, and classify the customer's intention based on the demand information and transaction intention, and divide the customer into different transaction probability groups; Step S3 includes the following sub-steps:

[0114] Step S301, extracting the customer's transaction intention based on keyword annotation and semantic analysis;

[0115] Step S301 includes the following sub-steps:

[0116] Step S3011, the number of times that customers inquire about product details based on the demand name is named product inquiry number, and the number of times that customers inquire about the performance, appearance and function of the product is regarded as inquiring about product details; the number of times that customers inquire about the price of the product based on the demand name is named price inquiry number; the number of times that customers inquire about the time based on the demand name is named time inquiry number;

[0117] Step S3012, analyzing the text keywords and text content by NLP technology, and counting the number of product inquiries, price inquiries, and time inquiries;

[0118] Step S3013, obtaining historical inquiry records, which store historical data on the number of product inquiries, price inquiries, and time inquiries, respectively named as product historical inquiry number, price historical inquiry number, and time historical inquiry number, and also store whether the customer finally completed the transaction;

[0119] Step S3014: Obtain the historical number of inquiries about products, the historical number of inquiries about prices, and the historical number of inquiries about time for the customers who finally made a purchase, and mark them as HT1, HT2, and HT3 respectively; obtain the historical number of inquiries about products, the historical number of inquiries about prices, and the historical number of inquiries about time for the customers who finally did not make a purchase, and mark them as HF1, HF2, and HF3 respectively.

[0120] Step S3015: Obtain the median of HT1 and the median of HF1, and mark them as ZA1 and ZA2 respectively. Then calculate the median of ZA1 and ZA2, and mark it as the product inquiry threshold, represented by the symbol A; obtain the median of HT2 and the median of HF2, and mark them as ZB1 and ZB2 respectively. Then calculate the median of ZB1 and ZB2, and mark it as the price inquiry threshold, represented by the symbol B; obtain the median of HT3 and the median of HF3, and mark them as ZC1 and ZC2 respectively. Then calculate the median of ZC1 and ZC2, and mark it as the time inquiry threshold, represented by the symbol C.

[0121] Step S3016: Calculate A / (A + B + C), and mark the calculation result as the product inquiry weight, represented by the symbol K1; calculate B / (A + B + C), and mark the calculation result as the price inquiry weight, represented by the symbol K2; calculate C / (A + B + C), and mark the calculation result as the time inquiry weight, represented by the symbol K3.

[0122] Step S3017: Obtain max(HF1), and eliminate the HT1 that is less than or equal to max(HF1). Then obtain min(HT1), and mark it as U1, where max() is the maximum operator and min() is the minimum operator; obtain max(HF2), and eliminate the HT2 that is less than or equal to max(HF2). Then obtain min(HT2), and mark it as U2; obtain max(HF3), and eliminate the HT3 that is less than or equal to max(HF3). Then obtain min(HT3), and mark it as U3.

[0123] Step S3018: Calculate U1×K1 + U2×K2 + U3×K3, and mark the calculation result as the transaction index, represented by the symbol R.

[0124] Step S3019: Calculate the purchase intention of each customer in real time based on the calculated product inquiry weight, price inquiry weight, time inquiry weight, and transaction index.

[0125] Step S3019 includes the following sub-steps:

[0126] Step S3019.1: Obtain the customer's number of inquiries about products, number of inquiries about prices, and number of inquiries about time, and mark them as Y1, Y2, and Y3 respectively.

[0127] Step S3019.2: Set the first increase threshold, denoted by symbol S. Calculate Y1 / A, mark the integer part of the calculation result as X1, increase K1 by S×X1 to obtain P1; calculate Y2 / B, mark the integer part of the calculation result as X2, increase K2 by S×X2 to obtain P2; calculate Y3 / C, mark the integer part of the calculation result as X3, increase K3 by S×X3 to obtain P3;

[0128] Step S3019.3: Calculate the customer's transaction intention through the formula where the transaction intention represents the transaction probability, and L is the transaction intention;

[0129] Step S302: Based on the demand information and the transaction intention, classify the customers and divide them into different transaction probability groups;

[0130] Step S302 includes the following sub-steps:

[0131] Step S3021: Set a high-probability transaction group, a medium-probability transaction group, and a low-probability transaction group;

[0132] Step S3022: Set the first probability threshold and the second probability threshold, where the first probability threshold is greater than the second probability threshold;

[0133] Step S3023: Compare the customer's transaction intention with the first probability threshold and the second probability threshold. If the transaction intention is greater than or equal to the first probability threshold, output a high-probability signal; if the transaction intention is less than the first probability threshold and greater than or equal to the second probability threshold, output a medium-probability signal; if the transaction intention is less than the second probability threshold, output a low-probability signal;

[0134] Step S3024: If a high-probability signal is output, classify the customer into the high-probability transaction group; if a medium-probability signal is output, classify the customer into the medium-probability transaction group; if a low-probability signal is output, classify the customer into the low-probability transaction group.

