Intelligent indexing method and system for radiation effect database based on AI large model
By introducing AI large models into the radiation effect database and using intelligent indexing methods to summarize and classify research data, the problem of inconvenient data sorting in the existing technology is solved, and the effect of quickly and efficiently obtaining research data is achieved.
Patent Information
- Application Number
- CN202411619724.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-13
- Publication Date
- 2025-03-07
AI Technical Summary
The current field of electronic reliability lacks a public radiation effect database with comprehensive data storage and orderly sorting, making it difficult to obtain research data quickly and efficiently.
Using the intelligent indexing method of radiation effect database based on AI large model, by setting up multiple tag databases, users upload research data and extract keywords to form tags, and the data is stored according to tags. After the user enters the search information, the matching data is listed or explained through the matching degree comparison and cosine similarity algorithm.
It improves data processing and induction capabilities, can intelligently answer user questions, provide the research data required by users, and improves the efficiency and user experience of the database.
Smart Images
Figure CN119577057A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of information technology, and in particular to an intelligent indexing method for a radiation effect database based on an AI big model. Background Art
[0002] There is a complex radiation environment in space composed of a large number of radiation particles, including electrons, alpha particles, protons, neutrons, gamma rays, pions, muons, etc. These particles can cause single particle effects, total ionization dose effects, and displacement damage effects in electronic equipment in space, causing damage to the devices and affecting their functions.
[0003] Every year, various radiation research units around the world publish a large amount of research data in the field of electronic reliability, including experimental reports, journal articles, experimental data, etc. These research data are scattered on the Internet, distributed on various network platforms, and lack effective organization and summary. Today, the field of electronic reliability lacks a comprehensive, well-organized and public radiation effect database to help it quickly and efficiently obtain research data for reference.
[0004] In recent years, AI big models have been popular due to their powerful data processing capabilities, efficient learning and reasoning capabilities, high flexibility and scalability. Through advanced algorithms such as deep learning, AI big models can process massive amounts of data and extract valuable information from them. Introducing AI big models into the radiation effect database can improve data processing capabilities, process complex and disordered research data in a short time, and extract keywords from research data for classification; its powerful learning and reasoning capabilities can help users quickly filter out the research data they need based on their keywords or questions when indexing.
[0005] There is an urgent need for a public radiation effect database combined with an AI big model to collect and organize research data in the field of electronic reliability that is publicly available on the Internet. These research data are collected, processed, and classified and stored, so that users can accurately index the required research data based on key information when accessing the radiation effect database. Summary of the invention
[0006] The present disclosure provides a radiation effect database intelligent indexing method and system based on an AI big model to solve one of the technical problems recognized by the inventor.
[0007] The present disclosure provides a radiation effect database intelligent indexing method based on an AI big model, comprising the following steps:
[0008] S1. The administrator sets up multiple tag databases, which together form a corpus;
[0009] S2, users upload text data containing research data to the corpus;
[0010] S3, extracting keywords from the text data, and forming corresponding tags according to the extracted keywords. One text data includes one or more tags. According to the formed tags, the text data are respectively stored in the corresponding tag database;
[0011] S4, the user inputs search information, and the search information input by the user is compared with the text data in the corpus for matching degree;
[0012] S5. Set a threshold for the matching degree of the text data. When the matching degree is higher than the threshold, list the data with the same label in descending order of matching degree for the user to choose. When the matching degree is lower than the threshold, answer the user's search information through the text data explained by the label.
[0013] Preferably, the matching comparison in step S4 includes the following steps:
[0014] S41, splitting the search information input by the user into one or more keywords;
[0015] S42, performing word frequency matching between the split keywords and the labels of the text data in the corpus;
[0016] S43, calculating the matching degree by using the cosine similarity algorithm.
