Patent transaction behavior identification method and system based on Lelvanstein distance algorithm

Abstract

The invention provides a patent transaction behavior recognition method and system based on a Levinstein distance algorithm, and the method comprises the steps: carrying out the preprocessing of a to-be-recognized patent text, screening out the patent text of a non-marketization transfer behavior in the to-be-recognized patent text, and obtaining a first type of patent text; based on an external investment relationship between a transferor and a transferee of the first type of patent texts, screening patent texts with behaviors waiting for external investment in the first type of patent texts to obtain a second type of patent texts; and determining the text similarity of the names and / or addresses of the transferor and the transferee of the second type of patent texts based on a Lelwstein distance algorithm, and confirming the patent transaction behavior corresponding to the second type of patent texts of which the text similarity meets a preset threshold condition as an effective patent transaction behavior. According to the method, the patent texts to be recognized are screened out step by step by comprehensively applying text word segmentation, sorting and a Lelwstein distance algorithm, and a systematized screening process is constructed in combination with enterprise relationship verification.

Description

Technical Field

[0001] The present invention relates to the field of text recognition technology, and in particular to a patent transaction behavior recognition method and system based on the Levenshtein distance algorithm. Background Art

[0002] With the rapid development of global science and technology and economy, technological innovation has become the key driving force for promoting industrial upgrading and enhancing national competitive advantages. In this process, patents, as an important part of intellectual property rights, play a vital role in promoting corporate technological progress and enhancing the country's overall innovation capabilities.

[0003] Although patent transactions have gradually become an important means to promote technology transfer, at present, the research on patent transaction behavior usually focuses on the research on the value and / or similarity of patent texts, that is, determining whether patent transactions are patent transaction behaviors that generate economic value from the perspective of patent texts. Such a method of determining patent transaction behavior usually requires a more complex calculation process and is less efficient. Summary of the invention

[0004] The present invention provides a patent transaction behavior identification method and system based on the Levenshtein distance algorithm, which is used to solve the defect of low efficiency of the valuable patent transaction behavior identification method in the prior art, and realize an accurate and efficient patent transaction behavior identification method for identifying patent transaction behaviors that generate actual economic value.

[0005] The present invention provides a patent transaction behavior identification method based on the Levenshtein distance algorithm, comprising: Preprocess the patent texts to be identified, filter out patent texts with non-market transfer behaviors from the patent texts to be identified, and obtain the first category of patent texts; Based on the external investment relationship between the transferor and the transferee of the first category of patent texts, the patent texts in the first category of patent texts that are waiting for external investment are screened out to obtain the second category of patent texts; The text similarity of the name and / or address of the transferor and the transferee of the second-category patent text is determined based on the Levenshtein distance algorithm, and the patent transaction behavior corresponding to the second-category patent text whose text similarity meets the preset threshold condition is confirmed as a valid patent transaction behavior.

[0006] According to a patent transaction behavior identification method based on the Levenshtein distance algorithm provided by the present invention, the Levenshtein distance algorithm includes one or more of a Levenshtein distance algorithm based on a simple ratio, a Levenshtein distance algorithm based on a partial ratio, a Levenshtein distance algorithm based on a Token Sort Ratio, and a Levenshtein distance algorithm based on a Token Set Ratio.

[0007] According to a patent transaction behavior identification method based on the Levenshtein distance algorithm provided by the present invention, the step of determining the text similarity of the name and / or address of the transferor and the transferee of the second-category patent text based on the Levenshtein distance algorithm, and confirming the patent transaction behavior corresponding to the second-category patent text whose text similarity meets the preset threshold condition as a valid patent transaction behavior specifically includes: At least one of the Levenshtein distance algorithms is used to calculate and obtain at least one candidate text similarity of the name and / or address of the assignor and the assignee of the second category of patent text; Confirming the candidate text similarity with the highest similarity calculation value as the text similarity of the second category of patent text; Determine the second category of patent texts whose text similarity is less than the first preset threshold as second category of patent texts whose text similarity meets the preset threshold condition; The patent transaction behavior corresponding to the second type of patent text whose text similarity meets the preset threshold condition is confirmed as a valid patent transaction behavior.

[0008] According to a patent transaction behavior identification method based on the Levenshtein distance algorithm provided by the present invention, the step of determining the text similarity of the name and / or address of the transferor and the transferee of the second-category patent text based on the Levenshtein distance algorithm, and confirming the patent transaction behavior corresponding to the second-category patent text whose text similarity meets the preset threshold condition as a valid patent transaction behavior specifically includes: At least one of the Levenshtein distance algorithms is used to calculate and obtain at least one candidate text similarity of the name and / or address of the assignor and the assignee of the second category of patent text; Configuring a preset similarity threshold corresponding to each algorithm in the Levenshtein distance algorithm; Determine at least one second-category patent text whose candidate text similarity is less than the corresponding preset similarity threshold as a second-category patent text whose text similarity meets the preset threshold condition; The patent transaction behavior corresponding to the second type of patent text whose text similarity meets the preset threshold condition is confirmed as a valid patent transaction behavior.

