A book information identification direct loan management system and method based on LLM

Abstract

The present invention discloses a book information identification direct borrowing management system and method based on LLM, and relates to the technical field of large language model information management. The present invention analyzes historical borrowing orders and historical book search and retrieval records generated by users in a smart library, identifies and analyzes book information that is difficult for users to search and has low retrieval efficiency, judges and identifies book information with high overlap between feature information libraries that need to be traversed when performing corresponding matching retrieval, determines warning book information, and then selects an interference book information range for the warning book information. By prompting to adjust the retrieval weight of feature data corresponding to the warning book information when performing matching analysis with the warning book information, the accuracy of the search result output by the smart library according to the description information of the book input by the user is improved, thereby driving the improvement of the efficiency of users in finding book information in the smart library and the book borrowing rate.

Description

Technical Field

[0001] The present invention relates to the technical field of large language model information management, and in particular to a book information recognition direct loan management system and method based on LLM. Background Art

[0002] LLM (Large Language Model) refers to a deep learning model trained with a large amount of text data, which can generate natural language text or understand the meaning of language text. Large language models can handle a variety of natural language tasks, such as text classification, question answering, and dialogue, and are an important path to artificial intelligence.

[0003] Large language models have had a significant impact on traditional information retrieval technology and application paradigms, and the two are closely complementary in terms of technical paths. Among them, large language models have powerful language understanding, reasoning and generation capabilities, which can help build a more intelligent information retrieval system, and the information retrieval system can efficiently obtain the required relevant information from the outside world, providing more accurate and reliable contextual information for large language models.

[0004] When using a large language model to search for book information, the description needs to be as accurate as possible, because large language models often face challenges in computing resource requirements and generalization capabilities. In practice, the scope of book information that needs to be searched is too large. Even if its powerful language understanding and generation capabilities are fully utilized, the final search results may still be inaccurate. Summary of the invention

[0005] The purpose of the present invention is to provide a book information identification direct loan management system and method based on LLM to solve the problems raised in the prior art.

[0006] To achieve the above object, the present invention provides the following technical solution: a book information identification direct loan management method based on LLM, the method comprising:

[0007] Step S1: extract historical borrowing orders and historical book search and retrieval records from the borrowing management terminal of the smart library, identify the characteristic order sequence corresponding to each borrowing user, cluster and aggregate the historical book search and retrieval records that meet the temporal logical order relationship with each characteristic order sequence, and obtain a search and retrieval record set corresponding to each characteristic order sequence;

[0008] Step S2: identifying the target search node by combing the identical corresponding distribution of the book information presented in the feature order sequence and the corresponding search and retrieval record set, and extracting the associated search record group according to the target search node distribution presented in the corresponding search and retrieval record set;

[0009] Step S3: identifying and extracting the retrieval association structure and evaluating the characteristic attribute value according to the associated retrieval record group, and evaluating and calculating the associated retrieval characteristic value of the retrieval association structure in combination with the extraction frequency distribution of the retrieval association structure;

[0010] Step S4: The search association structure whose corresponding associated search feature value is greater than the threshold is screened and set as the target structure to identify the warning book information, and at the same time, the range of the interfering book information is extracted for the warning book information. In the feature database that needs to be traversed when searching for matching warning book information, the feature data that needs to adjust the search weight when performing matching analysis with the warning book information is pushed to the management personnel for warning.

