Knowledge Question Answering System Based on Airport Safety Management

Abstract

The invention discloses a knowledge question-and-answer system based on airport safety management, relates to the technical field of knowledge question-and-answer, and solves the technical problems that, when analyzing the relevance of airport clauses, the potential influence of the relevance of problems solved by clauses on the relevance of clauses is ignored, and it is difficult to reasonably combine the relevance of problems with the relevance of clauses; the invention obtains key information of each clause in airport regulations, performs reverse reasoning on the key information to obtain relevant problems of each clause, performs similarity fitting based on the relevant problems to obtain similarity factors between each relevant problem; performs analysis based on the similarity factors to obtain similarity values ​​between clauses, judges each clause based on the similarity values ​​to obtain similar situations of clauses, and bundles the clauses based on the similar situations and stores them in a database; the invention combines questions and clauses to analyze the relevance, so that the system can better adapt to question-and-answer situations with many or complex clauses, thereby improving the efficiency and accuracy of system analysis.

Description

Technical Field

[0001] The invention belongs to the field of airport safety management, relates to a knowledge question answering technology, and specifically is a knowledge question answering system based on airport safety management. Background Art

[0002] With the rapid development of the aviation industry, airports, as the core nodes of air transportation, have become increasingly complex in terms of safety management. Airport safety management not only involves the safety of the flight area, but also includes terminal buildings, baggage security checks, airport transportation and other aspects. Airport safety has become a matter of great public concern. Traditional airport safety management Q&A often relies on manual answers, which is inefficient and prone to errors. In this context, an efficient and convenient Q&A method is needed to answer various safety and management issues encountered in daily operations. The knowledge Q&A system for airport safety management, as an auxiliary tool, can provide users with the information they need according to the corresponding rules and regulations.

[0003] At present, most knowledge question-and-answer systems based on airport safety management, when analyzing the relevance of airport clauses, ignore the potential impact of the relevance of the problems solved by the clauses on the relevance of the clauses, and often classify and associate the clauses only based on the similarity between the clauses. This may result in the failure to provide users with a more comprehensive answer when dissimilar clauses can solve the same problem, which may result in the answer not meeting user expectations and reducing user expectations. At the same time, most knowledge question-and-answer systems based on airport safety management find it difficult to reasonably combine the relevance of questions with the relevance of clauses, which may cause the problem of mismatch between the analyzed relevance and reality, resulting in inaccurate answers from the question-and-answer system.

[0004] Therefore, the present invention discloses a knowledge question and answer system based on airport safety management to solve the above technical problems. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a knowledge question and answer system based on airport safety management, which is used to solve the technical problems that when analyzing the relevance of airport clauses, the potential impact of the relevance of problems solved by clauses on the relevance of clauses is ignored, and it is difficult to reasonably combine the relevance of problems with the relevance of clauses. The present invention obtains the key information of each clause in the airport rules and regulations, reversely infers the key information to obtain the relevant problems of each clause, and performs similarity fitting based on the relevant problems to obtain the similarity factors between the relevant problems; analyzes based on the similarity factors to obtain the similarity values ​​between the clauses, judges each clause based on the similarity values ​​to obtain the similarity of the clauses, and bundles the clauses based on the similarity of the clauses and stores them in a database to solve the above problems.

[0006] To achieve the above-mentioned object, the first aspect of the present invention provides a knowledge question-answering system based on airport safety management, comprising: a knowledge analysis module, and an information acquisition module, an answer selection module and a database connected thereto;

[0007] The information acquisition module is used to obtain key information of each clause in the airport regulations and obtain the user's question text through the information collection device; wherein the key information includes definitions and keywords;

[0008] The knowledge analysis module is used to perform reverse reasoning on key information to obtain relevant issues of each clause, perform similarity fitting based on the relevant issues to obtain similarity factors between the relevant issues; perform analysis based on the similarity factors to obtain similarity values ​​between the clauses, judge each clause based on the similarity values ​​to obtain similarity situations of the clauses, and bundle the clauses based on the similarity situations of the clauses and store them in a database;

[0009] The answer selection module is used to extract keywords from the question text, query similar questions from the database based on the keywords, and extract corresponding terms based on the similar questions to answer.

