Semantic search method and device based on legal knowledge graph and medium

Abstract

The invention discloses a semantic search method and device based on a legal knowledge graph and a medium, relates to the technical field of legal semantic search, and solves the problems that when a current legal search system is used, people need to provide relatively accurate key phrases, and once query phrases with a relatively large range are input, the query phrases cannot be accurately searched. In order to solve the problem that people are difficult to obtain wanted query results from numerous retrieval results, the semantic search method specifically comprises the following steps: S1, obtaining an original query statement input by a user terminal; s2, carrying out primary processing on the input original query statement; s3, performing mapping association on the primarily processed original query statement by utilizing a legal knowledge graph; and S4, performing cross matching on the query elements and the query result set, and obtaining a query result of the original query statement based on a cross matching result. The method realizes accurate search of legal provisions and legal knowledge based on the legal knowledge graph.

Description

Technical Field

[0001] The present invention belongs to the field of semantic retrieval technology, and specifically to a semantic search method, device and medium based on a legal knowledge graph. Background Art

[0002] At present, when people seek legal advice, providing them with an efficient and accurate legal search system can effectively alleviate the shortage of manual legal service resources. Although some existing legal search systems require people to provide relatively accurate keyword phrases when searching for legal information, once a large query phrase is entered, the legal search system will feedback a large number of search results to the people. It is difficult for people to get the query results they want from the numerous search results, which runs counter to the idea of ​​providing accurate online legal assistance to the general public. To this end, the present invention proposes a semantic search method, device and medium based on a legal knowledge graph. Summary of the invention

[0003] In view of the above problems, the present application aims to provide a semantic search method, device and medium based on a legal knowledge graph, so as to solve the problem that the current legal search system in the prior art requires the public to provide relatively accurate keyword phrases when using it. Once a query phrase with a larger range is input, a large number of search results will be fed back to the public. It is difficult for the public to get the query results they want from the numerous search results, which runs counter to the idea of ​​providing accurate legal online assistance to the general public.

[0004] In the first aspect, the purpose of the present invention can be achieved by the following technical solution: a semantic search method based on a legal knowledge graph, the semantic search method is as follows: Step S1, obtaining an original query statement input by a user terminal; Step S2, preliminarily processing the input original query statement; Step S3, using the legal knowledge graph to map and associate the original query statements after preliminary processing; Step S4, cross-matching the query elements with the query result set, and obtaining the query result of the original query statement based on the cross-matching result.

[0005] In a possible implementation manner, the preliminary processing in step S2 includes the following specific processes: Step S21, removing punctuation marks and special symbols in the original query statement; Step S22, segmenting the original query sentence after removing punctuation marks and special symbols; Step S23, deleting repeated characters in the original query statement after word segmentation.

[0006] In a possible implementation, the word segmentation algorithm of the original query statement is specifically: a word segmentation algorithm based on conditional random fields; The specific algorithm for deleting repeated characters is: double-ended queue repeated statement screening algorithm.

[0007] In a possible implementation, the legal knowledge graph includes an upper-level index, a middle-level index, and a bottom-level index, and each of the upper-level index, the middle-level index, and the bottom-level index is provided with a corresponding index character.

[0008] In a possible implementation manner, step S3 includes the following sub-steps: Step S31, marking each character of the original query sentence after preliminary processing as a query element; Step S32, the query element is matched in the legal knowledge graph using a text morphology similarity algorithm to obtain a corresponding query result; wherein the query element may correspond to one query result or may correspond to multiple query results; Step S33, summarizing the query results of the query elements into a query result set, and establishing a mapping link between the query elements and the query result set, so as to realize the mapping association between the original query statement and the query result set; Step S34: According to the above steps, a query result set corresponding to all query elements is obtained.

[0009] In a possible implementation, the query result is obtained as follows: Step S321, comparing each character in the query element with the corresponding index character in the index layer one by one; Step S322, if the similarity between each character in the query element and the index character corresponding to the index layer exceeds a set threshold, the index character is used as the query result of the query element; Step S323: if the similarity between each character in the query element and the index character corresponding to the index layer does not exceed the set threshold, the corresponding index character is removed.

[0010] In a possible implementation, step S4 includes the following sub-steps: Step S41, cross-matching the query element with the query result sets corresponding to other query elements one by one; Step S42, if the query element matches any query result in the query result set corresponding to other query elements, the corresponding query result set is retained, and the query result in the query result set corresponding to other query elements that matches is marked as an approximate query result; If the query element does not match any query result in the query result set corresponding to other query elements, the corresponding query result set will be removed; Step S43: Using the approximate query result as the query result after the query element is combined with other query elements.

