Document indexing method and device, storage medium and electronic device

Abstract

The present disclosure discloses a document indexing method and device, a standardized configuration and local deployment method and device, and relates to the document indexing technology field. The document indexing method comprises: based on a target category label, determining a first document set that matches the target category label; based on the first document set, training and generating an indexing model; using the indexing model, indexing a second document set to obtain the indexing results of each document included in the second document set. The present disclosure obtains the first document set through the target category label, which can reduce the complexity of the indexing process and the time spent on indexing, completes the indexing using the indexing model, obtains the indexing results of each document, and improves the accuracy of indexing and the utilization rate of document resources.

Description

Technical Field

[0001] The present disclosure relates to the technical field of text processing, and in particular to a document indexing method and device, a storage medium and an electronic device. Background Art

[0002] As time goes by, the number of documents continues to increase, especially the number of patent documents. Patent document indexing can help people use patent documents more effectively, and can also help enterprises improve their competitive advantages and grasp the development of the industry. Therefore, patent document indexing plays an irreplaceable and important role. Patent document indexing is to classify patent documents according to a certain dimension according to user needs. The number of patent documents under one dimension is tens of thousands. Patent document indexing is a complicated and time-consuming task. Therefore, a more efficient and fast document indexing method is urgently needed to help enterprises complete patent indexing work. Summary of the invention

[0003] In view of this, the present disclosure provides a document indexing method and device, a storage medium and an electronic device to solve the problem that document indexing is time-consuming and inefficient.

[0004] In a first aspect, an embodiment of the present disclosure provides a document indexing method, the method comprising: determining a first document set matching a target category label based on a target category label; training and generating an indexing model based on the first document set; and using the indexing model to index a second document set to obtain indexing results for each document included in the second document set.

[0005] In combination with the first aspect, in certain implementations of the first aspect, based on the target category label, determining a first document set that matches the target category label includes: performing word segmentation on the target category label to obtain a word segmentation result of the target category label; based on a second document set, determining document feature information of each document included in the second document set; based on the word segmentation result of the target category label and the document feature information of each document included in the second document set, generating matching degree information between each document included in the second document set and the target category label; based on the matching degree information between each document included in the second document set and the target category label, determining the first document set.

[0006] In combination with the first aspect, in certain implementations of the first aspect, based on the matching information between each document included in the second document set and the target category label, determining the first document set includes: based on the matching information between each document included in the second document set and the target category label, determining the first M documents to be indexed that have the highest matching degree with the target category label as the first document set; or, based on the matching information between each document included in the second document set and the target category label, determining the first S documents to be indexed that have the highest matching degree with the target category label, where S is a positive integer greater than M; and verifying and eliminating the S documents to be indexed to obtain the first document set.

[0007] In combination with the first aspect, in certain implementations of the first aspect, based on the target category label, determining a first set of documents that matches the target category label includes: obtaining document data that is manually screened based on the target category label and matches the target category label, and determining the first set of documents that matches the target category label.

[0008] In combination with the first aspect, in certain implementations of the first aspect, after using the indexing model to index the second document set to obtain the indexing results of each document included in the second document set, it also includes: obtaining the verification and elimination records corresponding to the first document set; and correcting the indexing results of each document included in the second document set based on the verification and elimination records.

[0009] In combination with the first aspect, in certain implementations of the first aspect, an indexing model is trained and generated based on the first document set, including: generating a training data set and a verification data set based on the label confidence of each document included in the first document set; training and verifying a neural network classification model based on the training data set and the verification data set to obtain an indexing model.

[0010] In combination with the first aspect, in certain implementations of the first aspect, a training data set and a verification data set are generated based on the label confidence of each document contained in the first document set, including: based on the label confidence of each document contained in the first document set, determining the documents in the first document set that fall within a preset confidence trust interval; based on the documents in the first document set that fall within the preset confidence trust interval, generating a training data set and a verification data set.

[0011] In combination with the first aspect, in certain implementations of the first aspect, a training data set and a verification data set are generated based on documents in the first document set that fall within a preset confidence trust interval, including: if there are documents in the first document set that fall within a preset confidence question interval, the documents in the first document set that fall within the preset confidence question interval are sent to target users for screening to obtain documents in the first document set that meet the training requirements; a training data set and a verification data set are generated based on documents in the first document set that meet the training requirements and documents in the first document set that fall within the preset confidence trust interval.

[0012] In combination with the first aspect, in certain implementations of the first aspect, the indexing model is used to index the second document set to obtain indexing results for each document included in the second document set, including: using the indexing model to index some of the documents to be indexed in the second document set to obtain indexing results for each document in the some of the documents to be indexed; based on correction information for the indexing results of each document in the some of the documents to be indexed, correcting the indexing results of each document in the some of the documents to be indexed; based on the corrected some of the documents to be indexed as the documents in the new first document set, continuing to train the indexing model; and using the continuously trained indexing model to index the unindexed documents to be indexed in the second document set.

[0013] In combination with the first aspect, in certain implementations of the first aspect, before training and generating an indexing model based on the first document set, it also includes: obtaining indexing request information; if the indexing request information meets the preset requirements, sending request success information to the target client; if the indexing request information does not meet the preset requirements, sending request failure information to the target client; wherein, training and generating an indexing model based on the first document set includes: after sending the request success information to the target client, training and generating an indexing model based on the first document set.

[0014] In combination with the first aspect, in certain implementations of the first aspect, the first document set includes seed patent documents, and the second document set includes patent documents to be indexed; and / or, the first document set includes seed technical documents, and the second document set includes technical documents to be indexed.

[0015] In a second aspect, an embodiment of the present disclosure provides another document indexing method, which obtains a document database to be indexed; and uses the document indexing method mentioned in the first aspect to perform document indexing operations on the document database to be indexed.

[0016] In a third aspect, an embodiment of the present disclosure provides another document indexing method, which obtains a document indexing requirement; based on the document indexing requirement, determines a plurality of documents to be indexed in a target technical field corresponding to the document indexing requirement; uses the method mentioned in the first aspect to perform indexing operations on the plurality of documents to be indexed, and obtains indexing results for each of the plurality of documents to be indexed.

[0017] In a fourth aspect, an embodiment of the present disclosure provides another document indexing method, which obtains a first document set that is manually screened based on a target category label and matches a target category label; sends the first document set to an indexing end so that the indexing end generates an indexing model based on training of the first document set, and uses the indexing model to index a second document set to obtain indexing results for each of the documents included in the second document set.