[0010] In the present invention, reverse reasoning is performed on key information to obtain relevant questions of each clause. This step is the first key point of the present invention. When the similarities of each clause are analyzed in the subsequent analysis, the relevant questions of each clause are analyzed to increase the accuracy of the analysis; in the present invention, similarity fitting is performed based on related questions to obtain similarity factors between related questions. This step obtains the correlation between related questions by performing similarity fitting on questions, which provides data support for the subsequent correlation of clauses obtained through related questions; in the present invention, similarity values ​​between clauses are obtained by analysis based on similarity factors. This step can obtain clauses with greater correlation when the clause text is complex, so that the system can better adapt to question-and-answer situations with many or complex clauses, thereby improving the efficiency and accuracy of system analysis.

[0011] Preferably, the obtaining of key information of each clause in the airport rules and regulations includes:

[0012] Obtain the latest rules and regulations from the rules and regulations publishing website, obtain the content of each clause in the latest rules and regulations, and add each clause in the latest rules and regulations to the standard rules and regulations library through manual recognition; use natural language recognition technology to obtain the definition and keywords of each clause in the standard rules and regulations library; among them, the rules and regulations publishing website is manually selected and set.

[0013] Preferably, obtaining the user's question text through the information collection device includes:

[0014] The user's questions are obtained by a question-and-answer robot installed in a suitable area of ​​the airport, and the user asks questions by scanning a QR code installed in a suitable area of ​​the airport with a mobile phone, and the questions asked through the QR code are marked as questions. The system uses text recognition technology to obtain the question text of the user's question. Among them, the suitable area of ​​the airport includes the waiting area, security check area, and boarding gate area.

[0015] Preferably, the reverse reasoning of the key information to obtain relevant questions of each clause includes:

[0016] Extracting the definitions and keywords of historical clauses from a database, using the definitions and keywords of historical clauses as training data, and using the relevant questions corresponding to the definitions and keywords of historical clauses as test data; using the training data to train an artificial intelligence model, using the test data to test the trained artificial intelligence model, and adjusting the parameters of the artificial intelligence model according to the test results to obtain a problem matching model with definitions and keywords as input and relevant questions as output; wherein the artificial intelligence model includes a Transformer neural network and a BP neural network;

[0017] The definitions and keywords of each clause in the standard regulations database are input into the question matching model to obtain the relevant questions of each clause.

[0018] It should be noted that the training data adopts the definitions and keywords of historical terms that exist in the database; the test data corresponding to the training data is obtained by manual selection.

[0019] The present invention obtains a question matching model by performing intelligent model simulation on the definitions and keywords of historical terms, thereby providing model support for subsequent acquisition of related questions of the terms.

[0020] Preferably, the similarity fitting based on the related questions to obtain similarity factors between the related questions includes:

[0021] A1: Extract relevant questions of each clause in the standard rules and regulations database, sort them to obtain relevant questions Wij, and obtain the text of relevant questions Wij through language recognition technology;

[0022] A2: Extract relevant questions Wij in sequence, compare the extracted relevant questions Wij with other relevant questions that have not been compared with them, and determine whether the ratio of the total number of similar words in the two compared relevant questions Wij to the total number of words in the two related questions Wij is 1; if yes, mark the similarity factor of the two compared relevant questions Wij as 1; if no, jump to A3; where i represents the item number, and j represents the item number corresponding to the item;

[0023] A3: Obtain the ratio BL of the total number of similar characters to the total number of characters in the two related questions Wij, obtain the common similarity factor CYZ of the two related questions Wij based on the formula CYZ=α×(exp(BL)-1), and assign the value of the common similarity factor SYZ to the similarity factors of the two related questions Wij respectively; wherein α is a ratio adjustment coefficient greater than 0.

