A business name alignment method, electronic device and storage medium

Abstract

The application provides an enterprise name alignment method, comprising: obtaining an enterprise alias in target text information, and an associated regional entity and an associated person entity related to the enterprise alias; obtaining a constituent element in the enterprise alias, and obtaining an enterprise abbreviation set corresponding to the enterprise alias based on the obtained constituent element, the constituent element comprising at least one of a business name, a region, an industry and an organization type; obtaining a candidate enterprise information list related to the enterprise alias based on the enterprise alias and the corresponding enterprise abbreviation set, the candidate enterprise information list comprising n candidate enterprise information, the candidate enterprise information comprising at least an enterprise name of a corresponding candidate enterprise; and obtaining a target enterprise name related to the enterprise alias based on the n candidate enterprise information. The application also provides an electronic device and a storage medium. The application can improve the accuracy of enterprise name alignment.

Description

Technical Field

[0001] This invention relates to the field of entity alignment, and in particular to a method for aligning company names, an electronic device, and a storage medium. Background Technology

[0002] Entity alignment aims to discover entities from two or more different information sources that point to the same real-world object and to establish alignment relationships between these entities. A company, as an organization, is a specific category of entity. The same company may have multiple different names in different data sources, such as its full name, abbreviation, nickname, or stock name, all of which can refer to the same company object. Furthermore, the same abbreviation or nickname may represent different company objects in different information contexts. These complex and diverse names pose significant challenges to enterprise information aggregation and enterprise knowledge graph construction. Therefore, aligning multiple company names with the company's unique full name is of great significance for enterprise information aggregation and enterprise knowledge graph construction.

[0003] One known method for aligning company names includes: performing abbreviation recognition processing on text containing the abbreviation to be identified to obtain a set of candidate abbreviations; obtaining the word frequency of each candidate abbreviation in a preset text library; determining the target abbreviation based on the word frequency of each candidate abbreviation; traversing the preset abbreviation library to obtain abbreviations that match the target abbreviation; obtaining the full name of the company corresponding to the abbreviation; and determining that the full name of the company and the target abbreviation are successfully matched when texts in which the target abbreviation and the full name of the company are found are found. Another known method for aligning company names includes: obtaining the company abbreviation; retrieving relevant candidate full names of companies through the company abbreviation; segmenting each candidate full name of a company into multiple words; comparing each segmented word with the company abbreviation; calculating the matching degree of all candidate full names of companies; selecting the highest matching degree; and if the highest matching degree is greater than or equal to a set threshold, selecting the candidate full name of the company corresponding to the highest matching degree and matching it with the company abbreviation.

[0004] Existing alignment methods primarily rely on abbreviation databases or the method of retrieving the full name of a company using its abbreviation. However, these methods still have limitations. First, due to the sheer size of the full company database and the complexity and diversity of company abbreviations, constructing a complete and comprehensive abbreviation database is extremely difficult, resulting in low recall rates for company name alignment based on abbreviation databases. Second, since companies may be referred to in public opinion as their stock names, abbreviations, nicknames, etc., which are not necessarily substrings of the full company name, the method of retrieving the full company name using abbreviations may not accurately retrieve the full name, leading to low recall rates. Finally, since the same company abbreviation or nickname may represent different company entities in different information contexts, the above two alignment methods cannot handle this situation, affecting the accuracy of company name alignment. Summary of the Invention

[0005] To address the aforementioned technical problems, the technical solution adopted by this invention is as follows:

[0006] This invention provides a method for aligning company names, the method comprising the following steps:

[0007] S100, Obtain the company name in the target text information, as well as the associated geographic entities and associated person entities related to the company name;

[0008] S200, obtain the constituent elements in the enterprise name, and obtain the set of enterprise abbreviations corresponding to the enterprise name based on the obtained constituent elements. The constituent elements include at least one of trade name, region, industry and organization type.