[0009] According to a patent transaction behavior identification method based on the Levenshtein distance algorithm provided by the present invention, the step of determining the text similarity of the name and / or address of the transferor and the transferee of the second category of patent text based on the Levenshtein distance algorithm specifically includes: Determine the first text similarity of the names of the assignor and the assignee of the second-category patent text based on the Levenshtein distance algorithm; Determine the second text similarity of the addresses of the assignor and the assignee of the second type of patent text based on the Levenshtein distance algorithm; The higher one of the first text similarity and the second text similarity is used as the text similarity.

[0010] According to a patent transaction behavior identification method based on the Levenshtein distance algorithm provided by the present invention, the step of preprocessing the patent text to be identified specifically includes: Screening out the patent texts to be identified whose patent rights have expired; Screening out the patent texts to be identified in which the transferor is one of the inventors of the transferee; The patent texts to be identified whose address of the transferor is the same as that of the transferee are screened out.

[0011] The present invention also provides a patent transaction behavior identification system based on the Levenshtein distance algorithm, comprising: The first screening module is used to pre-process the patent texts to be identified, screen out the patent texts with non-market transfer behavior in the patent texts to be identified, and obtain the first category of patent texts; A second screening module is used to screen out patent texts in the first category that are waiting for external investment based on the external investment relationship between the transferor and the transferee of the first category of patent texts, so as to obtain second category patent texts; The patent identification module is used to determine the text similarity of the name and / or address of the transferor and the transferee of the second-category patent text based on the Levenshtein distance algorithm, and confirm the patent transaction behavior corresponding to the second-category patent text whose text similarity meets the preset threshold condition as a valid patent transaction behavior.

[0012] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for identifying patent transaction behavior based on the Levenshtein distance algorithm as described above is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a patent transaction behavior identification method based on the Levenshtein distance algorithm as described in any one of the above.

[0014] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned patent transaction behavior identification methods based on the Levenshtein distance algorithm.

[0015] The patent transaction behavior identification method and system based on the Levenshtein distance algorithm provided by the present invention gradually screen out the patent texts to be identified by comprehensively applying text segmentation, sorting and the Levenshtein distance algorithm, and construct a systematic screening process in combination with enterprise relationship verification, thereby screening out patents that represent real market transaction behaviors, laying a high-quality data foundation for subsequent empirical research. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 This is one of the flow charts of the patent transaction behavior identification method based on the Levenshtein distance algorithm provided by the present invention; Figure 2 This is the second flow chart of the patent transaction behavior identification method based on the Levenshtein distance algorithm provided by the present invention; Figure 3 This is the third flow chart of the patent transaction behavior identification method based on the Levenshtein distance algorithm provided by the present invention; Figure 4 This is the fourth flow chart of the patent transaction behavior identification method based on the Levenshtein distance algorithm provided by the present invention; Figure 5 It is a structural schematic diagram of a patent transaction behavior identification system based on the Levenshtein distance algorithm provided by the present invention; Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] First, the following contents are introduced: Although patent transactions have gradually become an important means to promote technology transfer, different patent entities, such as enterprises, universities, research institutions, and non-patent implementing entities, have significant differences in transaction behaviors, motivations, and characteristics.

[0020] Specifically, current patent transaction data includes a large number of internal transactions involving stakeholders, such as patent transfers between parent companies and subsidiaries or patent flows within inventors. Such transactions usually reflect the need for internal resource allocation or strategic adjustments rather than directly reflecting the true value of technology transfer. In contrast, patent transactions between unrelated parties are more substantial in terms of technology transfer and often represent innovation diffusion and market transfer.

[0021] Therefore, in the process of patent transactions, it is necessary to identify patents that are truly economically valuable and provide data support and analytical basis for subsequent patent transaction behaviors and analysis.

[0022] In the current analysis of patent transaction behavior in the patent field, researchers usually use traditional text similarity algorithms and deep learning methods to conduct research in applications such as patent recommendation, technology evolution path identification, and cooperation prediction.

[0023] For example, through traditional text similarity algorithms such as TF-IDF and LSI models, the technical similarities and market values ​​between patents can be identified; through deep learning models, the understanding of semantic understanding and complex technology identification can be improved to provide technical decision-making support for enterprises; through text similarity and semantic matching technology, the technology evolution path can be intelligently identified.

[0024] However, on the one hand, the text similarity of the above-mentioned patent transaction identification method is usually based on the patent text itself, that is, by semantically identifying and analyzing the features and technical solutions represented in the patent text, the value with transaction value is identified. It is understandable that since patent texts are usually of a certain length, the semantic analysis and similarity calculation of the text also require a large amount of calculation, the identification method is more complicated, and the recognition accuracy is low.

[0025] On the other hand, the above-mentioned patent transaction identification methods usually ignore the significant differences in motivations and behaviors among the subjects of patent transactions and lack analysis of the subjects of patent transactions.

[0026] Universities and research institutes are mainly driven by technology transfer and knowledge dissemination, showing regional dependence and application orientation; enterprises pay more attention to market value and competitive advantages, especially active in cross-regional and international transactions; non-patent implementing entities take economic interests as the core motivation and make profits through patent control and litigation strategies. In addition, the psychological expectations of university researchers play a key role in transaction decisions, which is different from the economic motivations of other commercial entities.