[0024] It is worth noting that the present invention uses the exp() function when obtaining the common similarity factor. This is because the larger the value of the ratio BL is, the stronger the correlation between the two related problems Wij is, and the greater the correlation increases as the value of the ratio BL increases.

[0025] Preferably, the similarity values ​​between the clauses obtained by analyzing based on similarity factors include:

[0026] Extract the clauses in the standard rules and regulations database based on clause sequence number i, mark the extracted clauses as comparison clauses, and establish several comparison groups for the comparison clauses and clauses in the standard rules and regulations database that have not been compared with them; extract the total number of related questions Wij between the two clauses in the comparison group;

[0027] Sequentially extract the relevant questions Wij of the comparison clauses in the comparison group, and respectively judge whether the similarity factor between the extracted relevant questions Wij and another clause relevant question Wij exceeds the set threshold; if yes, mark the corresponding two relevant questions Wij as the relevant questions to be selected; if no, do nothing; wherein the set threshold is obtained through experience;

[0028] Obtain the ratio ZBL of the total number of times marked as related questions to be selected in the comparison group to the total number of questions, and obtain the ratio TBL of the total number of words of similar keywords between two clauses in the comparison group to the total number of words of the two clauses; obtain the similarity value SZ between the two clauses based on the formula SZ=β×(exp(ZBL)-1)+γ×(exp(TBL)-1); wherein β and γ are ratio adjustment coefficients, and β+γ=1; wherein similar keywords include keywords with the same and similar meanings after passing the disambiguation algorithm, and the keywords of each clause are obtained by manual selection.

[0029] It is worth noting that most of the current fitting of clause similarity is to obtain similarity values ​​by analyzing the clauses themselves, but different clauses may also answer the same question. The present invention extracts features of the problems that can be solved by the clauses. When the problems that can be solved by two clauses are mostly the same, it proves that the environments in which the two clauses are applicable are also the same. Therefore, the two clauses are marked as similar answers, and the similar answers are used as auxiliary answers to supplement the answers. This can make the system better adapt to question-and-answer situations with many or complex clauses, and improve the efficiency and accuracy of system analysis.

[0030] Preferably, the determining of each clause based on the similarity value to obtain the clause similarity includes:

[0031] Extract the similarity value SZ between the two clauses in the comparison group, and determine whether the similarity value SZ exceeds the similarity threshold; if yes, mark the two clauses in the comparison group as similar clauses; if no, mark the two clauses in the comparison group as dissimilar clauses; wherein the similarity threshold is obtained through experience; clause similarity includes similar clauses and dissimilar clauses.

[0032] Preferably, bundling the clauses based on the similarity of the clauses includes:

[0033] The clauses marked as similar clauses are bundled and trigger instructions are set. When one of the similar clauses is marked as the first answer by the system, another clause is marked as an auxiliary answer to the first answer.

[0034] Preferably, the querying of similar questions from a database based on keywords includes:

[0035] Extract relevant questions containing keywords of question text from the database, and sort the relevant questions; obtain the ratio of the number of words of similar words between relevant questions and question text to the number of words of relevant questions GBLm in sequence, obtain the ratio of the number of words of the longest continuous identical sentences between relevant questions and question text to the number of words of relevant questions LBLm in sequence, and obtain the question similarity factor WSY of the relevant question based on the formula WSY=δ×(exp(GBLm)-1)+μ×ln(LBLm+1); wherein m is the serial number of the relevant question containing keywords of question text; δ and μ are ratio adjustment coefficients, and δ+μ=1; the longest continuous identical sentence is the sentence with the most words in the sentence that appears in both sentences; ln() is a logarithmic function with the natural number e as the base;

[0036] The related questions with the largest question similarity factor are marked as similar questions.

[0037] Preferably, the extracting corresponding clauses based on similar questions to answer them includes:

[0038] The clauses corresponding to similar questions are extracted, and the clauses are marked as the first answers, and the first answers are presented to the user first; the auxiliary answers corresponding to the first answers are provided to the user as subsequent answers.