[0017] Preferably, the formula of the cosine similarity algorithm is:
[0018] Preferably, the step of forming the interpretation text data of the label in step S5 includes:
[0019] S51, extracting explanation information for the label in each text data;
[0020] S52, performing similarity matching on the explanation information of the tags in all text data containing the same tags;
[0021] S53, generating an explanation text for each tag through similarity matching;
[0022] S54. The manager reviews the generated explanation text to determine whether it is reasonable. If it is reasonable, the explanation text data of the label is finally formed; if it is unreasonable, the manager modifies it and finally forms the explanation text data of the label.
[0023] Preferably, after the user's search is completed, an answer satisfaction rating is sent to the user, and the degree of match between the sent explanatory text data and the user's question is determined based on the user's satisfaction.
[0024] Preferably, the step of judging the matching degree between the user's satisfaction and the interpretation text data includes:
[0025] S61, setting a satisfaction threshold, if the user's satisfaction exceeds the threshold, it is judged as "satisfied", and if the user's satisfaction is lower than the threshold, it is judged as "dissatisfied";
[0026] S62. When multiple users ask the same question and provide the same explanatory text data, a ratio value is set. If the number of users who evaluate "satisfied" is higher than or equal to the ratio value, the explanatory text data is judged to be correct. If the number of users who evaluate "satisfied" is lower than the ratio value, the explanatory text data is judged to be wrong.
[0027] Preferably, the interpretation text data judged to be erroneous will prompt the administrator to make revisions.
[0028] An intelligent indexing system for radiation effect database based on AI big model, including:
[0029] A corpus, including a plurality of different tag databases, wherein the tag databases store text data containing corresponding tags;
[0030] Data transmission module, used to upload text data;
[0031] The data classification processing module is used to perform full-text recognition on the text data uploaded by the user, extract keywords or information from the text data, form corresponding keywords for each text data, and define labels for the text data based on the similarity between the keywords and the labels;
[0032] The data retrieval module is used to retrieve text data in the corpus and list the retrieved text data with the same label in descending order of matching degree for users to review.
[0033] Preferably, an intelligent learning module is also included to form a summary for each tag database and an explanatory text data for each tag.
[0034] Preferably, an AI question-and-answer module is also included to match the questions raised by the user with the text data in the corpus, and a similarity threshold is set. When the matching degree between the questions raised by the user and the text data in the corpus is lower than the threshold, the AI question-and-answer module is called, and the AI question-and-answer module calls the interpreted text data of the intelligent learning module to answer the user.
[0035] The beneficial effects of the present disclosure are mainly as follows: the present invention introduces an AI big model into the radiation effect database, and summarizes and classifies a large amount of text data by extracting keywords and tags. The radiation effect database accurately provides the information required by users based on the powerful learning ability of the AI big model and the existing research data in the database, improves the data processing and summarizing and organizing capabilities of the database, and can answer users' questions more intelligently.
[0036] It should be understood that both the foregoing general description and the following specific embodiments are for the purpose of example and illustration and do not necessarily limit the present disclosure. The accompanying drawings, which are incorporated into and constitute a part of the specification, illustrate the subject matter of the present disclosure. At the same time, the specification and the drawings are used to explain the principles of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] In order to more clearly illustrate the specific embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the specific embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0038] Figure 1 This is a flowchart of the steps of the indexing method according to an embodiment of the present disclosure;
[0039] Figure 2 A flowchart of the matching degree comparison step in the embodiment of the present disclosure;
[0040] Figure 3 A flowchart of steps for interpreting text data is provided for an embodiment of the present disclosure; DETAILED DESCRIPTION
[0041] The technical solution of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments.
[0042] Based on the embodiments in the present disclosure, 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 disclosure.
[0043] In the description of the present disclosure, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present disclosure and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present disclosure. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.
[0044] In the description of the present disclosure, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present disclosure can be understood according to specific circumstances.
