Information structuring and retrieval method and device

Abstract

The present invention provides an information structuring and retrieval method and device, wherein the information structuring method includes: obtaining a support policy text for an enterprise; structuring the support policy text to obtain the structural type of each paragraph in the support policy text; classifying the paragraph content of each paragraph to obtain the content type of each paragraph, wherein the content type includes at least one of support method, support direction, application conditions, policy party, and application time; based on the structural type and content type of each paragraph, determining the structured information of the support policy text. The information structuring and retrieval method and device provided by the present invention reliably and accurately realizes the information structuring of the support policy text, and the structured information obtained thereby facilitates information search and positioning and information matching.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an information structuring and retrieval method and device. Background Art

[0002] Policy documents related to enterprise support policies are published on the Internet from time to time. The sources are very scattered, and policy documents are usually in long text format, making it difficult for enterprises to quickly and accurately obtain policy analysis. At present, it is still necessary to manually track and review the published policy documents, which is time-consuming and labor-intensive, and it is very easy to miss or make mistakes, resulting in enterprises that meet the guidance direction not necessarily receiving the corresponding policy support.

[0003] How to achieve automatic monitoring of enterprise support policies or project plans and quickly obtain relevant documents has become a problem that needs to be solved urgently. Summary of the invention

[0004] The present invention provides an information structuring and retrieval method and device, which are used to solve the defects in the prior art that the monitoring of enterprise support policies or project plans is time-consuming, labor-intensive and difficult to find.

[0005] The present invention provides an information structuring method, comprising:

[0006] Obtain the text of support policies for enterprises;

[0007] Structuring the support policy text to obtain the structural type of each paragraph in the support policy text;

[0008] Classifying the content of each paragraph to obtain a content type of each paragraph, wherein the content type includes at least one of a support method, a support direction, an application condition, a policy party, and an application time;

[0009] Based on the structure type and content type of each paragraph, the structured information of the support policy text is determined.

[0010] According to the information structuring method provided by the present invention, the paragraph contents of each paragraph are classified to obtain the content type of each paragraph, and then the method further includes:

[0011] When the content type of any paragraph is a declaration condition, the paragraph keywords of the any paragraph are extracted, and based on the paragraph keywords, the restriction objects of the declaration condition indicated by the any paragraph are classified to obtain the restriction object type of the any paragraph, and the restriction object type is added to the content type of the any paragraph.

[0012] According to the information structuring method provided by the present invention, the step of obtaining the support policy text for the enterprise includes:

[0013] Obtain candidate policy texts;

[0014] Based on the text classifier, the candidate policy text is classified based on the content related to application conditions and application procedures in the support policy text;

[0015] The candidate policy texts whose text classification results are support policies are determined as support policy texts for enterprises.

[0016] The present invention also provides a retrieval method, comprising:

[0017] Receiving a target support direction sent by a user terminal;

[0018] From the structured information of each support policy text, determine the support policy text whose support direction matches the target support direction as the target support policy text, and return the structured information of the target support policy text to the user terminal, wherein the structured information is determined based on the information structuring method.

[0019] According to the retrieval method provided by the present invention, determining the support policy text whose support direction matches the target support direction from the structured information of each support policy text includes:

[0020] Based on the semantic model, semantic encoding is performed on the target support direction to obtain target support direction features;

[0021] Matching the target support direction feature with the support direction features of each support policy text to obtain a support policy text whose support direction feature matches the target support direction feature;

[0022] The support direction features of each support policy text are obtained by semantically encoding the paragraphs with the content type of support direction in each support policy text based on the semantic model.

[0023] According to the retrieval method provided by the present invention, the semantic model is obtained by performing unsupervised training on the initial semantic model based on positive samples and negative samples;

[0024] The positive example sample includes a first supporting direction segment and a second supporting direction segment, wherein the second supporting direction segment is obtained by fine-tuning the first supporting direction segment;

[0025] The counterexample sample includes the first support direction segment and a third support direction segment, and the third support direction segment and the first support direction segment represent different support directions.