[0018] In a fifth aspect, an embodiment of the present disclosure provides a document indexing device, which includes: a determination module, used to determine a first document set that matches a target category label based on the target category label; a training module, used to train and generate an indexing model based on the first document set; and an indexing module, used to use the indexing model to index a second document set to obtain indexing results for each document contained in the second document set.

[0019] In a sixth aspect, an embodiment of the present disclosure provides another document indexing device, which includes: an acquisition module for acquiring a document database to be indexed; and an indexing module for performing document indexing operations on the document database to be indexed using the document indexing method mentioned in the first aspect.

[0020] In a seventh aspect, an embodiment of the present disclosure provides another document indexing device, which includes: an acquisition module for acquiring document indexing requirements; a determination module for determining, based on the document indexing requirements, a plurality of documents to be indexed in a target technical field corresponding to the document indexing requirements; and an indexing module for performing indexing operations on the plurality of documents to be indexed using the method mentioned in the first aspect to obtain indexing results for each of the plurality of documents to be indexed.

[0021] In an eighth aspect, an embodiment of the present disclosure provides another document indexing device, which comprises: an acquisition module, used to acquire a first document set that is manually screened based on a target category label and matches a target category label; a sending module, used to send the first document set to an indexing end, so that the indexing end generates an indexing model based on training of the first document set, and uses the indexing model to index a second document set to obtain indexing results for each document included in the second document set.

[0022] In a ninth aspect, an embodiment of the present disclosure provides an electronic device, comprising a processor; and a memory for storing instructions executable by the processor, wherein the processor is used to execute the method mentioned in the first aspect above.

[0023] In a tenth aspect, an embodiment of the present disclosure provides a computer-readable storage medium, which stores a computer program, and the computer program is used to execute the method mentioned in the first aspect above.

[0024] The document indexing method provided by the present invention obtains documents in a first document set that match the target category label through a target category label, obtains an indexing model through the documents in the first document set, and uses the model to index the second document set to be indexed. Only the target category label is needed to index the patent documents to be indexed, which can shorten the indexing time. Indexing through the trained model can improve the indexing accuracy, improve the indexing efficiency and resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The above and other purposes, features and advantages of the present disclosure will become more apparent by describing the embodiments of the present disclosure in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and do not constitute a limitation of the present disclosure.

[0026] Figure 1 Shown is a schematic diagram of an application scenario provided by an embodiment of the present disclosure.

[0027] Figure 2 Shown is a flow chart of a document indexing method provided by an embodiment of the present disclosure.

[0028] Figure 3 FIG. 1 is a flow chart of determining a first document set matching a target category label based on a target category label provided by an embodiment of the present disclosure.

[0029] Figure 4 The figure is a flow chart of determining the first document set based on the matching degree information between each document contained in the second document set and the target category label provided by an embodiment of the present disclosure.

[0030] Figure 5 Shown is a flow chart of another document indexing method provided by an embodiment of the present disclosure.

[0031] Figure 6 The figure is a schematic diagram of a process of training and generating an indexing model based on a first document set provided by an embodiment of the present disclosure.

[0032] Figure 7 The figure shows a flow chart of generating a training data set and a validation data set based on the label confidence of each document included in the first document set according to an embodiment of the present disclosure.

[0033] Figure 8 The figure shows a flow chart of generating a training data set and a validation data set based on documents in a first document set falling within a preset confidence interval, provided by an embodiment of the present disclosure.

[0034] Fig. 9 The figure shows a flow chart of using an indexing model to index a second document set and obtain indexing results of respective documents contained in the second document set, provided by an embodiment of the present disclosure.

[0035] Fig.10 Shown is a flow chart of another document indexing method provided by an embodiment of the present disclosure.

[0036] Fig.11 The figure is a flowchart of a standardized configuration method provided by an embodiment of the present disclosure.

[0037] Fig.12 Shown is a flow chart of a localized deployment method provided by an embodiment of the present disclosure.

[0038] Fig.13 Shown is a flow chart of another document indexing method provided by an embodiment of the present disclosure.

[0039] Fig.14 Shown is a flow chart of another document indexing method provided by an embodiment of the present disclosure.

[0040] Fig.15 Shown is a schematic structural diagram of a document indexing device provided by an embodiment of the present disclosure.

[0041] Fig.16 Shown is a schematic structural diagram of another document indexing device provided by an embodiment of the present disclosure.

[0042] Fig.17 Shown is a schematic structural diagram of another document indexing device provided by an embodiment of the present disclosure.

[0043] Fig.18 Shown is a schematic diagram of the structure of a standardized configuration device provided by an embodiment of the present disclosure.

[0044] Fig.19 Shown is a schematic diagram of the structure of a localized deployment device provided by an embodiment of the present disclosure.

[0045] Fig. 20 Shown is a schematic structural diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments.

[0047] As time goes by, the number of documents continues to increase, especially the number of patent documents. Therefore, searching, indexing and analyzing patent documents has become an important task. Patent document indexing can help people use patent documents effectively, and can help companies understand the position of a technology in the industry, grasp the future development trend of a field, the future development layout of the industry, and help companies improve their competitiveness. Patent document indexing is to classify patent documents according to a certain dimension according to user needs. The number of patent documents under one dimension is tens of thousands. Under the same dimension, a patent can have multiple labels. It can be seen that patent document indexing is a complicated and time-consuming task.

[0048] At present, common patent document indexing methods generally use keyword matching and TF-IDF matching algorithms. The keyword matching method often has a very low accuracy rate, and manual summary of keywords is not only time-consuming and labor-intensive, but also requires professional knowledge in the field to summarize high-quality keywords. Compared with keyword matching, the TF-IDF matching algorithm has improved accuracy, but it still cannot meet user needs. In addition, when the number of patents is hundreds of thousands or even millions, the indexing efficiency is relatively low. Therefore, there is an urgent need for a more efficient and fast document indexing method to help companies complete patent indexing work.

[0049] Combine the following Figure 1 The application scenario of an embodiment of the present disclosure is briefly introduced.

[0050] Figure 1 FIG. 1 is a schematic diagram of an application scenario of an embodiment of the present disclosure. Figure 1 As shown, this scenario is a scenario of indexing Category A documents. Specifically, the patent indexing scenario includes a server 110, a user terminal 120 and a data storage device 130 respectively connected to the server 110 for communication, and the server 110 is used to execute the document indexing method mentioned in the embodiment of the present disclosure.

