Automobile financial customer consultation method and system based on natural language processing

Abstract

The invention discloses an automobile financial customer consultation method and system based on natural language processing, and relates to the technical field of data processing, and the method comprises the steps: obtaining an automobile financial database, obtaining a consultation information set and keywords, and obtaining first consultation text information and second consultation text information; acquiring second word set data, acquiring a second occurrence frequency, acquiring a second effective occurrence frequency, and acquiring a first weight coefficient; obtaining first word set data, obtaining a first occurrence frequency, obtaining a first effective occurrence frequency, recording a keyword with the first effective occurrence frequency and a second effective occurrence frequency as a to-be-analyzed word, obtaining a word frequency difference value, obtaining a correction parameter according to the word frequency difference value, and obtaining a second weight coefficient according to the correction parameter and the first weight coefficient; and obtaining a consultation strategy according to each consultation information set and the corresponding second weight coefficient. The method has the advantages of being accurate and comprehensive, improving customer consultation experience and enhancing response efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to an automobile finance customer consultation method and system based on natural language processing. Background Art

[0002] With the booming development of the automobile finance industry, customers' demands for automobile finance-related consultations are increasing day by day. In order to improve the efficiency and quality of automobile finance customer consultations, the industry has begun to explore the use of natural language processing technology to assist or replace human customer service. Natural language processing technology can simulate the process of human language understanding and generation, realize the automatic analysis and processing of customer consultation texts, quickly and accurately extract the key information of customer consultations, and provide strong support for subsequent responses.

[0003] However, the existing automobile finance customer consultation methods based on natural language processing still have some deficiencies. On the one hand, these methods often only focus on the consultation text information sent by customers once, while ignoring the relevance and trend between multiple consultation texts continuously sent by customers within a certain period of time. In actual applications, customers may continuously send multiple related consultations because they have doubts about a certain problem or need further understanding. There are often internal logical connections and evolution laws between these consultations. If only single consultation texts are analyzed and processed while ignoring this relevance and trend, it may lead to inaccurate or incomplete understanding of customers' consultation needs, thus affecting the accuracy and effectiveness of subsequent responses. On the other hand, when processing customer consultation texts, the existing methods usually simply determine the most relevant consultation information set based on keyword matching or semantic similarity calculation and give responses accordingly. Although this method can meet customers' consultation needs to a certain extent, it lacks in-depth understanding and analysis of customers' consultation intentions, resulting in instability and unreliability of subsequent response results.

[0004] Therefore, there is an urgent need for an automobile finance customer consultation method based on natural language processing that can fully consider the relevance and trend between multiple consultation texts continuously sent by customers within a certain period of time, and can deeply understand and analyze customers' consultation intentions. The present invention is precisely proposed in view of the above problems, aiming to improve the efficiency and quality of customer consultations and meet the actual needs of the automobile finance industry by means of a new consultation content conversion method to convert customers' continuous consultation texts into content convenient for subsequent analysis and response. Summary of the Invention

[0005] In view of the defects in the prior art, the present invention provides an automobile finance customer consultation method and system based on natural language processing.

[0006] A method for consulting automotive finance customers based on natural language processing, comprising: obtaining an automotive finance database, obtaining multiple consultation information sets and corresponding keywords according to the automotive finance database, and obtaining the first consultation text information and the second consultation text information successively sent by the automotive finance customer; obtaining second subset data according to the second consultation text information, obtaining the second occurrence times of the keywords corresponding to each consultation information set in the second subset data, obtaining the second occurrence times exceeding a preset number threshold and recording them as the second effective occurrence times, and obtaining the first weight coefficient corresponding to each consultation information set based on the weight model and the second effective occurrence times of the keywords corresponding to each consultation information set; obtaining first subset data according to the first consultation text information, obtaining the first occurrence times of the keywords corresponding to each consultation information set in the first subset data, obtaining the first occurrence times exceeding a preset number threshold and recording them as the first effective occurrence times, and recording the keywords with the first effective occurrence times and the second effective occurrence times as the words to be analyzed, obtaining the word frequency difference according to the first effective occurrence times and the second effective occurrence times of the words to be analyzed, obtaining a correction parameter according to the word frequency difference, and obtaining a second weight coefficient according to the correction parameter and the first weight coefficient; obtaining a consultation strategy according to each consultation information set and its corresponding second weight coefficient.

[0007] Optionally, obtaining a consultation strategy according to each consultation information set and its corresponding second weight coefficient includes: arranging each second weight coefficient in descending order and constructing a consultation content framework; forming multiple response data spaces corresponding to the second weight coefficients in the consultation content framework, wherein the proportion of each response data space is equal to the proportion of the corresponding second weight coefficient; adding the data in the consultation information set corresponding to each second weight coefficient to the response data space corresponding to each second weight coefficient.

[0008] Optionally, the weight model in obtaining the first weight coefficient corresponding to each consultation information set based on the weight model and the second effective occurrence times of the keywords corresponding to each consultation information set is expressed as: wherein, A i1 is the first weight coefficient corresponding to the i-th consultation information set, T i2 is the second effective occurrence time corresponding to the i-th consultation information set, T j2 is the second effective occurrence time corresponding to the j-th consultation information set, and n is the number of consultation information sets with the second effective occurrence time.

[0009] Optionally, obtaining the word frequency difference according to the first effective occurrence times and the second effective occurrence times of the words to be analyzed includes: obtaining the time interval between the sending times of the first consultation text information and the second consultation text information; matching a time coefficient according to the time interval; obtaining the word frequency difference according to the time coefficient, the first effective occurrence times and the second effective occurrence times of the words to be analyzed.

[0010] Optionally, the word frequency difference is obtained based on the time coefficient, the first effective occurrence times and the second effective occurrence times of the word to be analyzed, and is expressed as: D k =(T k2 -T k1 ) α ; where D k is the word frequency difference of the k-th word to be analyzed, T k1 is the first effective occurrence times of the k-th word to be analyzed, T k2 is the second effective occurrence times of the k-th word to be analyzed, and α is the time coefficient.