[0011] Further, step S1 includes:

[0012] Step S1-1: extract all historical borrowing orders of the corresponding borrowing user respectively, and arrange them in the order of order generation time, and successively extract several order sequences whose arrangement intervals are less than the time difference threshold, and set them as the characteristic order sequences of the corresponding borrowing user; extract the order generation time range [a, b] covered by each characteristic order sequence, where a and b are the order generation time of the first and last historical borrowing order in the corresponding characteristic order sequence respectively;

[0013] Step S1-2: When it is identified that the search user corresponding to a certain historical book search and retrieval record is the same user as a certain borrowing user, from each characteristic order sequence corresponding to the certain borrowing user, a characteristic order sequence P whose corresponding order generation time range is closest to the record generation time of the certain historical book search and retrieval record is captured, and it is determined that the certain historical book search and retrieval record and the characteristic order sequence P satisfy a temporal logical sequence relationship;

[0014] Step S1-3: Gather all historical book search and retrieval records that satisfy the time logic sequence relationship with each characteristic order sequence in chronological order to generate a search and retrieval record set corresponding to each characteristic order sequence;

[0015] Each characteristic order sequence outputted above is a borrowing order generated by the same borrowing user within a certain time fluctuation range, and the search and retrieval record set corresponding to each characteristic order sequence refers to the book information retrieval record generated by the above-mentioned borrowing user at the corresponding book retrieval device of the smart library before generating the above-mentioned borrowing order. The subsequent application is mainly to monitor whether the corresponding book retrieval device of the smart library provides good retrieval services for readers and whether it improves the readers' book search and borrowing efficiency.

[0016] Further, step S2 comprises:

[0017] Step S2-1: obtaining target borrowed book information in each historical borrowing order, and extracting feature description information of the relevant books input by the corresponding search user to the borrowing management terminal in each historical book search and retrieval record, as well as book information output by the borrowing management terminal after matching and searching through a large language model;

[0018] Step S2-2: Let the search and retrieval record set corresponding to a certain characteristic order sequence Y be X. When the book information corresponding to the i-th historical borrowing order in the certain characteristic order sequence Y is captured to be the same as the book information corresponding to the j-th historical book search and retrieval record in the search and retrieval record set X, a characteristic mark is made for the j-th historical book search and retrieval record;

[0019] The above-mentioned historical book search and retrieval records marked with features refer to the actual phenomenon that after the corresponding readers searched for book information based on the book retrieval device, they also found the target book in the smart library;

[0020] Step S2-3: traverse the distribution of characteristic marks presented in each search and retrieval record set in chronological order of the corresponding historical book search and retrieval records; set each historical book search and retrieval record containing the characteristic mark in each search and retrieval record set as the target search node;

[0021] Step S2-4: Assume that in a certain search record set D, the i-th historical book search record d i is the kth target search node in a search record set D, that is, the kth historical book search record with characteristic tags distributed in a related search record group D; when k=1, the related search record group {d1,...,d i}; When k≧2, if the k-1th target search node is the jth historical book search record d j , that is, the k-1th historical book search and retrieval record with characteristic tags distributed in a search and retrieval record set D, then the associated search record group {d j+1 ,...,d i}, where d j+1 Find the retrieval record for the j+1th historical book; wherein, based on a target retrieval node, a corresponding associated retrieval record group is extracted.

[0022] Further, step S3 includes:

[0023] Step S3-1: Gathering the associated search record groups extracted based on all search and retrieval record sets; extracting, for each associated search record group, the corresponding book information output in the corresponding historical book search and retrieval records contained therein and the feature description information of the relevant books input by the corresponding search user to the borrowing management terminal;

[0024] Step S3-2: Set the book information extracted and output from the last historical book search and retrieval record in each associated search record group as the target information; establish retrieval association relationships between other extracted and output book information in each associated search record group and the corresponding target information one by one, and construct and generate a number of retrieval association structures; at the same time, obtain the feature description information corresponding to other extracted and output book information from the corresponding historical book search and retrieval records, and the similarity between the feature description information corresponding to the extracted and output target information, and set the similarity as the feature attribute value of the corresponding constructed retrieval association structure;

[0025] In each associated search record group, the corresponding target information is the book information that the reader is actually looking for. This feature is assisted in the proof by the borrowing order finally generated in this application, because the book information that is presented continuously in time and extracted and output before the target information is often the search result output and feedback by the book search device based on the inaccurate description information input by the reader. This data phenomenon also proves that there is a certain similarity between the target information and the above book information.