[0051] Exemplarily, in actual application, the user uses the user terminal 120 to send a target tag and an instruction to index Category A documents to the server 110. After receiving the instruction, the server 110 extracts the Category A documents stored in the data storage device 130, and processes the Category A documents to obtain a first document set of documents in the first document set that matches the target tag, and then trains and generates an indexing model based on the first document set, uses the indexing model to index all Category A documents, and then outputs the Category A document indexing results to the user terminal 120 so that the user terminal 120 presents the results to the user.

[0052] Exemplarily, the user terminal 120 mentioned above includes but is not limited to computer terminals such as desktop computers and laptop computers and mobile terminals such as tablet computers and mobile phones. The data stored in the data storage device 130 mentioned above includes but is not limited to patent document data, technical document data, etc.

[0053] In another application scenario disclosed in this embodiment, it includes an Internet Protocol address (IP) end, a research and development end and an audit end connected to the IP end respectively; the research and development end is used to perform operations based on the target category label, determine the first document set matching the target category label and receive indexing instructions; the IP end is used to receive the first document set sent by the research and development end and train and generate an indexing model, and use the indexing model to index the second document set to obtain the indexing results of the second document set; the audit end is used to receive the indexing results of the second document indexing set from the IP end, audit the indexing results, and obtain the final indexing results of each document in the second document set. In some embodiment application scenarios, the audit end is also used to receive indexing instructions, receive the first document set from the research and development end, and train and generate an indexing model to obtain the final indexing results.

[0054] Combine the following Figures 2 to 12 The document indexing method disclosed in the present invention is briefly introduced.

[0055] Figure 2 FIG. 1 is a flow chart of a document indexing method provided by an embodiment of the present disclosure. Figure 2 As shown, the document indexing method provided by the embodiment of the present disclosure includes the following steps.

[0056] Step S210: Based on the target category label, determine a first document set matching the target category label.

[0057] Exemplarily, patent document data is selected as the document data to be indexed, and according to the target category label defined by the target user, the first document set matching the target category label is determined manually or by calculation. The target category label is the name of the different categories of document data divided by the user under a specific dimension, which reflects the unique characteristic information of the category and cannot be defined by symbols such as numbers or letters that have no distinguishing meaning. Specifically, the target category label cannot be defined by symbols such as "category 1" and "category 2" that have no distinguishing meaning. It should be understood that the above-mentioned document data to be indexed includes but is not limited to patent documents, scientific journals, conference documents, scientific reports, dissertations and other documentary materials.

[0058] Step S220: training and generating an indexing model based on the first document set.

[0059] Exemplarily, a training data set and a validation data set are generated based on the first document set obtained above, and then the neural network classification model is trained to obtain an indexing model.

[0060] Step S230: index the second document set using the indexing model to obtain indexing results of the documents included in the second document set.

[0061] Extract some documents from the second document set as documents in the first document set, use a small number of documents in the first document set to train the model, and obtain an indexing model. Use the indexing model to index a large number of second document sets. By obtaining a small number of documents in the first document set, a large number of documents to be indexed can be indexed.

[0062] According to the disclosed embodiment, the indexing result of the patent to be indexed can be obtained through the target category label defined by the user and the indexing model, thereby reducing the complexity of document indexing and improving the efficiency of indexing. In addition, since the indexing model is obtained by training the classification model using the documents in the first document set, and the indexing is performed using the indexing model, the utilization rate of the patent document data in the indexing process is improved.

[0063] Figure 3 FIG. 1 is a flow chart of determining a first document set matching a target category label based on a target category label provided by an embodiment of the present disclosure. Figure 3 As shown, based on the target category label, determining the first document set matching the target category label includes the following steps.

[0064] Step S310, segmenting the target category label to obtain a segmentation result of the target category label.

[0065] Exemplarily, the target category label defined by the target user is segmented to obtain a segmentation result of the target category label.

[0066] Step S320: Based on the second document set, determine that the second document set contains document feature information of each document.

[0067] Exemplarily, the document to be indexed is a patent document, and correspondingly, the document feature information includes information such as the patent name, abstract, claims, and keywords of the specification.

[0068] Step S330 , based on the word segmentation result of the target category label and the document feature information of each document contained in the second document set, generate the matching degree information of each document contained in the second document set and the target category label.

[0069] Exemplarily, based on the word segmentation results of the target category label and the second document set containing the document feature information of each document, the ES method or the Best Match 25 (BM25) method is used to generate the second document set containing the matching degree information of each document with the target category label. For example, all patents to be indexed are indexed into ES, where the information of the indexed patents includes the keywords of the patent name, abstract, claims and instructions, and then a request is generated based on the word segmentation results of the target category label, and the request is searched in ES to generate the matching degree information of each document with the target category label. Alternatively, the BM25 method is used to extract the keywords of the patent name, abstract, claims and instructions of each patent to be indexed based on the word segmentation results of the target category label, and the BM25 score is calculated for the keywords and category labels of each patent to be indexed, and the second document set is generated containing the matching degree information of each document with the target category label.

[0070] Step S340: determining the first document set based on the matching degree information between each document contained in the second document set and the target category label.

[0071] Exemplarily, M patents with relatively high scores in the ES algorithm or the BM25 algorithm are used as the first document set under the category label.

[0072] Since the documents in the first document set are obtained through an algorithm, documents with higher scores are selected as seed patents. When faced with a large number of indexed documents, it is possible to respond quickly and obtain seed patents quickly, thus shortening the time for document indexing and improving the efficiency of document indexing.

[0073] In another embodiment of the present application, the algorithm is configured in the cold start module to determine the first document set that matches the target category label based on the target category label. The algorithm set in the cold start module can index some documents and label the documents with category labels according to the category label name defined by the user, without the need for the user to mark the document data, that is, there is no need to manually index the data for each category label from 0, only the category label name defined by the user can be used to find the patent closest to the label from a large number of seed patents. The cold start module can help users quickly and conveniently obtain seed data, saving users' time, and can provide data support for subsequent model training.

[0074] Another document indexing method provided in one embodiment of the present disclosure includes, based on a target category label, determining a first document set that matches the target category label, including: based on the target category label, obtaining document data that matches the target category label and is manually screened based on the target category label, and determining the first document set that matches the target category label. It should be understood that obtaining document data that matches the target category label and is manually screened based on the target category label can be screened by manual indexing or manually screened by an algorithm, and the present disclosure does not limit the specific method of screening; the first document set obtained can be document data in the document to be indexed or document data in a related field, and the present disclosure does not make specific limitations on the source of the first document set data.