[0026] According to the retrieval method provided by the present invention, the receiving of the target support direction sent by the user terminal includes:

[0027] Receive target support directions and target declaration conditions sent by user terminals;

[0028] The step of determining, from the structured information of each support policy text, a support policy text whose support direction matches the target support direction as the target support policy text, and returning the structured information of the target support policy text to the user terminal includes:

[0029] Determining the matching degree of the target support policy text based on the target application condition, the application condition formula of the target support policy text and the application condition restriction type;

[0030] Returning the target support policy text and the matching degree to the user terminal;

[0031] The application condition formula is obtained by dividing the paragraphs in the target support policy text whose content type is application condition into sentences, and converting the clauses obtained by the sentence division into formulas based on a preset template;

[0032] The application condition restriction type is determined based on keywords in a paragraph whose content type is application conditions in the target support policy text.

[0033] The present invention also provides an information structuring device, comprising:

[0034] A text acquisition unit, used to acquire the text of support policies for enterprises;

[0035] A structuring unit, used for structuring the support policy text to obtain the structural type of each paragraph in the support policy text;

[0036] A classification unit, used to classify the content of each paragraph to obtain a content type of each paragraph, wherein the content type includes at least one of a support method, a support direction, an application condition, a policy party, and an application time;

[0037] The information determination unit is used to determine the structured information of the support policy text based on the structure type and content type of each paragraph.

[0038] The present invention also provides a retrieval device, comprising:

[0039] A direction receiving unit, used for receiving a target support direction sent by a user terminal;

[0040] A text determination unit is used to determine, from the structured information of each support policy text, a support policy text whose support direction matches the target support direction as the target support policy text, and return the structured information of the target support policy text to the user terminal, wherein the structured information is determined based on the information structuring method.

[0041] According to the retrieval device provided by the present invention, the direction receiving unit includes a conditional receiving subunit, which is used to:

[0042] Receive target support directions and target declaration conditions sent by user terminals;

[0043] Accordingly, the text determination unit includes a matching degree determination subunit, which is used to:

[0044] Determining the matching degree of the target support policy text based on the target application condition, the application condition formula of the target support policy text and the application condition restriction type;

[0045] Returning the target support policy text and the matching degree to the user terminal;

[0046] The application condition formula is obtained by dividing the paragraphs in the target support policy text whose content type is application condition into sentences, and converting the clauses obtained by the sentence division into formulas based on a preset template;

[0047] The application condition restriction type is determined based on keywords in a paragraph whose content type is application conditions in the target support policy text.

[0048] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the information structuring method and the retrieval method described above are implemented.

[0049] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the information structuring method and the retrieval method described above are implemented.

[0050] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any one of the information structuring methods and retrieval methods described above.

[0051] The information structuring and retrieval method and device provided by the present invention structures the support policy text, and classifies the content from the dimensions of support mode, support direction, application conditions, policy party and application time, and determines the structured information of the support policy text based on the structure type and content type obtained by classification. Thus, the information structuring of the support policy text is reliably and accurately realized, and the structured information obtained thereby facilitates information search, positioning and information matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] Figure 1 It is a flow chart of the information structuring method provided by the present invention;

[0054] Figure 2 is a schematic diagram of structured information provided by the present invention;

[0055] Figure 3 It is a schematic diagram of the flow of the retrieval method provided by the present invention;

[0056] Figure 4 This is the second flow chart of the search method provided by the present invention;

[0057] Figure 5 is a schematic diagram of the structure of the information structuring device provided by the present invention;

[0058] Figure 6 It is a schematic diagram of the structure of the retrieval device provided by the present invention;

[0059] Figure 7 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] Policy documents related to enterprise support policies are published on the Internet from time to time. The sources are very scattered, and the policy documents are usually in long text format. If the enterprise does not arrange for someone to track these websites every day, it is easy to miss the project application. However, since the release dates of these projects vary greatly and are very low frequency, it is particularly inefficient and not cost-effective to check various related websites every day.

[0062] Therefore, monitoring enterprise support policies or project plans and quickly obtaining relevant documents have become urgent issues that need to be addressed.

[0063] The embodiment of the present invention provides an information structuring method, which extracts and classifies structured information from support policy texts, and converts support policy texts with complex contents into structured information to facilitate information search and project application.