[0011] Optionally, the correction parameter is obtained based on the word frequency difference, and is expressed as: where P kd is the correction parameter of the k-th word to be analyzed, and D k is the word frequency difference of the k-th word to be analyzed.

[0012] Optionally, the second weight coefficient is obtained based on the correction parameter and the first weight coefficient, and is expressed as: A k2 =P kd ·A k1 ; where A k2 is the second weight coefficient corresponding to the consultation information set of the k-th word to be analyzed, P kd is the correction parameter of the k-th word to be analyzed, and A k1 is the first weight coefficient corresponding to the consultation information set of the k-th word to be analyzed.

[0013] There is also provided an automotive finance customer consultation system based on natural language processing. The system includes: an acquisition module, configured to acquire an automotive finance database, and based on the automotive finance database, acquire multiple consultation information sets and the keywords corresponding to the consultation information sets, and acquire the first consultation text information and the second consultation text information successively sent by an automotive finance customer; a first language processing module, configured to acquire second character set data according to the second consultation text information, and acquire the second occurrence times of the keywords corresponding to each consultation information set in the second character set data, acquire the second occurrence times exceeding a preset number threshold and record them as second effective occurrence times, and based on a weight model and the second effective occurrence times of the keywords corresponding to each consultation information set, acquire the first weight coefficient corresponding to each consultation information set; a second language processing module, configured to acquire first character set data according to the first consultation text information, and acquire the first occurrence times of the keywords corresponding to each consultation information set in the first character set data, acquire the first occurrence times exceeding a preset number threshold and record them as first effective occurrence times, and record the keywords having the first effective occurrence times and the second effective occurrence times as the words to be analyzed, acquire a word frequency difference according to the first effective occurrence times and the second effective occurrence times of the words to be analyzed, acquire a correction parameter according to the word frequency difference, and acquire a second weight coefficient according to the correction parameter and the first weight coefficient; a consultation generation module, configured to acquire a consultation strategy according to each consultation information set and its corresponding second weight coefficient.

[0014] Optionally, the consultation generation module is further configured to: arrange each second weight coefficient in descending order and construct a consultation content framework; form multiple response data spaces corresponding to the second weight coefficients in the consultation content framework, wherein the proportion of each response data space is equal to the proportion of the corresponding second weight coefficient; and add the data in the consultation information set corresponding to each second weight coefficient to the response data space corresponding to each second weight coefficient.

[0015] Optionally, the second language processing module is further configured to: acquire the time interval between the sending times of the first consultation text information and the second consultation text information; match a time coefficient according to the time interval; and acquire a word frequency difference according to the time coefficient, the first effective occurrence times and the second effective occurrence times of the words to be analyzed.

[0016] The beneficial effects of the present invention are embodied in:

[0017] In the entire natural language processing-based automotive finance customer consultation method, not only the customer's single consultation text is concerned, but also the relevance and trend between multiple consultation text messages continuously sent by the customer within a certain period of time are deeply analyzed, which helps to more accurately understand the customer's consultation needs and intentions. Further, by obtaining the customer's first consultation text information and second consultation text information, and respectively performing keyword matching and occurrence frequency statistics with the consultation information set in the automotive finance database, this solution can initially determine which consultation information sets in the database are relevant to the customer's consultation and calculate the first weight coefficient. Then, by comparing the change in the occurrence frequency of keywords in the consultation texts sent by the customer successively, and combining with the time coefficient to calculate the correction parameter, the first weight coefficient is corrected to obtain the second weight coefficient. This step fully considers the dynamic change and continuity of the customer's consultation needs, making the weight coefficient more accurately reflect the customer's actual focus of attention. Further, according to each consultation information set and its corresponding second weight coefficient, a comprehensive and accurate consultation strategy is constructed. This strategy not only covers the issues that the customer is most concerned about, but also organizes the answer content according to the importance and logical relationship of the issues, ensuring the coherence and understandability of the answer. It not only improves the accuracy and comprehensiveness of understanding the customer's consultation needs, but also enhances the stability and reliability of the subsequent response, thus greatly improving the customer consultation experience and meeting the actual needs of the automotive finance industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. In all the drawings, similar elements or parts are generally denoted by similar reference numerals. In the drawings, the elements or parts do not necessarily draw according to the actual scale.

[0019] Figure 1 It is a schematic diagram of the steps of the automotive finance customer consultation method based on natural language processing of the present invention;

[0020] Figure 2 It is a partial schematic diagram of the steps of S4 in the automotive finance customer consultation method based on natural language processing of the present invention;

[0021] Figure 3 It is a partial schematic diagram of the steps of S3 in the automotive finance customer consultation method based on natural language processing of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and illustrated herein generally can be arranged and designed in a variety of different configurations.

[0023] Therefore, the detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. 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 scope of protection of the present invention.

[0024] It should be noted that: like reference numerals and letters denote like items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. In addition, the terms "first", "second", etc. are only used for descriptive distinction and cannot be construed as indicating or implying relative importance.

[0025] As Figure 1 shown, a method for automobile finance customer consultation based on natural language processing is provided, including:

[0026] S1. Obtain an automobile finance database, obtain multiple consultation information sets and corresponding keywords according to the automobile finance database, and obtain the first consultation text information and the second consultation text information successively sent by an automobile finance customer;

[0027] S2. Obtain second subset data according to the second consultation text information, obtain the second occurrence times of the keywords corresponding to each consultation information set in the second subset data, obtain the second occurrence times exceeding a preset number threshold and record them as second effective occurrence times, and obtain the first weight coefficient corresponding to each consultation information set based on a weight model and the second effective occurrence times of the keywords corresponding to each consultation information set;

[0028] S3. Obtain first subset data according to the first consultation text information, obtain the first occurrence times of the keywords corresponding to each consultation information set in the first subset data, obtain the first occurrence times exceeding a preset number threshold and record them as first effective occurrence times, and record the keywords having the first effective occurrence times and the second effective occurrence times as words to be analyzed. Obtain a word frequency difference according to the first effective occurrence times and the second effective occurrence times of the words to be analyzed, obtain a correction parameter according to the word frequency difference, and obtain a second weight coefficient according to the correction parameter and the first weight coefficient;

[0029] S4. Obtain a consultation strategy according to each consultation information set and its corresponding second weight coefficient.