[0045] Example
[0046] like Figure 1-3 As shown, this embodiment provides a radiation effect database intelligent indexing method based on an AI large model, comprising the following steps:
[0047] S1. The administrator sets up multiple tag databases, which together form a corpus;
[0048] The administrator sets multiple tags, such as "single particle effect", "CMOS", etc., sets a database for each tag, and stores text data with the tag in the corresponding tag database. Multiple tag databases together constitute a corpus.
[0049] S2, users upload text data containing research data to the corpus;
[0050] Users upload text data to the corpus. The text data can be experimental reports, journal articles, experimental data, etc., and match them to the corresponding label database.
[0051] S3, extracting keywords from the text data, and forming corresponding tags according to the extracted keywords. One text data includes one or more tags. According to the formed tags, the text data are respectively stored in the corresponding tag database;
[0052] The text data uploaded by the user is fully recognized, and the keywords or information in the text data are extracted. The keywords or information of each text data are converted into its own label. One data can have multiple labels. The text data is stored in the corresponding label database according to different labels, that is, one text data can be stored in multiple label databases. Through the AI large model, multiple text data can be processed and summarized at one time to improve data processing efficiency. For example, a user uploads a text data named "Study on Electronic Single Particle Effect of 45nm CMOS Chip". The text data uploaded by the user is fully scanned, and keywords such as "single particle effect", "electronic", "45nm", and "CMOS" are extracted. Correspondingly, the text data has four keywords; traverse the corpus, assuming that the corpus contains two label databases of "single particle effect" and "CMOS", then the text data uploaded by the user is stored in these two label databases respectively.
[0053] S4, the user inputs search information, and the search information input by the user is compared with the text data in the corpus for matching degree;
[0054] The user inputs the search information, extracts the keywords or information in the search information, compares the extracted keywords with the text data in the corpus, and retrieves the matching text data.
[0055] S5. Set a threshold for the matching degree of the text data. When the matching degree is higher than the threshold, list the data with the same label in descending order of matching degree for the user to choose. When the matching degree is lower than the threshold, answer the user's search information through the text data explained by the label.
[0056] The matching degree of the search information and text data input by the user is calculated, a matching degree threshold is set, all the text data in the corpus that exceeds the matching degree threshold is extracted, and all the text data that exceeds the matching degree threshold is listed in a list from high to low according to the matching degree for the user to select and consult; if the matching degree of all the text data in the corpus and the search information input by the user is lower than the threshold, the keywords of the search information are extracted, and the keywords are matched with the tags. If the keywords are the same or similar to the tags, the explanation text of the tags is sent as the answer. If there are no keywords that are the same or similar to the tags, the user is answered "no explanation".
[0057] For example, the question that the user searched for was: Research on the mechanism of electron-induced single-event effect in 45nm CMOS SRAM;
[0058] Assume that the corpus contains the following text data:
[0059] Text A: 28nm, CMOS type, SRAM, electronics, induced, single event upset, mechanism study;
[0060] Text B: 45nm,and,28nm,CMOS-type,SRAM,electronic,induced,single-event effects;
[0061] Text C: Electronic, Induction, Single Particle Effects, Mechanistic Studies;
[0062] Text D: Neutron, proton, single particle effect, mechanism study;
[0063] Step 1: Split the texts in the user and corpus (text A, B, C, D), and the split results are:
[0064] User question: 45nm, CMOS, SRAM, electronics, induction, single event effect, mechanism study
[0065] Text A: 28nm, CMOS type, SRAM, electronics, induced, single event upset, mechanism study
[0066] Text B: 45nm, and, 28nm, CMOS, SRAM, electronics, induced, single event effects
[0067] Text C: Electronics, Induction, Single Particle Effects, Mechanistic Studies
[0068] Text D: Neutrons, protons, single particle effects, mechanism studies
[0069] Step 2: List all the words in the user question and the texts in the corpus (text A, B, C, D) (taking text A as an example), user question and text A: 45nm, 28nm, CMOS type, SRAM, electronics, induced, single-particle effect, single-particle upset, mechanism research;
[0070] Step 3: Calculate word frequency,
[0071] User questions: 45nm1, 28nm0, CMOS type1, SRAM1, electronics1, induced1, single event effect1, single event upset0, mechanism study1;
[0072] Text A: 45nm0, 28nm1, CMOS type1, SRAM1, electronics1, induction1, single event effect0, single event upset1, mechanism study1;
[0073] Step 4. Write out the vectors: user question (1, 0, 1, 1, 1, 1, 0, 1) and text A (0, 1, 1, 1, 1, 0, 1, 1).