[0011] In a possible implementation manner, step S4 further includes the following sub-steps: Step S44, taking the query element of the retained query result set as the next cross-matching object, and matching the next query element with the approximate query result in the retained query result set; Step S45, if the next query element does not match the approximate query result in the retained query result set, the approximate query result is eliminated; if the next query element matches the approximate query result in the retained query result set, cross matching is performed again until the query element cross matching is completed; Step S46: When the approximate query result meets all query elements, the approximate query result is marked as the query result of the original query statement.

[0012] In a second aspect, the present invention further provides an electronic device, the electronic device comprising: A memory storing a computer program; A processor is communicatively connected to the memory, and when the computer program is executed by the processor, the semantic search method is implemented.

[0013] In a third aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the semantic search method when executed by a processor.

[0014] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: The present invention first obtains the original query statement input by the user terminal, and then performs preliminary processing on the input original query statement. After the preliminary processing, the original query statement after preliminary processing is mapped and associated using the legal knowledge graph, and the query elements are cross-matched with the query result set. Finally, the query result of the original query statement is obtained based on the cross-matching result. The present invention realizes accurate search of legal provisions and legal knowledge based on the legal knowledge graph. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0016] Figure 1 It is a scene application diagram in the present invention; Figure 2 is a schematic diagram of the legal knowledge graph in the present invention; Figure 3 is a flow chart of the method of the present invention; Figure 4It is a simplified flow chart of the present invention; Figure 5 It is a schematic diagram of the structure of the electronic device in the present invention. DETAILED DESCRIPTION

[0017] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Example 1: Please refer to Figure 1-Figure 4 As shown, the technical solution provided by the present invention is: a semantic search method based on legal knowledge graph, the method is as follows: Step S1, obtaining an original query statement input by a user terminal; In practice, user terminals include but are not limited to mobile phones, personal computers, etc. The original query statement input by the user terminal is sent to the corresponding server through the network, and subsequent analysis and search work is performed by the server.

[0019] Step S2, preliminarily processing the input original query statement; In this embodiment, the preliminary processing in step S2 includes the following specific processes: Step S21, removing punctuation marks and special symbols in the original query statement by using regular expressions; Step S22, segmenting the original query sentence after removing punctuation marks and special symbols by a word segmentation algorithm based on conditional random fields; In practice, the original query statement can be segmented not only based on the word segmentation algorithm of the hidden Markov model, but also based on the word segmentation algorithm of the n-gram grammar. It is understandable that the above word segmentation algorithms are all existing technologies and will not be described in detail here. Step S23, using a repeated sentence elimination algorithm to delete repeated characters in the original query sentence after word segmentation: In this embodiment, the duplicate statement filtering algorithm can complete the deletion of duplicate characters through the stack duplicate statement filtering algorithm, specifically: initialize a stack, traverse each character in the string, if the stack is empty or the top element of the stack is different from the current character, then push the current character into the stack, if the top element of the stack is the same as the current character, then pop the top element of the stack, after the traversal is completed, the remaining elements in the stack are non-duplicate elements. However, it should be noted that the order of elements in the stack is opposite to the order required by the question, so it is necessary to pop the elements in the stack and assign a new string at the same time; at the same time, the duplicate characters can also be processed through the double-ended queue duplicate statement filtering algorithm, specifically: traverse each character in the string, if the double-ended queue is empty or the top element of the stack is not equal to the current character, then push the current character into the double-ended queue, if the top element of the stack is equal to the current character, then remove the top element of the stack, after the traversal is completed, the characters in the double-ended queue are converted into the final string result; Exemplarily, the original query sentence input is "basic theory of economic law and basic theory of criminal law", which is converted into "economic law", "criminal law", "foundation" and "theory" after preliminary processing.