[0030] In this embodiment, it should be noted that in S1, first, an auto finance database is obtained. The auto finance database contains a large number of consultation information sets, and each consultation information set covers specific problems in the field of auto finance and their relevant answers or explanations. In order to accurately match and analyze the customer's consultation text subsequently, it is also necessary to extract the keywords corresponding to each consultation information set from the database. These keywords are the core identifiers of the consultation information sets and can reflect the themes and contents of the consultation information sets. At the same time, the first consultation text information and the second consultation text information successively sent by the auto finance customer are also obtained. These information are the specific problems or requirements put forward by the customer during the actual consultation process and are also the main objects for subsequent analysis and processing.

[0031] Specifically, after obtaining the auto finance database, representative keywords are extracted from each consultation information set through data mining and text analysis techniques. For example, for the four consultation information sets of "auto loan amount", "conditions required for auto loan", "auto loan application process", and "auto loan term", the keywords of "amount", "conditions", "application process", and "term" are extracted respectively. Then, after obtaining the first consultation text information and the second consultation text information successively sent by the auto finance customer. For example, the customer may first send a consultation text about "auto loan interest rate", and then send a consultation text about "time required for loan application". There is a certain correlation and trend between these two consultation texts. Through the acquisition and processing in step S1, it can provide strong data support for subsequent analysis and response.

[0032] In S2, the second consultation text information is mainly analyzed to determine its degree of association with each consultation information set in the auto finance database. First, all the words in the second consultation text information are extracted to form the second word set data. Then, each consultation information set in the auto finance database is traversed. For each consultation information set, the number of occurrences of its corresponding keyword in the second word set data is counted, which is the second occurrence count. To filter out accidentally occurring keywords, a preset count threshold is set. Only when the number of occurrences of a keyword in the second word set data exceeds this threshold is it considered valid and recorded as the second valid occurrence count. Then, based on a weight model, the second valid occurrence counts of the keywords corresponding to each consultation information set are used to calculate the first weight coefficient for each consultation information set. This weight model evaluates the degree of relevance between the consultation information set and the second consultation text information according to the number of occurrences of the keyword. For example, if keywords related to "auto loan amount" frequently appear in the second consultation text information, then the consultation information set related to "auto loan amount" will obtain a higher first weight coefficient. In this way, it can be preliminarily determined which consultation information sets the second consultation text information is closer to, providing a strong basis for subsequent analysis and response.

[0033] For example, assume that the second consultation text information sent by the customer is: "I want to know about the content of auto loans. May I ask about the loan amount... application process... loan time... required conditions...". In step S2, first, all the words in this text are extracted to form the second word set data. Then, each consultation information set in the auto finance database is traversed, and it is found that there are two consultation information sets related to "auto loan amount" and "auto loan application process" respectively, and the second occurrence counts of their keywords in the second word set data both exceed the preset count threshold. Therefore, these two consultation information sets are considered to have a high degree of relevance to the second consultation text information, and their first weight coefficients are calculated respectively. If the keyword related to "auto loan amount" appears more times, then the consultation information set related to "auto loan amount" will obtain a higher first weight coefficient, indicating that the customer may be more concerned about this aspect of the problem.

[0034] In S3, the first consultation text information is mainly analyzed, and combined with the analysis results of the second consultation text information, to further correct and optimize the weight coefficients of each consultation information set. First, according to the first consultation text information, all the words in it are extracted to form the first word set data. Then, traverse each consultation information set that has been determined to be related to the second consultation text information in S2. For each consultation information set, count the number of occurrences of its corresponding keyword in the first word set data, that is, the first occurrence frequency. Similarly, in order to filter out keywords that appear accidentally and are not substantially helpful for analysis, a preset frequency threshold is set. Only when the number of occurrences of a keyword in the first word set data exceeds this threshold is it considered valid, and it is recorded as the first valid occurrence frequency. Then, the keywords that have both the first valid occurrence frequency and the second valid occurrence frequency are recorded as the words to be analyzed. These words to be analyzed are the key links connecting the first consultation text information and the second consultation text information, and they can reflect the continuity and trend of the customer's consultation needs. Next, according to the first valid occurrence frequency and the second valid occurrence frequency of the words to be analyzed, calculate the word frequency difference. The word frequency difference reflects the change in the customer's attention to the same problem at different time periods. To more accurately evaluate the impact of this change on the weight coefficient, a time coefficient is also introduced. The time coefficient is determined according to the time interval between the first consultation text information and the second consultation text information, and it reflects the time urgency and continuity of the customer's consultation needs. Finally, according to the time coefficient, the first valid occurrence frequency and the second valid occurrence frequency of the words to be analyzed, calculate the correction parameter, and according to the correction parameter and the first weight coefficient, obtain the second weight coefficient. The second weight coefficient more comprehensively considers the relevance and trend between multiple consultation text information continuously sent by the customer within a certain time period, providing strong support for formulating more accurate consultation strategies in the future.