[0009] S300, based on the set of enterprise abbreviations and corresponding enterprise short names, obtain a list of candidate enterprise information related to the enterprise abbreviation, the list of candidate enterprise information includes n candidate enterprise information, and the candidate enterprise information includes at least the enterprise name of the corresponding candidate enterprise;

[0010] S400, based on the information of the n candidate companies, obtain the name of the target company related to the company alias.

[0011] Another embodiment of the present invention provides a non-transitory computer-readable storage medium storing at least one instruction or at least one program segment, characterized in that the at least one instruction or the at least one program segment is loaded and executed by a processor to implement the method described above.

[0012] Another embodiment of the present invention provides an electronic device, including a processor and the aforementioned non-transitory computer-readable storage medium.

[0013] The present invention has at least the following beneficial effects:

[0014] The enterprise name alignment method provided in this invention first extracts enterprise aliases and related entities from the information. Based on the extracted enterprise aliases, it extracts abbreviations, trade names, regions, industries, and organization types to construct a set of enterprise abbreviations. Then, it performs multi-level enterprise information retrieval based on the enterprise aliases and abbreviations to obtain a candidate enterprise list. Next, it calculates the relevance between the enterprise aliases in the information and the candidate enterprises, and then sorts the candidate enterprises according to the relevance. Finally, it obtains the full name of the enterprise aligned with the enterprise alias. This invention can effectively solve the alignment mapping problem of different aliases of the same enterprise to the full name of the enterprise, and the alignment mapping problem of the same enterprise alias to the full names of different enterprises under different information backgrounds. It can improve the accuracy and recall rate of enterprise name alignment and is of great significance for enterprise information aggregation and enterprise knowledge graph construction. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart of a company name alignment method provided in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] In this embodiment of the invention, enterprise name alignment refers to mapping the enterprise name in text information such as enterprise public opinion information or internal documents to the enterprise entity in the industrial and commercial enterprise information database in order to find the corresponding enterprise entity.

[0019] Figure 1 A flowchart of a company name alignment method provided in an embodiment of the present invention.

[0020] This invention provides a method for aligning company names, such as... Figure 1 As shown, the method may include the following steps:

[0021] S100, obtain the company name in the target text information, as well as the associated geographical entities and associated person entities related to the company name.

[0022] In this embodiment of the invention, the target text information can be corporate public opinion information or internal documents, etc. The corporate abbreviation can be the full name of the company, its abbreviation, a stock name, or an alias—a proprietary name that can refer to a specific company in the information. Corporate abbreviation extraction can be performed using a named entity recognition model. The named entity recognition model is trained on a sample dataset with entity annotations. Based on the feature vector of the text, the identified named entities are abbreviations in the text, including personal names, place names, organization names, proper nouns, etc. Organization names often refer to companies, government agencies, etc. Therefore, the named entity recognition model can be used to extract organization names to obtain corporate abbreviations. Commonly used named entity recognition models such as BiLSTM-CRF and IDCNN-CRF can be used. For example, for the text information {[China, Shenzhen, July 18, 2022] During the 2022 Huawei Win-Win Innovation Week, A delivered a keynote speech entitled "Continuous Innovation, Fully Moving Towards the 5.5G Era", proposing the concept of "fully moving towards the 5.5G era" and discussing with operators and industry partners the direction of the overall generational evolution and innovative development of the industry in the next 5 to 10 years, working together to move towards the 5.5G era and innovate for a better future}, the enterprise affix that can be extracted using the named entity model is Huawei.

[0023] In this embodiment of the invention, the associated geographic entities and associated person entities can be obtained through the following steps:

[0024] S101, obtain n1 Chinese characters, n2 sentences, or n3 paragraphs before and after the location of the enterprise alias in the target text information, as the context text of the enterprise alias.

[0025] In this embodiment of the invention, n1 to n3 can be customized. In one illustrative embodiment, n1 can be 100 to 300, n2 can be 1 to 10, and n3 can be 1 to 3.