[0020] Step S3, using the legal knowledge graph to map and associate the original query statements after preliminary processing; Among them, Figure 2 As shown, the legal knowledge graph includes an upper index, a middle index and a bottom index, and each upper index, middle index and bottom index is provided with a corresponding index character; In this embodiment, step S3 includes the following sub-steps: Step S31, marking each character of the original query sentence after preliminary processing as a query element; Step S32, the query element is matched in the legal knowledge graph using a text morphology similarity algorithm to obtain a corresponding query result; wherein the query element may correspond to one query result or may correspond to multiple query results; Among them, text morphology similarity algorithms include but are not limited to cosine similarity algorithm, longest common subsequence algorithm, BM25 algorithm, etc.; In the specific implementation, the process of obtaining the query result is as follows: Step S321, comparing each character in the query element with the corresponding index character in the index layer one by one; Step S322, if the similarity between each character in the query element and the index character corresponding to the index layer exceeds a set threshold, the index character is used as the query result of the query element; Step S323, if the similarity between each character in the query element and the index character corresponding to the index layer does not exceed the set threshold, the corresponding index character is removed; Step S33, summarizing the query results of the query elements into a query result set, and establishing a mapping link between the query elements and the query result set, so as to realize the mapping association between the original query statement and the query result set; In practice, corresponding matching codes may be added between the query elements and the query result set to achieve unique matching between the query elements and the query result set; Step S34, according to the above steps, obtain the query result set corresponding to all query elements.

[0021] Step S4, cross-matching the query elements with the query result set, and obtaining the query result of the original query statement based on the cross-matching result; In this embodiment, step S4 includes the following sub-steps: Step S41, cross-matching the query element with the query result sets corresponding to other query elements one by one; Step S42, if the query element matches any query result in the query result set corresponding to other query elements, the corresponding query result set is retained, and the query result in the query result set corresponding to other query elements that matches is marked as an approximate query result; If the query element does not match any query result in the query result set corresponding to other query elements, the corresponding query result set will be removed; Step S43, using the approximate query result as the query result after the query element is combined with other query elements; Step S44, taking the query element of the retained query result set as the next cross-matching object, and matching the next query element with the approximate query result in the retained query result set; Step S45, if the next query element does not match the approximate query result in the retained query result set, the approximate query result is eliminated; if the next query element matches the approximate query result in the retained query result set, cross matching is performed again until the query element cross matching is completed; Step S46, when the approximate query result meets all query elements, the approximate query result is marked as the query result of the original query statement; For example, the query result set corresponding to "Economic Law" includes {Basic Theory of Economic Law, Economic Law}; the query result set corresponding to "Criminal Law" includes {Criminal Law, Basic Theory of Criminal Law, Criminal Law Theory}; the query result set corresponding to "Foundation" includes {Legal Foundation, Basic Theory of Law, Basic Theory of Civil Law, Basic Theory of Economic Law, Basic Theory of Criminal Law}; the query result set corresponding to "Theory" includes {Basic Theory of Law, Basic Theory of Economic Law, Basic Theory of Criminal Law, Basic Theory of Civil Law}; When the query element "Economic Law" is first cross-matched, the process of matching "Economic Law" with the query results in the query result sets corresponding to "Criminal Law", "Theory", and "Foundation" is cross-matching. When "Economic Law" matches the "Basic Theory of Economic Law" in the query result set corresponding to "Foundation", the "Basic Theory of Economic Law" in the query result set corresponding to "Foundation" is marked as an approximate query result. At this time, the approximate query result satisfies the combined query of "Economic Law + Foundation". When the query results in the query result sets corresponding to "Economic Law" and "Criminal Law" do not match, the query result set corresponding to "Criminal Law" is eliminated, and the query result sets corresponding to "Foundation" and "Theory" are retained; Based on the retained query result set, at this time, the query element "foundation" is cross-matched, and the query element "foundation" is matched with the approximate query results in the retained query result set, and so on. Since the approximate query results at this time have met the combined query of "economic law + foundation", if the approximate query results meet the query element "foundation", then the approximate query results cover the three elements of "economic law", "foundation" and "theory", and the approximate query results are the query results of the original query statement; Similarly, the query results of “Basic Theory of Criminal Law” can also be obtained according to the above method.

[0022] Due to the adoption of the technical solution in this embodiment, the original query statement input by the user terminal can be first obtained, and then the input original query statement can be preliminarily processed. After the preliminary processing, the original query statement after the preliminary processing is mapped and associated using the legal knowledge graph, and the query elements are cross-matched with the query result set. Finally, the query result of the original query statement is obtained based on the cross-matching result. The present invention realizes accurate search of legal provisions and legal knowledge based on the legal knowledge graph.

[0023] In this application, if corresponding calculation formulas appear, the above calculation formulas are all dimensionless and take their numerical calculations. The weight coefficients, proportional coefficients and other coefficients in the formulas are set to a result value obtained by quantifying each parameter. The size of the weight coefficient and the proportional coefficient can be determined as long as it does not affect the proportional relationship between the parameter and the result value.

