KR102578899B1 - Method of expanding queries for searching medical content - Google Patents

Abstract

The query expansion method for retrieving medical content includes the following steps: (a) comparing the total search volume reflecting the number of primary searches as a search result using an already selected primary query with a predetermined passing standard; (b) comparing the total search volume to a predetermined passing standard; If not satisfied, (c) selecting a secondary query field containing a secondary query derived from the primary query, (c) calculating a new total search volume reflecting the number of secondary searches by the secondary query field, and (d) ) It may include comparing the total search amount in step (c) above with the passing criteria.

Description

{METHOD OF EXPANDING QUERIES FOR SEARCHING MEDICAL CONTENT}

The present invention relates to a query expansion method that can improve search quality in the process of searching medical-related content.

With the spread of the Internet, people can find useful information through Internet searches. For example, people access portal sites to search for information about famous restaurants, comfortable hotels, the latest home appliances, etc., search blogs and user reviews, and refer to the number of 'likes' to find items they think are useful. Get information.

However, information in certain professional fields contains elements that cannot be evaluated through information such as user reviews or 'likes', and when evaluated in that way, there are things that are disadvantageous or dangerous to consumers. One such example is information about doctors or hospitals. This is because evaluation of disease diagnosis, treatment, procedure, or surgery is in a specialized area where the general public cannot provide advice, and sometimes such evaluation is not related to the actual skills or capabilities of the doctor or hospital, but is based on completely unrelated factors. For example, it can be influenced by the friendliness of the doctor, the temporary feeling after medical treatment, the gender of the doctor, the attitude of the nurse, the time of the hospital appointment, and the cleanliness of the hospital.

According to an overseas survey, it was reported that insurance coverage, hospital location, etc. are considered more important than the number of surgeries performed by the surgeon and the reliability of the medical staff when selecting a surgeon (jamasurgery.com, 2014, 2) month).

Recently, there are websites (Physician Rating Websites: PRWs) or organizations that allow you to rate doctors and surgeons, examples of which include Yelp.com, Healthgrade.com, Insurance Plan, and Angie's List. However, these websites or organizations may also have limitations in that they only consider reviews, evaluation scores from experienced people, and regional distance when evaluating doctors and their abilities.

Looking at Chinese Published Patent CN 104680458 A, a method for recommending a hospital or doctor is disclosed. The Chinese published patent collects all prescriptions for each doctor, obtains treatment effects through various methods such as phone contact, networks, and visits after a certain period of time, and integrates them into a database to recommend medical subjects to the public. However, this method requires a lot of time and control, it is impossible to collect all prescriptions, and the possibility of personal information leakage is high, making it impossible to implement.

Korean Patent No. 10-1928533 relates to “a method of recommending medical subjects based on quantitative physician evaluation using academic achievements,” and includes information about searching paper data, including the number of papers and publication time data for each paper. , it includes the use of information such as first author information, corresponding or responsible author information, co-author information, included journal information, journal impact factor, and physician h-index.

However, when the number of retrieved papers is relatively small, evaluating and selecting medical subjects with only a small number of papers may be inaccurate, and when related fields are segmented, such as the medical field, the search results do not meet minimum expectations. In some cases, this problem may occur from time to time.

The present invention provides a query expansion method for quantitatively obtaining searched medical contents in a certain amount or above a certain standard in the process of utilizing search results according to a query.

The present invention provides a method of expanding a query according to reasonable criteria rather than simple expansion.

According to an exemplary embodiment of the present invention for achieving the purposes of the present invention described above, the query expansion method for retrieving medical content includes (a) reflecting the number of primary searches as a search result using an already selected primary query; Comparing the search total amount with a predetermined passing standard, (b) If the search total amount does not meet the passing standard, selecting a secondary query field containing the secondary query derived from the primary query, (c) 2 It may include the step of calculating a new total search amount reflecting the number of secondary searches by the secondary query field, and (d) comparing the total search amount in step (c) with the passing criteria.

For reference, in this specification, a 'search word' may be a word or a combination of words or characters provided directly by a user connected to the server or another server to obtain results, and a 'query' may be a word or a combination of words provided by a user accessing the server or another server to obtain results, and a 'query' may be a word or a combination of words provided by a user connected to the server or another server to obtain results. Alternatively, the database may be words, terms, or combinations of words selected as medical terms that the database can recognize. Depending on the server, search terms can be used directly as queries, but in others, other queries can be added, changed, or selected to use the database.

