Patent document database construction method and device based on technical field

Abstract

The invention provides a patent document database construction method and device based on the technical field, and the method comprises the steps: carrying out the named entity recognition of a technical field part in a patent document based on a named entity recognition model, so as to recognize a main body tag and a field tag of the patent document; according to the named entity recognition result, dividing all patent files into a first type of patent files and a second type of patent files; constructing a mapping dictionary of the first type of patent files; determining a first main body label most similar to the second main body label in a mapping dictionary, and distributing a first domain label corresponding to the first main body label to the second main body label; and establishing a corresponding relationship between the patent number of each patent file and the corresponding first field label, and forming a patent file database based on the technical field. According to the method and the device, the patent file is analyzed by utilizing the field information of the fine granularity, so that the correlation between the complete patent number and the technical field is realized, and the correspondingly formed database has the field division with the finer granularity.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of data processing, and in particular to a patent file database construction method and device based on technical fields, a storage medium and an electronic device. BACKGROUND

[0002] Currently, the information in a patent that customers are concerned about and has high value is technical field information, technical feature information, technical problem information and technical effect information.

[0003] In the existing patent database construction and retrieval process, it is mainly based on the key technical features in the patent file. The patent database is formed by identifying and associating the key words in the patent file, and the corresponding patent is retrieved based on the key words input by the user. The database formed based on the above process cannot effectively utilize the key words input by the user, especially in the process of patent semantic retrieval. The user simply inputs text and then performs background retrieval, and cannot distinguish and separately extract different content modules in the original text, resulting in that part of the semantic content is not completely understood and utilized, affecting the retrieval accuracy and use effect. SUMMARY

[0004] The purpose of the embodiments of the present disclosure is to provide a patent file database construction method and device based on technical fields, a storage medium and an electronic device, to solve the problem that the database in the prior art cannot effectively utilize the key words input by the user.

[0005] The embodiments of the present disclosure adopt the following technical solutions: a patent file database construction method based on technical fields, comprising: performing named entity recognition on a technical field part in a patent file based on a named entity recognition model, to recognize a subject label and a field label of each patent file; according to the named entity recognition result, dividing all the patent files into first type patent files and second type patent files, wherein the first type patent files are patent files in which the field label and the subject label are recognized, and the second type patent files are patent files in which only the subject label is recognized; constructing a mapping dictionary between all first subject labels appearing in the first type patent files and at least one first field label corresponding to the first subject labels; based on each second subject label in the second type patent files, determining a first subject label most similar to the second subject label in the mapping dictionary, and assigning at least one first field label corresponding to the first subject label to the second subject label; according to the first field label corresponding to the second subject label, establishing a corresponding relationship between a patent number of each patent file and the first field label corresponding thereto, to form a patent file database based on technical fields.

[0006] This disclosure also provides a patent document database construction apparatus based on the technical field, comprising: a named entity recognition module, used to perform named entity recognition on the technical field portion of the patent documents based on a named entity recognition model, to identify the subject tag and field tag of each patent document; a classification module, used to classify all the patent documents into a first type of patent documents and a second type of patent documents according to the named entity recognition results, wherein the first type of patent documents are patent documents in which both field tags and subject tags are identified, and the second type of patent documents are patent documents in which only subject tags are identified; a dictionary construction module, used to construct a mapping dictionary between all first subject tags appearing in the first type of patent documents and at least one first field tag corresponding to the first subject tag; a tag allocation module, used to determine, based on each second subject tag in the second type of patent documents, the first subject tag most similar to the second subject tag in the mapping dictionary, and allocate at least one first field tag corresponding to the first subject tag to the second subject tag; and a database construction module, used to establish a correspondence between the patent number of each patent document and its corresponding first field tag according to the first field tag corresponding to the second subject tag, forming a patent document database based on the technical field.

[0007] This disclosure also provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for constructing a patent document database based on the technical field.

[0008] This disclosure also provides an electronic device, including at least a memory and a processor. The memory stores a computer program, and the processor, when executing the computer program in the memory, implements the steps of the above-described patent document database construction method based on the technical field.

