Identifying method for non-standard drugstore names

By building a rule library in a computer and using natural language processing and decision tree algorithms, the standardization of pharmacy names is solved, and the problems of low efficiency and poor accuracy of traditional Chinese medicine store name recognition in the existing technology are solved, and rapid and accurate irregularities are achieved. Pharmacy name identification.

CN119067119AInactive Publication Date: 2024-12-03北京健易保科技有限公司
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202411080180.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-07
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the prior art, the identification of pharmacy names mainly relies on manual judgment, which leads to high work intensity and prone to errors, making it difficult to quickly identify irregular pharmacy names.

Method used

By building a rule library in a computer, performing word frequency analysis and natural language processing, combining decision tree algorithms, automatically identifying the normative nature of pharmacy names, initially screening public irregular names and deeply identifying niche names.

Benefits of technology

It reduces the labor intensity of manual identification and detection, improves the accuracy of identification and judgment, and is suitable for the identification and judgment of a large number of pharmacies.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067119A_ABST
    Figure CN119067119A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of pharmacy supervision, in particular to a non-standard pharmacy name identification method, which comprises the following identification steps that a rule base is constructed in a computer, and the rule base comprises standard specifications of pharmacy names; the method comprises the following steps: acquiring a nonstandard drugstore name sample set, performing word frequency analysis on the sample set to obtain high-frequency vocabularies in the sample set, adding the acquired vocabularies into a rule base, perfecting the rule base, and dividing into a preliminary screening rule and a deep screening rule; and based on a natural language processing model, inputting a to-be-identified pharmacy name and the preliminary screening rule into the model. According to the method, preliminary identification and judgment are performed on the drugstore names, non-standard drugstore names of the public on the market can be identified and screened out through preliminary identification, deep identification is performed on drugstore names of some small people based on the decision tree, the non-standard drugstore names are screened out, the labor intensity of manual identification and detection can be reduced, and the detection efficiency is improved. And the recognition and judgment accuracy is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of drugstore supervision, and specifically relates to a method for identifying an irregular drugstore name. Background Art

[0002] Pharmacies are retail stores that specialize in selling medicines and other health-related products. They not only provide various types of prescription and over-the-counter drugs, but may also include health products, medical devices, personal care products, vitamins, mineral supplements, traditional Chinese medicine, and health consultation services. Pharmacies play an important role in protecting public health because they provide consumers with convenient channels to obtain necessary medicines and health information. Nowadays, with the popularity of pharmacies, large and small pharmacy names have appeared on the streets. The choice of pharmacy name is crucial to the brand image and customer appeal of the pharmacy. A good pharmacy name often attracts more customers. When applying for a pharmacy name, it is necessary to go to the industrial and commercial administrative department to register. When the staff registers, it is necessary to identify the registered pharmacy name to determine whether it is a standardized pharmacy name. In the existing technology, the identification method is based on manual identification and judgment. When a large number of pharmacy names are required, the human workload is high, and it is difficult to quickly determine the non-standard pharmacy names. After a long period of work, it is easy to make mistakes. In this regard, we propose a method for identifying non-standard pharmacy names. Summary of the invention

[0003] In view of the deficiencies of the prior art, the present invention provides a method for identifying non-standard drugstore names to solve the above technical problems.

[0004] To achieve the above object, the present invention provides the following technical solution: a method for identifying irregular drugstore names, the identification steps are:

[0005] Building a rule base in a computer, which includes standard specifications for drugstore naming;

[0006] Obtain a sample set of non-standard drugstore names, perform word frequency analysis on the sample set, obtain high-frequency words in the sample set, add the obtained words to the rule base, improve the rule base, and divide it into preliminary screening rules and in-depth screening rules;

[0007] Based on the natural language processing model, the name of the pharmacy to be identified and the preliminary screening rules are input into the model, and compared with the vocabulary in the rule library based on cosine similarity calculation, and similar glyphs are output after recognition, and the output is judged as the irregular name to be determined and the irregular name;

[0008] Based on the decision tree, in-depth screening and identification are carried out, and the output is continuously analyzed. If it meets the naming standards, the output is a qualified pharmacy name. If it does not meet the naming standards, the output is an irregular pharmacy name, and the identification results are stored in the computer.

[0009] Preferably, the standard specifications for drugstore naming include industry specifications and drugstore naming specifications, and the standard specifications are obtained based on big data and input into the rule base.