[0026] Step S3-3: classify the two corresponding retrieval association structures containing exactly the same book information into one type of retrieval association structure, obtain the average characteristic attribute value δ of each type of retrieval association structure, and calculate the associated retrieval characteristic value β=[sum(Q) / sum(M)]×δ for each type of retrieval association structure, wherein sum(Q) represents the frequency of constructing each type of retrieval association structure in the associated retrieval record group extracted from all search retrieval record sets, and sum(M) represents the total number of retrieval association structures constructed in the associated retrieval record group extracted from all search retrieval record sets;

[0027] The larger the average feature attribute value is, the higher the similarity between the two book information contained in a certain type of search association structure is. The similarity proposed here is more inclined to mean that the higher the information overlap rate required to traverse the feature information library when performing a matching search between two book information based on the description information of the book input by the user; the larger the sum(Q) / sum(M) value is, the greater the probability that another book information will be pushed out when the user targets any book information contained in a certain type of search association structure, and this is often caused by the small number of distinguishing features between the two book information.

[0028] Further, step S4 includes:

[0029] Step S4-1: Capture the book information whose frequency of appearance in all target structures is greater than the frequency threshold, and set it as the warning book information, and set other book information that constitutes the target structure with the warning book information as the interference book information of the warning book information;

[0030] Warning book information has two data characteristics: first, there are a large number of other book information with high similarity to it that needs to be traversed in the feature information database when matching and searching; second, based on the distribution of historical book search and retrieval records, when users are looking for the warning book information, in most cases, they often need to adjust the description information a few or many times;

[0031] Step S4-2: extract from the borrowing management terminal of the smart library the feature database that the corresponding retrieval device needs to traverse when performing matching analysis between the feature description information of the book input by the user and the warning book information and the book information interfering with the warning book information, and mark the feature data in the feature database corresponding to the warning book information that is different from the feature data contained in the feature database corresponding to the corresponding interfering book information; prompt the management personnel to adjust the retrieval weight when performing matching analysis with the warning book information for the feature data marked with the feature data of the warning book information.

[0032] In order to better implement the above method, a book information identification direct loan management system is also proposed, which includes: a book information retrieval data management module, a related retrieval record group identification and extraction module, a related retrieval feature value calculation module, and an early warning prompt management module;

[0033] The book information retrieval data management module is used to extract historical borrowing orders and historical book search and retrieval records from the borrowing management terminal of the smart library, identify the characteristic order sequence corresponding to each borrowing user, cluster and aggregate the historical book search and retrieval records that meet the temporal logical order relationship with each characteristic order sequence, and obtain the search and retrieval record set corresponding to each characteristic order sequence;

[0034] The associated search record group identification and extraction module is used to identify the target search node by combing the corresponding distribution of the same search book information presented in the feature order sequence and the corresponding search and retrieval record set, and extract the associated search record group according to the distribution of the target search node presented in the corresponding search and retrieval record set;

[0035] The associated search feature value calculation module is used to sort out the distribution of the book information corresponding to the search output of the borrowing management terminal in each associated search record group, perform feature recognition and feature attribute value evaluation on the search association structure, and evaluate and calculate the associated search feature value of the search association structure in combination with the extraction frequency distribution of the search association structure;

[0036] The early warning prompt management module is used to screen and set as the target structure all the retrieval association structures whose corresponding associated retrieval feature values ​​are greater than the threshold, extract the early warning book information from it, and at the same time extract the range of the interfering book information from the early warning book information. In the feature database that needs to be traversed when searching for matching early warning book information, the feature data that needs to adjust the retrieval weight when performing matching analysis with the early warning book information is pushed to the management personnel.

[0037] Further, the associated search record group identification and extraction module includes a target search node setting unit and an associated search record group extraction unit;

[0038] A target search node setting unit is used to identify a target search node by combing the corresponding distribution of the same search book information presented in the feature order sequence and the corresponding search record set;

[0039] The associated search record group extraction unit is used to extract the associated search record group according to the target search node distribution presented in the corresponding search record set.