[0135] Example 3, please refer to Figure 4 as shown Figure 4The structure schematic diagram of an electronic device is exemplified. The electronic device may include: a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete communication with each other through the communication bus. The memory stores computer-readable instructions. The processor can call the instructions in the memory. When the computer-readable instructions are executed by the processor, the steps in the business management method based on AI keyword annotation are run to achieve the following functions: obtaining the text content output by the customer in the business platform; performing keyword annotation and semantic analysis on the text content to extract the customer's demand information; based on the keyword annotation and semantic analysis, extracting the customer's transaction intention, and at the same time, based on the demand information and the transaction intention, classifying the customers, and dividing the customers into different transaction probability groups.

[0136] In addition, when the logical instructions in the above-mentioned memory are implemented in the form of software functional units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs, Read-Only Memories), random access memories (RAMs, Random Access Memories), magnetic disks, or optical discs that can store program codes.

[0137] Embodiment 4, the present application also provides a computer-readable storage medium. The present application provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps in the business management method based on AI keyword annotation are run to achieve the following functions: obtaining the text content output by the customer in the business platform; performing keyword annotation and semantic analysis on the text content to extract the customer's demand information; based on the keyword annotation and semantic analysis, extracting the customer's transaction intention, and at the same time, based on the demand information and the transaction intention, classifying the customers, and dividing the customers into different transaction probability groups.

[0138] Through the description of the above embodiments, the embodiments of the present invention can be provided as a method, a system or a computer program product. Based on such an understanding, the above technical solution, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0139] In the embodiments provided in the present application, it should be understood that the disclosed system or method can be implemented in other ways. The above-described embodiments are merely illustrative. For example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation. For another example, multiple modules or units can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some communication interfaces. The indirect coupling or communication connection of systems, modules, and units can be electrical, mechanical, or other forms.

[0140] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A business management method based on AI keyword annotation, characterized in that: The steps include: Obtain the text content input by the customer in the business platform; Perform keyword tagging and semantic analysis on text content to extract customer demand information; Based on keyword tagging and semantic analysis, the customer's transaction intention is extracted. At the same time, based on demand information and transaction intention, the customer's intention is classified and divided into different transaction probability groups.

2. The business management method based on AI keyword annotation according to claim 1 is characterized in that: Obtaining the text content output by the customer in the business platform includes the following sub-steps: Obtain the voice data output by the customer in the business platform; Convert voice data into text, convert language data into text content through voice-to-text conversion; Extract the text data output by customers in the business platform and mark it as text content.

3. The business management method based on AI keyword annotation according to claim 2 is characterized in that: Keyword tagging and semantic analysis of text content to extract customer demand information and transaction intentions include the following sub-steps: The text content is segmented by Jieba word segmentation to obtain text segmentation. By comparing it with the stop word database, the stop words in the text segmentation are screened out and removed, and the retained text segmentation is marked as text keywords. Perform semantic analysis on text keywords and identify demand keywords in text keywords through NLP technology. Demand keywords are defined as keywords that can represent or imply that customers have demand for a certain product. Get the first punctuation mark after the required keyword, excluding quotation marks, and mark the sentence composed of the text keywords between the required keyword and the punctuation mark as a sentence to be confirmed; Perform semantic analysis on the statement to be confirmed and use NLP technology to analyze whether the statement to be confirmed is fluent. If so, mark the statement to be confirmed as the requirement name. Otherwise, output a signal that the statement is not fluent. If the output statement does not pass the signal, remove the text keywords at the end of the statement to be confirmed and repeat the analysis until the requirement name is obtained; Check whether there is a time keyword in the text keyword, where the time keyword is defined as a keyword representing time or date. If there is a time keyword in the text keyword, the time keyword is marked as the required time, otherwise a signal of no time requirement is output; The requirement name and the requirement time are marked together as the requirement information.

4. The business management method based on AI keyword annotation according to claim 3 is characterized in that: Based on keyword annotation and semantic analysis, the customer's transaction intention is extracted. At the same time, based on the demand information and transaction intention, the customer's intention is classified. The customers are divided into different transaction probability groups, which includes the following sub-steps: Extract the customer's transaction intention based on keyword annotation and semantic analysis; Based on demand information and transaction intention, customers are classified and divided into different transaction probability groups.