In this specification, the server can select a primary query from the search word, and if it is determined that the total search volume reflecting the search by the primary query is insufficient, it can additionally select a secondary query derived from the primary query.

In the present invention, the server can use its own database or an external database, and the database external to the server is Pubmed, Google scholar, SCOPUS, Naver full-text search, and NDSL, which can be connected through a network. , KCI, RISS, Altmetric, Mendeley, PlumX metrics and other metadata services, and may also use external databases with similar purposes and functions.

If the total search volume in step (d) above satisfies the passing criteria, query selection for medical content search can be completed.

However, in some cases, the total search volume in step (c) may not be sufficient. If the total search volume in step (d) does not meet the passing criteria, the secondary query is defined as a new primary query ( Steps a) to (d) can be repeated.

Secondary queries can be extended or derived based on various criteria.

As an example, the secondary query may be selected within a range where the primary query is classified according to its root and is a combination of the classified roots. For example, if the total number of searches reflecting the number of primary searches for “arthroscopic meniscus restoration” as the primary query does not meet the predetermined passing criteria, the root of the primary query is changed to arthroscopic, meniscus, restoration, etc. They can be classified together, and secondary query fields such as arthroscopic meniscus, arthroscopic reconstruction, meniscus restoration, arthroscopy, meniscus, reconstruction, etc. can be selected as secondary queries according to the classified roots. .

As another example, the secondary query classifies the primary query to include at least one of symptoms, diagnosis, treatment, and area, and is selected within the range consisting of a combination of similar words or superordinate concepts of the classified symptom, diagnosis, treatment, or area. You can. For example, if the total number of searches reflecting the number of primary searches for "finger sprain" as the primary query does not meet the predetermined passing criteria, the primary query can be classified according to symptoms, diagnosis, treatment, or area, ' 'Finger' is judged as a part, and 'finger', 'index finger', 'thumb', 'hand', etc. are classified as analogs or superordinate concepts of the part, and 'sprain' is judged as a symptom and 'sprain' is classified as a synonym or superordinate concept of the symptom. After classifying 'damage', etc., secondary query fields such as finger sprain, hand sprain, finger damage, hand damage, etc., which are combinations of these, can be selected as secondary queries.

High-level concepts of symptoms, diagnosis, and treatment are included in known classification systems (e.g., International Classification of Diseases, ICD-11; Diagnostic and Statistical Manual of Mental Disorders. DSM-5; International Classification of Primary Care (ICPC); International Classification of Procedures in Medicine (ICPM); International Classification of Health Interventions (ICHI); SNOMED International, SNOMED Clinical Terms, etc.) or independently developed systems can be used, and the higher-level concept of the region is anatomical system, etc. can be used.

The predetermined passing standard is the number of searches that can provide reasonable and objective results as search results. The same standard may be applied to all queries, but may have different standards depending on the query.

For example, as a simple passing criterion, it can be determined whether the search total amount is above or around a certain constant (ex. 500, 1000).

As another example, the passing standard may be determined by reflecting the trend of the total search volume, and whether or not the passing standard has been reached can be determined based on changes in the total search volume. Possible methods include when the trend straight line of total search volume increases significantly, when the trend changes from a straight line to a curve (J curve, etc.), when the trend curve changes back to a straight line, or when a steep straight line or curve changes to a gentle one. You can.

For example, depending on changes in searched content over time, expanded search scope, or changes in symptoms, diseases, treatments, and stages of expanded queries, the total number of searches may increase to a certain number (e.g. 500, 1000 or more) or a change of more than a certain percentage (eg. 50%) compared to before expansion can be set as a passing criterion.

Even if the total search volume is expanded by secondary queries, it may be desirable not to judge it equally. Therefore, when the query is expanded according to the present invention, for the medical content retrieved by the expanded secondary query, the score of the medical content is calculated by considering the number of secondary searches or the characteristics of the secondary query field and the score of the medical content by the primary query. It may be necessary to calculate it differently.

For example, the score of medical content retrieved by secondary query uses a search range correction weight (r) that reflects the number of secondary searches, a query field characteristic weight (f) that reflects the characteristics of the secondary query field, and a combination thereof. It can be calculated as follows.

The score of the medical content retrieved by the secondary query can be calculated using the secondary query correction coefficient (QF _i ) below, where QF _i is the correction coefficient of the ith secondary query among the secondary query fields, r _i may be the search range correction weight of the ith secondary query, f _i may be the query field correction weight of the ith secondary query, and α _i may be other weights.