[0039] It should be noted that if a similar question corresponds to multiple clauses, all clauses will be marked as the first answer.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. The present invention obtains key information of each clause in the airport rules and regulations, performs reverse reasoning on the key information to obtain related issues of each clause, performs similarity fitting based on related issues to obtain similarity factors between related issues; performs analysis based on similarity factors to obtain similarity values ​​between clauses, judges each clause based on similarity values ​​to obtain clause similarity, and bundles the clauses based on clause similarity and stores them in a database, thereby solving the technical problems of ignoring the potential impact of the relevance of the problems solved by the clauses on the relevance of the clauses when analyzing the relevance of airport clauses, and the difficulty in reasonably combining the relevance of problems with the relevance of clauses. The present invention can increase the adaptability of the system to complex clauses by combining issues and clauses to analyze the relevance.

[0042] 2. Most of the current fitting of clause similarity is to obtain similarity values ​​by analyzing the clauses themselves, but different clauses may also answer the same question. The present invention extracts features of the problems that can be solved by the clauses. When the problems that can be solved by two clauses are mostly the same, it proves that the environments in which the two clauses are applicable are also the same. Therefore, the two clauses are marked as similar answers, and the similar answers are used as auxiliary answers to supplement the answers. This can make the system better adapt to question-and-answer situations with many or complex clauses, and improve the efficiency and accuracy of system analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 It is a schematic diagram of the operation steps of the present invention;

[0045] Figure 2 It is a schematic diagram of the system module of the present invention;

[0046] Figure 3 Schematic diagram of the operation steps for obtaining similarity values ​​between clauses in the present invention. DETAILED DESCRIPTION

[0047] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] See also Figure 1-Figure 2 , the first aspect of the present invention provides a knowledge question-answering system based on airport security management, including: a knowledge analysis module, and an information acquisition module, an answer selection module and a database connected thereto;

[0049] Information acquisition module: used to obtain key information of each clause in the airport regulations and obtain the user's question text through the information collection device; wherein the key information includes definitions and keywords;

[0050] Knowledge analysis module: used to reversely reason key information to obtain relevant issues of each clause, perform similarity fitting based on relevant issues to obtain similarity factors between relevant issues; perform analysis based on similarity factors to obtain similarity values ​​between clauses, judge clauses based on similarity values ​​to obtain clause similarity, and bundle clauses based on clause similarity and store them in the database;

[0051] Answer selection module: used to extract keywords from the question text, query similar questions from the database based on keywords, and extract corresponding terms based on similar questions to answer.

[0052] Exemplarily, in this embodiment, the question raised by the user is: What should I do if the checked baggage contains prohibited items?

[0053] If the airport rules and regulations state:

[0054] Clause 1: If the checked baggage contains prohibited items, you should report it to the staff immediately and cooperate with the relevant departments to handle it;

[0055] Clause 2: Passengers whose checked baggage contains prohibited items and who do not cooperate with airport operations will be handed over to the public security department for handling;

[0056] The system obtains the key information of clauses 1 and 2 in the airport regulations and reversely infers the key information to obtain relevant questions of each clause, such as:

[0057] Questions related to Article 1:

[0058] Related questionW11: What should I do if my checked baggage contains prohibited items?

[0059] Related question W12: What should I do if there are prohibited items in my checked baggage?

[0060] Related question W13: What should I do if prohibited items are found in checked baggage?

[0061] Questions related to Article 2:

[0062] Related questionW21: What should I do if my checked baggage contains prohibited items?

[0063] Related question W22: What should I do if I carry prohibited items in my checked baggage?

[0064] Extract the relevant questions of clause 1 and clause 2 in the standard rules and regulations database, and sort them to obtain the relevant questions Wij, and the sorting method is: W11, W12, W13, W21, W22;

[0065] First, extract W11, and compare W11 with W21 and W22 respectively; because the total number of similar words between W11 and W21 accounts for 1 of the total number of words in the two related questions, the similarity factor between W11 and W21 is marked as 1;

[0066] What should I do if my checked baggage contains prohibited items due to W11?