[0040] Further, the associated retrieval feature value calculation module includes a retrieval associated structure information management unit and an associated retrieval feature value evaluation unit;

[0041] The search association structure information management unit is used to sort out the distribution of the book information corresponding to the search output of the borrowing management terminal in each associated search record group, and perform feature recognition and feature attribute value evaluation on the search association structure;

[0042] The associated retrieval feature value evaluation unit is used to evaluate and calculate the associated retrieval feature value of the retrieval associated structure in combination with the extraction frequency distribution of the retrieval associated structure.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention analyzes the historical borrowing orders and historical book search and retrieval records generated by users in the smart library, identifies and analyzes the book information that is difficult for users to search and has low retrieval efficiency, judges and identifies the book information that has a high overlap between the feature information libraries that need to be traversed when performing corresponding matching retrieval, determines the warning book information, and then selects the interference book information range for the warning book information, and by prompting to adjust the retrieval weight of the feature data corresponding to the warning book information when performing matching analysis with the warning book information, the accuracy of the search results output by the smart library according to the description information of the book input by the user is improved, thereby driving the improvement of the efficiency of users in finding book information in the smart library and the book borrowing rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of a method for direct borrowing management of book information identification based on LLM of the present invention;

[0045] Figure 2 The present invention is a structural schematic diagram of a book information identification direct loan management system based on LLM. DETAILED DESCRIPTION

[0046] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without making any creative work shall fall within the scope of protection of the present invention.

[0047] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution, a book information identification direct loan management method based on LLM, the method comprising:

[0048] Step S1: extract historical borrowing orders and historical book search and retrieval records from the borrowing management terminal of the smart library, identify the characteristic order sequence corresponding to each borrowing user, cluster and aggregate the historical book search and retrieval records that meet the temporal logical order relationship with each characteristic order sequence, and obtain a search and retrieval record set corresponding to each characteristic order sequence;

[0049] Wherein, step S1 comprises:

[0050] Step S1-1: extract all historical borrowing orders of the corresponding borrowing user respectively, and arrange them in the order of order generation time, and successively extract several order sequences whose arrangement intervals are less than the time difference threshold, and set them as the characteristic order sequences of the corresponding borrowing user; extract the order generation time range [a, b] covered by each characteristic order sequence, where a and b are the order generation time of the first and last historical borrowing order in the corresponding characteristic order sequence respectively;

[0051] Step S1-2: When it is identified that the search user corresponding to a certain historical book search and retrieval record is the same user as a certain borrowing user, from each characteristic order sequence corresponding to the certain borrowing user, a characteristic order sequence P whose corresponding order generation time range is closest to the record generation time of the certain historical book search and retrieval record is captured, and it is determined that the certain historical book search and retrieval record and the characteristic order sequence P satisfy a temporal logical sequence relationship;

[0052] Among them, in the process of judging which order generation time range the record generation time of a historical book search and retrieval record is closest to, the record generation time of the historical book search and retrieval record can be compared with the minimum time and the maximum time in each order generation time range, and the time difference can be calculated, and the average time difference can be used as a measurement basis. For example, the record generation time of the historical book search and retrieval record is 16:08, and the order generation time range is [18:12, 18:14]. Two time differences are calculated: the time difference between 18:12 and 16:08 is 2 hours and 4 minutes, and the time difference between 18:14 and 16:08 is 2 hours and 6 minutes. The average time difference is 2 hours and 5 minutes, and 2 hours and 5 minutes is the shortest time difference. Therefore, it is judged that the order generation time range [18:12, 18:14] is the order generation time range closest to the record generation time of the historical book search and retrieval record.

[0053] Step S1-3: All historical book search and retrieval records that satisfy the temporal logical order relationship with each characteristic order sequence are collected in chronological order to generate a search and retrieval record set corresponding to each characteristic order sequence.