As an example, QF _i can be calculated by multiplying by r _i xf _i xα _i .

The score of the medical content retrieved by the secondary query can _be calculated as follows, where P _i is the score before correction of the medical content retrieved by the ith secondary query, and It may be a score after correction of medical content.

As an example, X _QFi can be calculated by multiplying by QF _i x P _i xα _i .

Other weights (α _i ) may reflect at least one of the completeness of the root, the type of medical content, the user's characteristics, and whether the query used is expanded into MeSH terms.

The completeness of the root can correspond to the completeness of the search for the query (Q). For example, when "shoulder tendon rupture" is entered as a search term by the search subject, the server, etc. may select "rotator cuff rupture" as the first or second query (Q) among the queries that can be selected as medical terms, In the medical content retrieved by the above query (Q), the case where the roots 'rotator cuff' and 'rupture' exist separately, and the case where they exist as a combination of the roots 'rotator cuff tear' should be evaluated differently. . In this case, when weighting (α _i ) is given according to the integrity or root completeness, the case where 'rotator cuff tear' is searched more completely than when 'rotator cuff' and 'rupture' are searched separately (ex. α _i = 1) ((ex. α _i =10) can be given higher.

For example, the type weight of medical content may be defined the same or differently depending on whether the medical content is a paper, patent, abstract, presentation, opinion paper, research plan, editorial, or contribution.

If the medical content is a paper, it is judged differently where the first and second queries were searched among the section information that constitutes the medical content, that is, title, keywords, MeSH terms, abstract, data, results of the main text, discussion, methods, and introduction. This can be reflected.

According to the query expansion method of the present invention, the objectivity of the evaluation score can be improved by securing a sufficient amount of medical contents.

However, considering that the medical field and medical terminology are unfamiliar to general users, you may go through a process of selecting the primary query rather than using the search term as is, and in this process, you may use medical terminology or disease categories. .

According to the query expansion method of the present invention, in addition to simply expanding the query, the query can be expanded by reasonable standards according to medical fields such as symptoms, treatment, diagnosis, and site.

The query expansion method of the present invention is a standard for selecting medical content, and the evaluation score of the medical content can also reflect the standard when expanding the query so that the results of the medical content related to the query are more substantially reflected.

According to this query expansion method, medical target information can be provided accurately, and the reliability of medical content evaluation can be increased by assigning weight to highly relevant results or relatively low weight to search results resulting from expansion. .

Figure 1 is a diagram illustrating a server and network relationship for performing a query expansion method and a method of providing medical subject information using the same according to an embodiment of the present invention.
Figure 2 is a diagram for explaining the process of a query expansion method and a method of providing medical subject information using the same according to an embodiment of the present invention.
Figure 3 is a diagram for explaining a query expansion method and a medical content evaluation process using the same according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings, but the present invention is not limited or limited by the embodiments. For reference, in this description, the same numbers refer to substantially the same elements, and under the above rules, content written in other drawings may be cited and explained, and content deemed obvious to those skilled in the art or repeated may be omitted.

Figure 1 is a diagram illustrating a server and network relationship for performing a query expansion method and a method of providing medical subject information using the same according to an embodiment of the present invention, and Figure 2 is a query expansion according to an embodiment of the present invention. This is a diagram for explaining the method and the process of providing medical subject information using the same, and the query expansion method according to an embodiment of the present invention in Figure 3 is a diagram for explaining the process of evaluating medical content using the method.

For reference, in explaining the query expansion method according to the present invention, the process of providing medical information is used, but the query expansion method of the present invention is not limited to providing medical information, and is not limited to other processes or systems that search for meaningful medical content. It can also be applied.

Referring to FIGS. 1 to 3, the user or accessor can access the server 110 through a smartphone 10 or a PC computer 20 by logging in or not logging in (S120), and the accessor can access the server 110 (S120). 110), a search term can be entered while connected to the network, and the server 110 can receive a search term from the user (S130).

The server 110 may store the search term received from the user in the database 114, etc., and the search unit 113 of the server may use the search term as is, but may select a primary query to collect specialized information from the search term. there is.

For reference, in this specification, a 'search word' may be a word or a combination of words or characters provided directly by a user connected to the server or another server to obtain results, and a 'query' may be a word or a combination of words provided by a user accessing the server or another server to obtain results, and a 'query' may be a word or a combination of words provided by a user connected to the server or another server to obtain results. Alternatively, the database may be words, terms, or combinations of words selected as medical terms that the database can recognize. Depending on the server, search terms can be used directly as queries, but in others, other queries can be added, changed, or selected to use the database.

