Keyword clustering method and device and nuclear industry semantic library optimization method and device

Abstract

The invention belongs to the technical field of data mining, and particularly relates to a keyword clustering method and device and a nuclear industry semantic library optimization method and device. Comprising the following steps: S1, establishing a keyword set; s2, respectively calculating a first similarity value between each first keyword and each second keyword related to the first keyword; s3, correcting the first similarity value to obtain a second similarity value; and S4, a plurality of clustering clusters are established based on the second similarity values, each first keyword corresponds to one clustering cluster, and each clustering cluster further comprises a second keyword, and the second similarity value between the second keyword and the corresponding first keyword is larger than or equal to a threshold value. The method has the beneficial effects that the similarity value corresponding to the second keyword related to the first keyword is larger, the keyword mining is more accurate in the clustering processing of the keyword clustering method, the clustering relationship can be fully and comprehensively mined, and the optimized nuclear industry semantics can improve the retrieval efficiency and recall ratio of nuclear power information and reduce leak detection.

Description

Technical Field

[0001] The present invention belongs to the technical field of data mining, and in particular relates to a keyword clustering method and device, and a nuclear industry semantic library optimization method and device. Background Art

[0002] When searching for information, users generally obtain corresponding search results by submitting keywords in various search engines. However, for some special fields such as nuclear power, there are fewer commonly used keywords and some keywords are not well understood by ordinary users due to their high level of professionalism. Therefore, in order to improve the completeness of nuclear power information and prevent missed detections, it is necessary to cluster the commonly used keywords in the nuclear power industry.

[0003] The most commonly used keyword clustering method is: for the keywords in a specific semantic library, keywords with a large literal similarity are retrieved from the keywords of all existing semantic libraries, and the keywords in the specific semantic library are clustered together with the keywords matched by the search. In this way, when the user retrieves the corresponding nuclear power information through the search term, the nuclear power information corresponding to the keyword matching the search term and the nuclear power information corresponding to the keyword clustered with the matching keyword can be displayed to the user.

[0004] The existing nuclear power keyword clustering method does not perform correction after calculating the similarity values ​​between keywords, resulting in low accuracy of the obtained similarity values. As a result, the clustering relationship cannot be fully and comprehensively mined when the keywords are clustered subsequently, resulting in poor clustering effect. Summary of the invention

[0005] The purpose of the present invention is to provide a keyword clustering method and device, and a nuclear industry semantic library optimization method and device. The keywords formed by the present invention can be used to mine keywords more accurately during clustering, and the clustering relationship can be mined more fully and comprehensively. The optimized nuclear industry semantics can improve the retrieval efficiency and recall rate of nuclear power information and reduce missed detections.

[0006] The technical solution of the present invention is as follows: A keyword clustering method comprises the following steps:

[0007] S1: Establish a keyword set;

[0008] S2: respectively calculating a first similarity value between each first keyword and each second keyword related thereto;

[0009] S3: Correcting the first similarity value to obtain a second similarity value;

[0010] S4: establishing a plurality of clusters based on the second similarity value, each first keyword corresponding to a cluster, each cluster also including a second keyword having a second similarity value with the corresponding first keyword greater than or equal to a threshold.

[0011] The keyword set in step 1 includes a plurality of first keywords and a plurality of second keywords, and each second keyword is related to at least one first keyword.

[0012] In the step 1, the first keyword is a keyword provided by the nuclear industry semantic library; the second keyword includes a keyword matching the first keyword and a keyword having a mapping relationship with the first keyword;

[0013] The keyword in the second keyword that matches the first keyword may be a synonym or near synonym of the first keyword; or may be a keyword that is determined based on actual experience and is often used together with the first keyword;

[0014] The second keyword also includes keywords that have a mapping relationship with the first keyword;

[0015] The keywords matching the first keyword and the keywords having a mapping relationship with the first keyword are all related to the corresponding first keyword, that is, each second keyword in the keyword set is related to at least one first keyword.

[0016] In the step 2, the first similarity value between each first keyword and each second keyword related thereto is calculated based on the text features or semantic features of the first keyword, wherein the first similarity value is specifically calculated based on various existing machine learning models.

[0017] The step S3 specifically includes:

[0018] S31: Calculate the correction coefficient corresponding to the first similarity value

[0019] The correction coefficient corresponding to the first similarity value is calculated by formula (1):

[0020] C=Ln(1+ M / N) (1)

[0021] Wherein, C represents the correction coefficient, Ln represents the natural logarithm function, M represents the number of second keywords related to the first keyword, and N represents the number of second keywords in the keyword set;

[0022] S32: Based on the first similarity value and the correction coefficient corresponding thereto, a second similarity value is calculated: the second similarity value=the first similarity value×the correction coefficient.