[0067] W21: What should I do if I have prohibited items in my checked baggage?

[0068] Similar words for W11 and W22 are: checked baggage, prohibited items, what to do, included and brought;

[0069] Therefore, the ratio of the total number of similar words in W11 and W22 to the total number of words in the two related questions Wij is BL = 24 / 25 = 0.96

[0070] Based on the formula CYZ=α×(exp(0.96)-1), the common similarity factor CYZ of the two related problems to be compared is obtained; the value of the common similarity factor SYZ is assigned to the similarity factors of W11 and W22 respectively; then the similarity factors of W12 and W21, W12 and W22, W13 and W21, and W13 and W22 are obtained respectively;

[0071] Extract clause 1 and clause 2 in turn, mark clause 1 as a comparison clause, create a comparison group between the comparison clause and clause 2, and extract the total number of related questions Wij between the two clauses in the comparison group, which is 5;

[0072] The relevant issues due to the comparison clauses are: W11, W12, W13;

[0073] Therefore, relevant questions of the comparison clauses in the comparison group are extracted one by one, and it is judged whether the similarity factor between the extracted relevant questions and the relevant questions of another clause exceeds the set threshold:

[0074] 1. Extract W11 first: determine whether the similarity factor between W11 and W21 exceeds the set threshold; if yes, mark W11 and W21 as related issues to be selected; if no, do nothing;

[0075] Determine whether the similarity factor between W11 and W22 exceeds the set threshold; if yes, mark W11 and W22 as related issues to be selected; if no, do nothing;

[0076] 2. Extract W12 again: determine whether the similarity factor between W12 and W21 exceeds the set threshold; if yes, mark W12 and W21 as related issues to be selected; if no, do nothing;

[0077] Determine whether the similarity factor between W12 and W22 exceeds the set threshold; if yes, mark W12 and W22 as related issues to be selected; if no, do nothing;

[0078] 3. Extract W13 again: determine whether the similarity factor between W13 and W21 exceeds the set threshold; if yes, mark W13 and W21 as related issues to be selected; if no, do nothing;

[0079] Determine whether the similarity factor between W13 and W22 exceeds the set threshold; if yes, mark W13 and W22 as related issues to be selected; if no, do nothing;

[0080] Obtain the ratio ZBL of the total number of times marked as related questions to be selected in the comparison group to the total number of questions, and obtain the ratio TBL of the total number of words of similar keywords between two clauses in the comparison group to the total number of words of the two clauses; obtain the similarity value SZ between the two clauses based on the formula SZ=β×(exp(ZBL)-1)+γ×(exp(TBL)-1).

[0081] This application provides key information on each section of the airport regulations, including:

[0082] Obtain the latest rules and regulations from the rules and regulations publishing website, obtain the content of each clause in the latest rules and regulations, and add each clause in the latest rules and regulations to the standard rules and regulations library through manual recognition; use natural language recognition technology to obtain the definition and keywords of each clause in the standard rules and regulations library; among them, the rules and regulations publishing website is manually selected and set.

[0083] It should be noted that when manually identifying and adding the clauses in the latest rules and regulations to the standard rules and regulations library, if there is a clause in the standard rules and regulations library that is the same as the latest rules and regulations, this clause will not be added to the standard rules and regulations library; if there is a clause in the standard rules and regulations library that is inconsistent with the latest rules and regulations, the clause in the standard rules and regulations library that is inconsistent with the latest rules and regulations will be removed. This can ensure that the standard rules and regulations library used by the system is in the latest information state.