[0035] For example, assume that the first consultation text information sent by the customer is: "I want to know about the loan amount of auto loans..., amount..., application..., ... application..., ... amount..., process...", and the second consultation text information sent subsequently is: "If it is an auto loan..., ... the application process will be..., so... how long will it take... the process..., then... the process...". In S3, first, all words are extracted from the first consultation text information to form the first set of word data. Then, the two consultation information sets of "auto loan amount" and "auto loan application process" that have been determined to be relevant to the second consultation text information in S2 are traversed. For the "auto loan amount" consultation information set, it is found that the first occurrence times of its keywords "auto", "loan", "amount", "application", and "process" in the first set of word data exceed the preset number threshold, so it is recorded as the first effective occurrence times. Next, the keywords that have both the first effective occurrence times and the second effective occurrence times (such as "application") are recorded as the words to be analyzed, and their word frequency differences are calculated. Assume that the occurrence times of "amount", "application", and "process" in the first set of word data are relatively high, while their occurrence times in the second set of word data are relatively low. This may mean that the customer was more concerned about the application aspect during the initial consultation, and may have had a more specific understanding or need during the subsequent consultation. To more accurately reflect the impact of this change on the weight coefficient, the time coefficient is matched according to the time interval between the first consultation text information and the second consultation text information, and the correction parameter is calculated accordingly. Finally, the second weight coefficient is obtained based on the correction parameter and the first weight coefficient. In this way, the relevance and trend between multiple consultation text information continuously sent by the customer within a certain period of time can be more comprehensively considered, providing strong support for formulating more accurate consultation strategies in the future.

[0036] In S4, based on the analysis results of the previous S1 to S3, the final consultation strategy is formulated. In S4, first, each consultation information set and its corresponding second weight coefficient are obtained. These second weight coefficients are derived after fully considering the relevance and trend among multiple consultation text messages continuously sent by the customer within a certain period of time, as well as deeply understanding and analyzing the customer's consultation intention. They reflect the degree of relevance and importance of each consultation information set to the customer's current consultation needs. Next, it is necessary to construct a consultation strategy based on these second weight coefficients, which can comprehensively and accurately reflect the customer's consultation needs. This consultation strategy will serve as the basis for subsequent responses, ensuring that the answers can closely follow the customer's questions and meet their consultation needs. To construct such a consultation strategy, the content of each consultation information set, the magnitude of the second weight coefficient, and their mutual relationships are comprehensively considered to ensure that the consultation strategy is both comprehensive and accurate. In specific implementation, the second weight coefficients may be sorted first to find the consultation information sets with higher weights, which are usually the issues that the customer is most concerned about. Then, based on the content and characteristics of these consultation information sets, the answers are organized to ensure that all aspects that the customer cares about are covered in the answers. At the same time, the logical relationships among the consultation information sets are also considered to ensure that the content of the answers is well-organized and easy to understand. In addition, according to the magnitude of the second weight coefficient, the level of detail of each part in the answers is adjusted to ensure that the key content is fully elaborated while the secondary content is appropriately simplified. In this way, a comprehensive and accurate consultation strategy can be formulated to provide strong support for subsequent responses.

[0037] For example, assume that the customer has continuously sent two consultation text messages within a period of time. The first one is about the amount and application conditions of auto loans, and the second one is about the application process and required time of auto loans. After the analysis in steps S1 to S3, two consultation information sets are obtained: "auto loan amount" and "auto loan application process", and their corresponding second weight coefficients are 0.6 and 0.4 respectively. In step S4, the consultation strategy is formulated based on these two second weight coefficients. Specifically, the two weight coefficients are sorted first, and it is found that the weight of "auto loan amount and application conditions" is higher. Therefore, this part of the content will be the focus of the answer. In the answer, the amount and application conditions of auto loans will be elaborated in detail, including the range of the amount and the specific requirements of the application conditions. At the same time, the application process and required time of auto loans will also be mentioned, but this part of the content will be relatively simplified and only used as supplementary information. In this way, a comprehensive and accurate consultation strategy can be formulated to ensure that the answers can closely follow the customer's questions and meet their consultation needs.

[0038] In summary, in the entire automotive finance customer consultation method based on natural language processing, not only the customer's single consultation text is concerned, but also the relevance and trend between multiple consultation text messages continuously sent by the customer within a certain period of time are deeply analyzed, which helps to more accurately understand the customer's consultation needs and intentions. Further, by obtaining the customer's first consultation text information and second consultation text information, and respectively performing keyword matching and occurrence frequency statistics with the consultation information set in the automotive finance database, this solution can initially determine which consultation information sets in the database are relevant to the customer's consultation, and calculate the first weight coefficient. Then, by comparing the change in the occurrence frequency of keywords in the consultation texts sent by the customer successively, and combining the time coefficient to calculate the correction parameter, the first weight coefficient is corrected to obtain the second weight coefficient. This step fully considers the dynamic change and continuity of the customer's consultation needs, making the weight coefficient more accurately reflect the customer's actual focus of attention. Further, according to each consultation information set and its corresponding second weight coefficient, a comprehensive and accurate consultation strategy is constructed. This strategy not only covers the issues that the customer is most concerned about, but also organizes the answer content according to the importance and logical relationship of the issues, ensuring the clarity and comprehensibility of the answer. It not only improves the accuracy and comprehensiveness of understanding the customer's consultation needs, but also enhances the stability and reliability of subsequent responses, thus greatly improving the customer consultation experience and meeting the actual needs of the automotive finance industry.

[0039] As Figure 2 shown, in one embodiment, obtaining the consultation strategy according to each consultation information set and its corresponding second weight coefficient in S4 includes:

[0040] S41. Arrange each second weight coefficient in descending order and construct a consultation content framework;

[0041] S42. Form multiple response data spaces corresponding to the second weight coefficients in the consultation content framework, where the proportion of each response data space is equal to the proportion of the corresponding second weight coefficient;

[0042] S43. Add the data in the consultation information set corresponding to each second weight coefficient to the response data space corresponding to each second weight coefficient.