[0027] Patent intermediaries provide transaction support, helping small and medium-sized enterprises overcome the challenges of lack of experience and improving transaction success rates.

[0028] Although these studies reveal the transaction characteristics of different entities, they do not delve into the specific types and differences of transaction behaviors, especially in distinguishing internal stakeholder transactions from external market-based transactions.

[0029] Secondly, in the application of text similarity analysis, existing research mainly focuses on the identification of technical relevance of patent texts, patent recommendation and technology trend analysis.

[0030] Traditional text similarity algorithms (such as TF-IDF and LSI) perform well in capturing text features and basic similarity calculations, while deep learning models (such as BERT and Doc2Vec) further improve the accuracy of semantic understanding, especially when processing long texts and complex semantics. Existing research has enriched the theoretical and practical results in the field of patent transactions, but there are still certain limitations.

[0031] Previous studies have mostly focused on improving algorithm performance, lacked a systematic analysis of the types of patent transaction behaviors, and failed to use text similarity calculations to reveal the differences behind different types of transaction behaviors.

[0032] This makes it difficult for existing analytical methods to accurately distinguish between internal transactions (such as patent transfers between parent companies and subsidiaries) and external market-based transactions, making it impossible to deeply understand the real market dynamics and technology transfer paths of patent transactions.

[0033] In order to solve the above defects, the present invention provides a patent transaction behavior identification method based on the Levenshtein distance algorithm, which is used to identify patent behaviors in market-oriented transactions.

[0034] Combine the following Figures 1 to 4 The present invention introduces a patent transaction behavior identification method based on the Levenshtein distance algorithm. Figure 1 As shown, including: Step 101, pre-processing the patent texts to be identified, filtering out patent texts with non-market transfer behaviors from the patent texts to be identified, and obtaining the first category of patent texts; Obtain the patent text to be identified from various public databases, wherein the patent text includes relevant information of the patent, such as the patent name, the original applicant of the patent, the current patentee, the inventor, the address of the original applicant of the patent and the address of the current patentee, etc. Collect the relevant information and organize it into a text form to obtain the patent text to be identified.

[0035] Among them, the original applicant of the patent represents the transferor of the patent, and the current patent owner of the patent represents the assignee of the patent.

[0036] It is understandable that if the patent corresponding to the patent text to be identified is an invalid patent, it has no economic value for patent transactions. If the original applicant's address in the patent text to be identified is the same as the current patent owner's address but the name is different, it means that the change in patent information is caused by the original applicant's name change, and no actual transaction behavior has occurred. If the original applicant and the current patent owner in the patent text to be identified have the same name but different addresses, it means that the change in patent information is caused by the original applicant's address relocation, and no actual transaction behavior has occurred.

[0037] Therefore, based on the above characteristics, corresponding screening rules are constructed to clean the patent texts to be identified, and preprocess the patent texts to be identified to obtain the first category of patent texts, eliminating directly clear invalid patents, patent ownership adjustments between inventors, and non-market transfer records within the company, so as to ensure that the data set focuses on real market-based patent transactions, providing clear data boundaries for subsequent analysis.

[0038] Step 102, based on the external investment relationship between the transferor and the transferee of the first category of patent texts, screen out the patent texts in the first category that are waiting for external investment, and obtain the second category of patent texts; On this basis, it is also necessary to further screen out patent transaction records involving external investment in the first category of patent texts, such as screening out patent transactions resulting from mergers and acquisitions.

[0039] Optionally, the external investment relationship between the transferor and the transferee in the first category of patent texts can be queried through a public enterprise information query platform to eliminate transactions between parent and subsidiary companies or other related parties that usually only involve capital-level integration rather than technology.

[0040] Through the above method, further screening is carried out on the basis of the first category of patent texts to obtain the second category of patent texts, and the patent texts corresponding to the patent transaction behaviors that have not generated actual economic value are further screened out.

[0041] Step 103, based on the Levenshtein distance algorithm, determine the text similarity of the name and / or address of the transferor and the transferee of the second category of patent text, and confirm the patent transaction behavior corresponding to the second category of patent text whose text similarity meets the preset threshold condition as a valid patent transaction behavior.

[0042] Furthermore, there are still some patent text changes caused by the reconfiguration of resources within the industry. The transferors and transferees corresponding to such changes usually have similar but not identical names and / or addresses. Therefore, the text similarity of the names and / or addresses of the transferors and transferees of the second type of patent texts is determined by the Levenshtein distance algorithm. If the text similarity meets the preset threshold conditions, it is considered that the patent transaction behavior belongs to the reconfiguration of resources within the industry, and does not generate patent transactions with economic value.

[0043] Optionally, the preset threshold condition is predetermined according to a calculation method of writing style similarity.

[0044] For example, the text similarity between the names of the transferor and the transferee of the second patent text can be calculated separately. When the calculated text similarity is less than a preset name threshold, it is considered that the text similarity of the second patent text meets the preset threshold condition.