[0023] The threshold in step S4 is set by the user.

[0024] include:

[0025] S5: When a new first keyword appears, determine a new second keyword related to the new first keyword, add the new first keyword and the new second keyword to the keyword set, and repeat steps S2 to S4.

[0026] Keyword clustering device, including:

[0027] A first establishing module, used to establish a keyword set, wherein the keyword set includes a plurality of first keywords and a plurality of second keywords, each second keyword being related to at least one first keyword;

[0028] A calculation module, used to respectively calculate a first similarity value between each first keyword and each second keyword related thereto;

[0029] a correction module, configured to correct the first similarity value to obtain a second similarity value;

[0030] The second establishing module is used to establish multiple clusters based on the second similarity value, each first keyword corresponds to a cluster, and each cluster also includes a second keyword whose second similarity value with the corresponding first keyword is greater than or equal to a threshold.

[0031] The method for optimizing the nuclear industry semantic database includes the following steps:

[0032] K1: Clustering the keywords in the nuclear industry semantic library according to any keyword clustering method of claims 1 to 5;

[0033] K2: construct word vectors for the clustered keywords to achieve correlation between the keywords;

[0034] K3: Build a knowledge graph based on the keywords constructed after the word vectors are constructed.

[0035] Nuclear industry semantic library optimization device, including:

[0036] A clustering module, used for clustering keywords in the nuclear industry semantic library according to any keyword clustering method of claims 1 to 5;

[0037] A word vector construction module is used to construct word vectors for the clustered keywords to achieve correlation between the keywords;

[0038] A construction module is used to construct a knowledge graph based on the keywords constructed after the word vectors are constructed.

[0039] The beneficial effects of the present invention are as follows: by calculating the correction coefficient corresponding to the first similarity value based on the formula of natural logarithm and performing non-uniform correction on the first similarity value, the similarity value corresponding to the second keyword related to the first keyword is made larger, so that the keyword clustering method can mine keywords more accurately during clustering processing, and can more fully and comprehensively mine clustering relationships. The optimized nuclear industry semantics can improve the retrieval efficiency and recall rate of nuclear power information and reduce missed detections. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The figure is a flow chart of a keyword clustering method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The specific implementation of the present invention is described in detail below. It should be understood that the implementation of the present invention is not limited to the embodiments shown in the accompanying drawings, and the protection scope of the present invention is not limited by the specific implementation. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, similar words such as "one", "one" or "the" do not indicate quantity restrictions, but indicate that there is at least one. Unless otherwise expressly indicated, throughout the specification and claims, the term "include" or its changes such as "include" or "include" will be understood to include the stated elements or components, without excluding other elements or other components. Similar words such as "connect" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0042] Unless otherwise defined, the meanings of all technical terms and scientific terms used in the present invention are the same as those commonly understood by ordinary technicians in the field. In addition, the meanings of the technical terms and scientific terms used in the present invention should be interpreted as having the same meaning as the corresponding terms defined in commonly used technical manuals, and should not be interpreted as having idealized or excessively formal meanings, unless the present invention clearly defines it in this way.

[0043] Figure 1 A flowchart of a keyword clustering method according to an embodiment of the present invention is shown. Figure 1 ,The keyword clustering method includes the following steps:

[0044] S1: Create a keyword set

[0045] The keyword set includes a plurality of first keywords and a plurality of second keywords, and each second keyword is related to at least one first keyword.

[0046] Keyword mining is performed on the massive data to establish a keyword set, where the keyword set includes a plurality of first keywords and a plurality of second keywords, and each second keyword is related to at least one first keyword.

[0047] Specifically, the first keyword may be a keyword provided by the nuclear industry semantic library; the second keyword includes a keyword that matches the first keyword and a keyword that has a mapping relationship with the first keyword.

[0048] The keyword in the second keyword that matches the first keyword can be a synonym or near-synonym of the first keyword; or, it can be a keyword that is determined based on actual experience and is often used together with the first keyword. For example, the first keyword is "reactor". According to actual experience, in the field of nuclear power, "reactor" is often used together with "moderator", "control rod", "reflector", "coolant", "reactor vessel", "steam generator", "steam turbine", "condenser", etc., so these keywords can be determined as the second keyword that matches "reactor".

[0049] The second keyword also includes keywords that have a mapping relationship with the first keyword. For example, when the first keyword is "moderator", according to actual experience, in the field of nuclear power, commonly used "moderators" are "water", "heavy water", "beryllium", "graphite", and "helium", so these keywords can be determined as second keywords that have a mapping relationship with "moderator"; similarly, when the first keyword is "control rod", the second keywords that have a mapping relationship with it include "cadmium rod", "hafnium rod", and "boron rod"; when the first keyword is "reflector", the second keywords that have a mapping relationship with it include "water", "carbon", "graphite", and "beryllium"; when the first keyword is "coolant", the second keywords that have a mapping relationship with it include "heavy water", "light water", "helium", "carbon dioxide", and "hydrogen".