[0010] Preferably, a sample set is constructed based on the non-standard drugstore names screened out historically, the sample set is cleaned to remove duplicate drugstore names, and then the names of the sample set are segmented. After segmentation, the frequency of occurrence of the words is analyzed by word frequency analysis.

[0011] Preferably, the word frequency analysis formula is:

[0012]

[0013] Among them, TF is the word frequency, d is the specific word, dt is the number of times the specific word appears in the sample set, and dz is the total number of words that appear in the sample set. The word frequency threshold in the sample set is based on the independent definition of the supervisor, and a specific threshold is directly given. The threshold is adjusted according to the number of samples in the sample set. The word frequency value of each word in the sample set is calculated and compared with the established threshold. Words greater than the threshold are defined as popular words, and words less than the threshold are defined as niche words. The obtained popular words are added to the rule base to form preliminary screening rules.

[0014] Preferably, the names of the pharmacies to be identified are preprocessed, stop words and word segmentation are removed, and the preprocessed words are vectorized to obtain vector words A (A1, A2, ... An) of the words to be analyzed and vector words B (B1, B2, ... Bn) to be compared, which are calculated by the following formula:

[0015]

[0016] Where cosY is the similarity between two vector words, A·B is the dot product of two vector words, and ||A|| and ||B|| represent the lengths of vector word A and vector word B.

[0017] Preferably, the numerical value learned from the text data represents the characteristics of the vocabulary, wherein the calculated cosY value is equal to 1, indicating that the two vector vocabularies are completely equal, and if the cosY value is not equal to 1, it indicates that the two vector vocabularies are not equal, and a comparison is performed. If they are consistent, the output is a non-standard name, and if they are inconsistent, the output is a non-standard name to be determined.

[0018] Preferably, the in-depth screening rules include avoiding the use of uncommon characters, ambiguous words and exaggerated words. Based on the decision tree algorithm, in-depth identification is performed. The screening rule word library in the rule library is input into the node on the decision tree as a matching basis. The name of the non-standard pharmacy to be determined is input into the decision tree, and the information entropy value at each node is calculated. The formula is:

[0019]

[0020] Where HD is the information entropy value, p(x i ) represents vocabulary x i The probability in D, where D is the vocabulary in the rule base and n is the number of vocabulary categories.

[0021] Preferably, the information entropy value of the vocabulary in the rule base is obtained based on calculation, wherein the smaller the information entropy is, the lower the probability that the vocabulary in the non-standard name to be determined is retrieved in the rule base is, and a threshold of the node is established based on the confidence level, and the information entropy calculated at the node is compared with the threshold. If it is greater than the threshold, the output is a non-standard vocabulary. If it is less than the threshold, the calculation and comparison are continued until the output is a standardized pharmacy name after passing through each node.

[0022] Preferably, the confidence calculation formula is:

[0023]

[0024] Where K is the confidence interval of information entropy, under is the lower bound of confidence, upper is the upper bound of confidence, a is the mean of sample vocabulary in the vocabulary library, c is the standard deviation of sample vocabulary in the vocabulary library, and N is the square root of the number of samples. A sample set of pharmacy names with standardized names is obtained, and the confidence interval of the vocabulary in the sample set is calculated to establish the threshold of the node.

[0025] Preferably, the screened non-standard drugstore names are stored based on cloud storage, and the data is distributed on storage nodes through data distribution strategy. The stored non-standard drugstore names are used as sample sets for subsequent word frequency analysis to obtain words that appear frequently in the non-standard drugstore names.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] The present application obtains a sample set of non-standard drugstore names in advance, filters out high-frequency words, and performs preliminary recognition and judgment on the drugstore names. The preliminary recognition can identify and filter out the non-standard drugstore names on the market, and for some niche drugstore names, in-depth recognition is performed based on decision trees to filter out non-standard drugstore names. This can reduce the labor intensity of manual recognition and detection, and improve the accuracy of recognition and judgment. It is suitable for occasions where a large number of drugstore names are recognized and judged. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A flowchart of the framework of identification steps of the present invention;

[0029] Figure 2 Establishing a flow chart for the preliminary screening rules of the present invention;

[0030] Figure 3 This is a flowchart of the preliminary screening of the names of pharmacies to be identified in the present invention;

[0031] Figure 4 This is a decision tree diagram for the non-standard drugstore names to be determined in the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not 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.