[0009] The beneficial effects of this disclosure are as follows: the patent documents are analyzed using more granular field information recorded in the technical field section of the patent documents, and field tags are assigned to patent documents that have not been associated with field information, so as to realize the complete association between patent number and technical field, and make the corresponding database have more granular field division, which is more conducive to improving the accuracy and professionalism of retrieval in subsequent searches. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart of a patent document database construction method based on the technical field in the first embodiment of this disclosure;

[0012] Figure 2 This is a schematic diagram of the structure of a patent document database construction device based on the technical field in the second embodiment of this disclosure. Detailed Implementation

[0013] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0014] Currently, the information in patents that customers care about most and that has the highest information value includes technical field information, technical feature information, technical problem information, and technical effect information.

[0015] Existing patent database construction and retrieval processes primarily rely on key technical features from patent documents. The database is formed by identifying and associating keywords within these documents, and searches are conducted based on user-input keywords. However, databases built upon this process often fail to effectively utilize user-input keywords, particularly during semantic patent searches. Users simply input text and the search proceeds without distinguishing and extracting different content modules from the original text. This results in some semantic content not being fully understood and utilized, impacting search accuracy and usability.

[0016] To address the aforementioned problems, the first embodiment of this disclosure provides a method for constructing a patent document database based on the technical field, the flowchart of which is shown below. Figure 1 As shown, it includes at least steps S10 to S50:

[0017] S10, Based on the named entity recognition model, perform named entity recognition on the technical field part of the patent document to identify the main tag and field tag of each patent document.

[0018] A patent document typically comprises the following parts: the technical field section, the technical features section, the technical problem section, and the technical effect section. The technical field section refers to the technical field to which the patent document belongs, as indicated in the specification. This technical field can be described by specific field information or by entities with specific functions included in the patent document, such as methods or devices. The technical features section describes the actual technical solution of the patent document, such as the claims or the invention description in the specification. The technical problem section describes the technical problem that the patent document aims to solve, usually described in the background section of the specification. The technical effect section describes the beneficial effects achieved by the technical solution, typically found in the latter part of the invention description. During patent searches, users generally expect search results to match their expectations, i.e., to obtain results with the most similar fields, technical solutions, technical problems, or technical effects to their search results. Therefore, a more refined division of patent documents is necessary.

[0019] When establishing a patent database based on the technical field, unlike traditional methods that rely on broad field information such as IPC or CPC classifications, this embodiment uses a Named Entity Recognition (NER) model to identify the technical field portion of the patent document. This includes the field information and descriptions of the specific devices or methods involved in the patent document. In this embodiment, the NER model can identify the corresponding field tags and subject tags, which correspond to the technical field information and specific device or method information of the patent, respectively.

[0020] Specifically, a NER named entity recognition model with a SpanBERT structure can be constructed first. Its key structure involves building two network layers after the BERT output sequence to identify and train the start and end position indices of entities. Subsequently, the NER named entity recognition model is trained to recognize the start and end position indices of entities within the technical field. Finally, the trained NER named entity recognition model is used to perform named entity recognition on the technical field portion of the patent document, and the identified entities are then labeled with subject tags or domain tags. In this embodiment, the domain tags are mainly generated based on the domain information recorded in the patent document, such as semiconductor, computer, communication, and medical fields. The subject tags are devices and / or methods with specific functions, or they can be specific equipment or systems, typically corresponding to the patent name, such as transistor fabrication methods, laptops, mobile terminals with camera functions, and fixing clamps.

[0021] S20. Based on the named entity recognition results, all patent documents are divided into Class I patent documents and Class II patent documents.

[0022] In practical applications, some patent documents may not describe the field information in the technical field section. Therefore, the corresponding field label may not be identifiable in these patent documents. However, the technical field section will definitely record the subject of the disclosed solution. Therefore, each patent document should be able to identify the subject label. Based on the named entity recognition results, all patent documents are divided into Class I patent documents and Class II patent documents. Class I patent documents are those that identify both the field label and the subject label, while Class II patent documents are those that only identify the subject label.

[0023] S30, construct a mapping dictionary between all first subject tags appearing in the first type of patent documents and at least one first field tag corresponding to the first subject tag.