[0054] Step S2: identifying the target search node by combing the identical corresponding distribution of the book information presented in the feature order sequence and the corresponding search and retrieval record set, and extracting the associated search record group according to the target search node distribution presented in the corresponding search and retrieval record set;

[0055] Wherein, step S2 comprises:

[0056] Step S2-1: obtaining target borrowed book information in each historical borrowing order, and extracting feature description information of the relevant books input by the corresponding search user to the borrowing management terminal in each historical book search and retrieval record, as well as book information output by the borrowing management terminal after matching and searching through a large language model;

[0057] Step S2-2: Let the search and retrieval record set corresponding to a certain characteristic order sequence Y be X. When the book information corresponding to the i-th historical borrowing order in the certain characteristic order sequence Y is captured to be the same as the book information corresponding to the j-th historical book search and retrieval record in the search and retrieval record set X, a characteristic mark is made for the j-th historical book search and retrieval record;

[0058] Step S2-3: traverse the distribution of characteristic marks presented in each search and retrieval record set in chronological order of the corresponding historical book search and retrieval records; set each historical book search and retrieval record containing the characteristic mark in each search and retrieval record set as the target search node;

[0059] Step S2-4: Assume that in a certain search record set D, the i-th historical book search record d i is the kth target search node in a search record set D; when k=1, the associated search record group {d1,...,d i}; When k≧2, if the k-1th target search node is the jth historical book search record d j , then based on the kth target retrieval node, the associated retrieval record group {d j+1 ,...,d i}, where d j+1 Find the retrieval record for the j+1th historical book; wherein, based on a target retrieval node, a corresponding associated retrieval record group is extracted.

[0060] Step S3: identifying and extracting the retrieval association structure and evaluating the characteristic attribute value according to the associated retrieval record group, and evaluating and calculating the associated retrieval characteristic value of the retrieval association structure in combination with the extraction frequency distribution of the retrieval association structure;

[0061] Wherein, step S3 comprises:

[0062] Step S3-1: Gathering the associated search record groups extracted based on all search and retrieval record sets; extracting, for each associated search record group, the corresponding book information output in the corresponding historical book search and retrieval records contained therein and the feature description information of the relevant books input by the corresponding search user to the borrowing management terminal;

[0063] Step S3-2: Set the book information extracted and output from the last historical book search and retrieval record in each associated search record group as the target information; establish retrieval association relationships between other extracted and output book information in each associated search record group and the corresponding target information one by one, and construct and generate a number of retrieval association structures; at the same time, obtain the feature description information corresponding to other extracted and output book information from the corresponding historical book search and retrieval records, and the similarity between the feature description information corresponding to the extracted and output target information, and set the similarity as the feature attribute value of the corresponding constructed retrieval association structure;

[0064] Step S3-3: classify the two corresponding retrieval association structures containing exactly the same book information into one type of retrieval association structure, obtain the average characteristic attribute value δ of each type of retrieval association structure, and calculate the associated retrieval characteristic value β=[sum(Q) / sum(M)]×δ for each type of retrieval association structure, wherein sum(Q) represents the frequency of constructing each type of retrieval association structure in the associated retrieval record group extracted from all search retrieval record sets, and sum(M) represents the total number of retrieval association structures constructed in the associated retrieval record group extracted from all search retrieval record sets;

[0065] Step S4: The search association structure with the corresponding associated search feature value greater than the threshold is selected and set as the target structure, the warning book information is identified, and the range of the interfering book information is extracted for the warning book information. In the feature database that needs to be traversed when searching for matching warning book information, the feature data that needs to adjust the search weight when performing matching analysis with the warning book information is pushed to the management personnel for warning;

[0066] Wherein, step S4 comprises:

[0067] Step S4-1: Capture the book information whose frequency of appearance in all target structures is greater than the frequency threshold, and set it as the warning book information, and set other book information that constitutes the target structure with the warning book information as the interference book information of the warning book information;

[0068] Step S4-2: extract from the borrowing management terminal of the smart library the feature database that the corresponding retrieval device needs to traverse when performing matching analysis between the feature description information of the book input by the user and the warning book information and the book information interfering with the warning book information, and mark the feature data in the feature database corresponding to the warning book information that is different from the feature data contained in the feature database corresponding to the corresponding interfering book information; prompt the management personnel to adjust the retrieval weight when performing matching analysis with the warning book information for the feature data marked with the feature data of the warning book information.