[0026] S102, obtain the person and region from the context text of the enterprise alias, and use them as the person entity and region entity of the enterprise alias, respectively.

[0027] In this embodiment of the invention, a named entity recognition model can be used. The named entity recognition model is used to extract the people and places in the context text of the company's alias as the people and places entities of the company's alias. The extraction method is similar to that used for extracting the company's alias. From the above text information, the company-related person entity – Wang Tao – and the company-related place entities – China and Shenzhen – can be extracted.

[0028] S200, obtain the constituent elements of the enterprise name, and obtain the set of enterprise abbreviations corresponding to the enterprise name based on the obtained constituent elements. The constituent elements include at least one of trade name, region, industry and organization type.

[0029] In one embodiment of the present invention, the constituent elements of the enterprise designation can be obtained through the following steps:

[0030] S201, the company name is segmented into m words.

[0031] S202, match each of the m words with the set regional dictionary, the set industry dictionary, and the set organization type dictionary to obtain the region, industry, and organization type in the enterprise name.

[0032] In this embodiment of the invention, "region" refers to the place name portion contained in the company name, such as Shanghai, Beijing, Shenzhen, etc. "Organization type" refers to the portion of the company name indicating the type of organization, such as a joint-stock company, a group company, a limited liability company, etc. "Industry" refers to the portion of the company name indicating the industry, such as technology, healthcare, pharmaceutical manufacturing, etc.

[0033] If a word is included in a dictionary, it belongs to the constituent elements defined by that dictionary. For example, if a word is included in a designated regional dictionary, then that word is the region in the company's name.

[0034] In this embodiment of the invention, the method for constructing the regional dictionary is as follows: Based on the "Administrative Division Code of the People's Republic of China," the regions within it are included in the regional dictionary, while the generic names of administrative divisions are removed, and the remaining proper names are also included in the regional dictionary. Alternatively, structured place name information can be crawled from social networking sites and e-commerce websites using web crawlers, and the regions within it are included in the regional dictionary, while the generic names of administrative divisions are removed, and the remaining proper names are also included in the regional dictionary.

[0035] The method for constructing the dictionary of organization types is as follows: it is constructed based on the "Comparison Table of Enterprise Registration Types" and then manually screened and expanded. Specifically, it is screened and expanded according to commonly used registration types. If the "Comparison Table of Enterprise Registration Types" does not contain some commonly used registration types, then these registration types are added.

[0036] The method for constructing the industry dictionary is as follows: construct a dataset of company names, segment all company names, remove words in the segmentation results that are identical to words in the regional dictionary and the organization type dictionary, and then count all remaining words according to word frequency. For words with a high frequency of more than 1,000, they are manually reviewed and verified to remove words that are not industry-related and words with inaccurate segmentation, thus forming the industry dictionary.

[0037] Those skilled in the art will know that a company name may include one or more of the following: region, industry, and type of organization, or none of them.

[0038] S203, Remove the region, industry and organization type from the enterprise name to obtain the trade name.

[0039] Those skilled in the art will know that if a company name does not include a region, industry, or type of organization, then the trade name is the company name.

[0040] In this embodiment of the invention, the set of enterprise abbreviations can be formed by combining trade names, regions, industries, and organizational types. It can include enterprise abbreviations based on trade names and enterprise abbreviations based on trade names and combined elements. The combined elements include region, industry, and organizational type. The positions of the trade name and combined elements in the enterprise abbreviation based on trade names and combined elements are the same as their positions in the original enterprise name. That is, the words are combined according to their order in the original enterprise name to form the set of enterprise abbreviations. For example, if the enterprise name is "Beijing Zhongke Wenge," the region is "Beijing," and the trade name is "Zhongke Wenge," then the resulting set of abbreviations is {Zhongke Wenge, Beijing Zhongke Wenge}. As another example, if the enterprise name is "Wenge China Technology Co., Ltd.", then the resulting set of abbreviations is {Wenge, Wenge China, Wenge Technology, Wenge Co., Ltd.}.