[0043] In this embodiment, it should be noted that in S41, the second weight coefficients corresponding to each consultation information set need to be arranged in ascending order. The purpose of this step is to determine the priority or importance of each aspect in the customer's consultation needs. Through sorting, it can be clearly seen which consultation information sets are of the greatest concern to the customer and which are secondary. This sorting is not only based on the customer's interests and concerns shown in the continuous consultation text but also takes into account the second weight coefficients calculated in the previous steps. These coefficients comprehensively reflect the degree of association between the customer's consultation text and the consultation information sets in the database as well as the dynamic changes in the customer's consultation needs. After completion of the sorting, a consultation content framework is constructed based on these sorted consultation information sets. This framework will serve as the basis for organizing the subsequent response content, ensuring that the answer can be organized according to the priority that the customer cares about, thereby improving the pertinence and effectiveness of the answer. For example, assume that after the previous analysis, three consultation information sets are obtained: "Automobile loan amount", "Automobile loan application process", and "Conditions required for automobile loan", and their corresponding second weight coefficients are 0.5, 0.3, and 0.2 respectively. In step S41, first, these weight coefficients are arranged in ascending order to obtain "Automobile loan amount" (0.5) > "Automobile loan application process" (0.3) > "Conditions required for automobile loan" (0.2). Then, a consultation content framework is constructed based on this sorting. In the framework, the part of "Automobile loan amount" will be placed at the front, followed by "Automobile loan application process", and finally "Conditions required for automobile loan".

[0044] In S42, multiple response data spaces corresponding to the second weight coefficients are formed in the consultation content framework constructed in S41. These response data spaces are prepared for filling in specific response content later, and their sizes or proportions are equal to the proportions of the corresponding second weight coefficients. This means that for the consultation information set that the customer cares more about, the corresponding response data space is larger, and more detailed information and explanations will be filled in it. Through such a design, it can be ensured that the final response content can accurately reflect the customer's consultation needs and is organized according to the degree of concern of the customer, thereby improving the coherence and comprehensibility of the response. Continuing with the above example, in S42, according to the second weight coefficients (0.5, 0.3, 0.2) of the three consultation information sets of "Automobile loan amount", "Automobile loan application process", and "Conditions required for automobile loan", 50%, 30%, and 20% of the response data spaces are allocated to them respectively in the consultation content framework. This means that in the final response content, 50% of the space will be used to explain in detail the issues related to "Automobile loan amount", 30% of the space to explain "Automobile loan application process", and 20% of the space to explain "Conditions required for automobile loan".

[0045] In S43, the data within the consultation information sets corresponding to each second weight coefficient will be added to the corresponding response data space formed in S42. This data may include specific answers, explanations, examples, etc., all of which are prepared to meet the customer's consultation needs. Through this step, it can be ensured that the final response content not only covers all aspects that the customer cares about, but also is organized according to the degree of concern of the customer, and the information for each aspect is detailed and accurate enough. Such a response content will greatly improve the customer's satisfaction and consultation experience.

[0046] Continuing with the above example, in S43, specific data and answers related to the three consultation information sets of "auto loan amount", "auto loan application process", and "conditions required for auto loan" will be extracted from the auto finance database and added to the 50%, 30%, and 20% response data spaces allocated in S42 respectively. For example, in the response data space of "auto loan amount", detailed information such as the amount range, calculation method, influencing factors, etc. may be added; in the response data space of "auto loan application process", specific descriptions such as application steps, required materials, review time, etc. may be added; in the response data space of "conditions required for auto loan", requirements regarding age, income, credit record, etc. may be added. In this way, a comprehensive and accurate consultation strategy can be constructed to provide strong support for subsequent responses.

[0047] In one implementation, the weight model for obtaining the first weight coefficient corresponding to each consultation information set based on the weight model and the second effective occurrence times of the keywords corresponding to each consultation information set in S2 is expressed as:

[0048] Where,

[0049] A i1 is the first weight coefficient corresponding to the i-th consultation information set, T i2 is the second effective occurrence times corresponding to the i-th consultation information set, T j2 is the second effective occurrence times corresponding to the j-th consultation information set, and n is the number of consultation information sets with second effective occurrence times.

[0050] In this implementation, it should be noted that A i1 represents the first weight coefficient corresponding to the i-th consultation information set; this coefficient reflects the degree of association between the i-th consultation information set and the customer's second consultation text information, and is calculated based on the occurrence times of the keywords in the second consultation text information. T i2It represents the second effective occurrence times of the keyword corresponding to the i-th consultation information set in the second subset data; the effective occurrence times refer to the number of times the keyword appears in the second consultation text information exceeding a preset threshold, which is considered effective and can reflect the customer's attention to the relevant content of the consultation information set. It represents the sum of the second effective occurrence times of all consultation information sets with the second effective occurrence times; n is the number of consultation information sets with the second effective occurrence times; this summation operation is to summarize the keyword occurrence times of all relevant consultation information sets as the denominator to calculate the relative weight of each consultation information set.

[0051] By dividing the second effective occurrence times of each consultation information set by the sum of the second effective occurrence times of all consultation information sets, normalization processing is achieved; the normalized weight coefficient A i1 Is between 0 and 1, representing the degree of association of each consultation information set with the customer's second consultation text information relative to other consultation information sets; this processing method makes the weight coefficients between different consultation information sets comparable, facilitating subsequent analysis and decision-making. The weight model evaluates the degree of association between the consultation information set and the customer's consultation needs by calculating the number of times the keyword appears in the second consultation text information; the more times it appears, the more frequently the customer mentions the content related to the consultation information set in the second consultation text information, so the weight coefficient of this consultation information set is larger; this setting method helps to accurately identify the issues or fields that the customer is most concerned about in the current consultation. Through the weight model, it is possible to initially determine which consultation information sets the customer's second consultation text information is closer to, providing a strong basis for subsequent analysis and response; the weight coefficient reflects the degree of association between the customer's consultation needs and the consultation information sets in the database, helping to more accurately understand the customer's consultation intention and needs.

[0052] Such as Figure 3 As shown, in one embodiment, obtaining the word frequency difference according to the first effective occurrence times and the second effective occurrence times of the word to be analyzed in S3 includes:

[0053] S31. Obtain the time interval between the sending times of the first consultation text information and the second consultation text information;

[0054] S32. Match the time coefficient according to the time interval;

[0055] S33. Obtain the word frequency difference according to the time coefficient, the first effective occurrence times and the second effective occurrence times of the word to be analyzed.