[0024] Based on another concept of the same invention, the present invention also proposes a semantic search device based on a legal knowledge graph, comprising: A user input module, used to obtain the original query sentence input by the user terminal; A statement processing module is used to perform preliminary processing on the input original query statement; In this embodiment, the working process of the statement processing module is as follows: Punctuation marks and special symbols in the original query statement are removed by regular expressions, and then the original query statement without punctuation marks and special symbols is segmented by a word segmentation algorithm based on conditional random fields. Finally, repeated characters in the original query statement after segmentation are deleted by a repeated statement screening algorithm.

[0025] A mapping and association module is used to map and associate the original query statements after preliminary processing using the legal knowledge graph; In this embodiment, the mapping association process of the mapping association module is specifically as follows: Use regular expressions to remove punctuation and special symbols in the original query statement; Step S22, segmenting the original query sentence after removing punctuation marks and special symbols by a word segmentation algorithm based on conditional random fields; In practice, the original query statement can be segmented not only based on the word segmentation algorithm of the hidden Markov model, but also based on the word segmentation algorithm of the n-gram grammar. It is understandable that the above word segmentation algorithms are all existing technologies and will not be described in detail here. Step S23, using a repeated sentence elimination algorithm to delete repeated characters in the original query sentence after word segmentation: In this embodiment, the duplicate statement filtering algorithm can complete the deletion of duplicate characters through the stack duplicate statement filtering algorithm, specifically: initialize a stack, traverse each character in the string, if the stack is empty or the top element of the stack is different from the current character, then push the current character into the stack, if the top element of the stack is the same as the current character, then pop the top element of the stack, after the traversal is completed, the remaining elements in the stack are non-duplicate elements. However, it should be noted that the order of elements in the stack is opposite to the order required by the question, so it is necessary to pop the elements in the stack and assign a new string at the same time; at the same time, the duplicate characters can also be processed through the double-ended queue duplicate statement filtering algorithm, specifically: traverse each character in the string, if the double-ended queue is empty or the top element of the stack is not equal to the current character, then push the current character into the double-ended queue, if the top element of the stack is equal to the current character, then remove the top element of the stack, after the traversal is completed, the characters in the double-ended queue are converted into the final string result; Exemplarily, the original query sentence input is "basic theory of economic law and basic theory of criminal law", which is converted into "economic law", "criminal law", "foundation" and "theory" after preliminary processing.

[0026] Step S3, using the legal knowledge graph to map and associate the original query statements after preliminary processing; Each character of the original query statement after preliminary processing is marked as a query element, and each character in the query element is compared with the index character corresponding to the index layer one by one. If the similarity between each character in the query element and the index character corresponding to the index layer exceeds the set threshold, the index character is used as the query result of the query element. If the similarity between each character in the query element and the index character corresponding to the index layer does not exceed the set threshold, the corresponding index character is removed. That is, the query element is matched in the legal knowledge graph using the text morphology similarity algorithm to obtain the corresponding query result, and the query results of the query element are summarized into the query result set, and a mapping link is established between the query element and the query result set to realize the mapping association between the original query statement and the query result set. According to the above steps, the query result set corresponding to all query elements is obtained.

[0027] A cross-matching module, used to cross-match the query elements with the query result set, and obtain the query result of the original query statement based on the cross-matching result; In this embodiment, the working process of the cross-matching module is specifically as follows: The query result sets corresponding to the query element and other query elements are cross-matched one by one. If the query element matches any query result in the query result sets corresponding to the other query elements, the corresponding query result set is retained, and the matching query results in the query result sets corresponding to the other query elements are marked as approximate query results. If all query results in the query result sets corresponding to the query element and other query elements do not match, the corresponding query result set is eliminated, and the approximate query result is used as the query result after the query element and the other query elements are combined. The query element of the retained query result set is used as the next cross-matching object, and the next query element is matched with the approximate query result in the retained query result set. If the next query element does not match the approximate query result in the retained query result set, the approximate query result is eliminated. If the next query element matches the approximate query result in the retained query result set, cross-matching is performed again until the query element cross-matching is completed. When the approximate query result meets all query elements, the approximate query result is marked as the query result of the original query statement.

[0028] Embodiment 3: Figure 5 An example of a structural diagram of an electronic device is shown in FIG. Figure 4As shown, the electronic device may include: a processor, a communications interface, a memory, and a system bus, wherein the processor, the communications interface, and the memory communicate with each other through the system bus. The processor may call the logic instructions in the memory to execute a semantic search method based on a legal knowledge graph, the method comprising: obtaining an original query statement input by a user terminal; performing preliminary processing on the input original query statement; mapping and associating the initially processed original query statement using the legal knowledge graph; cross-matching the query elements with the query result set, and obtaining the query result of the original query statement based on the cross-matching result.