[0041] In another embodiment of the present invention, the constituent elements of the enterprise abbreviation and the set of enterprise abbreviations can be obtained based on a trained AI model. Specifically, the type of each character in multiple enterprise abbreviations used as training samples can be labeled, and the type can include abbreviation, region, industry, organization type, etc., and then input into an AI model, such as a deep neural network, for training to obtain a trained AI model. The obtained enterprise abbreviations can be input into the trained AI model to obtain the corresponding enterprise abbreviation, region, industry, and organization type, etc.

[0042] S300, based on the set of enterprise aliases and corresponding enterprise abbreviations, obtain a list of candidate enterprise information related to the enterprise aliases. The list of candidate enterprise information includes n candidate enterprise information, and the candidate enterprise information includes at least the enterprise name of the corresponding candidate enterprise.

[0043] In one embodiment of the present invention, the candidate enterprise information list includes at least two of a first search result, a second search result, and a third search result. For example, it may include the first search result and the second search result, or it may include the first search result and the third search result, or it may include the second search result and the third search result, or it may include all three search results. Preferably, it may include all three search results to maximize the number of candidate enterprises.

[0044] The first search result is obtained based on the company name recall. Specifically, the first search result is obtained through the following steps:

[0045] S301, the company name is used as a search term to search in the set company information database to obtain the corresponding candidate company information, which is used as the first search result.

[0046] In this embodiment of the invention, the enterprise information database can be an existing database, such as the enterprise information database of the industrial and commercial administration department. The enterprise information database includes Q records, each record including the corresponding enterprise name field, securities name field, and attribute information, such as registered address, registered capital, paid-in capital, number of employees, core personnel, etc.

[0047] Specifically, the system uses the company's pseudonym as a search term to retrieve company names and stock names from the company information database. If a record contains the company's pseudonym or its stock name is equal to the pseudonym, then the condition is met, and that record is considered a candidate company. For example, if the pseudonym is Huawei Technology, and the company information database is stored in Elasticsearch, the query statement would be {"size":1000, "query": {"bool": {"should": [{"match_phrase": {"name":"Huawei"}},{"term": {"bond.keyword": "Huawei"}}]}}}. The query results will be used as candidate company information. In this way, all candidate company information related to the company's pseudonym can be obtained.

[0048] The second search result is obtained based on the set of company abbreviations. Specifically, the second search result is obtained through the following steps:

[0049] S302, take the i-th abbreviation of the enterprise abbreviation set as the search term, search in the set enterprise information database, and obtain h(i) candidate enterprise information corresponding to the i-th enterprise abbreviation, where i is from 1 to k and k is the number of enterprise abbreviations in the enterprise abbreviation set; obtain N candidate enterprise information as the second search result, where N=h(1)+h(2)+…+h(i)+…+h(k).

[0050] Specifically, each abbreviation in the set of company abbreviations is used as a search term to retrieve company names from the company information database. If a record contains a company name with that abbreviation, the condition is met, and that record is selected as a candidate company. For example, if the set of company abbreviations includes Huawei, Huawei Technologies, Huawei Shares, and Huawei Group, the query condition is that the company information database is stored in Elasticsearch, and the query statement is {"size":1000, "query": {"bool": {"should": [{"match_phrase": {"name": "Huawei"}}, {"match_phrase": {"name": "Huawei Technologies"}}, {"match_phrase": {"name": "Huawei Shares"}}, {"match_phrase": {"name": "Huawei Group"}}]}}}. The query results are then used as candidate companies. In this way, candidate company information related to each abbreviation can be obtained, resulting in N candidate company records.

[0051] The third search result is obtained based on word segmentation of the enterprise abbreviation set. Specifically, it can be obtained through the following steps:

[0052] S303, perform word segmentation on the i-th abbreviation in the set of abbreviations to obtain f(i) words, and use each of the f(i) words as a search term to search in the set enterprise information database to obtain g(i) candidate enterprise information corresponding to the i-th abbreviation; obtain M candidate enterprise information as the third search result, where M=g(1)+g(2)+…+g(i)+…+g(k).