[0074] Step 5: Calculate the matching degree between the user question and the radiation effect database using the cosine similarity algorithm:
[0075]
[0076] User question and text A
[0077] The larger the calculated COS(θ) value is, the more similar the user question is to text A.
[0078] Similarly, the COS(θ) of texts B, C, and D are calculated to be 0.802, 0.756, and 0.365 respectively.
[0079] Considering the matching degree between the user question and the text, a COS(θ) threshold is set to 0.7. When the matching degree between the user question and the text data is calculated to be lower than 0.7, the text is not displayed.
[0080] Finally, the texts are arranged from high to low according to the matching degree between the user and the text as shown below:
[0081] User question: Research on the mechanism of electron-induced single event effects in 45nm CMOS SRAM.
[0082] Systematic listing: Text B: 45nm and 28nm CMOS SRAM electron-induced single event effects:
[0083] Text C: Study on the mechanism of electron-induced single-particle effect;
[0084] Text A: Study on the mechanism of electron-induced single-event upset in 28nm CMOS SRAM.
[0085] Furthermore, the matching comparison in step S4 includes the following steps:
[0086] S41, splitting the search information input by the user into one or more keywords;
[0087] S42, performing word frequency matching between the split keywords and the labels of the text data in the corpus;
[0088] S43, calculating the matching degree by using the cosine similarity algorithm.
[0089] The formula of the cosine similarity algorithm is:
[0090] For example, the user inputs "single particle effect of integrated circuits caused by electrons in space environment" (recorded as text X). Based on the cosine similarity of the similarity algorithm, the "single particle effect of integrated circuits caused by electrons in space radiation environment" is first split into words such as "space radiation environment", "electrons", "single particle effect", and "integrated circuit" (text X); the above keywords are matched with the keywords of the articles in the corpus. The keywords of "Research on electronic single particle effects of 45nm CMOS chips" (recorded as text Y) in the tag database are "single particle effect", "electrons", "45nm", and "CMOS". This is a word frequency match of text X and text Y. Text X: space radiation environment 0, electronics 1, integrated circuit 0, single particle effect 1; text Y: 45nm 0, CMOS 0, electronics 1, single particle effect 1; write out the word frequency, text X (0, 1, 0, 1), text Y (0, 0, 1, 1), and finally through the algorithm formula Calculate the similarity of two texts. The larger the calculated match value, the higher the similarity between the texts. According to the size of the match, a text list is listed for users to browse.
[0091] Furthermore, the step of forming the interpretation text data of the label in step S5 includes:
[0092] S51, extracting explanation information for the label in each text data;
[0093] S52, performing similarity matching on the explanation information of the tags in all text data containing the same tags;
[0094] S53, generating an explanation text for each tag through similarity matching;
[0095] S54. The manager reviews the generated explanation text to determine whether it is reasonable. If it is reasonable, the explanation text data of the label is finally formed; if it is unreasonable, the manager modifies it and finally forms the explanation text data of the label.
[0096] Each text data will contain explanatory information for the label. The AI big model will be used to extract the explanatory information for the label in each text data, and the explanatory information for the same label in different text data will be matched for similarity. Through the similarity value, a short explanatory text will be formed for each label. The formed explanatory text will eventually be sent to the background for manual judgment by the administrator whether it is reasonable. If it is reasonable, the explanatory text data of the label will be finally formed; if it is unreasonable, the administrator will modify it and finally form the explanatory text data of the label.