[0069] In order to better implement the above method, a book information identification direct loan management system is also proposed, which includes: a book information retrieval data management module, a related retrieval record group identification and extraction module, a related retrieval feature value calculation module, and an early warning prompt management module;

[0070] The book information retrieval data management module is used to extract historical borrowing orders and historical book search and retrieval records from the borrowing management terminal of the smart library, identify the characteristic order sequence corresponding to each borrowing user, cluster and aggregate the historical book search and retrieval records that meet the temporal logical order relationship with each characteristic order sequence, and obtain the search and retrieval record set corresponding to each characteristic order sequence;

[0071] The associated search record group identification and extraction module is used to identify the target search node by combing the corresponding distribution of the same search book information presented in the feature order sequence and the corresponding search and retrieval record set, and extract the associated search record group according to the distribution of the target search node presented in the corresponding search and retrieval record set;

[0072] The associated search record group identification and extraction module includes a target search node setting unit and an associated search record group extraction unit;

[0073] A target search node setting unit is used to identify a target search node by combing the corresponding distribution of the same search book information presented in the feature order sequence and the corresponding search record set;

[0074] The associated search record group extraction unit is used to extract the associated search record group according to the target search node distribution presented in the corresponding search record set;

[0075] The associated search feature value calculation module is used to sort out the distribution of the book information corresponding to the search output of the borrowing management terminal in each associated search record group, perform feature recognition and feature attribute value evaluation on the search association structure, and evaluate and calculate the associated search feature value of the search association structure in combination with the extraction frequency distribution of the search association structure;

[0076] Among them, the associated retrieval feature value calculation module includes a retrieval associated structure information management unit and an associated retrieval feature value evaluation unit;

[0077] The search association structure information management unit is used to sort out the distribution of the book information corresponding to the search output of the borrowing management terminal in each associated search record group, and perform feature recognition and feature attribute value evaluation on the search association structure;

[0078] The associated retrieval characteristic value evaluation unit is used to evaluate and calculate the associated retrieval characteristic value of the retrieval associated structure in combination with the extraction frequency distribution of the retrieval associated structure;

[0079] The early warning prompt management module is used to screen and set as the target structure all the retrieval association structures whose corresponding associated retrieval feature values ​​are greater than the threshold, extract the early warning book information from it, and at the same time extract the range of the interfering book information from the early warning book information. In the feature database that needs to be traversed when searching for matching early warning book information, the feature data that needs to adjust the retrieval weight when performing matching analysis with the early warning book information is pushed to the management personnel.