In the process of converting the user's search term into a primary query, medical-related high-frequency search terms can be extracted and secured from search engines such as Naver, Kakao, and Google, and Dabin can be used in the medical object search service using medical object patents using academic achievements. Search terms can also be secured, and a sufficient number of secured search words can be converted into possible queries as primary queries. For high-frequency search words that are continuously accumulated, servers, etc. can use the relationship between secured search words and queries. It can be converted manually, semi-automatically, and automatically using methods including deep learning.

For example, when a user enters a search term such as “shoulder tendon rupture,” “shoulder rotator cuff tear surgery,” or “shoulder injury surgery,” these terms may include natural terms or expressions that are not used in medical terminology. .

Servers can select various combinations of primary queries based on medical terms and roots, such as rotator cuff tear, from search terms.

Based on this selected primary query, patient information can be searched or paper data can be searched and extracted through an external database. The calculation unit 112 of the server can use patient information to reflect the weights described later or calculate academic scores in advance by taking paper data into consideration. Then, an evaluation score is calculated in consideration of the academic score, target score, review score, and their weights stored in the database 114, and at least one medical target is recommended to the accessor or a medical target according to the final or temporary evaluation score. Evaluation information can be provided.

For this purpose, the server 110 may include a calculation unit 112, a search unit 113, and a database 114. The server 110 may further include a communication unit 115 for transmitting and receiving data to and from the outside, and the communication unit 115 receives a search term from a connecter or transmits a selected primary query and then performs a search from the external database 30. You can receive the results. Additionally, the communication unit 115 can transmit search keywords according to the primary query to the external database 30, and provide medical target recommendation results according to evaluation scores to the connected person in real time.

For reference, in this embodiment, the user connects to the server 110 and enters a search term, but in other embodiments, another server or program interface, such as an API, transmits the search term to the server to provide medical treatment. It can also be used in a method of providing information on medical subjects using medical content in the process of requesting a score or a list of medical subjects according to the scores and providing the requested results.

The external database 30 can be a public data server, a private data server, or a blockchain, and if it is to view patient information, etc., the target is the National Health Insurance Service, Health Insurance Review and Assessment Service, private insurance company, national/national/medical insurance company, etc. It can be a public hospital, a private hospital, a private or commercial application, a cloud database, or a blockchain. To search medical-related papers, databases with real-time access include Pubmed, Scopus, and Google Scholar. , SCOPUS, etc., and as domestic search sites, you can use Naver specialized search, NDSL, KCI, RISS, etc., and you can also use Altmetric and Mendeley metadata services. there is. In addition, to calculate academic scores, data such as citations, usage, capture, mentions, and social media usage from newspapers, magazines, radio, TV, Internet portals, social media, and social network services can be used, as well as Altmetric and PlumX metrics. You can also use the same metadata service.

In this specification, a medical subject can be understood as including a doctor, physician, surgeon, medical staff, or hospital, and the score evaluated according to the conditions desired by the user and the results processed accordingly can be used to refer to a specific doctor, medical staff, or hospital. Alternatively, it can be provided by the name of the hospital, etc.

If the medical subject is an individual, the final evaluation score can be compiled based on the academic performance of the individual doctor. If the medical subject is a medical staff, the academic performance of the doctors included in the medical staff can be collected. In the case of a hospital, the academic achievements of doctors belonging to the hospital can be collected.

Referring to FIG. 2, according to the method of providing medical subject information according to this embodiment, the server 110 creates a database 114 that distinguishes academic scores (AP), subject scores (PP), and review scores (RP) for medical subjects. ) can be maintained (S110). When a user accesses the server 110 (S120) and enters a search term, the server 110 receives the search term (S130), and the server 110 selects a primary query suitable for medical terms, etc. based on the search term. (S140).

Based on the primary query, medical content stored in a database included in the server or a database external to the server is searched, and the total number of searches (1) can be calculated by reflecting the number of primary searches (S150). Here, the search total (1) reflecting the number of primary searches can be simply added to the number of primary searches, but in some cases, duplicate searched medical content is excluded or the weight according to the primary query is reflected in the search total (1). can do.

When the search total amount (1) is calculated, the server 110 can check whether the search total amount (1) satisfies a specific passing criterion (S160). If the total search volume (1) satisfies the passing criteria just by reflecting the number of primary searches, secondary query selection in relation to medical content search can be omitted.