[0056] In this embodiment, it should be noted that in S31, the main task is to obtain the time interval between the first consultation text information and the second consultation text information sent by the customer. This time interval is an important basis for subsequent calculation of the time coefficient, evaluation of the change speed and urgency of the customer's consultation needs. Specifically, the timestamp of each consultation text sent by the customer will be recorded. When the customer sends two or more consultation texts consecutively, the time interval between them will be calculated by comparing the timestamps of adjacent consultation texts. For example, if the customer sends the first consultation text at 10 am and the second consultation text at 3 pm, then the time interval between these two consultation texts is 5 hours. This time interval will be used in the subsequent step S32 to match the corresponding time coefficient, and then affect the calculation of the word frequency difference in step S33, and finally adjust the weight coefficients of each consultation information set to make the consultation strategy more in line with the actual consultation needs and intention changes of the customer.

[0057] In S32, a time coefficient will be matched according to the time interval calculated in S31. This time coefficient is a parameter used to reflect the continuity and urgency of the customer's consultation needs. It is usually a value that varies according to the length of the time interval. The setting of the time coefficient can be based on the actual business scenario and experience. For example, a rule can be set: if the time interval is very short (such as within a few minutes), the time coefficient is relatively large, indicating that the customer is very concerned about a certain issue and may urgently need an answer; if the time interval is relatively long (such as several hours or days), the time coefficient is relatively small, indicating that the customer may not have a high sense of urgency for the problem, or different questions are raised at different stages. Through such a setting, it is possible to more flexibly respond to the changes in the customer's consultation needs and ensure that the consultation strategy can accurately reflect the actual intention of the customer.

[0058] In S33, the time coefficient obtained in S32, as well as the first effective occurrence times and the second effective occurrence times of the word to be analyzed (i.e., the keyword that appears in both the first consultation text information and the second consultation text information and the number of occurrences exceeds the preset threshold), will be used to calculate the word frequency difference. The word frequency difference is an indicator used to measure the change in the customer's attention to the same problem at different time periods. Specifically, the difference in the number of occurrences of the word to be analyzed in the two consultation texts (i.e., the second effective occurrence times minus the first effective occurrence times) will be calculated first, and then the word frequency difference will be calculated based on this time difference and the time coefficient. The size of the word frequency difference will directly affect the calculation of the subsequent correction parameter, and then adjust the weight coefficients of each consultation information set. For example, if the number of occurrences of a certain word to be analyzed in the second consultation text is significantly more than that in the first consultation text, and the time interval is short, then the word frequency difference will be relatively large, indicating that the customer's attention to this problem has increased significantly in a short period of time, and more emphasis will be placed on this aspect when formulating the consultation strategy.

[0059] In one embodiment, the word frequency difference obtained according to the time coefficient, the first effective occurrence times and the second effective occurrence times of the word to be analyzed in S33 is expressed as:

[0060] D k =(T k2 -T k1 ) α ; where

[0061] D k is the word frequency difference of the kth word to be analyzed, T k1 is the first effective occurrence times of the kth word to be analyzed, T k2 is the second effective occurrence times of the kth word to be analyzed, and α is the time coefficient.

[0062] In this embodiment, it should be noted that the word frequency difference D k is calculated based on the difference in the effective occurrence times of the word to be analyzed in two consecutive consultation texts (T k2 -T k1 ). This difference reflects the change in the customer's attention to the same problem at different time periods. By calculating the difference, the changing trend of the customer's consultation needs can be quantified, providing data support for subsequent adjustment of the weight coefficient.

[0063] The time coefficient α is a parameter used to reflect the continuity and urgency of the customer's consultation needs. It is determined according to the time interval between the sending times of the first consultation text information and the second consultation text information. The shorter the time interval, the larger α is, indicating that the customer's attention to a certain problem has increased significantly in a short period of time, and the weight coefficient needs to be adjusted more quickly to respond to the customer's needs. By introducing the time coefficient, the changes in the customer's consultation needs can be more flexibly addressed, ensuring that the consultation strategy can accurately reflect the customer's actual intentions.

[0064] Expressing the word frequency difference as an exponential form of the difference (T k2 -T k1 ) α is to amplify the influence of the word frequency difference on the change in the customer's consultation needs. When the word frequency difference is large (i.e., the customer's attention to a certain problem has increased or decreased significantly in a short period of time), the exponential operation will further increase the word frequency difference, thus having a greater impact when subsequently correcting the weight coefficient. This setting method helps to more accurately capture the changing trend of the customer's consultation needs, ensuring that the consultation strategy can timely respond to the changes in the customer's needs.

[0065] In one embodiment, the correction parameter obtained according to the word frequency difference in S3 is expressed as:

[0066] where

[0067] P kdis the correction parameter for the k-th word to be analyzed, D k is the difference in word frequency of the k-th word to be analyzed.

[0068] In this embodiment, it should be noted that the correction parameter P kd is used to adjust the first weight coefficient to reflect the changing trend of customer consultation needs; by introducing the correction parameter, the changing degree of customer attention to the same problem at different time periods can be more accurately reflected, so as to optimize the final weight coefficient and make the consultation strategy more in line with the actual consultation needs of customers. The difference in word frequency D k is calculated based on the difference in the effective occurrence times of the word to be analyzed in two consecutive consultation texts, reflecting the change in customer attention; by introducing the difference in word frequency into the calculation of the correction parameter, this change can be quantified and the weight coefficient can be adjusted accordingly.

[0069] The exponential functions exp(-D k ) and exp(-D k ) can map the difference in word frequency D k to a non-linear space, making the correction parameter P kd more sensitive to the change in the difference in word frequency. Ensures that the value range of the correction parameter P kd is between (0, 2), avoiding the appearance of extreme values and making the adjustment of the weight coefficient smoother and more reasonable.