[0045] For another example, the text similarity between the addresses of the transferor and the transferee of the second patent text may be calculated separately. When the calculated text similarity is less than a preset address threshold, it is considered that the text similarity of the second patent text meets the preset threshold condition.

[0046] For another example, the names and addresses of the transferor and transferee of the second patent text can be integrated into the same text according to the same rules, and then the text similarity between the two integrated texts can be calculated. When the calculated text similarity is less than the preset comprehensive threshold, the text similarity of the second patent text is considered to meet the preset threshold conditions.

[0047] On this basis, the patent transaction behavior corresponding to the second category of patent texts that meet the preset threshold conditions, that is, patent transactions that generate real economic value, are confirmed as valid patent transaction behaviors.

[0048] In this implementation, the text similarity of the names and / or addresses of the assignor and assignee of the second category of patent texts is calculated based on the Levenshtein distance algorithm.

[0049] Among them, the Levenshtein Distance algorithm is a method for calculating string similarity, which is suitable for processing situations where the content is the same but the order is different. The algorithm first segmented and sorted the string to eliminate the order difference and standardized the result into a unified format. Then, the similarity is measured by calculating the number of steps to convert one string into another with the least editing operation (insertion, deletion or replacement). The calculation formula is as follows: ; In the formula, i and j They are strings a and b The length of i and j When any one of them is zero, the Levenshtein distance is max ( i , j ), the length of another non-zero string; when i and j When all are not zero, the Levenshtein distance is obtained by calculating the minimum number of edits for the three edit operations, i.e., deleting a A character in a Insert a b The character that matches the current character in , and the replacement a and b By calculating the minimum value of these three operations, we can get the string a Convert to b The minimum number of editing steps required. Using the Levenshtein distance, this study calculated the similarity score between the transferor's name and address, with the score ranging from 0 to 100, with higher scores indicating more similarity between the two strings.

[0050] In a specific implementation, the name and / or address information of the transferor and the transferee is first decomposed into tokens using a word segmentation tool, and then sorted in a preset order to eliminate word segmentation differences. Optionally, the preset order may be a dictionary order.

[0051] Subsequently, the sorted tags are reassembled into a standardized string to provide a unified comparison basis for similarity calculation. Finally, the similarity between the two parties is calculated by the Levenshtein distance algorithm as text similarity to identify possible internal resource adjustment records.

[0052] Optionally, the Levenshtein distance algorithm in the present invention includes a directly applied Levenshtein distance algorithm, namely the Levenshtein algorithm shown in the above formula, and other extended algorithms based on the Levenshtein distance algorithm.

[0053] Optionally, to ensure the accuracy of the screening results, the screened-out second patent text may be further reviewed manually to confirm whether it involves a reallocation of internal industry resources.

[0054] The present invention uses a comprehensive application of text segmentation, sorting and Levenshtein distance algorithms to gradually screen out patent texts to be identified, and combines corporate relationship verification to build a systematic screening process, thereby screening out patents that represent real market transaction behaviors, laying a high-quality data foundation for subsequent empirical research.

[0055] In the patent transaction behavior identification method based on the Levenshtein distance algorithm of the present invention, the Levenshtein distance algorithm includes one or more of the Levenshtein distance algorithm based on a simple ratio, a Levenshtein distance algorithm based on a partial ratio, a Levenshtein distance algorithm based on a Token Sort Ratio, and a Levenshtein distance algorithm based on a Token Set Ratio.

[0056] In this implementation, the Levenshtein distance algorithm based on the simple ratio is a directly applied Levenshtein distance algorithm, which is suitable for overall comparison of two text segments.

[0057] The Levenshtein distance algorithm based on partial ratio is an extended algorithm based on Simple Ratio. Specifically, it compares the similarity of all possible substrings of one string with another string, and takes the highest similarity as the calculation result. It is suitable for focusing on local matching.

[0058] The Levenshtein distance algorithm based on the Token Sort Ratio is an extended algorithm based on the Simple Ratio. Specifically, the Simple Ratio is calculated after sorting the segmentation results of the two strings. It is suitable for dealing with the problem of character order differences.

[0059] The Levenshtein distance algorithm based on Token Set Ratio is an extended algorithm based on Simple Ratio. Specifically, it segments two strings, calculates the intersection and union of the segmentation sets, and calculates the similarity based on the intersection and union. It is suitable for dealing with character order and set difference problems.

[0060] In the process of calculating the text similarity of the name and / or address of the transferor and the transferee of the second category of patent text, the present invention selects at least one Simple Ratio and an extended algorithm based on Simple Ratio as the Levenshtein distance algorithm, and comprehensively evaluates the similarity between the transferor and the transferee by combining and applying these methods.

[0061] In the patent transaction behavior identification method based on the Levenshtein distance algorithm of the present invention, the step of determining the text similarity of the name and / or address of the transferor and the transferee of the second-category patent text based on the Levenshtein distance algorithm, and confirming the patent transaction behavior corresponding to the second-category patent text whose text similarity meets the preset threshold condition as a valid patent transaction behavior specifically includes: At least one of the Levenshtein distance algorithms is used to calculate and obtain at least one candidate text similarity of the name and / or address of the assignor and the assignee of the second category of patent text; In a specific implementation, Simple Ratio, Partial Ratio, Token Sort Ratio and Token Set Ratio are selected together as the Levenshtein distance algorithm used when calculating text similarity.