[0080] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for managing direct borrowing of books by searching book information and identifying books based on LLM, characterized in that: The method comprises: Step S1: extract historical borrowing orders and historical book search and retrieval records from the borrowing management terminal of the smart library, identify the characteristic order sequence corresponding to each borrowing user, cluster and aggregate the historical book search and retrieval records that meet the temporal logical order relationship with each characteristic order sequence, and obtain a search and retrieval record set corresponding to each characteristic order sequence; Step S2: identifying the target search node by combing the identical corresponding distribution of the book information presented in the feature order sequence and the corresponding search and retrieval record set, and extracting the associated search record group according to the target search node distribution presented in the corresponding search and retrieval record set; Step S3: identifying and extracting the retrieval association structure and evaluating the characteristic attribute value according to the associated retrieval record group, and evaluating and calculating the associated retrieval characteristic value of the retrieval association structure in combination with the extraction frequency distribution of the retrieval association structure; Step S4: The search association structure with the corresponding associated search feature value greater than the threshold is selected and set as the target structure, the warning book information is identified, and the range of the interfering book information is extracted for the warning book information. In the feature database that needs to be traversed when searching for matching warning book information, the feature data that needs to adjust the search weight when performing matching analysis with the warning book information is pushed to the management personnel for warning; The step S2 comprises: Step S2-1: obtaining target borrowed book information in each historical borrowing order, and extracting feature description information of the relevant books input by the corresponding search user to the borrowing management terminal in each historical book search and retrieval record, as well as book information output by the borrowing management terminal after matching and searching through a large language model; Step S2-2: Let the search and retrieval record set corresponding to a certain characteristic order sequence Y be X. When the book information corresponding to the i-th historical borrowing order in the certain characteristic order sequence Y is captured to be the same as the book information corresponding to the j-th historical book search and retrieval record in the search and retrieval record set X, a characteristic mark is made for the j-th historical book search and retrieval record; Step S2-3: traverse the distribution of characteristic marks presented in each search and retrieval record set in chronological order of the corresponding historical book search and retrieval records; set each historical book search and retrieval record containing the characteristic mark in each search and retrieval record set as the target search node; Step S2-4: Assume that in a certain search record set D, the i-th historical book search record d i is the kth target search node in a search record set D; when k=1, the associated search record group {d1,...,d i }; When k≧2, if the k-1th target search node is the jth historical book search record d j , then based on the kth target retrieval node, the associated retrieval record group {d j+1 ,...,d i }, where d j+1 Find the retrieval record for the j+1th historical book; wherein, based on a target retrieval node, a corresponding associated retrieval record group is extracted.

2. According to claim 1, a method for managing direct borrowing based on LLM for searching book information is characterized by: The step S1 comprises: Step S1-1: extract all historical borrowing orders of the corresponding borrowing user respectively, and arrange them in the order of order generation time, and successively extract several order sequences whose arrangement intervals are less than the time difference threshold, and set them as the characteristic order sequences of the corresponding borrowing user; extract the order generation time range [a, b] covered by each characteristic order sequence, where a and b are the order generation time of the first and last historical borrowing order in the corresponding characteristic order sequence respectively; Step S1-2: When it is identified that the search user corresponding to a certain historical book search and retrieval record is the same user as a certain borrowing user, a characteristic order sequence P whose corresponding order generation time range is closest to the record generation time of the certain historical book search and retrieval record is captured from each characteristic order sequence corresponding to the certain borrowing user, and it is determined that the certain historical book search and retrieval record and the characteristic order sequence P satisfy a temporal logical sequence relationship; Step S1-3: All historical book search and retrieval records that satisfy the temporal logical order relationship with each characteristic order sequence are collected in chronological order to generate a search and retrieval record set corresponding to each characteristic order sequence.

3. The method for managing direct borrowing based on LLM retrieval and identification of book information according to claim 2 is characterized by: The step S3 comprises: Step S3-1: Gathering the associated search record groups extracted based on all the search and retrieval record sets; extracting, for each associated search record group, the corresponding book information output in the corresponding historical book search and retrieval records contained therein and the feature description information of the relevant books input by the corresponding search user to the borrowing management terminal; Step S3-2: Set the book information extracted and output from the last historical book search and retrieval record in each associated search record group as the target information; establish retrieval association relationships between other extracted and output book information in each associated search record group and the corresponding target information one by one, and construct and generate a number of retrieval association structures; at the same time, obtain the feature description information corresponding to the other extracted and output book information from the corresponding historical book search and retrieval record, and the similarity between the feature description information corresponding to the extracted and output target information, and set the similarity as the feature attribute value of the corresponding constructed retrieval association structure; Step S3-3: Classify the two corresponding retrieval association structures that contain exactly the same book information into one type of retrieval association structure, obtain the average characteristic attribute value δ of each type of retrieval association structure, and calculate the associated retrieval characteristic value β=[sum(Q) / sum(M)]×δ for each type of retrieval association structure, wherein sum(Q) represents the frequency of constructing each type of retrieval association structure in the associated retrieval record group extracted from all search retrieval record sets, and sum(M) represents the total number of retrieval association structures constructed in the associated retrieval record group extracted from all search retrieval record sets.