[0084] In this application, similarity fitting is performed based on related questions to obtain similarity factors between related questions, including:

[0085] A1: Extract relevant questions of each clause in the standard rules and regulations database, sort them to obtain relevant questions Wij, and obtain the text of relevant questions Wij through language recognition technology;

[0086] A2: Extract relevant questions Wij in sequence, compare the extracted relevant questions Wij with other relevant questions that have not been compared with them, and determine whether the ratio of the total number of similar words in the two compared relevant questions Wij to the total number of words in the two related questions Wij is 1; if yes, mark the similarity factor of the two compared relevant questions Wij as 1; if no, jump to A3; where i represents the item number, and j represents the item number corresponding to the item;

[0087] A3: Obtain the ratio BL of the total number of similar characters to the total number of characters in the two related questions Wij, obtain the common similarity factor CYZ of the two related questions Wij based on the formula CYZ=α×(exp(BL)-1), and assign the value of the common similarity factor SYZ to the similarity factors of the two related questions Wij respectively; wherein α is a ratio adjustment coefficient greater than 0.

[0088] It should be noted that the present invention performs sorting in step A1 to obtain related questions Wij, wherein the sorting method is: the clause listed in the standard rules and regulations library is the most forward, and the i number corresponding to the group of questions corresponding to the clause is smaller, and in the same group of questions, the fewer the number of words, the smaller the corresponding j number.

[0089] It should be noted that the similarity factor of the two related problems Wij compared in step A2 of the present invention is marked as 1, which means that the two related problems Wij currently being compared are exactly the same problems, and there is no need to calculate the common similarity factor for the two related problems, thereby reducing the amount of computation of the system.

[0090] It should be noted that similar characters are characters that contain the same characters, synonyms, and characters with similar meanings.

[0091] See also Figure 3 , in this application, similarity values ​​between clauses are obtained by analyzing based on similarity factors, including:

[0092] Extract the clauses in the standard rules and regulations database based on clause sequence number i, mark the extracted clauses as comparison clauses, and establish several comparison groups for the comparison clauses and clauses in the standard rules and regulations database that have not been compared with them; extract the total number of related questions Wij between the two clauses in the comparison group;

[0093] Sequentially extract the relevant questions Wij of the comparison clauses in the comparison group, and judge whether the similarity factor between the extracted relevant questions Wij and the relevant question Wij of another clause exceeds the set threshold; if yes, mark the corresponding two relevant questions Wij as the relevant questions to be selected; if no, do nothing; wherein the set threshold is obtained through experience;

[0094] Obtain the ratio ZBL of the total number of times marked as related questions to be selected in the comparison group to the total number of questions, and obtain the ratio TBL of the total number of words of similar keywords between two clauses in the comparison group to the total number of words of the two clauses; obtain the similarity value SZ between the two clauses based on the formula SZ=β×(exp(ZBL)-1)+γ×(exp(TBL)-1); wherein β and γ are ratio adjustment coefficients, and β+γ=1; wherein similar keywords include keywords with the same and similar meanings after passing the disambiguation algorithm, and the keywords of each clause are obtained by manual selection.

[0095] It should be noted that if the same related question Wij is marked as a related question to be selected twice, the total number of times it is marked as a related question to be selected is increased by 2.

[0096] In this application, similar questions are queried from the database based on keywords, including:

[0097] Extract relevant questions containing keywords of question text from the database, and sort the relevant questions; obtain the ratio of the number of words of similar words between relevant questions and question text to the number of words of relevant questions GBLm in sequence, obtain the ratio of the number of words of the longest continuous identical sentences between relevant questions and question text to the number of words of relevant questions LBLm in sequence, and obtain the question similarity factor WSY of the relevant question based on the formula WSY=δ×(exp(GBLm)-1)+μ×ln(LBLm+1); wherein m is the serial number of the relevant question containing keywords of question text; δ and μ are ratio adjustment coefficients, and δ+μ=1; the longest continuous identical sentence is the sentence with the most words in the sentence that appears in both sentences; ln() is a logarithmic function with the natural number e as the base;

[0098] The related questions with the largest question similarity factor are marked as similar questions.

[0099] It should be noted that the longest consecutive identical sentences are illustrated as follows:

[0100] Statement 1: Littering is not allowed in the airport, and offenders will be warned through the broadcast;

[0101] Sentence 2: Advise people who litter for the first time, and broadcast a warning to those who do not listen;

[0102] Therefore, the consecutive identical sentences are: littering, broadcast warnings; the longest consecutive identical sentence is: littering.