[0064] Figure 1 It is a flow chart of the information structuring method provided by the present invention, such as Figure 1 As shown, the method includes:

[0065] Step 110, obtaining the support policy text for the enterprise.

[0066] Here, the support policy text for the enterprise, that is, part of the text or all of the text in the policy document that needs to be structured, can be obtained by crawling from relevant websites using a web crawler, or by taking an image or scanning the paper version of the support policy text. The embodiment of the present invention does not make any specific limitations on this.

[0067] The support policy texts may be obtained at regular intervals or by real-time monitoring of the support policy publishing website and obtaining the corresponding support policy texts after new support policies appear.

[0068] Step 120 , structuring the support policy text to obtain the structural type of each paragraph in the support policy text.

[0069] Specifically, since the support policy text describes information related to the support policy in the form of text, which is not convenient for reference, structuring the support policy text can obtain the correspondence between various levels in the support policy text, such as the attribution of titles at each level, and the attribution of titles at each level and the text, so that the text content that needs to be consulted can be quickly located.

[0070] The support policy text can be structurally classified based on the semantics of each paragraph in the support policy text, or based on the position, font, font size, whether it is bolded, and other features of each paragraph in the support policy text. The structural type of each paragraph thus obtained can be a main title, a secondary title, a tertiary title, or the main text.

[0071] When performing structured classification, the structure type of each segment may be obtained by matching rules according to pre-set rules, or by inputting into a pre-trained structured classification model, which is not specifically limited in the embodiment of the present invention.

[0072] In one embodiment, the support policy text can be structurally classified through a hierarchical template, which can include common hierarchical tags, such as (first, second...), (1., 2....), (one, two...), ((1), (2)...), etc. The support policy text can be abstracted into a tree structure through the hierarchical template.

[0073] Step 130, classify the content of each paragraph to obtain the content type of each paragraph, and the content type includes at least one of support method, support direction, application conditions, policy party and application time.

[0074] Specifically, the text of the support policy usually contains at least one of the following relevant information: support method, support direction, application conditions, policy party and application time.

[0075] Among them, the support methods may include: at least one of financial support, talent support, honor support, land support, and financing support; it is mainly used to provide financial, talent, office space and other support or rewards to enterprises or projects that meet the application conditions.

[0076] The direction of support may include the industry direction targeted by the support policy, which can be classified according to the national economic industry classification, such as emerging information industry, bio-industry, new energy, high-end equipment manufacturing, new materials, etc.

[0077] The application conditions may include: at least one of the company's basic information indicators, economic indicators, and intellectual property indicators. Furthermore, the company's basic information indicators may include staff size, years of establishment, etc., economic indicators may include R&D investment, financing amount, net profit, operating income, asset-liability ratio, etc., and intellectual property indicators may include invention patents, design patents, utility model patents, trademarks, software copyrights, etc.

[0078] Policy makers may include specific provinces, cities, issuing authorities, etc.

[0079] The application time may include specific application time in previous years.

[0080] The content of each paragraph in the support policy text can be classified to obtain the content type of each paragraph. According to the common typesetting method of the support policy text, the first-level title, that is, the paragraph of the main title is usually used to explain the content type of the paragraph, or carries keywords of the content type. Based on the above typesetting method, after obtaining the structural type of each paragraph, the paragraph with the structural type of the main title can be screened out, and the content of each screened paragraph can be classified.

[0081] Specifically, content classification may be performed by matching rules according to pre-set rules, or may be input into a pre-trained text classification model to obtain the content type of each paragraph, which is not specifically limited in the embodiment of the present invention.

[0082] Step 140 , determining the structured information of the support policy text based on the structure type and content type of each paragraph.

[0083] Specifically, after completing the structural classification and content classification of the support policy text, the structural type and content type of each paragraph obtained by classification can be structured in a pre-set form to obtain the structural information of the support policy text. The structural information here can be a tree structure, in which each node is the text content under the structural type, and the structural type corresponding to the parent node is one level higher than the structural type corresponding to the child node.

[0084] The method provided by the embodiment of the present invention structures the support policy text, and classifies the content from multiple dimensions such as support method, support direction, application conditions, policy party and application time, and determines the structured information of the support policy text based on the structure type and content type obtained by classification. Thus, the information structuring of the support policy text is reliably and accurately achieved, and the structured information obtained thereby facilitates information search, positioning and information matching.