[0050] It can be seen from the above that the keywords matching the first keyword and the keywords having a mapping relationship with the first keyword are all related to the corresponding first keyword, that is, each second keyword in the keyword set is related to at least one first keyword.

[0051] S2: Calculate the first similarity value between each first keyword and each second keyword related thereto

[0052] Specifically, according to the text features or semantic features of the first keyword, the first similarity value between each first keyword and each second keyword related thereto is calculated respectively, wherein the first similarity value can be specifically calculated based on various existing machine learning models, etc., and the present invention does not impose any limitation thereto.

[0053] S3: Correct the first similarity value to obtain a second similarity value

[0054] Step S3 specifically includes:

[0055] S31: Calculate the correction coefficient corresponding to the first similarity value

[0056] Specifically, the correction coefficient corresponding to the first similarity value is calculated by the following formula 1:

[0057] C=Ln(1+ M / N) (1)

[0058] Wherein, C represents a correction coefficient, Ln represents a natural logarithmic function, M represents the number of second keywords related to the first keyword, and N represents the number of second keywords in the keyword set.

[0059] The inventors have found through actual testing and fitting that the correction coefficient calculated by formula (1) has a better correction effect on the first similarity value. The reasons are analyzed as follows:

[0060] I. According to formula (1), M represents the number of second keywords related to the first keyword, N represents the number of second keywords in the keyword set (a constant), M / N is a value greater than 0 and less than 1, and the corresponding correction coefficient obtained after adding 1 and logarithmic operation is still a value greater than 0 and less than 1;

[0061] II. More importantly, for the first keyword, the more second keywords related to it, the larger M is, the larger the value of M / N is, and the natural logarithm function is a concave function. The larger the calculated correction coefficient is, the larger the corresponding similarity value is. That is, the correction of the first similarity value by the correction coefficient obtained by formula 1 is a non-uniform correction. The more second keywords related to the first keyword are, the more important the keyword is, and the larger the corresponding similarity value is, which also conforms to the rules and requirements of highlighting the key points of information retrieval.

[0062] S32: Based on the first similarity value and the correction coefficient corresponding thereto, a second similarity value is calculated. Specifically, the second similarity value=the first similarity value×the correction coefficient.

[0063] S4: establishing a plurality of clusters based on the second similarity value, each first keyword corresponding to a cluster, each cluster also including a second keyword having a second similarity value with the corresponding first keyword greater than or equal to a threshold.

[0064] Specifically, the threshold is set by the user.

[0065] Furthermore, the keyword clustering method further includes the following steps:

[0066] S5: When a new first keyword appears, determine a new second keyword related to the new first keyword, add the new first keyword and the new second keyword to the keyword set, and repeat steps S2 to S4.

[0067] The present invention also provides a keyword clustering device, which comprises:

[0068] A first establishing module, used to establish a keyword set, wherein the keyword set includes a plurality of first keywords and a plurality of second keywords, each second keyword being related to at least one first keyword;

[0069] A calculation module, used to respectively calculate a first similarity value between each first keyword and each second keyword related thereto;

[0070] a correction module, configured to correct the first similarity value to obtain a second similarity value;

[0071] The second establishing module is used to establish multiple clusters based on the second similarity value, each first keyword corresponds to a cluster, and each cluster also includes a second keyword whose second similarity value with the corresponding first keyword is greater than or equal to a threshold.

[0072] In summary, the present invention provides a keyword clustering method and device, which calculates the correction coefficient corresponding to the first similarity value through a specific formula and then obtains the second similarity value, so that a more accurate similarity value can be obtained, the keyword mining is more accurate, and the clustering relationship can be more fully and comprehensively mined during clustering processing, and the clustering effect is good.

[0073] Furthermore, the present invention also provides a method for optimizing a nuclear industry semantic library, comprising the following steps:

[0074] K1: Cluster the keywords in the nuclear industry semantic database according to any of the aforementioned keyword clustering methods;

[0075] K2: construct word vectors for the clustered keywords to achieve correlation between the keywords;

[0076] K3: Build a knowledge graph based on the keywords constructed after the word vectors are constructed.