[0033] Reference Figure 1 As shown, a method for identifying irregular drugstore names, the identification steps are:

[0034] Building a rule base in a computer, which includes standard specifications for drugstore naming;

[0035] Obtain a sample set of non-standard drugstore names, perform word frequency analysis on the sample set, obtain high-frequency words in the sample set, add the obtained words to the rule base, improve the rule base, and divide it into preliminary screening rules and in-depth screening rules;

[0036] Based on the natural language processing model, the name of the pharmacy to be identified and the preliminary screening rules are input into the model, and compared with the vocabulary in the rule library based on cosine similarity calculation, and similar glyphs are output after recognition, and the output is judged as the irregular name to be determined and the irregular name;

[0037] Based on the decision tree, in-depth screening and identification are carried out, and the output is continuously analyzed. If it meets the naming standards, the output is a qualified pharmacy name. If it does not meet the naming standards, the output is an irregular pharmacy name, and the identification results are stored in the computer.

[0038] The standard specifications for pharmacy naming include industry specifications and pharmacy naming specifications. The standard specifications are obtained based on big data and input into the rule base.

[0039] The present application pre-constructs a rule library of naming standards, obtains the naming standards of drugstores through the Internet, and conducts word frequency analysis on non-standard drugstore names by constructing a sample set to obtain high-frequency words, and conducts comparative analysis based on the high-frequency words to preliminarily screen the drugstore names. This step can screen out most of the non-standard words, thereby reducing the workload of screening, and further conducts in-depth screening for some niche drugstore names, screening through decision trees, and further identifying and screening out the niche drugstore names, which can improve the efficiency of recognition, reduce the inconvenience of manual recognition, and is suitable for large-scale drugstore name recognition and judgment.

[0040] Reference Figure 2 As shown, a sample set is constructed based on the irregular drugstore names screened out historically, the data of the sample set is cleaned to remove duplicate drugstore names, and then the names of the sample set are segmented. After segmentation, the frequency of occurrence of the words is analyzed through word frequency analysis.

[0041] The word frequency analysis formula is:

[0042]

[0043] Among them, TF is the word frequency, d is the specific word, dt is the number of times the specific word appears in the sample set, and dz is the total number of words that appear in the sample set. The word frequency threshold in the sample set is based on the independent definition of the supervisor, and a specific threshold is directly given. The threshold is adjusted according to the number of samples in the sample set. The word frequency value of each word in the sample set is calculated and compared with the established threshold. Words greater than the threshold are defined as popular words, and words less than the threshold are defined as niche words. The obtained popular words are added to the rule base to form preliminary screening rules.

[0044] The data cleaning, deduplication and word segmentation processing steps in this application are all existing technologies and will not be elaborated on here. The establishment of the threshold is based on the needs of the regulator to set a reasonable word frequency threshold. The threshold is dynamic and can be dynamically adjusted according to the situation of the sample set. By obtaining high-frequency words and performing preliminary screening, the non-standardized pharmacy names can be identified more quickly, and high-frequency non-standardized pharmacy names can be screened out to reduce the workload of manual review.

[0045] Reference Figure 3 As shown, the names of the pharmacies to be identified are preprocessed, stop words and word segmentation are removed, and the preprocessed words are vectorized to obtain vector words A (A1, A2, ... An) of the words to be analyzed and vector words B (B1, B2, ... Bn) to be compared, which are calculated by the following formula:

[0046]

[0047] Where cosY is the similarity between two vector words, A·B is the dot product of two vector words, and ||A|| and ||B|| represent the lengths of vector word A and vector word B.

[0048] The numerical values ​​learned from the text data represent the characteristics of the vocabulary. If the calculated cosY value is equal to 1, it means that the two vector vocabularies are completely equal. If the cosY value is not equal to 1, it means that the two vector vocabularies are not equal. A comparison is performed. If they are consistent, the output is a non-standard name. If they are inconsistent, the output is a non-standard name to be determined.