[0070] The constant 2 in the expression is used to adjust the scale of the correction parameter P kd ; by setting 2, it can be ensured that when the difference in word frequency D k is 0 (i.e., the customer continuously maintains attention), the correction parameter P kd is 1 and has no impact on the first weight coefficient. When the difference in word frequency D k increases, the correction parameter P kd gradually approaches 2, indicating that the customer's attention has increased significantly and the weight coefficient needs to be adjusted more substantially; conversely, when the difference in word frequency D k is negative and decreasing (i.e., the customer is not paying attention or reducing attention), the correction parameter P kd gradually approaches 0, indicating that the customer's attention has decreased and the weight coefficient needs to be reduced accordingly.

[0071] In one embodiment, obtaining the second weight coefficient according to the correction parameter and the first weight coefficient in S3 is expressed as:

[0072] A k2 = P kd · A k1 ; where

[0073] A k2is the second weight coefficient corresponding to the consultation information set corresponding to the k-th word to be analyzed, P kd is the correction parameter of the k-th word to be analyzed, A k1 is the first weight coefficient corresponding to the consultation information set corresponding to the k-th word to be analyzed.

[0074] In this embodiment, it should be noted that the second weight coefficient A k2 is based on the first weight coefficient A k1 On this basis, by introducing the correction parameter P kd After correction. It aims to more comprehensively consider the relevance and trend between multiple consultation text messages continuously sent by customers within a certain period of time, as well as the dynamic changes in customers' consultation needs, so as to ensure that the consultation strategy can more accurately reflect customers' actual consultation needs.

[0075] The correction parameter P kd is calculated based on the word frequency difference D k Calculated to reflect the change in customers' attention to the same problem at different times. By introducing the correction parameter, the first weight coefficient can be dynamically adjusted to better adapt to the changes in customers' consultation needs. Multiply the correction parameter P kd with the first weight coefficient A k1 to get the second weight coefficient A k2 . This operation method is simple and intuitive, and can directly reflect the influence degree of the correction parameter on the first weight coefficient.

[0076] When the correction parameter P kd is close to 2 (indicating a significant increase in customer attention), the second weight coefficient A k2 will increase significantly, indicating that the importance of this consultation information set in the current consultation has increased; on the contrary, when the correction parameter P kd is close to 0 (indicating a significant decrease in customer attention), the second weight coefficient A k2 will decrease accordingly, indicating that the importance of this consultation information set has decreased.

[0077] It also provides an automotive finance customer consultation system based on natural language processing. The system is used to implement the automotive finance customer consultation method based on natural language processing in any of the above embodiments. The system includes:

[0078] An acquisition module, configured to acquire an automotive finance database, and acquire multiple consultation information sets and keywords corresponding to the consultation information sets according to the automotive finance database, and acquire the first consultation text information and the second consultation text information successively sent by the automotive finance customer;

[0079] The first language processing module is used to obtain the second character set data according to the second consultation text information, obtain the second occurrence times of the keywords corresponding to each consultation information set in the second character set data, obtain the second occurrence times exceeding the preset times threshold and record them as the second effective occurrence times, and obtain the first weight coefficient corresponding to each consultation information set based on the weight model and the second effective occurrence times of the keywords corresponding to each consultation information set;

[0080] The second language processing module is used to obtain the first character set data according to the first consultation text information, obtain the first occurrence times of the keywords corresponding to each consultation information set in the first character set data, obtain the first occurrence times exceeding the preset times threshold and record them as the first effective occurrence times, and record the keywords with the first effective occurrence times and the second effective occurrence times as the words to be analyzed, obtain the word frequency difference according to the first effective occurrence times and the second effective occurrence times of the words to be analyzed, obtain the correction parameter according to the word frequency difference, and obtain the second weight coefficient according to the correction parameter and the first weight coefficient;

[0081] The consultation generation module is used to obtain the consultation strategy according to each consultation information set and its corresponding second weight coefficient.

[0082] In one embodiment, the consultation generation module is further used to: arrange the second weight coefficients in descending order and construct a consultation content framework; form multiple response data spaces corresponding to the second weight coefficients in the consultation content framework, where the proportion of each response data space is equal to the proportion of the corresponding second weight coefficient; add the data in the consultation information set corresponding to each second weight coefficient to the response data space corresponding to each second weight coefficient.

[0083] In one embodiment, the second language processing module is further used to: obtain the time interval between the sending times of the first consultation text information and the second consultation text information; match the time coefficient according to the time interval; obtain the word frequency difference according to the time coefficient, the first effective occurrence times and the second effective occurrence times of the words to be analyzed.

[0084] In this embodiment, it should be noted that for the above-mentioned automotive finance customer consultation system based on natural language processing, the specific implementation manner of the execution operation has been described in detail in the embodiments of the automotive finance customer consultation method based on natural language processing, and will not be elaborated here.

[0085] The preferred embodiments of the present disclosure have been described in detail above with reference to the accompanying drawings. However, the present disclosure is not limited to the specific details in the above embodiments. Within the technical concept of the present disclosure, various simple modifications can be made to the technical solutions of the present disclosure, and these simple modifications all fall within the protection scope of the present disclosure.

[0086] In addition, it should be noted that, in the above specific embodiments, the various specific technical features described can be combined in any appropriate manner without contradiction. To avoid unnecessary repetition, the present disclosure will not separately describe various possible combination methods.

[0087] Furthermore, any combination can be made among the various different embodiments of the present disclosure, as long as it does not violate the idea of the present disclosure, and it should also be regarded as the content disclosed by the present disclosure.

[0088] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention 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 recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered by the scope of the claims and the specification of the present invention.