[0053] Specifically, each abbreviation in the set of enterprise abbreviations is segmented into words, and each word obtained after segmentation is used as a search term to search for enterprise names in the enterprise information database. When an enterprise name in a certain record contains all the segmented words of the enterprise abbreviation, the condition is met, and that record is used as candidate enterprise information for that enterprise abbreviation. For example, if the set of company abbreviations includes Huawei, Huawei Technologies, Huawei Shares, and Huawei Group, then the query condition is that the company information database is stored in Elasticsearch, and the query statement is: {"size": 1000, "query": {"bool": {"should": [{"bool": {"must": [{"match_phrase": {"name":"Huawei"}}]}}, {"bool": {"must": [{"match_phrase": {"name": "Huawei"}}, {"match_phrase": {"name": "Technology"}}]}}, {"bool": {"must": [{"match_phrase": {"name": "Huawei"}}, {"match_phrase": {"name": "Shares"}}]}}, {"bool":{"must": [{"match_phrase": {"name": "Huawei"}},{"match_phrase": {"name": The query results for "Group" will be used as candidate companies.

[0054] In this way, we can obtain candidate company information corresponding to each abbreviation, that is, we can obtain information on M candidate companies.

[0055] Furthermore, in another embodiment of the present invention, recall can be carried out according to a certain priority. If the number of candidate companies to be recalled exceeds, for example, 100, then no other recall methods will be used. Specifically, S300 may specifically include:

[0056] S310, using the company name as a search term, search in the set company information database to obtain H corresponding candidate company information as the first search result; if H≥d, then form the candidate company information list based on the first search result, otherwise, execute S312; d is a set quantity threshold, for example, d=100.

[0057] S312, take the i-th abbreviation of the enterprise abbreviation set as the search term, and search in the set enterprise information database to obtain h(i) candidate enterprise information corresponding to the i-th enterprise abbreviation, where i is from 1 to k and k is the number of enterprise abbreviations in the enterprise abbreviation set; obtain N candidate enterprise information as the second search result, where N=h(1)+h(2)+…+h(i)+…+h(k); if H+N≥d, then form the candidate enterprise information list based on the first search result and the second search result; otherwise, execute S314.

[0058] S314, perform word segmentation on the i-th abbreviation in the set of abbreviations to obtain f(i) words, and use each of the f(i) words as a search term to search in the set enterprise information database to obtain g(i) candidate enterprise information corresponding to the i-th abbreviation; obtain M candidate enterprise information as the third search result, where M=g(1)+g(2)+…+g(i)+…+g(k); execute S316.

[0059] S316, Based on the first search result, the second search result and the third search result, the candidate enterprise information list is obtained, that is, the H+N+M candidate enterprise information is merged to form the candidate enterprise information list.

[0060] S400, based on the information of the n candidate companies, obtain the name of the target company related to the company alias.

[0061] Furthermore, in one embodiment of the present invention, S400 may specifically include:

[0062] S401, sort the company names of n candidate company information based on the set sorting rule table to obtain the target sorting list; wherein, the u-th row of the set sorting rule table includes (Iu, Pu), where Iu is the u-th sorting attribute information in the set sorting rule table, Pu is the sorting priority corresponding to Iu, the value of u is from 1 to X, and X is the number of sorting attribute information in the set sorting rule table, P1 < P2 < ... < Pu < ... < PX; where Iu ∈ {name relevance, business name relevance, regional relevance, registered address relevance, organization type relevance, ... The ranking attributes can be selected from the set {name relevance, business name relevance, regional relevance, registered address relevance, organization type relevance, overall relevance, regional attribute value, organization type attribute value, registered capital, paid-in capital, and number of employees}. The specific selections can be made based on actual needs.