5. The business management method based on AI keyword annotation according to claim 4 is characterized in that: Based on keyword annotation and semantic analysis, extracting the customer's transaction intention includes the following sub-steps: The number of times customers inquire about product details based on the name of their needs is named product inquiries. When customers inquire about the performance, appearance, and function of a product, it is considered as inquiring about product details. The number of times customers inquire about the price of a product based on the name of their needs is named price inquiries. The number of times customers inquire about the time based on the name of their needs is named time inquiries. Analyze text keywords and text content through NLP technology, and count the number of product inquiries, price inquiries, and time inquiries; Obtaining historical inquiry records, wherein the historical inquiry records store historical data of the number of product inquiries, the number of price inquiries, and the number of time inquiries, respectively named as the number of product historical inquiries, the number of price historical inquiries, and the number of time historical inquiries, and also store whether the customer finally completed the transaction; Get the historical product inquiry times, price inquiry times, and time inquiry times of the customers who finally made a deal, and mark them as HT1, HT2, and HT3 respectively; get the historical product inquiry times, price inquiry times, and time inquiry times of the customers who finally did not make a deal, and mark them as HF1, HF2, and HF3 respectively; Get the median of HT1 and the median of HF1, mark them as ZA1 and ZA2 respectively, then calculate the median of ZA1 and ZA2, mark them as the commodity inquiry threshold, represented by symbol A; get the median of HT2 and the median of HF2, mark them as ZB1 and ZB2 respectively, then calculate the median of ZB1 and ZB2, mark them as the price inquiry threshold, represented by symbol B; get the median of HT3 and the median of HF3, mark them as ZC1 and ZC2 respectively, then calculate the median of ZC1 and ZC2, mark them as the time inquiry threshold, represented by symbol C; Calculate A / (A+B+C), mark the result as the commodity inquiry weight, represented by symbol K1, calculate B / (A+B+C), mark the result as the price inquiry weight, represented by symbol K2, calculate C / (A+B+C), mark the result as the time inquiry weight, represented by symbol K3; Get max(HF1), remove HT1 that is less than or equal to max(HF1), and then get min(HT1), marked as U1, where max() is the maximum value operator and min() is the minimum value operator; get max(HF2), remove HT2 that is less than or equal to max(HF2), and then get min(HT2), marked as U2; get max(HF3), remove HT3 that is less than or equal to max(HF3), and then get min(HT3), marked as U3; Calculate U1×K1+U2×K2+U3×K3, and mark the result as the transaction index, represented by the symbol R; The transaction intention of each customer is calculated in real time based on the calculated product inquiry weight, price inquiry weight, time inquiry weight and transaction index.

6. The business management method based on AI keyword annotation according to claim 5 is characterized in that: The real-time calculation of each customer's transaction intention based on the calculated product inquiry weight, price inquiry weight, time inquiry weight and transaction index includes the following sub-steps: Get the number of product inquiries, price inquiries, and time inquiries of customers, marked as Y1, Y2, and Y3 respectively; Set the first increase threshold, represented by the symbol S, calculate Y1 / A, mark the integer digit of the calculation result as X1, increase K1 by S×X1, and obtain P1; calculate Y2 / B, mark the integer digit of the calculation result as X2, increase K2 by S×X2, and obtain P2; calculate Y3 / C, mark the integer digit of the calculation result as X3, increase K3 by S×X3, and obtain P3; By formula Calculate the customer's transaction intention, which represents the transaction probability, where L is the transaction intention.

7. The business management method based on AI keyword annotation according to claim 6 is characterized in that: Based on demand information and transaction intention, customers are classified into different transaction probability groups, including the following sub-steps: Set up high probability transaction groups, medium probability transaction groups, and low probability transaction groups; Setting a first probability threshold and a second probability threshold, wherein the first probability threshold is greater than the second probability threshold; The customer's transaction intention is compared with the first probability threshold and the second probability threshold. If the transaction intention is greater than or equal to the first probability threshold, a high probability signal is output; if the transaction intention is less than the first probability threshold and greater than or equal to the second probability threshold, a medium probability signal is output; if the transaction intention is less than the second probability threshold, a low probability signal is output; If a high probability signal is output, the customer is classified into a high probability transaction group. If a medium probability signal is output, the customer is classified into a medium probability transaction group. If a low probability signal is output, the customer is classified into a low probability transaction group.

8. A business management system based on AI keyword annotation, used to implement the business management method based on AI keyword annotation according to any one of claims 1 to 7, characterized in that: It includes a text acquisition module, a semantic analysis module and a transaction probability division module; the text acquisition module and the semantic analysis module are respectively connected to the transaction probability division module data; The text acquisition module is used to acquire the text content output by the customer in the business platform; The semantic analysis module is used to perform keyword tagging and semantic analysis on the text content to extract customer demand information; The transaction probability classification module is used to extract the customer's transaction intention based on keyword tagging and semantic analysis, and at the same time, classify the customer's intention based on demand information and transaction intention, and divide the customer into different transaction probability groups.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the steps in the method according to any one of claims 1 to 7 are executed.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps in the method according to any one of claims 1 to 7 are executed.

Images