Claims

1. A method for automobile finance customer consultation based on natural language processing, characterized in that: include: Obtaining an automobile finance database, and obtaining a plurality of consultation information sets and keywords corresponding to the consultation information sets according to the automobile finance database, and obtaining first consultation text information and second consultation text information successively sent by the automobile finance customer; Obtaining second character set data according to the second consulting text information, and obtaining the second number of occurrences of the keyword corresponding to each consulting information set in the second character set data, obtaining the second number of occurrences exceeding a preset number threshold and recording it as the second effective number of occurrences, and obtaining the first weight coefficient corresponding to each consulting information set based on the weight model and the second effective number of occurrences of the keyword corresponding to each consulting information set; Acquire first character set data according to first consultation text information, and acquire first occurrence times of keywords corresponding to each consultation information set in the first character set data, acquire first occurrence times exceeding a preset number threshold and record them as first valid occurrence times, and record keywords with first valid occurrence times and second valid occurrence times as characters to be analyzed, acquire word frequency difference values ​​according to the first valid occurrence times and the second valid occurrence times of the characters to be analyzed, acquire correction parameters according to the word frequency difference values, and acquire second weight coefficients according to the correction parameters and the first weight coefficients; A consulting strategy is obtained according to each consulting information set and its corresponding second weight coefficient.

2. The automobile finance customer consultation method based on natural language processing according to claim 1, characterized in that: The obtaining of the consulting strategy according to each consulting information set and its corresponding second weight coefficient comprises: Arrange each second weight coefficient in order of magnitude and construct a consultation content framework; A plurality of response data spaces corresponding to the second weight coefficient are formed in the consultation content framework, wherein the proportion of each response data space is equal to the proportion of the corresponding second weight coefficient; The data in the consultation information set corresponding to each second weight coefficient is added to the response data space corresponding to each second weight coefficient.

3. The automobile finance customer consultation method based on natural language processing according to claim 1, characterized in that: The weight model in obtaining the first weight coefficient corresponding to each consulting information set based on the weight model and the second effective number of occurrences of the keyword corresponding to each consulting information set is expressed as: in, A i1 is the first weight coefficient corresponding to the i-th consulting information set, T i2 is the second valid occurrence number corresponding to the i-th consultation information set, T j2 is the second valid occurrence number corresponding to the j-th consulting information set, and n is the number of consulting information sets with the second valid occurrence number.

4. The automobile finance customer consultation method based on natural language processing according to claim 1, characterized in that: The step of obtaining the word frequency difference value according to the first effective number of occurrences and the second effective number of occurrences of the word to be analyzed comprises: Obtaining the time interval between the sending of the first consultation text message and the sending of the second consultation text message; Match the time coefficient according to the time interval of issuance; The word frequency difference is obtained according to the time coefficient, the first effective number of occurrences and the second effective number of occurrences of the word to be analyzed.

5. The automobile finance customer consultation method based on natural language processing according to claim 4 is characterized in that: The word frequency difference value obtained according to the time coefficient, the first effective number of occurrences and the second effective number of occurrences of the word to be analyzed is expressed as: D k =(T k2 -T k1 ) α ;in, D k is the frequency difference of the kth word to be analyzed, T k1 is the first valid occurrence number of the kth word to be analyzed, T k2 is the second effective number of occurrences of the kth word to be analyzed, and α is the time coefficient.

6. The automobile finance customer consultation method based on natural language processing according to claim 1, characterized in that: The correction parameter obtained according to the word frequency difference is expressed as: in, P kd is the correction parameter of the kth word to be analyzed, D k is the frequency difference of the kth word to be analyzed.

7. The automobile finance customer consultation method based on natural language processing according to claim 1, characterized in that: The second weight coefficient is obtained according to the correction parameter and the first weight coefficient as follows: A k2 =P kd ·A k1 ; in, A k2 is the second weight coefficient corresponding to the consultation information set corresponding to the kth word to be analyzed, P kd is the correction parameter of the kth word to be analyzed, A k1 is the first weight coefficient corresponding to the consultation information set corresponding to the kth word to be analyzed.

8. An automobile finance customer consultation system based on natural language processing, characterized in that: The system is used to implement the automobile finance customer consultation method based on natural language processing as described in any one of claims 1 to 7, and the system includes: An acquisition module is used to acquire an automobile finance database, and acquire multiple consultation information sets and keywords corresponding to the consultation information sets according to the automobile finance database, and acquire first consultation text information and second consultation text information successively sent by automobile finance customers; A first language processing module is used to obtain second character set data according to the second consulting text information, and obtain a second number of occurrences of a keyword corresponding to each consulting information set in the second character set data, obtain a second number of occurrences exceeding a preset number threshold and record it as a second valid number of occurrences, and obtain a first weight coefficient corresponding to each consulting information set based on a weight model and the second valid number of occurrences of the keyword corresponding to each consulting information set; A second language processing module is used to obtain first character set data according to the first consultation text information, and obtain the first occurrence number of the keyword corresponding to each consultation information set in the first character set data, obtain the first occurrence number exceeding a preset number threshold and record it as a first effective occurrence number, and record the keyword with the first effective occurrence number and the second effective occurrence number as a word to be analyzed, obtain a word frequency difference value according to the first effective occurrence number and the second effective occurrence number of the word to be analyzed, obtain a correction parameter according to the word frequency difference value, and obtain a second weight coefficient according to the correction parameter and the first weight coefficient; The consultation generating module is used to obtain the consultation strategy according to each consultation information set and its corresponding second weight coefficient.

9. The automobile finance customer consultation method based on natural language processing according to claim 8, characterized in that: The consultation generation module is also used for: Arrange each second weight coefficient in order of magnitude and construct a consultation content framework; A plurality of response data spaces corresponding to the second weight coefficient are formed in the consultation content framework, wherein the proportion of each response data space is equal to the proportion of the corresponding second weight coefficient; The data in the consultation information set corresponding to each second weight coefficient is added to the response data space corresponding to each second weight coefficient.

10. The automobile finance customer consultation method based on natural language processing according to claim 8, characterized in that: The second language processing module is also used for: Obtaining the time interval between the sending of the first consultation text message and the sending of the second consultation text message; Match the time coefficient according to the time interval of issuance; The word frequency difference is obtained according to the time coefficient, the first effective number of occurrences and the second effective number of occurrences of the word to be analyzed.

Images