[0049] The preprocessing of the drugstore name in this application is a prior art and will not be elaborated on here. The vector lexicalization processing is based on SVD singular value decomposition. This method maps the words into a low-dimensional vector space by analyzing the co-occurrence relationship of the words in a large amount of text data, thereby capturing the semantic and grammatical relationship between the words and generating the vector of the words. The dot product of vector words A and B can be used to measure the similarity or correlation between the two words. Specifically, the larger the result of the dot product, the closer or more related the two words are semantically. This is because the dot product reflects to some extent the degree of proximity of the two vectors in direction and the sum of the product of their values ​​in their respective dimensions. The length of the vector word does not directly reflect the semantic relationship between the words, but reflects the "size" or "strength" of the vector in the multidimensional space. However, in some cases, the length of the vector may be related to certain attributes of the words, such as frequency, importance or activity. The similarity of the two groups of words is measured by the formula to determine the standardization of the drugstore name.

[0050] Reference Figure 4 As shown, the in-depth screening rules include avoiding the use of rare characters, ambiguous words and exaggerated words. Based on the decision tree algorithm, in-depth identification is performed. The screening rule word library in the rule library is input into the nodes on the decision tree as the matching basis. The name of the non-standard pharmacy to be determined is input into the decision tree, and the information entropy value at each node is calculated. The formula is:

[0051]

[0052] Where HD is the information entropy value, p(x i ) represents vocabulary x i The probability in D, where D is the vocabulary in the rule base and n is the number of vocabulary categories.

[0053] The information entropy value of the vocabulary in the rule base is obtained based on calculation, where the smaller the information entropy, the lower the probability of the vocabulary in the non-standard name to be determined being retrieved in the rule base. The threshold of the node is established based on the confidence level, and the information entropy calculated at the node is compared with the threshold. If it is greater than the threshold, the output is a non-standard vocabulary. If it is less than the threshold, the calculation and comparison will continue until the output is a standardized pharmacy name after passing through each node.

[0054] The confidence calculation formula is:

[0055]

[0056] Where K is the confidence interval of information entropy, under is the lower bound of confidence, upper is the upper bound of confidence, a is the mean of sample vocabulary in the vocabulary library, c is the standard deviation of sample vocabulary in the vocabulary library, and N is the square root of the number of samples. A sample set of pharmacy names with standardized names is obtained, and the confidence interval of the vocabulary in the sample set is calculated to establish the threshold of the node.

[0057] Based on cloud storage, the screened out non-standard drugstore names are stored, and the data is distributed on the storage nodes through data distribution strategy. The stored non-standard drugstore names are used as sample sets for subsequent word frequency analysis to obtain the words that appear frequently in the non-standard drugstore names.

[0058] In this application, the screening rules are converted into specific matching criteria and input into each node of the decision tree. Each node represents a specific screening condition or rule. Through step-by-step matching and judgment, the standardization of the pharmacy name is finally determined. The pharmacy name to be determined is input into the decision tree, and step-by-step matching and judgment are performed starting from the root node. After matching and judging all nodes, if no non-standard words are found, the standardized pharmacy name is output. The mathematical method of the decision tree algorithm can more accurately identify and standardize the pharmacy name. The automated screening process reduces the workload of manual review and improves work efficiency. It can be updated and adjusted according to new screening rules or data to maintain the timeliness and accuracy of the screening rules. Based on data analysis results such as word frequency analysis, more scientific and reasonable pharmacy name naming specifications and management strategies can be formulated. The cloud storage platform provides efficient and secure data storage and access services, which provides strong support for subsequent data analysis and application.

[0059] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0060] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for identifying irregular drugstore names, characterized in that: The identification steps are: Building a rule base in a computer, which includes standard specifications for drugstore naming; Obtain a sample set of non-standard drugstore names, perform word frequency analysis on the sample set, obtain high-frequency words in the sample set, add the obtained words to the rule base, improve the rule base, and divide it into preliminary screening rules and in-depth screening rules; Based on the natural language processing model, the name of the pharmacy to be identified and the preliminary screening rules are input into the model, and compared with the vocabulary in the rule library based on cosine similarity calculation, and similar glyphs are output after recognition, and the output is judged as the irregular name to be determined and the irregular name; Based on the decision tree, in-depth screening and identification are carried out, and the output is continuously analyzed. If it meets the naming standards, the output is a qualified pharmacy name. If it does not meet the naming standards, the output is an irregular pharmacy name, and the identification results are stored in the computer.