[0075] Figure 4 FIG. 1 is a flow chart of determining a first document set based on the matching degree information between each document contained in the second document set and the target category label provided by an embodiment of the present disclosure. Figure 4 As shown, the step of determining the first document set based on the matching degree information between each document contained in the second document set and the target category label provided by the embodiment of the present disclosure includes the following steps.

[0076] Step S410, based on the matching degree information between each document in the second document set and the target category label, determine the first S documents to be indexed that have the highest matching degree with the target category label.

[0077] Exemplarily, based on the matching degree information between each document contained in the second document set and the target category label, the top S patent documents to be indexed with relatively high scores in the ES method and the BM25 algorithm are obtained.

[0078] Step S420, verify and eliminate the S documents to be indexed to obtain a first document set.

[0079] Exemplarily, S patent documents to be indexed are verified and eliminated to achieve manual correction, thereby eliminating patents that do not belong to the category under the target category label.

[0080] Alternatively, in this embodiment, the step of determining the first document set is to determine the first M documents to be indexed with the highest matching degree with the target category label as the first document set based on the matching degree information between each document contained in the second document set and the target category label, where M is less than S.

[0081] The embodiment of the present disclosure improves the accuracy of data by means of verification and elimination processing. In addition, for patents with incorrect algorithm recalls, the embodiment of the present disclosure can eliminate them, thereby improving the accuracy of indexing.

[0082] Figure 5FIG. 2 is a flow chart of another document indexing method provided by an embodiment of the present disclosure. Figure 5 As shown, in the embodiment of the present disclosure, after the indexing model is used to index the second document set to obtain the indexing results of the documents contained in the second document set, the following steps are also included.

[0083] Step S510, obtaining the verification and rejection record corresponding to the first document set.

[0084] Exemplarily, the result of the above manual correction is obtained, that is, the verification and elimination record corresponding to the first document set.

[0085] Step S520: correct the indexing results of the documents included in the second document set by eliminating records based on the verification.

[0086] Exemplarily, based on the manually corrected verification and rejection records, the indexing results of the respective documents included in the second document set are corrected.

[0087] The indexing results are corrected again based on the verification and elimination records. It is necessary to correct the erroneous patents in the indexing results based on the manually eliminated records, so as to ensure that the manually modified results will not be overwritten, thereby improving the accuracy of the indexing results.

[0088] Figure 6 FIG. 1 is a flow chart of training and generating an indexing model based on a first document set according to an embodiment of the present disclosure. Figure 6 As shown, based on the first document set, training and generating an indexing model includes the following steps.

[0089] Step S610: Generate a training data set and a validation data set based on the confidence of each label of the documents included in the first document set.

[0090] Exemplarily, the first document set is selected to contain document data with high confidence of each document label. Specifically, the document data with confidence of 70% or more of the document and label in the first document set are divided to generate a training data set and a validation data set of the model. It should be understood that as long as the range of the selected confidence can meet the training of the model, the present disclosure does not specifically limit the range of the selected confidence.

[0091] Step S620: training and verifying a neural network classification model based on the training data set and the verification data set to obtain an indexing model.

[0092] Exemplarily, a neural network classification model is trained and verified based on a training data set and a verification data set. Specifically, the neural network classification model can be any model of TextCNN, FastText, and BLSTM. Exemplarily, the training process of the model is the process of fitting the training samples.

[0093] Selecting the document data of the first document set containing the document labels with high confidence as the training data set and the verification data set can improve the accuracy of the indexing model, thereby further ensuring the accuracy of the indexing results.

[0094] Figure 7 FIG. 1 is a flow chart of generating a training data set and a validation data set based on the confidence of each label of the documents included in the first document set according to an embodiment of the present disclosure. Figure 7 As shown, based on the confidence of each label of the documents contained in the first document set, a training data set and a verification data set are generated, including the following steps.

[0095] Step S710, based on the confidence of each label of the documents included in the first document set, determine the documents in the first document set that fall within a preset confidence trust interval.

[0096] Exemplarily, the preset confidence trust interval can be determined according to the actual situation, or can be set according to the user's requirements. Specifically, the preset confidence trust interval is set according to the actual obtained confidence of each label of the documents contained in the first document set. It can be set according to the actual obtained confidence of each label of the documents contained in the first document set. For example, if the label confidence of the first document set is 80% as the highest confidence, and the label confidence of the documents in the first document set is more than 70%, the confidence trust interval is set to [0.7, 0.8]; if the user requires that the confidence of the documents in the first document set is more than 80%, the confidence trust interval is set to [0.8, 1]. It should be understood that the confidence trust interval can be set according to the actual situation and different requirements.

[0097] Step S720: Generate a training data set and a verification data set based on documents in the first document set that fall within a preset confidence interval.

[0098] The pre-set confidence interval can improve the reliability of the documents in the first document set, and can further improve the accuracy of the subsequent indexing model, thereby further improving the accuracy of the indexing results.

[0099] Figure 8 FIG. 1 is a flow chart of generating a training data set and a validation data set based on the confidence of each label of the documents included in the first document set according to an embodiment of the present disclosure. Figure 8 As shown, based on the confidence of each label of the documents contained in the first document set, a training data set and a verification data set are generated, including the following steps.

[0100] Step S810: If there are documents in the first document set that fall within the preset credibility question interval, the documents in the first document set that fall within the preset credibility question interval are sent to the target user for screening to obtain documents in the first document set that meet the training requirements.

[0101] Exemplarily, the confidence of the first document set is determined, and a confidence question interval is set. Assuming that there are documents in the first document set that fall within the preset confidence question interval, the documents in the first document set that fall within the preset confidence question interval are sent to target users for screening. The target users screen the received documents in the first document set to obtain the screened documents in the first document set, that is, the documents in the first document set that meet the training requirements.

[0102] Step S820: Generate a training data set and a verification data set based on the documents in the first document set that meet the training requirements and the documents in the first document set that fall within a preset confidence interval.

[0103] Exemplarily, a preset confidence interval is set, and the documents in the first document set in the interval and the documents in the first document set that meet the training requirements are divided according to a certain ratio, such as the ratio of the training data set to the verification data set is 8:2, to generate a training data set and a verification data set. Exemplarily, according to user requirements, the confidence of the document labels in the first document set is required to be above 90%, the confidence trust interval is set to [0.9, 1], and the confidence query interval is set to [0.70, 0.89]. The documents in the first document set whose confidence in the first document set is in the interval of [0.7, 0.8] are sent to the target user for screening to obtain the document data in the first document set that meet the training requirements; the documents in the first document set in the trust interval [0.9, 1] are obtained, and together with the document data in the first document set that meet the requirements, form a training data set and a verification data set. It should be understood that the confidence query interval can be adjusted according to the requirements, and the requirements for the confidence can also be adjusted according to the requirements. The present disclosure does not specifically limit the values ​​of the confidence, the confidence trust interval, and the confidence query interval.