[0024] For all first subject tags that have been identified as domain tags and subject tags, a mapping dictionary is constructed based on the extracted domain and corresponding specific device / method information. That is, a mapping dictionary is established based on the mapping relationship between each first subject tag appearing in the first type of patent document and at least one corresponding first domain tag.

[0025] It is important to note that in a single patent document, the subject label identified by the model typically corresponds to only one domain label. However, in multiple patent documents, the same subject label may correspond to different domain labels. Therefore, when building the mapping dictionary, it is essential to first determine in the first type of patent documents whether there are other first subject labels with the same function as the first subject label. If other first subject labels with the same function exist, a list of first domain labels corresponding to the first subject label can be constructed in the mapping dictionary. This list should include at least the first domain labels corresponding to the first subject label, as well as the first domain labels corresponding to the other subject labels. For example, the mapping dictionary format could be: {“Mobile terminal with shooting function”: [communication domain, photography domain, mobile terminal domain, etc.]}.

[0026] S40, based on each second subject tag in the second type of patent document, determine the first subject tag that is most similar to the second subject tag in the mapping dictionary, and assign at least one first domain tag corresponding to the first subject tag to the second subject tag.

[0027] Once the mapping dictionary for all Class I patent documents is constructed, the corresponding domain labels for each second subject label in Class II patent documents can be determined based on the mapping dictionary, so as to allocate specific domain information for each patent document.

[0028] Specifically, based on the constructed mapping dictionary, for each second subject label, the most similar first subject label can be determined from the mapping dictionary. Given the similarity in their subject labels, their corresponding domain labels should also be similar. Therefore, at least one first domain label corresponding to the most similar first subject label in the mapping dictionary is assigned to that second subject label as at least one second domain label. After performing the above processing on all second-class patent documents, the applicable domain information for each patent document can be determined.

[0029] In some embodiments, similarity tag lookup can be performed quickly in a mapping dictionary through encoding. Specifically, firstly, all first subject tags in the mapping dictionary are converted into corresponding first word vectors, which can be done using BERT (Browser Interchangeable Vector) with unsupervised fine-tuning based on domain information data. Then, all first word vectors are clustered and converted into indexes to form an index library, which can be implemented using the Faiss library tool. Then, using the same conversion method, each second subject tag in all second-class patent documents is converted into a corresponding second word vector. Subsequently, the second word vectors are sequentially input into the constructed index library for retrieval, determining the index most similar to the second word vector, and converting the index into the corresponding first word vector. At this point, the index most similar to the second word vector is obtained. The first word vector can be used to determine the first subject label most similar to the second subject label corresponding to the second word vector. The mapping dictionary is then searched based on the first subject label to determine at least one first domain label corresponding to the first subject label. After that, at least one first domain label is assigned to the second subject label, thus completing the assignment of domain labels for the current second subject label. This enables the process of finding the domain that a device or method that has not extracted domain information should originally correspond to based on the device or method with existing domain information. At the same time, based on clustering and indexing, more comprehensive specific domain information can also be obtained for patents whose domains have been extracted but whose similar devices are also applied in other specific domains.

[0030] S50, based on the first field label corresponding to the second subject label, establish the correspondence between the patent number of each patent document and its corresponding first field label, forming a patent document database based on the technical field.

[0031] After all patent documents have been assigned their corresponding field labels, a mapping between the patent number of each patent document and its corresponding first field label can be established, forming a patent document database based on the technical field. In the actual establishment process, the patent number can be the application number or publication number of the patent document, and the mapping between each patent number and the field can be stored in matrix form.

[0032] This embodiment analyzes patent documents using more granular field information recorded in the technical field section of the patent documents, and assigns field tags to patent documents that have not been associated with field information, so as to achieve a complete association between patent numbers and technical fields. This results in a database with more granular field division, which is more conducive to improving the accuracy and professionalism of subsequent searches.