2. The method for identifying irregular drugstore names according to claim 1, characterized in that: The standard specifications for pharmacy naming include industry specifications and pharmacy naming specifications. The standard specifications are obtained based on big data and input into the rule base.

3. The method for identifying irregular drugstore names according to claim 1, characterized in that: A sample set is constructed based on the irregular drugstore names screened out historically. The data in the sample set is cleaned to remove duplicate drugstore names. The names in the sample set are then segmented. After segmentation, the frequency of word occurrence is analyzed through word frequency analysis.

4. The method for identifying irregular drugstore names according to claim 3, characterized in that: The formula for word frequency analysis is: Among them, TF is the word frequency, d is the specific word, dt is the number of times the specific word appears in the sample set, and dz is the total number of words that appear in the sample set. The word frequency threshold in the sample set is based on the independent definition of the supervisor, and a specific threshold is directly given. The threshold is adjusted according to the number of samples in the sample set. The word frequency value of each word in the sample set is calculated and compared with the established threshold. Words greater than the threshold are defined as popular words, and words less than the threshold are defined as niche words. The obtained popular words are added to the rule base to form preliminary screening rules.

5. The method for identifying irregular drugstore names according to claim 1, characterized in that: The names of the pharmacies to be identified are preprocessed to remove stop words and word segmentation, and the preprocessed words are vectorized to obtain vector words A (A1, A2, ... An) of the words to be analyzed and vector words B (B1, B2, ... Bn) to be compared, which are calculated using the following formula: Where cosY is the similarity between two vector words, A·B is the dot product of two vector words, and ||A|| and ||B|| represent the lengths of vector word A and vector word B.

6. The method for identifying irregular drugstore names according to claim 5, characterized in that: The numerical values ​​learned from the text data represent the characteristics of the vocabulary. If the calculated cosY value is equal to 1, it means that the two vector vocabularies are completely equal. If the cosY value is not equal to 1, it means that the two vector vocabularies are not equal. A comparison is performed. If they are consistent, the output is a non-standard name. If they are inconsistent, the output is a non-standard name to be determined.

7. The method for identifying irregular drugstore names according to claim 1, characterized in that: In-depth screening rules include avoiding the use of rare characters, ambiguous words, and exaggerated words. Based on the decision tree algorithm, in-depth identification is performed. The screening rule word library in the rule library is input into the nodes on the decision tree as the matching basis. The name of the non-standard pharmacy to be determined is input into the decision tree, and the information entropy value at each node is calculated. The formula is: Where HD is the information entropy value, p(x i ) represents the word x i The probability in D, where D is the vocabulary in the rule base and n is the number of vocabulary categories.

8. The method for identifying irregular drugstore names according to claim 7, characterized in that: The information entropy value of the vocabulary in the rule base is obtained based on calculation, where the smaller the information entropy, the lower the probability of the vocabulary in the non-standard name to be determined being retrieved in the rule base. The threshold of the node is established based on the confidence level, and the information entropy calculated at the node is compared with the threshold. If it is greater than the threshold, the output is a non-standard vocabulary. If it is less than the threshold, the calculation and comparison will continue until the output is a standardized pharmacy name after passing through each node.

9. The method for identifying irregular drugstore names according to claim 8, characterized in that: The confidence calculation formula is: Where K is the confidence interval of information entropy, under is the lower bound of confidence, upper is the upper bound of confidence, a is the mean of sample vocabulary in the vocabulary library, c is the standard deviation of sample vocabulary in the vocabulary library, and N is the square root of the number of samples. A sample set of pharmacy names with standardized names is obtained, and the confidence interval of the vocabulary in the sample set is calculated to establish the threshold of the node.

10. The method for identifying irregular drugstore names according to claim 1, characterized in that: Based on cloud storage, the screened out non-standard drugstore names are stored, and the data is distributed on the storage nodes through data distribution strategy. The stored non-standard drugstore names are used as sample sets for subsequent word frequency analysis to obtain the words that appear frequently in the non-standard drugstore names.

Citation Information

Patent Citations

  • Disease name standardization specification method and specification system

    CN108922633A

  • power equipment name identification method

    CN109800416A

  • Entity recognition model hot updating method and device, equipment, medium and product

    CN114186013A

  • Security analysis and automatic evaluation method based on index threshold value and semantic analysis

    CN114528848A

  • Standardized name determination method, system, equipment and medium

    CN116484854A