[0104] The document data in the first document set after being screened by the target customers is more accurate. At the same time, the target users screen the documents of the questioned interval seeds, and there is no need to index all the document data, which saves the target users' time, improves the accuracy of the indexing results, reduces the complexity of indexing, and improves the efficiency of indexing.

[0105] Fig. 9 FIG. 1 is a flow chart of using an indexing model to index a second document set to obtain indexing results of each document included in the second document set, provided by an embodiment of the present disclosure. Fig. 9 As shown, the indexing model is used to index the second document set to obtain indexing results of the documents contained in the second document set, including the following steps.

[0106] Step S910, using the indexing model, indexing some of the documents to be indexed in the second document set, and obtaining indexing results for the indexed documents of the documents contained in the partial second document set.

[0107] Exemplarily, by using the trained indexing model, some of the documents to be indexed are selected for indexing, so that indexing results of the documents included in the second document set of some documents can be obtained.

[0108] Step S920: correcting the indexing results of the documents included in the partial second document set based on the correction information for the indexing results of the documents included in the partial second document set.

[0109] Exemplarily, the manually corrected indexing results are recorded, the indexing results of the documents included in the second document set are corrected, corrections are made according to the manually corrected indexing results, and the results of the model indexing are corrected again.

[0110] Step S930: Continue training the indexing model based on the corrected part of the documents to be indexed as documents of the new first document set.

[0111] Exemplarily, based on the result after the second correction, the corrected document indexing result is selected, and the corrected marked document data is used as the document in the new first document set to continue training the indexing model.

[0112] Step S940: use the indexing model that has been continuously trained to index the unindexed documents to be indexed in the second document set.

[0113] Exemplarily, the model after training is used to index the documents in the second document set to complete the indexing and obtain the indexing result. Exemplarily, in order to determine the accuracy of the trained model, an experiment is conducted on the accuracy of the indexing model, and the accuracy is tested when the number of seeds of each tag is 10, 50, 100, 200, and 500. The experimental results are shown in the following Table 1-1.

[0114] Table 1-1 Experimental results data table

[0115] Number of seed patents per category 10 50 100 200 500 Accuracy 66.7 73.3 75.6 79.4 85.0 Recall 45.2 67.5 70.5 76.0 82.1

[0116] It can be seen from Table 1-1 that when the number of seeds is within a reasonable range, the accuracy and recall of the indexing model can be maintained at a high level at the same time, ensuring the accuracy and recall of the indexing model.

[0117] For example, to determine the time consumption problem of indexing, an experimental test is conducted on the indexing model, and the experimental results are shown in Table 1-2 below.

[0118] It can be seen from Table 1-2 that when the total number of seed patents is 100 patents, it only takes 16 seconds to index 5,000 patents to be indexed using the CPU. In extreme cases, there are 50,000 seed patents, and then indexing 50,000 patents takes 10 minutes using the CPU, but if a GPU is used instead, it only takes 2 minutes and 44 seconds. This shows that the indexing method provided by the embodiment of the present disclosure is very fast and saves indexing time.

[0119] Table 1-2 Experimental results data table

[0120]

[0121] The indexing model was corrected twice based on the manual correction results, which further improved the accuracy of the model. At the same time, some of the labeled document data after the second correction was used as the documents in the new first document set to train the model, which improved the resource utilization of the documents. At the same time, the number of documents in the first document set was increased, the accuracy of the model was improved, and the indexing time was saved.

[0122] In some embodiments of the present disclosure, the manually indexed documents are distinguished from the model indexed documents. For example, different colors or different marks are selected to distinguish the manual indexing results from the model indexing results. In the manual correction process, the manual correction indexing results and the model indexing results are also marked differently to distinguish the indexing results.

[0123] Fig.10 FIG. 2 is a flow chart of another document indexing method provided by an embodiment of the present disclosure. Fig.10 As shown, the document indexing method mentioned in the embodiment of the present disclosure includes the following steps.

[0124] Step S1010, obtaining a database of documents to be indexed.

[0125] Exemplarily, information requesting indexing of patent documents is obtained, and the indexing request information includes relevant exemplary data for training the indexing model, such as the indexed patent document input by the target user and the partially indexed patent document data obtained by the front end based on the algorithm.

[0126] Step S1020: If the indexing request information meets the preset requirements, a request success message is sent to the target client.

[0127] Exemplarily, according to the received request information, preset requirements of the request information are set, including requirements of relevant parameters, document data format requirements, etc. If the preset requirements are met, a request success message is sent to the target client.

[0128] Step S1021: If the indexing request information does not meet the preset requirements, a request failure message is sent to the target client.

[0129] Exemplarily, if the received request information does not meet the preset requirements, for example, the relevant parameter information is wrong, the data is empty, etc., which does not meet the preset requirements, a request failure message is sent to the target client.

[0130] Step S1030: after sending the request success information to the target client, training and generating an indexing model based on the first document set.

[0131] Exemplarily, after sending a request success message to the target client, the obtained relevant exemplary data is used as the documents in the first document set, and the neural network classification model is trained using the obtained first document set to obtain an indexing model.

[0132] In the indexing method provided by one embodiment of the present disclosure, the first document set includes seed patent documents, the second document set includes patent documents to be indexed, and / or the first document set includes seed technical documents, and the second document set includes technical documents to be indexed. Exemplarily, the documents to be indexed include patent documents to be indexed. The first document set is obtained by calculation or manual marking, and the first document set is a patent document set; the second document set is technical documents to be indexed, and the first document set is a technical document set.

[0133] Fig.11 FIG. 1 is a flow chart of a standardized configuration method provided by an embodiment of the present disclosure. Fig.11 A standardized configuration method is shown, which is used for standardized configuration of document indexing. The method includes the following steps.

[0134] Step S1110, obtaining a document indexing method.

[0135] Exemplarily, a document indexing method is obtained, where the document indexing method is the indexing method mentioned in any of the above embodiments.

[0136] Step S1120: Based on the document indexing method, a work guide for implementing document indexing is planned.

[0137] Exemplarily, based on the document indexing method, the work of document indexing is planned to implement the operation guidance of document indexing.