[0103] Part of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is a formula closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0104] Working principle of the present invention:

[0105] Obtain key information of each clause in the airport rules and regulations, and obtain the user's question text through the information collection device; reverse reasoning is performed on the key information to obtain relevant questions of each clause. This step is the first key point of the present invention. When analyzing the similarity of each clause in the subsequent analysis, the relevant questions of each clause are analyzed to increase the accuracy of the analysis; similarity fitting is performed based on the relevant questions to obtain the similarity factor between each related question. This step obtains the correlation between each related question by performing similarity fitting on the questions, which provides data support for the subsequent correlation of clauses obtained through related questions; similarity values ​​are obtained based on the similarity factor. This step can obtain clauses with a large correlation when the clause text is complex, so that the system can better adapt to question-and-answer situations with many or complex clauses, thereby improving the efficiency and accuracy of system analysis; each clause is judged based on the similarity value to obtain the clause similarity, and the clauses are bundled and stored in the database based on the clause similarity; keyword extraction is performed on the question text, similar questions are queried from the database based on the keywords, and corresponding clauses are extracted based on the similar questions for answering.

[0106] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. The knowledge question-answering system based on airport safety management is characterized by: include: A knowledge analysis module, and an information acquisition module, an answer selection module and a database connected thereto; The information acquisition module is used to obtain key information of each clause in the airport regulations and obtain the user's question text through the information collection device; wherein the key information includes definitions and keywords; The knowledge analysis module is used to perform reverse reasoning on key information to obtain relevant issues of each clause, perform similarity fitting based on the relevant issues to obtain similarity factors between the relevant issues; perform analysis based on the similarity factors to obtain similarity values ​​between the clauses, judge each clause based on the similarity values ​​to obtain similarity situations of the clauses, and bundle the clauses based on the similarity situations of the clauses and store them in a database; The answer selection module is used to extract keywords from the question text, query similar questions from the database based on the keywords, and extract corresponding terms based on the similar questions to answer; The similarity of the clauses is determined based on the similarity value, including: Extract the similarity value SZ between the two clauses in the comparison group, and determine whether the similarity value SZ exceeds the similarity threshold; if yes, mark the two clauses in the comparison group as similar clauses; if no, mark the two clauses in the comparison group as dissimilar clauses; wherein the clause similarity includes similar clauses and dissimilar clauses; The bundling of clauses based on similarity of clauses includes: Bundle the clauses marked as similar clauses and set trigger instructions. When one of the similar clauses is marked as the first answer by the system, mark another clause as an auxiliary answer to the first answer. The extracting corresponding clauses based on similar questions to answer the questions includes: The clauses corresponding to similar questions are extracted, and the clauses are marked as the first answers, and the first answers are presented to the user first; the auxiliary answers corresponding to the first answers are provided to the user as subsequent answers.

2. The knowledge question-answering system based on airport safety management according to claim 1 is characterized in that: The key information of each clause in the airport regulations is obtained, including: Obtain the latest rules and regulations from the rules and regulations publishing website, obtain the content of each clause in the latest rules and regulations, and add each clause in the latest rules and regulations to the standard rules and regulations library through manual recognition; use natural language recognition technology to obtain the definition and keywords of each clause in the standard rules and regulations library; among them, the rules and regulations publishing website is manually selected and set.

3. The knowledge question-answering system based on airport safety management according to claim 1 is characterized in that: The step of obtaining the user's question text through the information collection device includes: The user's questions are obtained by a question-and-answer robot installed in a suitable area of ​​the airport, and the user asks questions by scanning a QR code installed in a suitable area of ​​the airport with a mobile phone, and the questions asked through the QR code are marked as questions. The system uses text recognition technology to obtain the question text of the user's question. Among them, the suitable area of ​​the airport includes the waiting area, security check area, and boarding gate area.