However, as in this embodiment, if the total search amount (1) based on the number of primary searches does not meet the passing criteria, a secondary query field including the secondary query can be selected (S170), and the secondary query A new search total (2) can be calculated by reflecting the number of secondary searches (S180). As described above, this new total search amount (2) can reflect the number of secondary searches by simply adding the number of secondary searches or using other methods as described above.

If the total search volume (2) satisfies the passing criteria through the secondary query field, query selection for medical content search can be completed through secondary query selection.

However, in some cases, the total search volume (2) using the secondary query may not be sufficient, and in this case, an additional tertiary query may need to be selected. The process of selecting the third query may be similar to the method of selecting and searching the second query from the first query. Therefore, logically, the process of substituting the secondary query into a new primary query and selecting a new secondary query from the new primary query (S170) can be repeated. Through this process, it is possible to repeatedly select a new secondary query, calculate the total search volume, and compare the calculated total search volume with the passing criteria.

Secondary queries can be expanded or derived based on various criteria, but in this embodiment, the secondary query fields can be classified according to roots and selected as a combination of classified roots.

Referring to Figure 3, it can be assumed that "arthroscopic meniscus restoration" was used as the first query and that the total search volume reflecting the number of first searches did not meet the passing criteria.

In this case, the root of the first query can be classified as arthroscopy, meniscus, reconstruction, etc., and knee can be classified as a higher-level concept of the region. Depending on the classified roots and high-level concepts, secondary query fields such as meniscus restoration, arthroscopic restoration, meniscus, knee, etc. can be selected as secondary queries.

By using the secondary query field where meniscus restoration, arthroscopic restoration, meniscus, and knee are selected as secondary queries, medical content can be searched again through the external database mentioned above.

The number of searches for each secondary query can be searched as follows. When searching for 'arthroscopic meniscus restoration surgery' as the first query, about 900 results were found, and when searching for 'meniscus restoration surgery' as a second query under the same conditions, about 2,700 results were found for 'arthroscopic restoration surgery'. When you search, you can get about 6,300 results, when you search for 'meniscus', you can get about 15,000 results, and when you search for 'knee', which is the higher level concept of a simple part, you can get about 180,000 results.

As a result of the secondary query field, the number of secondary searches may be more than 200,000, which can be used as a result that exceeds the predetermined passing standard. However, when searching for 'knee', the approximately 180,000 medical contents retrieved were searched with the primary query 'arthroscopic meniscus restoration surgery', and the search results and the secondary query 'meniscus restoration surgery' or 'arthroscopic restoration surgery' ', it is necessary to judge based on different criteria such as the search results.

Even if the total search volume is expanded by secondary queries, it may be desirable not to judge it equally. Therefore, in this embodiment, when the secondary query is expanded, the score of the medical content is calculated based on the primary query by considering the number of secondary searches or the characteristics of the secondary query field for the medical content retrieved by the expanded secondary query. It can be calculated differently from medical content.

Looking at the content added to FIG. 3, the search range correction weight (r) that reflects the number of secondary searches in the score of the medical content searched by the secondary query field, and the query field characteristic weight (f) that reflects the characteristics of the secondary query field. can be reflected.

For example, as a result of searching for 'meniscus restoration surgery' as a query among the secondary queries, when about 2,700 cases are searched, the search range correction weight (r ₁ ) may be 1/(3-1), which is about 0.5, The search range correction weight (r ₂ ) for ‘arthroscopic reconstruction’ is approximately 0.166, the search range correction weight (r ₃ ) for ‘meniscus’ is approximately 0.063, and the search range correction weight (r ₄ ) for ‘knee’ is approximately 0.063. It can be calculated as 0.005.

In addition, the query field characteristic weight is also defined as relatively high (ex. 1) for specific names (ex. arthroscopic meniscus) and relatively low (ex. 0.3 or less) for general names (ex. restorative surgery) according to each order. Therefore, the query field characteristic weight (f) of the first query was calculated to be about 2.3, while the query field characteristic weights of 'meniscus restoration', 'arthroscopic restoration', 'meniscus', and 'knee' ( f ₁ ~f ₄ ) can also be calculated as 1.3, 1.3, 1, or 0.1.

The score (X _QFi ) of the medical content retrieved by the secondary query can be calculated using the secondary query correction coefficient (QF _i ) below.

Here, QF _i is the correction coefficient of the ith secondary query among the secondary query fields, r _i is the search range correction weight of the ith secondary query, f _i is the query field correction weight of the ith secondary query, and α _i is It can be other weights, and QF _i can also be calculated as a multiplication function as r _i xf _i xα _i .