[0033] Based on the same inventive concept, the second embodiment of this disclosure provides a patent document database construction device based on the technical field, the structural schematic diagram of which is shown below. Figure 2As shown, the system includes at least a named entity recognition module 10, used to perform named entity recognition on the technical field portion of the patent documents based on a named entity recognition model, to identify the subject tag and field tag of each patent document; a classification module 20, used to classify all patent documents into a first category of patent documents and a second category of patent documents according to the named entity recognition results, wherein the first category of patent documents are those whose field tags and subject tags are identified, and the second category of patent documents are those whose subject tags are identified only; a dictionary construction module 30, used to construct a mapping dictionary between all first subject tags appearing in the first category of patent documents and at least one first field tag corresponding to the first subject tag; a tag allocation module 40, used to determine the first subject tag most similar to the second subject tag in the mapping dictionary based on each second subject tag in the second category of patent documents, and to allocate at least one first field tag corresponding to the first subject tag to the second subject tag; and a database construction module 50, used to establish a correspondence between the patent number of each patent document and its corresponding first field tag according to the first field tag corresponding to the second subject tag, forming a patent document database based on the technical field.

[0034] In some embodiments, the tag allocation module 40 is specifically used to: convert all first subject tags in the mapping dictionary into first word vectors; cluster all first word vectors and convert them into indexes to form an index library; convert each second subject tag in all second-class patent documents into a second word vector; search for each second word vector in the index library, determine the index most similar to the second word vector, and determine the first word vector corresponding to the index; convert the first word vector into a first subject tag, and determine at least one first domain tag corresponding to the first subject tag in the mapping dictionary; and assign at least one first domain tag to the second subject tag.

[0035] In some embodiments, the named entity recognition module 10 is specifically used to: construct a NER named entity recognition model with a SpanBert structure; train the NER named entity recognition model to recognize the start and end position indices of entities in the technical field section; use the trained NER named entity recognition model to perform named entity recognition on the technical field section of the patent document, and recognize the main label or domain label on the recognized entities.

[0036] In some embodiments, the dictionary construction module 30 is specifically used to: determine in a first type of patent document whether there are other first subject tags with the same function as the first subject tag; if there are other first subject tags with the same function as the first subject tag, construct a list of first domain tags corresponding to the first subject tag in the mapping dictionary, wherein the list of first domain tags includes at least the first domain tag corresponding to the first subject tag and the first domain tags corresponding to other subject tags.

[0037] This embodiment analyzes patent documents using more granular field information recorded in the technical field section of the patent documents, and assigns field tags to patent documents that have not been associated with field information, so as to achieve a complete association between patent numbers and technical fields. This results in a database with more granular field division, which is more conducive to improving the accuracy and professionalism of subsequent searches.

[0038] Based on the same inventive concept, the third embodiment of this disclosure provides a storage medium storing a computer program that, when executed by a processor, implements the steps of the patent document database construction method based on the technical field described in the first embodiment of this disclosure.

[0039] Based on the same inventive concept, the fourth embodiment of this disclosure provides an electronic device, including at least a memory and a processor. The memory stores a computer program, and when the processor executes the computer program in the memory, it implements the steps of the patent document database construction method based on the technical field described in the first embodiment of this disclosure.

[0040] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this disclosure, and are not intended to limit them. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this disclosure.

Claims

1. A method for constructing a patent document database based on the technical field, characterized in that, include: Named entity recognition is performed on the technical field section of the patent documents based on the named entity recognition model to identify the main tag and field tag of each patent document; Based on the named entity recognition results, all the patent documents are divided into a first category of patent documents and a second category of patent documents. The first category of patent documents are those that identify both the domain label and the subject label, while the second category of patent documents are those that only identify the subject label. Construct a mapping dictionary between all first subject tags appearing in the first type of patent documents and at least one first domain tag corresponding to the first subject tag; Based on each second subject tag in the second type of patent document, a first subject tag that is most similar to the second subject tag is determined in the mapping dictionary, and at least one first domain tag corresponding to the first subject tag is assigned to the second subject tag; Based on the first field label corresponding to the second subject label, a correspondence is established between the patent number of each patent document and its corresponding first field label, forming a patent document database based on the technical field.