[0138] Through the standardized configuration of document indexing, the efficiency of document indexing work can be improved, the time consumption can be reduced, thereby reducing the time cost of document indexing and improving efficiency.

[0139] Fig.12 FIG. 1 is a flow chart of a localized deployment method provided by an embodiment of the present disclosure. Fig.12 A localized deployment method shown is used for localized deployment of document indexing, and the method includes the following steps.

[0140] Step S1210, obtaining a database of documents to be indexed, the database including document indexing results.

[0141] Exemplarily, according to the method mentioned in any of the above embodiments, a document indexing result is obtained.

[0142] Step S1220: Process the document indexing result using the document indexing processing system.

[0143] Localized deployment is based on indexing methods, which can ensure the accuracy and efficiency of indexing. At the same time, completing localized deployment can ensure the security of literature data, meet the needs of different industries and enterprises, expand the scope of use of indexing methods, and make indexing methods more adaptable.

[0144] Fig.13 FIG. 2 is a flow chart of another document indexing method provided by an embodiment of the present disclosure. Fig.13 As shown, the document indexing method mentioned in the embodiment of the present disclosure is used for document indexing based on the technical field, and the method includes the following steps.

[0145] Step S1310, obtaining document indexing requirements.

[0146] Exemplarily, the document indexing requirements are obtained, such as the document indexing requirements in the target technical field. The target technical field may be a technical field formulated according to the requirements of the target users, or different target technical fields may be preset according to the requirements of different industries.

[0147] Step S1320: Based on the document indexing requirement, a plurality of documents to be indexed in the target technical field corresponding to the document indexing requirement are determined.

[0148] Exemplarily, based on the target technical field obtained above, the indexing method mentioned in any of the above embodiments is used to index the documents to be indexed in the target technical field to obtain indexing results.

[0149] Step S1330, using a document indexing method, indexing the plurality of documents to be indexed to obtain indexing results for each of the plurality of documents to be indexed.

[0150] Exemplarily, the document indexing method mentioned in the first aspect is used to perform indexing operations on multiple documents to be indexed, and obtain indexing results for each of the multiple documents to be indexed.

[0151] In response to different industry needs and different user needs, through preset technical fields, the indexing method of this embodiment can perform pre-learning by industry to form indexing products of standard data, or, according to different user needs, pre-learning to form customized indexing products for the user industry, which can improve the efficiency of document indexing in the preset technical field and reduce the tediousness of document indexing.

[0152] Fig.14 FIG. 2 is a flow chart of another document indexing method provided by an embodiment of the present disclosure. Fig.14 As shown, the document indexing method mentioned in the embodiment of the present disclosure is used for document indexing based on manual indexing, and the method includes the following steps.

[0153] Step S1410, obtaining document indexing results.

[0154] Exemplarily, a document indexing result is obtained, where the document indexing result is obtained using the indexing method mentioned in any of the above embodiments.

[0155] Step S1420: Based on the document indexing result, the manual indexing result is verified to determine the document indexing result.

[0156] Exemplarily, manual indexing is performed on the document data before the document indexing to obtain manual indexing results, and based on the obtained document indexing results, the manual indexing results are verified to determine the final document indexing results.

[0157] Manual labeling often requires high-level professional knowledge to obtain high-quality document indexing results, and the results of manual indexing are affected by people's relevant knowledge reserves and cognitive scope. By comparing the model indexing results with the manual indexing results, the accuracy of the manual indexing results can be improved and the scope of the manual indexing results can be expanded. For example, by comparing the model indexing results after manual indexing, indexing results within the cognitive scope that has not been covered by manual labeling can be obtained, thereby expanding the scope of manual indexing results.

[0158] In some embodiments, a document includes multiple tags. The indexing results of the multiple tags can be obtained by performing multiple manual indexing, or after manual indexing, the indexing model indexes again to obtain multiple tags for a document. The manually indexed tags can be single or multiple, and the document tags indexed by the indexing model can be single or multiple.

[0159] Combination of the above Figures 2 to 14 , describes in detail the method embodiment of the present disclosure, and the following is combined with Figures 15 to 20 , describing the device embodiment of the present disclosure. It should be understood that the description of the method embodiment corresponds to the description of the device embodiment, and therefore, the parts not described in detail can refer to the previous method embodiment.

[0160] Fig.15 FIG. 1 is a schematic diagram of a document indexing device according to an embodiment of the present disclosure. Fig.15 As shown, the document determination device 1500 provided by the embodiment of the present disclosure includes a determination module 1510, a training module 1520 and an indexing module 1530. The determination module 1510 is used to determine a first document set matching a target category label based on a target category label. The training module 1520 is used to train and generate an indexing model based on the first document set. The indexing module 1530 is used to index a second document set using the indexing model to obtain indexing results of each document contained in the second document set.

[0161] In some embodiments, the determination module 1510 is also used to segment the target category label to obtain the segmentation result of the target category label; based on the second document set, determine that the second document set contains document feature information of each document; based on the segmentation result of the target category label and the document feature information of each document contained in the second document set, generate the second document set containing the matching degree information of each document with the target category label; based on the matching degree information of each document contained in the second document set with the target category label, determine the first document set.

[0162] In some embodiments, the determination module 1510 is also used to determine the first M documents to be indexed that have a high matching degree with the target category label as the first document set based on the matching degree information between each document contained in the second document set and the target category label; or, based on the matching degree information between each document contained in the second document set and the target category label, determine the first S documents to be indexed that have a high matching degree with the target category label, where S is a positive integer greater than M; and verify and eliminate the S documents to be indexed to obtain the first document set.

[0163] In some embodiments, the determination module 1510 is further used to obtain the verification and elimination records corresponding to the first document set; and correct the indexing results of the documents included in the second document set based on the verification and elimination records.

[0164] In some embodiments, the training module 1520 is further used to generate a training data set and a verification data set based on the confidence of each label of the documents contained in the first document set; train and verify the neural network classification model based on the training data set and the verification data set to obtain an indexing model.

[0165] In some embodiments, the training module 1520 is also used to determine the documents in the first document set that fall within a preset confidence interval based on the confidence of each label of the documents contained in the first document set; and generate a training data set and a verification data set based on the documents in the first document set that fall within the preset confidence interval.

[0166] In some embodiments, the training module 1520 is also used to, if there are documents in the first document set that fall within a preset credibility question interval, send the documents in the first document set that fall within the preset credibility question interval to the target user for screening, so as to obtain the documents in the first document set that meet the training requirements; and generate a training data set and a verification data set based on the documents in the first document set that meet the training requirements and the documents in the first document set that fall within the preset credibility trust interval.