[0077] Furthermore, the present invention also provides a nuclear industry semantic library optimization device, comprising:

[0078] A clustering module, used for clustering the keywords in the nuclear industry semantic library according to any of the aforementioned keyword clustering methods;

[0079] A word vector construction module is used to construct word vectors for the clustered keywords to achieve correlation between the keywords;

[0080] A construction module is used to construct a knowledge graph based on the keywords constructed after the word vectors are constructed.

[0081] The nuclear industry semantic library optimization method and device provided by the present invention can mine keywords more accurately when clustering, and can mine clustering relationships more fully and comprehensively. Therefore, the optimized nuclear industry semantics can improve the retrieval efficiency and recall rate of nuclear power information and reduce missed detections.

[0082] The foregoing description of specific exemplary embodiments of the present invention is for the purpose of illustration and demonstration. These descriptions are not intended to limit the present invention to the precise form disclosed, and it is clear that many changes and variations can be made in accordance with the above teachings. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the present invention and its practical application, so that those skilled in the art can realize and utilize various different exemplary embodiments of the present invention and various different selections and changes. The scope of protection of the present invention is defined by the claims and their equivalents.

Claims

1. The keyword clustering method is characterized by: The following steps are involved: S1: Establish a keyword set; S2: respectively calculating a first similarity value between each first keyword and each second keyword related thereto; S3: Correcting the first similarity value to obtain a second similarity value; S4: establishing a plurality of clusters based on the second similarity value, each first keyword corresponding to a cluster, each cluster also including a second keyword having a second similarity value with the corresponding first keyword greater than or equal to a threshold.

2. The keyword clustering method according to claim 1, characterized in that: The keyword set in step 1 includes a plurality of first keywords and a plurality of second keywords, and each second keyword is related to at least one first keyword.

3. The keyword clustering method according to claim 2, characterized in that: In the step 1, the first keyword is a keyword provided by the nuclear industry semantic library; the second keyword includes a keyword matching the first keyword and a keyword having a mapping relationship with the first keyword; The keyword in the second keyword that matches the first keyword may be a synonym or near synonym of the first keyword; or may be a keyword that is determined based on actual experience and is often used together with the first keyword; The second keyword also includes keywords that have a mapping relationship with the first keyword; The keywords matching the first keyword and the keywords having a mapping relationship with the first keyword are all related to the corresponding first keyword, that is, each second keyword in the keyword set is related to at least one first keyword.

4. The keyword clustering method according to claim 1, characterized in that: In the step 2, the first similarity value between each first keyword and each second keyword related thereto is calculated based on the text features or semantic features of the first keyword, wherein the first similarity value is specifically calculated based on various existing machine learning models.

5. The keyword clustering method according to claim 1, characterized in that: The step S3 specifically includes: S31: Calculate the correction coefficient corresponding to the first similarity value The correction coefficient corresponding to the first similarity value is calculated by formula (1): C=Ln(1+ M / N) (1) Wherein, C represents the correction coefficient, Ln represents the natural logarithm function, M represents the number of second keywords related to the first keyword, and N represents the number of second keywords in the keyword set; S32: Based on the first similarity value and the correction coefficient corresponding thereto, a second similarity value is calculated: the second similarity value=the first similarity value×the correction coefficient.

6. The keyword clustering method according to claim 1, characterized in that: The threshold in step S4 is set by the user.

7. The keyword clustering method according to claim 1, characterized in that: include: S5: When a new first keyword appears, determine a new second keyword related to the new first keyword, add the new first keyword and the new second keyword to the keyword set, and repeat steps S2 to S4.

8. A keyword clustering device, characterized in that: include: A first establishing module, used to establish a keyword set, wherein the keyword set includes a plurality of first keywords and a plurality of second keywords, each second keyword being related to at least one first keyword; A calculation module, used for respectively calculating a first similarity value between each first keyword and each second keyword related thereto; a correction module, configured to correct the first similarity value to obtain a second similarity value; The second establishing module is used to establish multiple clusters based on the second similarity value, each first keyword corresponds to a cluster, and each cluster also includes a second keyword whose second similarity value with the corresponding first keyword is greater than or equal to a threshold.

9. A method for optimizing the semantic library of nuclear industry, characterized in that: The following steps are involved: K1: Clustering the keywords in the nuclear industry semantic library according to any keyword clustering method of claims 1 to 5; K2: construct word vectors for the clustered keywords to achieve correlation between the keywords; K3: Build a knowledge graph based on the keywords constructed after the word vectors are constructed.

10. Nuclear industry semantic library optimization device, characterized in that: include: A clustering module, used for clustering keywords in the nuclear industry semantic library according to any keyword clustering method of claims 1 to 5; A word vector construction module is used to construct word vectors for the clustered keywords to achieve correlation between the keywords; A construction module is used to construct a knowledge graph based on the keywords constructed after the word vectors are constructed.

Images