2. The patent document database construction method according to claim 1, characterized in that, The step of determining the first subject tag most similar to the second subject tag in the mapping dictionary based on each second subject tag in the second type of patent documents, and assigning at least one first domain tag corresponding to the first subject tag to the second subject tag, includes: Transform all first subject labels in the mapping dictionary into first word vectors; After clustering all the first word vectors, they are converted into an index to form an index library; Transform each second subject tag in all second-class patent documents into a second word vector; For each of the second word vectors in the index library, the index most similar to the second word vector is determined, and the first word vector corresponding to the index is determined. The first word vector is converted into a first subject label, and at least one first domain label corresponding to the first subject label is determined in the mapping dictionary; Assign at least one of the first domain tags to the second body tag.

3. The patent document database construction method according to claim 1, characterized in that, The method of performing named entity recognition on the technical field portion of patent documents based on the named entity recognition model includes: Construct a NER named entity recognition model with a SpanBert structure; The NER named entity recognition model is trained to recognize the start and end position indices of entities in the technical field. The trained NER named entity recognition model is used to perform named entity recognition on the technical field part of the patent document, and the identified entities are then identified by subject label or domain label.

4. The patent document database construction method according to any one of claims 1 to 3, characterized in that, The construction of a mapping dictionary between all first subject tags appearing in the first type of patent documents and at least one first domain tag corresponding to the first subject tag includes: In the first type of patent documents, determine whether each first subject tag has other first subject tags that have the same function as the first subject tag; In the case where there are other first subject tags that have the same function as the first subject tag, a list of first domain tags corresponding to the first subject tag is constructed in the mapping dictionary. The list of first domain tags includes at least the first domain tag corresponding to the first subject tag and the first domain tag corresponding to the other first subject tags.

5. A patent document database construction device based on the technical field, characterized in that, include: The named entity recognition module is used to perform named entity recognition on the technical field portion of the patent document based on the named entity recognition model, so as to identify the main tag and field tag of each patent document; The classification module is used to classify all the patent documents into a first category of patent documents and a second category of patent documents based on the named entity recognition results. The first category of patent documents are those that have identified domain labels and subject labels, while the second category of patent documents are those that have only identified subject labels. A dictionary construction module is used to construct a mapping dictionary between all first subject tags appearing in the first type of patent documents and at least one first domain tag corresponding to the first subject tag; A tag allocation module is used to determine, based on each second subject tag in the second type of patent document, a first subject tag that is most similar to the second subject tag in the mapping dictionary, and to allocate at least one first domain tag corresponding to the first subject tag to the second subject tag. The database construction module is used to establish a correspondence between the patent number of each patent document and its corresponding first field label based on the first field label corresponding to the second subject label, thereby forming a patent document database based on the technical field.

6. The patent document database construction apparatus according to claim 5, characterized in that, The tag allocation module is specifically used for: Transform all first subject labels in the mapping dictionary into first word vectors; After clustering all the first word vectors, they are converted into an index to form an index library; Transform each second subject tag in all second-class patent documents into a second word vector; For each of the second word vectors in the index library, the index most similar to the second word vector is determined, and the first word vector corresponding to the index is determined. The first word vector is converted into a first subject label, and at least one first domain label corresponding to the first subject label is determined in the mapping dictionary; Assign at least one of the first domain tags to the second body tag.

7. The patent document database construction apparatus according to claim 5, characterized in that, The named entity recognition module is specifically used for: Construct a NER named entity recognition model with a SpanBert structure; The NER named entity recognition model is trained to recognize the start and end position indices of entities in the technical field. The trained NER named entity recognition model is used to perform named entity recognition on the technical field part of the patent document, and the identified entities are then identified by subject label or domain label.

8. The patent document database construction apparatus according to any one of claims 5 to 7, characterized in that, The dictionary construction module is specifically used for: In the first type of patent documents, determine whether each first subject tag has other first subject tags that have the same function as the first subject tag; In the case where there are other first subject tags that have the same function as the first subject tag, a list of first domain tags corresponding to the first subject tag is constructed in the mapping dictionary. The list of first domain tags includes at least the first domain tags corresponding to the first subject tag and the first domain tags corresponding to the other subject tags.

9. A storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the patent document database construction method based on the technical field as described in any one of claims 1 to 4.

10. An electronic device, comprising at least a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program on the memory, it implements the steps of the patent document database construction method based on any one of claims 1 to 4.

Images