[0062] Using the above four algorithms, the names and / or addresses of the transferor and transferee of the second category patent text are calculated respectively, and four candidate text similarities corresponding to each algorithm are obtained, including the first candidate text similarity, the second candidate text similarity, the third candidate text similarity and the fourth candidate text similarity.

[0063] Confirming the candidate text similarity with the highest similarity calculation value as the text similarity of the second category of patent text; Determine the second category of patent texts whose text similarity is less than the first preset threshold as second category of patent texts whose text similarity meets the preset threshold condition; The patent transaction behavior corresponding to the second type of patent text whose text similarity meets the preset threshold condition is confirmed as a valid patent transaction behavior.

[0064] For the four obtained candidate text similarities, the candidate text similarity with the highest similarity calculation value is used as the text similarity of the second category of patent text to which it belongs.

[0065] The first preset threshold is predetermined based on experience, and optionally, is set to 50 in this embodiment.

[0066] When the determined text similarity is less than the first preset threshold, it is considered that the similarity between the name and / or address of the transferor and the transferee of the current second-category patent text is low, that is, the correlation between the transferor and the transferee of the current second-category patent text is low, and the probability that the patent transaction corresponding to the current second-category patent text is a patent transaction behavior of reconfiguration of patent resources within the industry is also low.

[0067] Therefore, it is considered that the text similarity of the current second-category patent text meets the preset threshold condition, and the patent transaction phase corresponding to the current second-category patent text is confirmed as a valid patent transaction behavior.

[0068] In the patent transaction behavior identification method based on the Levenshtein distance algorithm of the present invention, the step of determining the text similarity of the name and / or address of the transferor and the transferee of the second-category patent text based on the Levenshtein distance algorithm, and confirming the patent transaction behavior corresponding to the second-category patent text whose text similarity meets the preset threshold condition as a valid patent transaction behavior specifically includes: At least one of the Levenshtein distance algorithms is used to calculate and obtain at least one candidate text similarity of the name and / or address of the assignor and the assignee of the second category of patent text; Configuring a preset similarity threshold corresponding to each algorithm in the Levenshtein distance algorithm; Determine at least one second-category patent text whose candidate text similarity is less than the corresponding preset similarity threshold as a second-category patent text whose text similarity meets the preset threshold condition; The patent transaction behavior corresponding to the second type of patent text whose text similarity meets the preset threshold condition is confirmed as a valid patent transaction behavior.

[0069] In this implementation, the method of calculating the similarity of the candidate texts is similar to that described above, and thus will not be described in detail.

[0070] On this basis, a corresponding preset similarity threshold is configured for each algorithm based on experience. Optionally, in this implementation, the preset similarity thresholds corresponding to the four algorithms are all set to 50.

[0071] That is to say, when the calculated similarity of the first candidate text, the second candidate text, the third candidate text and the fourth candidate text are all less than 50, it is considered that the correlation between the transferor and the transferee of the currently calculated second patent text is low, and the text similarity of the current patent meets the preset threshold conditions, and the corresponding patent transaction behavior is confirmed as a valid patent transaction behavior.

[0072] In the patent transaction behavior identification method based on the Levenshtein distance algorithm of the present invention, the step of determining the text similarity of the name and / or address of the transferor and the transferee of the second category of patent text based on the Levenshtein distance algorithm specifically includes: Determine the first text similarity of the names of the assignor and the assignee of the second-category patent text based on the Levenshtein distance algorithm; Determine the second text similarity of the addresses of the assignor and the assignee of the second type of patent text based on the Levenshtein distance algorithm; The higher one of the first text similarity and the second text similarity is used as the text similarity.

[0073] Since the names and addresses of the transferor and the transferee of the second category of patent texts are similar, it can be considered that it may be a patent transaction resulting from the reconfiguration of internal patent resources. Therefore, in this implementation, the names and addresses of the transferor and the transferee are calculated separately.

[0074] Optionally, the first text similarity of the names of the transferor and the transferee of the second category of patent text is determined based on the Levenshtein distance algorithm and its extended algorithm through the calculation method mentioned above, and then the second text similarity of the addresses of the transferor and the transferee of the second category of patent text is determined in the same way.

[0075] The higher one of the first text similarity and the second text similarity is used as the text similarity of the current second-category patent text. On this basis, when the text similarity of the second text meets the preset threshold condition, the patent transaction behavior of the current second-category patent text is confirmed as a valid patent transaction behavior.

[0076] In the patent transaction behavior identification method based on the Levenshtein distance algorithm of the present invention, the step of preprocessing the patent text to be identified specifically includes: Screening out the patent text to be identified based on the expiration date of the patent right; Alternatively, if Figure 2 As shown, the patent texts to be identified corresponding to the transfer records whose transfer dates are after the patent expiration date are excluded.

[0077] Screening out the patent texts to be identified in which the transferor is one of the inventors of the transferee; Optionally, the patent text to be identified corresponding to the assignment record in which the assignee is one of the inventors is excluded.