The score (X _QFi ) of the medical content retrieved by the secondary query can be calculated as follows.

Here, P _i is the score before correction of the medical content retrieved by the ith secondary query, and other variables can refer to the above-mentioned contents. Here too, X _QFi can be calculated as a multiplication function as QF _i x P _i xα _i .

For reference, other weights (α _i ) may reflect at least one of the completeness of the root, the type of medical content, the user's characteristics, and whether the query used is expanded to a MeSH term.

The completeness of the root can correspond to the completeness of the search for the query (Q). In this embodiment, when “knee cartilage restoration” is entered as a search term by the search subject, the server, etc. may select “arthroscopic meniscus restoration surgery” as the primary query among queries that can be selected as medical terms.

Even if “arthroscopic meniscus restoration” is selected as the first and second query, the roots of ‘arthroscopic’, ‘meniscus’, and ‘restorative surgery’ exist separately in the searched medical content, and ‘ As a combination of the root words ‘arthroscopic meniscus restoration,’ intact cases should be evaluated differently. In this case, when weighting (α _i ) is given according to integrity or root completeness, 'joint horn', 'meniscus', and 'restorative surgery' are searched separately (ex. α _i = 1). If the search term ‘horn meniscus restoration surgery’ is completely searched (ex. α _i = 10), other weights may be assigned higher.

In this embodiment, the passing criterion may be defined as an absolute value where the search total amount is a specific constant (ex. 10,000). However, in addition to this, the number of searches that can provide reasonable and objective results can be selected in various ways, and the same criteria may be used for all queries, but different criteria may be defined depending on the query.

As described above, although the present invention has been described with reference to preferred embodiments, those skilled in the art may make various modifications and changes to the present invention without departing from the spirit and scope of the present invention as set forth in the claims below. You will understand that you can do it.

10: Smartphone 20: PC
30: external database 110: server
112: calculation unit 113: search unit
113: database

Claims (11)

In a method of expanding a query for retrieving medical content in response to a selected primary query related to medical care,
(a) comparing the total search volume reflecting the number of primary searches as a search result using the primary query with a predetermined passing standard;
(b) if the search total does not meet the passing criteria, selecting a secondary query field including a secondary query derived from the primary query; and
(c) calculating a new search total amount reflecting the number of secondary searches by the secondary query field; and
(d) comparing the total search amount in step (c) with the passing criteria,
For the medical content searched by the secondary query, a score of the medical content is calculated in consideration of the number of secondary searches or characteristics of the secondary query field,
The score of the medical content searched by the secondary query includes a search range correction weight (r) reflecting the number of secondary searches, a query field characteristic weight (f) reflecting the characteristics of the secondary query field, and a combination thereof. A query expansion method for retrieving medical content, characterized in that it is calculated using: According to paragraph 1,
A query expansion method for medical content search, characterized in that if the total search amount in step (d) satisfies the pass criterion, query selection for medical content search is completed. According to paragraph 1,
Medical content, characterized in that, if the search total in step (d) does not meet the passing criteria, the secondary query is defined as a new primary query and the steps (a) to (d) are repeated. How to expand queries for search. According to paragraph 1,
A query expansion method for retrieving medical content, characterized in that the secondary query is selected within a range of classifying the primary query according to its roots and combining the classified roots. According to paragraph 1,
The secondary query classifies the primary query as including at least one of symptoms, diagnosis, treatment, and area, and is selected within a range consisting of a combination of similar words or superordinate concepts of the classified symptom, diagnosis, treatment, or area. Featured query expansion method for medical content retrieval. delete delete According to paragraph 1,
The score of the medical content retrieved by the secondary query is calculated using the secondary query correction coefficient (QF _i ) below,

Here, QF _i is the correction coefficient of the ith secondary query among the secondary query fields, r _i is the search range correction weight of the ith secondary query, f _i is the query field correction weight of the ith secondary query, α _i is a query expansion method for medical content retrieval, characterized in that other weights. According to clause 8,
The score of the medical content retrieved by the secondary query is calculated as follows,

Here, P _i is the score before correction of the medical _content retrieved by the ith secondary query, and method. According to paragraph 1,
A query expansion method for retrieving medical content, wherein the passing criterion is provided as a constant. According to paragraph 1,
A query expansion method for retrieving medical content, characterized in that the passing criterion is determined by reflecting the trend of the total search volume.