[0063] In this embodiment of the invention, name relevance is determined by whether the company's pseudonym is identical to the full name or securities name of the candidate company; if identical, it is 1, otherwise 0. Trade name relevance is calculated based on whether the trade name of the company's pseudonym is the same as the trade name of the candidate company; if identical, it is 1, otherwise 0. Geographic relevance is the number of regions included in the company's pseudonym that are identical to those included in the full name of the candidate company; for example, if the number of identical regions is 1, then the geographic relevance is 1. Registered address relevance is the number of regions included in the company's pseudonym that are identical to the registered address of the candidate company. Institution type relevance is calculated based on whether the institution type extracted from the company's pseudonym is the same as that extracted from the full name of the candidate company; if identical, it is 1, otherwise 0. Industry relevance is the number of industry keywords extracted from the company's pseudonym that overlap with those extracted from the full name of the candidate company. Related geographic entity relevance is the number of related geographic entities in the company's pseudonym that are identical to those included in the full name of the candidate company. Related entity relevance is the number of overlaps between the related entity (represented by a company) and the candidate company's legal representative, directors, supervisors, and other key personnel. Overall relevance equals... R s ∈{name relevance, trade name relevance, geographic relevance, registered address relevance, organization type relevance, industry relevance, related geographic entity relevance, related person entity relevance}, w s For R s The corresponding weights, the weights for each relevance can be obtained using machine learning, or they can be set based on expert experience.

[0064] In this embodiment of the invention, the abbreviation, trade name, region, industry, and organization type of the candidate enterprise may be the same as in S200.

[0065] Registered capital, paid-in capital, and number of employees refer to the registered capital, paid-in capital, and number of employees of the candidate company.

[0066] The geographic attribute value is determined based on the address and registered address extracted from the candidate company name. If the candidate company name contains "China," the geographic attribute value is 2; if it contains a major city such as Beijing, Shanghai, Shenzhen, or Guangzhou, the geographic attribute value is 1; otherwise, the geographic attribute value is 0. The organization type attribute value is determined based on the organization type extracted from the candidate company name. If the candidate company is a group or holding company, the organization type attribute value is 1; if the candidate company is a branch, subsidiary, or other branch office, the organization type attribute value is -1; otherwise, the organization type attribute value is 0.

[0067] Furthermore, S401 specifically includes:

[0068] S4011, Sort the company names in the n candidate company information according to I1 from high to low order to obtain the first sorted list.

[0069] S4012, sort the company names in the (r-1)th sorting list in descending order according to the sorting attribute information Ir corresponding to the rth sorting priority, to obtain the rth sorting list, that is, the order of the candidate company names in the current sorting is the order of the candidate company names after the previous sorting; the value of r is from 2 to X.

[0070] S4013, set r = r + 1; if r ≤ X, execute S4012, otherwise, use the current sorted list as the target sorted list.

[0071] S402, the top P company names in the target sorting list are used as the target company names corresponding to the company aliases. In this embodiment of the invention, P can be set based on actual needs, for example, P=3.

[0072] Furthermore, in another embodiment of the present invention, S400 may specifically include:

[0073] S410, input the ranking attribute information of each candidate enterprise information obtained based on the set ranking rule table into the trained entity relevance prediction model to obtain n prediction probabilities, each prediction probability being the probability that the enterprise name in the corresponding relevance information is the enterprise alias.

[0074] The trained entity relevance prediction model can be obtained by setting the ranking attribute information corresponding to the ranking rule table, and the specific training method can be any existing technology. The entity relevance prediction model can be a neural network model.

[0075] S420, sort the n predicted probabilities in descending order, and obtain the company names corresponding to the first P predicted probabilities from the n predicted probabilities in descending order as the target company name corresponding to the company alias.

[0076] Those skilled in the art will understand that the target text information may include one or more company aliases. The method of obtaining the target company name for each company alias is the same.