[0097] For example, there are 4 texts in the label library named "Single Particle Effect", each of which has an explanation for single particle effect.
[0098] Text A
[0099] Single particle effect: refers to the effect produced when a single particle with a certain amount of energy deposits energy in a semiconductor device and the generated electron-hole pairs are collected.
[0100] Text B
[0101] Single particle effect: refers to the effect produced when a single high-energy charged particle with a certain energy deposits energy in a semiconductor device and the generated electron-hole pairs are collected.
[0102] Text C
[0103] Single-particle effect: refers to the effect that a single charged particle with a certain energy deposits energy in a semiconductor device, and the generated electron-hole pairs cause the device to change its logical state.
[0104] Text D
[0105] Single particle effect: refers to the effect that charged particles with energy in space deposit energy in semiconductor devices, and the generated electron-holes are collected to produce transient current.
[0106] By segmenting the explanation of the single particle effect in texts A, B, C, and D, and finally looking at the frequency of each word, if the same word exists in each text data, the system will sum up these words to form a smooth explanation. This explanation is the explanation of the label library "single particle effect". The explanation of the "single particle effect" label will be sent to the background, and the manager will judge whether it meets the requirements. If not, the staff can make some modifications. The powerful data processing capabilities of the AI big model can reduce the pressure of manual work. Correspondingly, this link belongs to the ability attribute of the radiation effect data itself. It needs to rely on the combination of the AI big model and the professional manual team to finally complete the accurate interpretation of each label.
[0107] Furthermore, after the user search is completed, an answer satisfaction evaluation is sent to the user, and the matching degree between the sent explanation text data and the user question is determined based on the user's satisfaction. In this embodiment, the user's satisfaction evaluation is divided into scores of 1-10. After the user evaluation is completed, the matching degree between the sent explanation text data and the user question can be determined based on the user's evaluation, so as to facilitate the adjustment of the push of the answer.
[0108] Specifically, the steps of judging the matching degree between the user's satisfaction and the interpretation text data include:
[0109] S61, setting a satisfaction threshold, if the user's satisfaction exceeds the threshold, it is judged as "satisfied", and if the user's satisfaction is lower than the threshold, it is judged as "dissatisfied";
[0110] S62. When multiple users ask the same question and provide the same explanatory text data, a ratio value is set. If the number of users who evaluate "satisfied" is higher than or equal to the ratio value, the explanatory text data is judged to be correct. If the number of users who evaluate "satisfied" is lower than the ratio value, the explanatory text data is judged to be wrong.
[0111] For example, the satisfaction threshold is set to "8". When multiple users raise the same or similar search, the same explanatory text data is pushed. The satisfaction evaluation scores raised by multiple users are compared with the satisfaction threshold. If the score is higher than or equal to "8", it is judged as "satisfied". If it is lower than "8", it is judged as unsatisfied. Assuming that the number of users who have evaluated exceeds a certain number, for example, 100 users have evaluated the answer to the same question, a ratio value is set, which is the ratio of "satisfied" customers to the total number. Assuming that the set ratio value is "80%", if more than 80% of users are "satisfied", the answer is judged to be correct. If less than 80% of users are "satisfied", the explanatory text data is judged to be wrong. The wrong explanatory text data is pushed to the backend administrator for revision. The revised explanatory text data will be pushed the next time the same question is encountered.
[0112] An intelligent indexing system for radiation effect database based on AI big model, including:
[0113] A corpus, including a plurality of different tag databases, wherein the tag databases store text data containing corresponding tags;
[0114] Specifically, a plurality of label databases with different labels are set, and text data with corresponding labels are stored in corresponding label databases. The plurality of different label databases together constitute a corpus.