[0078] The patent texts to be identified whose address of the transferor is the same as that of the transferee are screened out.

[0079] Optionally, patent texts corresponding to assignment records where the name or address of the assignor and assignee are the same are excluded.

[0080] The above method is used to implement the preprocessing process of the patent text to be identified, and the first category of patent texts is screened.

[0081] The patent transaction behavior identification method based on the Levenshtein distance algorithm provided by the present invention is described below by taking a specific implementation as an example.

[0082] like Figure 3 and Figure 4 As shown in Figure 1, we first use a public patent database to establish the sample and feature space for the prediction task. The data collection period is from January 1, 2002 to December 31, 2011. The patent type is selected as "invention patent", the IPC classification number is selected as "G Division", the inventor's country is China, and the authorized patents are selected as the research object. Finally, 86,148 authorized patents were obtained as the data basis for this study.

[0083] Based on the research data collected in the early stage, 15,201 valid patents that have not yet been transferred were excluded first. Although these patents have not been traded at present, it is uncertain whether they will be traded in the future, so they cannot meet the research needs of this study.

[0084] Secondly, 104 transfer records that occurred after the patent expiration date were excluded, because these transfers after the end of the protection period cannot be regarded as valid transactions. Furthermore, 229 transferees were excluded, and 3,301 transferors and transferees had the same address or name. These transactions are usually normal rights transfers or internal transfers between inventors, which do not meet the conditions for transactions and are not real transactions.

[0085] Finally, the first category patent text is obtained.

[0086] Based on the tradable patent screening rules constructed in the previous article, 4,757 patent transaction records involving stakeholders were further screened and eliminated. These records are mainly divided into two categories: one involves external cross-industry investment, and the other may be caused by the reconfiguration of internal industry resources.

[0087] In order to confirm whether the above two types of relationships exist between the two parties to the transaction, two strategies are adopted to identify transfers involving interested parties.

[0088] First, for the first category, use business information query websites to check whether the two parties to the transaction are involved in external cross-industry investment relationships. If so, they will be excluded and the remaining patent texts will be regarded as the second category of patent texts.

[0089] For the latter category, the similarity calculation method based on the Levenshtein distance is used to calculate the similarity of the names and addresses of the transferors, including Simple Ratio, Partial Ratio, Token Sort Ratio and Token Set Ratio. In addition, the calculation results of the four algorithms are comprehensively compared, and the maximum value is selected as the basis for judging the similarity threshold.

[0090] Among them, the calculation result of the distance similarity between the addresses of a transferor and the transferee is shown as follows. In the second category of patent texts, the transferor's address is "No. 63, D Road, District C, City B, Province A", and the transferee's address is "No. 63, D Road, District C, City B, Province A, Building 4, Floor 6, District E". The similarity score between the two is 78, which is greater than the predetermined threshold. Therefore, it is considered to be a transaction behavior generated by internal resource reconfiguration and does not meet the preset threshold conditions.

[0091] The patent transaction behavior identification system based on the Levenstein distance algorithm provided by the present invention is described below. The patent transaction behavior identification system based on the Levenstein distance algorithm described below and the patent transaction behavior identification method based on the Levenstein distance algorithm described above can be referenced to each other.

[0092] like Figure 5 As shown, the patent transaction behavior identification system based on the Levenshtein distance algorithm includes a first screening module 501, a second screening module 502 and a patent identification module 503; The first screening module 501 is used to pre-process the patent texts to be identified, screen out the patent texts with non-market transfer behavior in the patent texts to be identified, and obtain the first category of patent texts; Obtain the patent text to be identified from various public databases, wherein the patent text includes relevant information of the patent, such as the patent name, the original applicant of the patent, the current patentee, the inventor, the address of the original applicant of the patent and the address of the current patentee, etc. Collect the relevant information and organize it into a text form to obtain the patent text to be identified.

[0093] Among them, the original applicant of the patent represents the transferor of the patent, and the current patent owner of the patent represents the assignee of the patent.

[0094] It is understandable that if the patent corresponding to the patent text to be identified is an invalid patent, it has no economic value for patent transactions. If the original applicant's address in the patent text to be identified is the same as the current patent owner's address but the name is different, it means that the change in patent information is caused by the original applicant's name change, and no actual transaction behavior has occurred. If the original applicant and the current patent owner in the patent text to be identified have the same name but different addresses, it means that the change in patent information is caused by the original applicant's address relocation, and no actual transaction behavior has occurred.

[0095] Therefore, based on the above characteristics, corresponding screening rules are constructed to clean the patent texts to be identified, and preprocess the patent texts to be identified to obtain the first category of patent texts, eliminating invalid patents, patent ownership adjustments between inventors, and non-market transfer records within the company, to ensure that the data set focuses on real market-based patent transactions, providing clear data boundaries for subsequent analysis.

[0096] A second screening module 502 is used to screen out patent texts in the first category that are waiting for external investment based on the external investment relationship between the transferor and the transferee of the first category of patent texts, so as to obtain second category patent texts; On this basis, it is also necessary to further screen out patent transaction records involving external investment in the first category of patent texts, such as screening out patent transactions resulting from mergers and acquisitions.