[0077] The enterprise name alignment method provided in this embodiment of the invention has at least the following advantages:

[0078] (1) Extracting company names from information, and extracting other company-related entities such as people and regions, and applying them to the calculation of the relevance between company names and candidate company entities and the ranking of candidate company entities can improve the accuracy of company name alignment.

[0079] (2) By extracting a set of abbreviations of enterprises based on their pseudonyms and using a multi-level method to recall enterprise information, the recall rate of enterprise name alignment can be improved. At the same time, the extracted regional and enterprise type information is used for relevance calculation, which can improve the accuracy of enterprise name alignment.

[0080] (3) Based on the candidate enterprise relevance calculation method and the multi-strategy candidate enterprise entity ranking method, the possibility that the enterprise alias and the candidate enterprise are the same entity object is evaluated and the candidate enterprises are ranked, which can improve the accuracy of enterprise name alignment.

[0081] Embodiments of the present invention also provide a non-transitory computer-readable storage medium that can be disposed in an electronic device to store at least one instruction or at least one program related to implementing a method in the method embodiments, wherein the at least one instruction or the at least one program is loaded and executed by the processor to implement the method provided in the above embodiments.

[0082] Embodiments of the present invention also provide an electronic device, including a processor and the aforementioned non-transitory computer-readable storage medium.

[0083] Embodiments of the present invention also provide a computer program product including program code, which, when the program product is run on an electronic device, causes the electronic device to perform the steps of the methods described above in various exemplary embodiments of the present invention.

[0084] While specific embodiments of the invention have been described in detail by way of example, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of the invention. It should also be understood that various modifications can be made to the embodiments without departing from the scope and spirit of the invention. The scope of this invention is defined by the appended claims.

Claims

1. A method for aligning company names, characterized in that, The method includes the following steps: S100, Obtain the company name in the target text information, as well as the associated geographic entities and associated person entities related to the company name; S200, obtain the constituent elements in the enterprise name, and obtain the set of enterprise abbreviations corresponding to the enterprise name based on the obtained constituent elements. The constituent elements include at least one of trade name, region, industry and organization type. S300, based on the set of enterprise abbreviations and corresponding enterprise short names, obtain a list of candidate enterprise information related to the enterprise abbreviation, the list of candidate enterprise information includes n candidate enterprise information, and the candidate enterprise information includes at least the enterprise name of the corresponding candidate enterprise; S400, based on the information of the n candidate companies, obtain the name of the target company related to the company alias; The candidate company information also includes the corresponding candidate company's registered capital, paid-in capital, registered address, number of employees, and core personnel; The S400 specifically includes: S401, sort the names of n candidate enterprise information based on the set sorting rule table to obtain the target sorting list; wherein, the u-th row of the set sorting rule table includes (Iu, Pu), where Iu is the u-th sorting attribute information in the set sorting rule table, Pu is the sorting priority corresponding to Iu, the value of u is from 1 to X, and X is the number of sorting attribute information in the set sorting rule table, P1 < P2 < ... < Pu < ... < PX; where Iu∈{name relevance, business name relevance, regional relevance, registered address relevance, organization type relevance, industry relevance, related regional entity relevance, related person entity relevance, comprehensive relevance, regional attribute value, organization type attribute value, registered capital, paid-in capital, number of personnel}, and the comprehensive relevance is equal to R s ∈{name relevance, trade name relevance, geographic relevance, registered address relevance, organization type relevance, industry relevance, related geographic entity relevance, related person entity relevance}, w s For R s The corresponding weights; S402, take the first P company names in the target sorting list as the target company names corresponding to the company aliases; S401 specifically includes: S4011, Sort the company names in the n candidate company information according to I1 from high to low order to obtain the first sorted list; S4012, sort the company names in the (r-1)th sorting list in descending order according to the sorting attribute information Ir corresponding to the rth sorting priority, to obtain the rth sorting list; the value of r is from 2 to X; S4013, set r = r + 1; if r ≤ X, execute S4012, otherwise, use the current sorted list as the target sorted list.