4. The knowledge question-answering system based on airport safety management according to claim 1 is characterized in that: The reverse reasoning of key information to obtain relevant issues of each clause includes: Extracting the definitions and keywords of historical clauses from a database, using the definitions and keywords of historical clauses as training data, and using the relevant questions corresponding to the definitions and keywords of historical clauses as test data; using the training data to train an artificial intelligence model, using the test data to test the trained artificial intelligence model, and adjusting the parameters of the artificial intelligence model according to the test results to obtain a problem matching model with definitions and keywords as input and relevant questions as output; wherein the artificial intelligence model includes a Transformer neural network and a BP neural network; The definitions and keywords of each clause in the standard regulations library are input into the question matching model to obtain the relevant questions of each clause.

5. The knowledge question-answering system based on airport safety management according to claim 4 is characterized in that: The similarity fitting based on the related questions to obtain the similarity factors between the related questions includes: A1: Extract relevant questions of each clause in the standard regulations database, sort them to obtain relevant questions Wij, and obtain the text of relevant questions Wij through language recognition technology; A2: Extract relevant questions Wij in sequence, compare the extracted relevant questions Wij with other relevant questions that have not been compared with them, and determine whether the ratio of the total number of similar words in the two compared relevant questions Wij to the total number of words in the two related questions Wij is 1; if yes, mark the similarity factor of the two compared relevant questions Wij as 1; if no, jump to A3; where i represents the item number, and j represents the item number corresponding to the item; A3: Obtain the ratio BL of the total number of similar characters to the total number of characters in the two related questions Wij, obtain the common similarity factor CYZ of the two related questions Wij based on the formula CYZ=α×(exp(BL)-1), and assign the value of the common similarity factor SYZ to the similarity factors of the two related questions Wij respectively; wherein α is a ratio adjustment coefficient greater than 0.

6. The knowledge question-answering system based on airport safety management according to claim 5 is characterized in that: The similarity values ​​between the clauses obtained by analyzing based on similarity factors include: Extract the clauses in the standard rules and regulations database based on clause sequence number i, mark the extracted clauses as comparison clauses, and establish several comparison groups for the comparison clauses and clauses in the standard rules and regulations database that have not been compared with them; extract the total number of related questions Wij between the two clauses in the comparison group; Sequentially extract the related questions Wij of the comparison clauses in the comparison group, and respectively determine whether the similarity factor between the extracted related questions Wij and another clause related question Wij exceeds a set threshold; if yes, mark the corresponding two related questions Wij as related questions to be selected; if no, do nothing; Obtain the ratio ZBL of the total number of times marked as related questions to be selected in the comparison group to the total number of questions, and obtain the ratio TBL of the total number of words of similar keywords between two clauses in the comparison group to the total number of words of the two clauses; obtain the similarity value SZ between the two clauses based on the formula SZ=β×(exp(ZBL)-1)+γ×(exp(TBL)-1); wherein β and γ are ratio adjustment coefficients, and β+γ=1; wherein similar keywords include keywords with the same or similar meanings after the disambiguation algorithm, and the keywords of each clause are obtained by manual selection.

7. The knowledge question-answering system based on airport safety management according to claim 1 is characterized in that: The querying of similar questions from the database based on keywords includes: Extract relevant questions containing keywords of question text from the database and sort the relevant questions; obtain the ratio of the number of similar words between relevant questions and question text to the number of words of relevant questions GBLm in sequence, obtain the ratio of the number of words of the longest continuous identical sentences between relevant questions and question text to the number of words of relevant questions LBLm in sequence, and obtain the question similarity factor WSY of the relevant questions based on the formula WSY=δ×(exp(GBLm)-1)+μ×ln(LBLm+1); wherein m is the serial number of the relevant question containing keywords of question text; δ and μ are ratio adjustment coefficients, and δ+μ=1; the longest continuous identical sentence is the sentence with the most words in the sentence that appears in both sentences; ln() is a logarithmic function with the natural number e as the base; The related questions with the largest question similarity factor are marked as similar questions.

Images