[0097] Optionally, the external investment relationship between the transferor and the transferee in the first category of patent texts can be queried through a public enterprise information query platform to eliminate transactions between parent and subsidiary companies or other related parties that usually only involve capital-level integration rather than technology.

[0098] Through the above method, further screening is carried out on the basis of the first category of patent texts to obtain the second category of patent texts, and the patent texts corresponding to the patent transaction behaviors that have not generated actual economic value are further screened out.

[0099] The patent identification module 503 is used to determine the text similarity of the name and / or address of the transferor and the transferee of the second-category patent text based on the Levenshtein distance algorithm, and confirm the patent transaction behavior corresponding to the second-category patent text whose text similarity meets the preset threshold condition as a valid patent transaction behavior.

[0100] Furthermore, there are still some patent text changes caused by the reconfiguration of resources within the industry. The transferors and transferees corresponding to such changes usually have similar but not identical names and / or addresses. Therefore, the text similarity of the names and / or addresses of the transferors and transferees of the second type of patent texts is determined by the Levenshtein distance algorithm. If the text similarity meets the preset threshold conditions, it is considered that the patent transaction behavior belongs to the reconfiguration of resources within the industry, and does not generate patent transactions with economic value.

[0101] Optionally, the preset threshold condition is predetermined according to a calculation method of writing style similarity.

[0102] For example, the text similarity between the names of the transferor and the transferee of the second patent text can be calculated separately. When the calculated text similarity is less than a preset name threshold, it is considered that the text similarity of the second patent text meets the preset threshold condition.

[0103] For another example, the text similarity between the addresses of the transferor and the transferee of the second patent text may be calculated separately. When the calculated text similarity is less than a preset address threshold, it is considered that the text similarity of the second patent text meets the preset threshold condition.

[0104] For another example, the names and addresses of the transferor and transferee of the second patent text can be integrated into the same text according to the same rules, and then the text similarity between the two integrated texts can be calculated. When the calculated text similarity is less than the preset comprehensive threshold, the text similarity of the second patent text is considered to meet the preset threshold conditions.

[0105] On this basis, the patent transaction behavior corresponding to the second category of patent texts that meet the preset threshold conditions, that is, patent transactions that generate real economic value, are confirmed as valid patent transaction behaviors.

[0106] In this implementation, the text similarity of the names and / or addresses of the assignor and assignee of the second category of patent texts is calculated based on the Levenshtein distance algorithm.

[0107] Optionally, the Levenshtein distance algorithm in the present invention includes a directly applied Levenshtein distance algorithm, namely the Levenshtein algorithm shown in the above formula, and other extended algorithms based on the Levenshtein distance algorithm.

[0108] Optionally, to ensure the accuracy of the screening results, the screened-out second patent text may be further reviewed manually to confirm whether it involves a reallocation of internal industry resources.

[0109] The present invention uses a comprehensive application of text segmentation, sorting and Levenshtein distance algorithms to gradually screen out patent texts to be identified, and combines corporate relationship verification to build a systematic screening process, thereby screening out patents that represent real market transaction behaviors, laying a high-quality data foundation for subsequent empirical research.

[0110] Figure 6 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the patent transaction behavior identification method based on the Levenstein distance algorithm, the method comprising: pre-processing the patent text to be identified, screening out the patent texts with non-market transfer behavior in the patent text to be identified, and obtaining the first category of patent texts; based on the external investment relationship between the transferor and the transferee of the first category of patent texts, screening out the patent texts with external investment behavior in the first category of patent texts, and obtaining the second category of patent texts; determining the text similarity of the name and / or address of the transferor and the transferee of the second category of patent texts based on the Levenstein distance algorithm, and confirming the patent transaction behavior corresponding to the second category of patent texts whose text similarity meets the preset threshold condition as a valid patent transaction behavior.

[0111] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. 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.

[0112] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the patent transaction behavior identification method based on the Levenshtein distance algorithm provided by the above methods, and the method includes: preprocessing the patent text to be identified, screening out the patent texts with non-market transfer behaviors in the patent text to be identified, and obtaining a first category of patent texts; based on the external investment relationship between the transferor and the transferee of the first category of patent texts, screening out the patent texts with pending external investment behaviors in the first category of patent texts, and obtaining a second category of patent texts; determining the text similarity of the name and / or address of the transferor and the transferee of the second category of patent texts based on the Levenshtein distance algorithm, and confirming the patent transaction behavior corresponding to the second category of patent texts whose text similarity meets the preset threshold condition as a valid patent transaction behavior.

[0113] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the patent transaction behavior identification method based on the Levenshtein distance algorithm provided by the above-mentioned methods, the method comprising: preprocessing the patent text to be identified, screening out the patent texts with non-market transfer behaviors in the patent text to be identified, and obtaining a first category of patent texts; based on the external investment relationship between the transferor and the transferee of the first category of patent texts, screening out the patent texts with pending external investment behaviors in the first category of patent texts, and obtaining a second category of patent texts; determining the text similarity of the name and / or address of the transferor and the transferee of the second category of patent texts based on the Levenshtein distance algorithm, and confirming the patent transaction behavior corresponding to the second category of patent texts whose text similarity meets the preset threshold condition as a valid patent transaction behavior.