[0029] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0030] Embodiment 4: On the other hand, the present application also provides a computer program product, which includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the semantic search method based on the legal knowledge graph provided by the above methods, and the method includes: obtaining an original query statement input by a user terminal; performing preliminary processing on the input original query statement; using the legal knowledge graph to map and associate the original query statement after preliminary processing; cross-matching the query elements with the query result set, and obtaining the query result of the original query statement based on the cross-matching result.

[0031] Embodiment 5: On the other hand, the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the above-mentioned semantic search method based on the legal knowledge graph, the method comprising: obtaining an original query statement input by a user terminal; performing preliminary processing on the input original query statement; mapping and associating the preliminary processed original query statement using the legal knowledge graph; cross-matching the query elements with the query result set, and obtaining the query result of the original query statement based on the cross-matching result.

[0032] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0033] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0034] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A semantic search method based on legal knowledge graph, characterized in that: The semantic search method is as follows: Step S1, obtaining an original query statement input by a user terminal; Step S2, preliminarily processing the input original query statement; Step S3, using the legal knowledge graph to map and associate the original query statements after preliminary processing; Step S4, cross-matching the query elements with the query result set, and obtaining the query result of the original query statement based on the cross-matching result.

2. The semantic search method based on legal knowledge graph according to claim 1 is characterized in that: The preliminary processing in step S2 includes the following specific processes: Step S21, removing punctuation marks and special symbols in the original query statement; Step S22, segmenting the original query sentence after removing punctuation marks and special symbols; Step S23, deleting repeated characters in the original query statement after word segmentation.

3. The semantic search method based on legal knowledge graph according to claim 2 is characterized in that: The word segmentation algorithm of the original query statement is specifically: a word segmentation algorithm based on conditional random fields; The specific algorithm for deleting repeated characters is: double-ended queue repeated statement screening algorithm.

4. The semantic search method based on legal knowledge graph according to claim 2 is characterized in that: The legal knowledge graph includes upper-level index, middle-level index and bottom-level index, and each upper-level index, middle-level index and bottom-level index is set with corresponding index characters.

5. The semantic search method based on legal knowledge graph according to claim 4 is characterized in that: The step S3 includes the following sub-steps: Step S31, marking each character of the original query sentence after preliminary processing as a query element; Step S32, the query element is matched in the legal knowledge graph using a text morphology similarity algorithm to obtain a corresponding query result; wherein the query element corresponds to one query result or to multiple query results; Step S33, summarizing the query results of the query elements into a query result set, and establishing a mapping link between the query elements and the query result set, so as to realize the mapping association between the original query statement and the query result set; Step S34: According to the above steps, a query result set corresponding to all query elements is obtained.

6. The semantic search method based on legal knowledge graph according to claim 5 is characterized in that: The specific process of obtaining the query results is as follows: Step S321, comparing each character in the query element with the corresponding index character in the index layer one by one; Step S322, if the similarity between each character in the query element and the index character corresponding to the index layer exceeds a set threshold, the index character is used as the query result of the query element; Step S323: if the similarity between each character in the query element and the index character corresponding to the index layer does not exceed the set threshold, the corresponding index character is removed.

7. The semantic search method based on legal knowledge graph according to claim 1 is characterized in that: The step S4 includes the following sub-steps: Step S41, cross-matching the query element with the query result sets corresponding to other query elements one by one; Step S42, if the query element matches any query result in the query result set corresponding to other query elements, the corresponding query result set is retained, and the query result in the query result set corresponding to other query elements that matches is marked as an approximate query result; If the query element does not match any query result in the query result set corresponding to other query elements, the corresponding query result set will be removed; Step S43: Using the approximate query result as the query result after the query element is combined with other query elements.

8. The semantic search method based on legal knowledge graph according to claim 7 is characterized in that: The step S4 also includes the following sub-steps: Step S44, taking the query element of the retained query result set as the next cross-matching object, and matching the next query element with the approximate query result in the retained query result set; Step S45, if the next query element does not match the approximate query result in the retained query result set, the approximate query result is eliminated; if the next query element matches the approximate query result in the retained query result set, cross matching is performed again until the query element cross matching is completed; Step S46: When the approximate query result meets all query elements, the approximate query result is marked as the query result of the original query statement.

9. An electronic device, characterized in that: The electronic device comprises: A memory storing a computer program; A processor is communicatively connected to the memory, and when the computer program is executed by the processor, the semantic search method described in any one of claims 1-8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the semantic search method described in any one of claims 1 to 8 is implemented.

Images