[0115] Data transmission module, used to upload text data;
[0116] Specifically, for example, a user uploads a text data titled “Study on Electronic Single Particle Effects of 45nm CMOS Chip” in the upload module;
[0117] The data classification processing module is used to perform full-text recognition on the text data uploaded by the user, extract keywords or information from the text data, form corresponding keywords for each text data, and define labels for the text data based on the similarity between the keywords and the labels;
[0118] Specifically, the system senses the text data of "Research on Electronic Single Particle Effects of 45nm CMOS Chips". After scanning the full text, the system extracts keywords such as "single particle effect", "electronic", "45nm", and "CMOS". The corresponding text contains four keywords; in the tag database set by the background administrator, there are two tag databases named "single particle effect" and "CMOS". The corresponding text data will be stored in the databases named "single particle effect" and "CMOS". Finally, multiple tag libraries are jointly formed into a radiation effect database; through the AI big model, multiple data texts can be processed and summarized at one time, thereby improving data processing efficiency.
[0119] The data retrieval module is used to retrieve text data in the corpus and list the retrieved text data with the same label in descending order of matching degree for users to review.
[0120] For example, the user inputs "single particle effect of integrated circuits caused by electrons in space environment" (recorded as text X). Based on the cosine similarity of the similarity algorithm, the "single particle effect of integrated circuits caused by electrons in space radiation environment" is first split into words such as "space radiation environment", "electrons", "single particle effect", and "integrated circuit" (text X); the above keywords are matched with the keywords of the articles in the corpus. The keywords of "Research on electronic single particle effects of 45nm CMOS chips" (recorded as text Y) in the tag database are "single particle effect", "electrons", "45nm", and "CMOS". This is a word frequency match of text X and text Y. Text X: space radiation environment 0, electronics 1, integrated circuit 0, single particle effect 1; text Y: 45nm 0, CMOS 0, electronics 1, single particle effect 1; write out the word frequency, text X (0, 1, 0, 1), text Y (0, 0, 1, 1), and finally through the algorithm formula Calculate the similarity of two texts. The larger the calculated match value, the higher the similarity between the texts. According to the size of the match, a text list is listed for users to browse.
[0121] Furthermore, it also includes an intelligent learning module, which forms a summary for each tag database and forms an explanation text data for each tag.
[0122] Each text data will contain explanatory information for the label. The AI big model will be used to extract the explanatory information for the label in each text data, and the explanatory information for the same label in different text data will be matched for similarity. Through the similarity value, a short explanatory text will be formed for each label. The formed explanatory text will eventually be sent to the background for manual judgment by the administrator whether it is reasonable. If it is reasonable, the explanatory text data of the label will be finally formed; if it is unreasonable, the administrator will modify it and finally form the explanatory text data of the label.
[0123] Furthermore, it also includes an AI question-and-answer module, which matches the questions raised by the user with the text data in the corpus and sets a similarity threshold. When the matching degree between the questions raised by the user and the text data in the corpus is lower than the threshold, the AI question-and-answer module is called, and the AI question-and-answer module calls the interpreted text data of the intelligent learning module to answer the user.
[0124] For example, the satisfaction threshold is set to "8". When multiple users raise the same or similar search, the same explanatory text data is pushed. The satisfaction evaluation scores raised by multiple users are compared with the satisfaction threshold. If the score is higher than or equal to "8", it is judged as "satisfied". If it is lower than "8", it is judged as unsatisfied. Assuming that the number of users who have evaluated exceeds a certain number, for example, 100 users have evaluated the answer to the same question, a ratio value is set, which is the ratio of "satisfied" customers to the total number. Assuming that the set ratio value is "80%", if more than 80% of users are "satisfied", the answer is judged to be correct. If less than 80% of users are "satisfied", the explanatory text data is judged to be wrong. The wrong explanatory text data is pushed to the backend administrator for revision. The revised explanatory text data will be pushed the next time the same question is encountered.
[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than to limit them. Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present disclosure.