4. The method for managing direct borrowing based on LLM retrieval and identification of book information according to claim 3 is characterized by: The step S4 comprises: Step S4-1: Capture book information whose frequency of appearance in all target structures is greater than a frequency threshold, and set it as warning book information, and set other book information that constitutes the target structure with the warning book information as interference book information of the warning book information; Step S4-2: extract from the borrowing management terminal of the smart library the feature database that the corresponding retrieval device needs to traverse when performing matching analysis between the feature description information of the book input by the user and the warning book information and the interference book information of the warning book information, and mark the feature data in the feature database corresponding to the warning book information as different from the feature data contained in the feature database corresponding to the corresponding interference book information; prompt the management personnel to adjust the retrieval weight when performing matching analysis with the warning book information for the feature data marked with the feature data of the warning book information.

5. A book information retrieval identification direct loan management system, used to implement a book information retrieval identification direct loan management method based on LLM as claimed in any one of claims 1 to 4, characterized in that: The system includes: a book information retrieval data management module, a related retrieval record group identification and extraction module, a related retrieval feature value calculation module, and an early warning prompt management module; The book information retrieval data management module is used to extract historical borrowing orders and historical book search and retrieval records from the borrowing management terminal of the smart library, identify the characteristic order sequence corresponding to each borrowing user, cluster and aggregate the historical book search and retrieval records that meet the temporal logical order relationship with each characteristic order sequence, and obtain a search and retrieval record set corresponding to each characteristic order sequence; The associated search record group identification and extraction module is used to identify the target search node by combing the corresponding distribution of the same search book information presented in the feature order sequence and the corresponding search and retrieval record set, and extract the associated search record group according to the distribution of the target search node presented in the corresponding search and retrieval record set; The associated search feature value calculation module is used to sort out the distribution of the book information corresponding to the search output of the lending management terminal in each associated search record group, perform feature recognition and feature attribute value evaluation on the search association structure, and evaluate and calculate the associated search feature value of the search association structure in combination with the extraction frequency distribution of the search association structure; The early warning prompt management module is used to screen all retrieval association structures with corresponding associated retrieval feature values ​​greater than a threshold value and set them as target structures, extract the early warning book information therefrom, and simultaneously perform range extraction of the interfering book information on the early warning book information, and push the feature data that needs to adjust the retrieval weight when performing matching analysis with the early warning book information to the management personnel in the feature database that needs to be traversed when searching for matching early warning book information.

6. A book information retrieval and identification direct loan management system according to claim 5, characterized in that: The associated search record group identification and extraction module includes a target search node setting unit and an associated search record group extraction unit; The target search node setting unit is used to identify the target search node by combing the corresponding distribution of the same search book information presented in the feature order sequence and the corresponding search record set; The associated search record group extraction unit is used to extract the associated search record group according to the target search node distribution presented in the corresponding search record set.

7. A book information retrieval and identification direct loan management system according to claim 5, characterized in that: The associated retrieval feature value calculation module includes a retrieval associated structure information management unit and an associated retrieval feature value evaluation unit; The retrieval association structure information management unit is used to sort out the distribution of the book information corresponding to the retrieval output of the lending management terminal in each associated retrieval record group, and perform feature recognition and feature attribute value evaluation on the retrieval association structure; The associated retrieval feature value evaluation unit is used to evaluate and calculate the associated retrieval feature value of the retrieval associated structure in combination with the extraction frequency distribution of the retrieval associated structure.

Images