[0167] In some embodiments, the indexing module 1530 is also used to use the indexing model to index some of the documents to be indexed in the second document set to obtain the indexing results of the documents included in the part of the second document set; based on the correction information of the indexing results of the documents in the part of the documents to be indexed, correct the indexing results of the documents included in the part of the second document set; based on the corrected documents in the part of the second document set as the documents in the new first document set, continue to train the indexing model; and use the indexing model that has been continuously trained to index the unindexed documents to be indexed in the second document set.

[0168] In some embodiments, the indexing module 1530 is further used to obtain indexing request information; if the indexing request information meets the preset requirements, a request success message is sent to the target client; if the indexing request information does not meet the preset requirements, a request failure message is sent to the target client. Wherein, training and generating an indexing model based on the first document set includes: after sending the request success message to the target client, training and generating an indexing model based on the first document set.

[0169] Fig.16 FIG. 1 is a schematic diagram of the structure of another document indexing device provided by an embodiment of the present disclosure. Fig.16 As shown, the document indexing device 1600 mentioned in the embodiment of the present disclosure is used for document indexing based on the technical field. The document indexing device 1600 includes an acquisition module 1610 and a determination module 1620. The acquisition module 1610 is used to acquire multiple documents to be indexed in the target technical field; the determination module 1620 is used to determine the document indexing result based on the target technical field.

[0170] Fig.17 FIG. 1 is a schematic diagram of the structure of another document indexing device provided by an embodiment of the present disclosure. Fig.17 As shown, another document indexing device 1700 mentioned in the embodiment of the present disclosure is used for document indexing based on manual indexing. The document indexing device 1700 includes an acquisition result module 1710 for acquiring document indexing results; and a determination module 1720 for verifying the manual indexing results based on the document indexing results to determine the document indexing results.

[0171] Fig.18FIG. 1 is a schematic diagram of a standardized configuration device structure provided by an embodiment of the present disclosure. Fig.18 As shown, a standardized configuration device 1800 provided by an embodiment of the disclosure includes an acquisition module 1810 for acquiring a document database to be indexed; and a standardized indexing module 1820 for planning and implementing a work guide for document indexing based on a document indexing method.

[0172] Fig.19 FIG. 1 is a schematic diagram of a local deployment device structure provided by an embodiment of the present disclosure. Fig.19 As shown, a localized deployment device structure 1900 provided by an embodiment of the present disclosure includes an acquisition module 1910 for acquiring document indexing results in a document database to be indexed; and a processing module 1920 for processing the document indexing results based on a document indexing processing system.

[0173] Fig. 20 Shown is a schematic diagram of the structure of an electronic device provided by an embodiment of the present disclosure. Fig. 20 Shown is a schematic structural diagram of an electronic device provided by an embodiment of the present disclosure. Fig. 20 The electronic device 2000 shown (the electronic device 2000 may be a computer device) includes a memory 2001, a processor 2002, a communication interface 2003 and a bus 2004. The memory 2001, the processor 2002 and the communication interface 2003 are connected to each other through the bus 2004.

[0174] The memory 2001 may be a read-only memory (ROM), a static storage device, a dynamic storage device or a random access memory (RAM). The memory 2001 may store a program. When the program stored in the memory 2001 is executed by the processor 2002, the processor 2002 and the communication interface 2003 are used to execute each step in the document indexing device of the embodiment of the present disclosure.

[0175] The processor 2002 can adopt a general central processing unit (CPU), a microprocessor, an application specific integrated circuit (ASIC), a graphics processing unit (GPU) or one or more integrated circuits to execute relevant programs to implement the functions required to be performed by each unit in the document indexing device of the embodiment of the present disclosure.

[0176] The processor 2002 may also be an integrated circuit chip with signal processing capability. In the implementation process, each step of the document indexing method disclosed in the present invention may be completed by an integrated logic circuit of hardware or software instructions in the processor 2002. The above-mentioned processor 2002 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The disclosed methods, steps and logic block diagrams in the embodiments of the present invention may be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present invention may be directly embodied as being executed by a hardware decoding processor, or may be executed by a combination of hardware and software modules in a decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 2001, and the processor 2002 reads the information in the memory 2001, and combines its hardware to complete the functions required to be executed by the units included in the document indexing device of the embodiment of the present disclosure, or executes the document indexing method of the method embodiment of the present disclosure.

[0177] The communication interface 2003 uses a transceiver such as but not limited to a transceiver to implement communication between the electronic device 2000 and other devices or a communication network. For example, the document data to be indexed can be obtained and processed through the communication interface 2003.

[0178] The bus 2004 may include a path for transmitting information between various components of the electronic device 2000 (eg, the memory 2001 , the processor 2002 , and the communication interface 2003 ).

[0179] It should be noted that although Fig. 20 The electronic device 2000 shown only shows a memory, a processor, and a communication interface, but in the specific implementation process, those skilled in the art should understand that the electronic device 2000 also includes other devices necessary for normal operation. At the same time, according to specific needs, those skilled in the art should understand that the electronic device 2000 may also include hardware devices for implementing other additional functions. In addition, those skilled in the art should understand that the electronic device 2000 may also only include the devices necessary for implementing the embodiments of the present disclosure, and does not necessarily include Fig. 20 All devices shown in .

[0180] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this disclosure.

[0181] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0182] In the several embodiments provided in the present disclosure, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0183] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0184] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0185] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present disclosure can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several 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 disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk.

[0186] The above is only a specific embodiment of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present disclosure, which should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A document indexing method, characterized in that: include: Based on the target category label, determining a first document set matching the target category label; Based on the first document set, training a neural network classification model to generate an indexing model; Using the indexing model, indexing the second document set to obtain indexing results for each document included in the second document set; The step of determining, based on the target category label, a first document set matching the target category label comprises: Performing word segmentation on the target category label to obtain a word segmentation result of the target category label; Based on the second document set, determining document feature information of each document included in the second document set; Based on the word segmentation result of the target category label and the document feature information of each document included in the second document set, generating the matching degree information between each document included in the second document set and the target category label; The first document set is determined based on the matching degree information between each document included in the second document set and the target category label.

2. The document indexing method according to claim 1, characterized in that: The determining the first document set based on the matching degree information between each document included in the second document set and the target category label includes: Based on the matching degree information between each document included in the second document set and the target category label, the first M documents to be indexed with the highest matching degree with the target category label are determined as the first document set, where M is a positive integer; or, Based on the matching degree information between each document included in the second document set and the target category label, the first S documents to be indexed with the highest matching degree with the target category label are determined, where S is a positive integer greater than M; the S documents to be indexed are verified and eliminated to obtain the first document set.