Claims
1. A radiation effect database intelligent indexing method based on AI big model, characterized in that: The following steps are included: S1. The administrator sets up multiple tag databases, which together form a corpus; S2, users upload text data containing research data to the corpus; S3, extracting keywords from the text data, and forming corresponding tags according to the extracted keywords. One text data includes one or more tags. According to the formed tags, the text data are respectively stored in the corresponding tag database; S4, the user inputs search information, and the search information input by the user is compared with the text data in the corpus for matching degree; S5. Set a threshold for the matching degree of the text data. When the matching degree is higher than the threshold, list the data with the same label in descending order of matching degree for the user to select. When the matching degree is lower than the threshold, the user's search information is answered through the interpreted text data of the tag.
2. According to claim 1, a radiation effect database intelligent indexing method based on AI large model is characterized in that: The matching degree comparison in step S4 The following steps are included: S41, splitting the search information input by the user into one or more keywords; S42, performing word frequency matching between the split keywords and the labels of the text data in the corpus; S43, calculating the matching degree by using the cosine similarity algorithm.
3. According to claim 2, a radiation effect database intelligent indexing method based on AI large model is characterized in that: The formula of the cosine similarity algorithm is:
4. According to claim 1, a radiation effect database intelligent indexing method based on AI large model is characterized in that: The step of forming the interpreted text data of the tag in step S5 includes: S51, extracting explanation information for the label in each text data; S52, performing similarity matching on the explanation information of the tags in all text data containing the same tags; S53, generating an explanation text for each tag through similarity matching; S54. The manager reviews the generated explanation text to determine whether it is reasonable. If it is reasonable, the explanation text data of the label is finally formed; if it is unreasonable, the manager modifies it and finally forms the explanation text data of the label.
5. According to claim 4, a radiation effect database intelligent indexing method based on AI large model is characterized in that: After the user's search is completed, an answer satisfaction evaluation is sent to the user, and the matching degree between the sent explanation text data and the user's question is determined based on the user's satisfaction.
6. According to claim 5, a radiation effect database intelligent indexing method based on AI large model is characterized in that: The steps for judging the user's satisfaction and the matching degree of the interpreted text data include: S61, setting a satisfaction threshold, if the user's satisfaction exceeds the threshold, it is judged as "satisfied", and if the user's satisfaction is lower than the threshold, it is judged as "dissatisfied"; S62. When multiple users ask the same question, they give the same explanatory text data and set a ratio value. If the number of users who evaluate "satisfied" is higher than or equal to the ratio value, the explanatory text data is judged to be correct. If the number of users who evaluate "satisfied" is lower than the ratio value, the explanatory text data is judged to be wrong.
7. The method for intelligent indexing of radiation effect database based on AI big model according to claim 6 is characterized in that: Interpretation text data that is judged to be erroneous will prompt managers to make revisions.
8. An intelligent indexing system for radiation effect database based on AI big model, characterized in that: include, A corpus, including a plurality of different tag databases, wherein the tag databases store text data containing corresponding tags; Data transmission module, used to upload text data; The data classification processing module is used to perform full-text recognition on the text data uploaded by the user, extract keywords or information from the text data, form corresponding keywords for each text data, and define labels for the text data based on the similarity between the keywords and the labels; The data retrieval module is used to retrieve text data in the corpus and list the retrieved text data with the same label in descending order of matching degree for users to review.
9. The radiation effect database intelligent indexing system based on AI big model according to claim 8 is characterized in that: It also includes an intelligent learning module that forms a summary for each tag database and an explanatory text data for each tag.
10. The radiation effect database intelligent indexing system based on AI big model according to claim 9 is characterized in that: It also includes an AI question-and-answer module, which matches the questions raised by users with the text data in the corpus and sets a similarity threshold. When the matching degree between the questions raised by users and the text data in the corpus is lower than the threshold, the AI question-and-answer module is called, and the AI question-and-answer module calls the interpreted text data of the intelligent learning module to answer the user.