2. The method according to claim 1, characterized in that, The associated geographic entities and associated person entities are obtained through the following steps: S101, obtain n1 Chinese characters, n2 sentences, or n3 paragraphs before and after the location of the enterprise alias in the target text information, as the context text of the enterprise alias; S102, obtain the person and region from the context text of the enterprise alias, and use them as the associated person entity and associated region entity of the enterprise alias, respectively.

3. The method according to claim 1, characterized in that, The set of enterprise abbreviations includes enterprise abbreviations based on trade names and enterprise abbreviations based on trade names and combined elements. The combined elements include region, industry, and organization type. The positions of the trade name and combined elements in the enterprise abbreviation based on trade names and combined elements are the same as the positions of the trade name and combined elements in the enterprise abbreviation.

4. The method according to claim 1, characterized in that, The constituent elements of the company name are obtained through the following steps: S201, perform word segmentation on the enterprise name to obtain m words; S202, match each of the m words with the set regional dictionary, the set industry dictionary, and the set organization type dictionary to obtain the region, industry, and organization type in the enterprise name; S203, Remove the region, industry and organization type from the enterprise name to obtain the trade name.

5. The method according to claim 1, characterized in that, The candidate enterprise information list includes at least two of the first search result, the second search result, and the third search result; The first search result is obtained based on the following steps: S301, Using the company name as a search term, a search is performed in the set company information database to obtain the corresponding candidate company information, which is then used as the first search result; The second search result was obtained based on the following steps: S302, take the i-th enterprise abbreviation in the enterprise abbreviation set as the search term, and search in the set enterprise information database to obtain h(i) candidate enterprise information corresponding to the i-th enterprise abbreviation, where i takes the value from 1 to k, and k is the number of enterprise abbreviations in the enterprise abbreviation set; obtain N candidate enterprise information as the second search result, where N=h(1)+h(2)+…+h(i)+…+h(k); The third search result is obtained based on the following steps: S303, perform word segmentation on the i-th abbreviation in the set of abbreviations to obtain f(i) words, and use each of the f(i) words as a search term to search in the set enterprise information database to obtain g(i) candidate enterprise information corresponding to the i-th abbreviation; obtain M candidate enterprise information as the third search result, where M=g(1)+g(2)+…+g(i)+…+g(k).

6. The method according to claim 1, characterized in that, The S300 specifically includes: S310, using the company name as a search term, a search is performed in the set company information database to obtain H corresponding candidate company information, which are taken as the first search result; if H≥d, the candidate company information list is formed based on the first search result, otherwise, S312 is executed; d is a set quantity threshold; S312, take the i-th company abbreviation in the set of company abbreviations as the search term, and search in the set company information database to obtain h(i) candidate company information corresponding to the i-th company abbreviation, where i is from 1 to k and k is the number of company abbreviations in the set of company abbreviations; obtain N candidate company information as the second search result, where N = h(1) + h(2) + ... + h(i) + ... + h(k); if H + N ≥ d, then form the candidate company information list based on the first search result and the second search result; otherwise, execute S314; S314, perform word segmentation on the i-th abbreviation in the set of enterprise abbreviations to obtain f(i) words, and use each of the f(i) words as a search term to search in the set enterprise information database to obtain g(i) candidate enterprise information corresponding to the i-th enterprise abbreviation; obtain M candidate enterprise information as the third search result, where M=g(1)+g(2)+…+g(i)+…+g(k); execute S316; S316, Based on the first search result, the second search result and the third search result, the candidate enterprise information list is obtained.

7. A non-transitory computer-readable storage medium, wherein the storage medium stores at least one instruction or at least one program segment, characterized in that, The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the method as described in any one of claims 1-6.

8. An electronic device, characterized in that, Includes a processor and the non-transitory computer-readable storage medium as described in claim 7.

Images