[0114] 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.

[0115] 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.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention 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 invention.

Claims

1. A patent transaction behavior identification method based on the Levenshtein distance algorithm, characterized in that: include: Preprocess the patent texts to be identified, filter out patent texts with non-market transfer behaviors from the patent texts to be identified, and obtain the first category of patent texts; Based on the external investment relationship between the transferor and the transferee of the first category of patent texts, the patent texts in the first category of patent texts that are waiting for external investment are screened out to obtain the second category of patent texts; The text similarity of the name and / or address of the transferor and the transferee of the second-category patent text is determined based on the Levenshtein distance algorithm, and the patent transaction behavior corresponding to the second-category patent text whose text similarity meets the preset threshold condition is confirmed as a valid patent transaction behavior.

2. The patent transaction behavior identification method based on the Levenshtein distance algorithm according to claim 1 is characterized in that: The Levenshtein distance algorithm includes one or more of a Levenshtein distance algorithm based on a simple ratio, a Levenshtein distance algorithm based on a partial ratio, a Levenshtein distance algorithm based on a Token Sort Ratio, and a Levenshtein distance algorithm based on a Token Set Ratio.

3. The patent transaction behavior identification method based on the Levenshtein distance algorithm according to claim 2 is characterized in that: The step of determining the text similarity of the names and / or addresses of the transferor and the transferee of the second-category patent text based on the Levenshtein distance algorithm, and confirming the patent transaction behavior corresponding to the second-category patent text whose text similarity meets the preset threshold condition as a valid patent transaction behavior specifically includes: At least one of the Levenshtein distance algorithms is used to calculate and obtain at least one candidate text similarity of the name and / or address of the assignor and the assignee of the second category of patent text; Confirming the candidate text similarity with the highest similarity calculation value as the text similarity of the second category of patent text; Determine the second category of patent texts whose text similarity is less than the first preset threshold as second category of patent texts whose text similarity meets the preset threshold condition; The patent transaction behavior corresponding to the second type of patent text whose text similarity meets the preset threshold condition is confirmed as a valid patent transaction behavior.

4. The patent transaction behavior identification method based on the Levenshtein distance algorithm according to claim 2 is characterized in that: The step of determining the text similarity of the names and / or addresses of the transferor and the transferee of the second-category patent text based on the Levenshtein distance algorithm, and confirming the patent transaction behavior corresponding to the second-category patent text whose text similarity meets the preset threshold condition as a valid patent transaction behavior specifically includes: At least one of the Levenshtein distance algorithms is used to calculate and obtain at least one candidate text similarity of the name and / or address of the assignor and the assignee of the second category of patent text; Configuring a preset similarity threshold corresponding to each algorithm in the Levenshtein distance algorithm; Determine at least one second-category patent text whose candidate text similarity is less than the corresponding preset similarity threshold as a second-category patent text whose text similarity meets the preset threshold condition; The patent transaction behavior corresponding to the second type of patent text whose text similarity meets the preset threshold condition is confirmed as a valid patent transaction behavior.

5. The patent transaction behavior identification method based on the Levenshtein distance algorithm according to claim 1 is characterized in that: The step of determining the text similarity of the names and / or addresses of the assignor and the assignee of the second category of patent text based on the Levenshtein distance algorithm specifically includes: Determine the first text similarity of the names of the assignor and the assignee of the second-category patent text based on the Levenshtein distance algorithm; Determine the second text similarity of the addresses of the assignor and the assignee of the second type of patent text based on the Levenshtein distance algorithm; The higher one of the first text similarity and the second text similarity is used as the text similarity.

6. The patent transaction behavior identification method based on the Levenshtein distance algorithm according to any one of claims 1 to 5, characterized in that: The step of preprocessing the patent text to be identified specifically includes: Screening out the patent texts to be identified whose patent rights have expired; Screening out the patent texts to be identified in which the transferor is one of the inventors of the transferee; The patent texts to be identified whose address of the transferor is the same as that of the transferee are screened out.

7. A patent transaction behavior identification system based on the Levenshtein distance algorithm, characterized in that: include: The first screening module is used to pre-process the patent texts to be identified, screen out the patent texts with non-market transfer behavior in the patent texts to be identified, and obtain the first category of patent texts; A second screening module is used to screen out patent texts in the first category that are waiting for external investment based on the external investment relationship between the transferor and the transferee of the first category of patent texts, so as to obtain second category patent texts; The patent identification module is used to determine the text similarity of the name and / or address of the transferor and the transferee of the second-category patent text based on the Levenshtein distance algorithm, and confirm the patent transaction behavior corresponding to the second-category patent text whose text similarity meets the preset threshold condition as a valid patent transaction behavior.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the patent transaction behavior identification method based on the Levenshtein distance algorithm as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the patent transaction behavior identification method based on the Levenshtein distance algorithm as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the patent transaction behavior identification method based on the Levenshtein distance algorithm as described in any one of claims 1 to 6 is implemented.

Images