3. The document indexing method according to claim 1, characterized in that: The step of determining, based on the target category label, a first document set matching the target category label comprises: The first document set that is manually screened based on the target category label and matches the target category label is obtained.

4. The document indexing method according to any one of claims 1 to 3, characterized in that: After indexing the second document set by using the indexing model to obtain indexing results of the documents included in the second document set, the method further includes: Obtaining a verification and rejection record corresponding to the first document set; The indexing results of the documents included in the second document set are corrected based on the verification and elimination records.

5. The document indexing method according to any one of claims 1 to 3, characterized in that: The step of training a neural network classification model based on the first document set to generate an indexing model includes: Generate a training data set and a validation data set based on the label confidence of each document included in the first document set; The neural network classification model is trained and verified based on the training data set and the verification data set to obtain the indexing model.

6. The document indexing method according to claim 5, characterized in that: The step of generating a training data set and a verification data set based on the label confidence of each document included in the first document set includes: Determining documents in the first document set that fall within a preset confidence interval based on the label confidence of each document included in the first document set; The training data set and the verification data set are generated based on the documents in the first document set that fall within a preset confidence interval.

7. The document indexing method according to claim 6, characterized in that: The step of generating the training data set and the verification data set based on the documents in the first document set that fall within the preset confidence interval includes: If there are documents in the first document set that fall within the preset credibility question interval, the documents in the first document set that fall within the preset credibility question interval are sent to the target user for screening to obtain documents in the first document set that meet the training requirements; The training data set and the verification data set are generated based on the documents in the first document set that meet the training requirements and the documents in the first document set that fall within a preset confidence interval.

8. The document indexing method according to any one of claims 1 to 3, characterized in that: The step of using the indexing model to index the second document set to obtain indexing results for each document included in the second document set includes: Using the indexing model, indexing some of the documents to be indexed in the second document set to obtain indexing results for each of the some of the documents to be indexed; Correcting the indexing results of the part of documents to be indexed based on the correction information of the indexing results of the part of documents to be indexed; Based on the corrected part of the documents to be indexed as documents in the new first document set, continue to train the indexing model; The indexing model that has been continuously trained is used to index the unindexed documents to be indexed in the second document set.

9. The document indexing method according to any one of claims 1 to 3, characterized in that: Before training and generating an indexing model based on the first document set, the method further includes: Get indexing request information; If the indexing request information meets the preset requirements, a request success message is sent to the target client; If the indexing request information does not meet the preset requirements, sending a request failure message to the target client; The step of training and generating an indexing model based on the first document set includes: After sending the request success information to the target client, an indexing model is trained and generated based on the first document set.

10. The document indexing method according to any one of claims 1 to 3, characterized in that: The first document set includes Seed patent documents, wherein the second document set includes patent documents to be indexed; and / or, The first document set includes seed technical documents, and the second document set includes technical documents to be indexed.

11. A document indexing method, characterized in that: include: Obtain the database of documents to be indexed; The document indexing method according to any one of claims 1 to 10 is used to perform document indexing operations on the document database to be indexed.

12. A document indexing method, characterized in that: include: Obtain document indexing requirements; Based on the document indexing requirement, determining a plurality of documents to be indexed in the target technical field corresponding to the document indexing requirement; The method according to any one of claims 1 to 10 is used to perform indexing operations on the multiple documents to be indexed, thereby obtaining indexing results for each of the multiple documents to be indexed.

13. A document indexing method, characterized in that: include: Obtaining a first document set that is manually screened based on the target category label and matches the target category label; The first document set is sent to an indexing end, so that the indexing end generates an indexing model based on the training of the first document set, and uses the indexing model to index the second document set to obtain indexing results of the documents included in the second document set; The method for determining the first document set that is screened based on the target type label and matches the target category label includes: Performing word segmentation on the target category label to obtain a word segmentation result of the target category label; Based on the second document set, determining document feature information of each document included in the second document set; Based on the word segmentation result of the target category label and the document feature information of each document included in the second document set, generating the matching degree information between each document included in the second document set and the target category label; The first document set is determined based on the matching degree information between each document included in the second document set and the target category label.

14. A document indexing device, characterized in that: include: A determination module, configured to determine, based on the target category label, a first document set matching the target category label; A training module, used for training and generating an indexing model based on the first document set; An indexing module, used to index the second document set using the indexing model to obtain indexing results for each document included in the second document set; The determination module is further used to segment the target category label to obtain a segmentation result of the target category label; Based on the second document set, determining document feature information of each document included in the second document set; Based on the word segmentation result of the target category label and the document feature information of each document included in the second document set, generating the matching degree information between each document included in the second document set and the target category label; The first document set is determined based on the matching degree information between each document included in the second document set and the target category label.

15. A document indexing device, characterized in that: include: An acquisition module is used to acquire a database of documents to be indexed; An indexing module is used to perform a document indexing operation on the document database to be indexed using the document indexing method according to any one of claims 1 to 10.

16. A document indexing device, characterized in that: include: Acquisition module, used to obtain document indexing requirements; A determination module, configured to determine, based on the document indexing requirement, a plurality of documents to be indexed in a target technical field corresponding to the document indexing requirement; The indexing module is used to perform indexing operations on the multiple documents to be indexed using the method described in any one of claims 1 to 10 to obtain indexing results for each of the multiple documents to be indexed.

17. A document indexing device, characterized in that: include: An acquisition module is used to acquire a first document set that is manually screened based on a target category tag and matches the target category tag, wherein the method for determining the first document set that is screened based on a target type tag and matches the target category tag comprises: performing word segmentation on the target category tag to obtain a word segmentation result of the target category tag; determining document feature information of each document included in the second document set based on a second document set; generating matching degree information of each document included in the second document set with the target category tag based on the word segmentation result of the target category tag and the document feature information of each document included in the second document set; determining the first document set based on the matching degree information of each document included in the second document set with the target category tag; The sending module is used to send the first document set to the indexing end, so that the indexing end generates an indexing model based on the training of the first document set, and uses the indexing model to index the second document set to obtain the indexing results of each document included in the second document set.

18. An electronic device, characterized in that: include: processor; memory for storing computer executable instructions; The processor is used to execute the computer executable instructions to implement the method as described in any one of claims 1 to 13 above.

19. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and the computer program is used to execute the method according to any one of claims 1 to 13.

Images