[0085] Based on the above embodiment, after step 130, the following steps are further included:

[0086] When the content type of any paragraph is a declaration condition, the paragraph keywords of the paragraph are extracted, and based on the paragraph keywords, the restriction objects of the declaration condition indicated by the paragraph are classified to obtain the restriction object type of the paragraph, and the restriction object type is added to the content type of the paragraph.

[0087] Specifically, since the content of the application conditions in the support policy text is relatively complex, in order to allow enterprises that meet the application conditions to accurately query the corresponding support policies, it is necessary to further classify the content of the application conditions. Therefore, if the content type of this paragraph in the support policy text is the application conditions, the restricted objects of the application conditions indicated by the paragraph can be further classified.

[0088] Here, the restriction objects of the application conditions indicated by the paragraph may include at least one of the company's basic information indicators, economic indicators, and intellectual property indicators; further, the application conditions indicated by the company's basic information indicators may be one or more, and accordingly, the restriction objects of the application conditions indicated by it may also be one or more. For example, taking the company's basic information indicators as an example, the restriction objects may include staff size, years of establishment, etc.; in economic indicators, the restriction objects may include R&D investment, financing amount, net profit, operating income, asset-liability ratio, etc.; in intellectual property indicators, the restriction objects may include invention patents, design patents, utility model patents, trademarks, software copyrights, etc.

[0089] Specifically, to classify the restricted objects, the keywords of the paragraphs whose content type is the declaration condition can be extracted first, the paragraph keywords of the paragraphs are obtained, and the restricted object type of the paragraphs is determined based on the keywords. Here, keyword extraction can be implemented by the mBert model.

[0090] After completing the classification of the restriction objects of the paragraph, the restriction object type thus obtained is added to the content type of the paragraph to obtain refined structured information related to the content of the application conditions in the support policy text. Figure 2 is a schematic diagram of the structured information provided by the present invention. The structured information determined for the support policy text can be displayed as follows Figure 2 The tree structure shown.

[0091] The method provided by the embodiment of the present invention classifies the restricted objects of the paragraph whose content type is the application conditions, obtains the restricted object type of the paragraph, and adds the restricted object type to the content type of any paragraph. Thus, refined structured information related to the application conditions content in the support policy text is obtained, which further facilitates information search, positioning and information matching.

[0092] Based on any of the above embodiments, step 110 specifically includes:

[0093] Obtain candidate policy texts;

[0094] Based on the text classifier, the candidate policy texts are classified based on the content related to application conditions and application procedures in the support policy texts;

[0095] The candidate policy texts whose text classification results are support policies are determined as support policy texts for enterprises.

[0096] Specifically, in order to accurately obtain the support policy texts for enterprises, we can first obtain a large number of candidate policy texts. The candidate policy texts are policy document information released by various departments or websites within a specified time. By passing in parameters, we can capture all policy information of a specified website within a specified time.

[0097] Next, the candidate policy texts can be classified by a text classifier, and the candidate policy texts with the text classification result of supporting policies can be determined as supporting policy texts for enterprises. Specifically, whether the supporting policy text contains relevant content of application conditions and application procedures is used as a classification basis to determine whether the text classification result is a supporting policy.

[0098] If the policy text contains relevant content about application conditions and application procedures, the text classification result is a support policy; if the policy text does not contain relevant content about application conditions and application procedures, the text classification result is that it is not a support policy text.

[0099] If the policy text directly contains keywords matching "application conditions" and "application procedures", it can be judged that there is content related to the application conditions and application procedures; however, different policy texts may have different descriptions, such as "collection and recommendation conditions", "application and recognition conditions", "declaration methods", "application and recognition procedures" and other related keywords, which can also be used to judge the existence of content related to the application conditions and application procedures.

[0100] The method provided by the embodiment of the present invention uses the relevant content of application conditions and application procedures in the support policy text as the classification basis, performs text classification on the candidate policy texts, and determines the candidate policy text whose text classification result is a support policy as a support policy text for the enterprise. The support policy text thus obtained is more targeted, thereby achieving more accurate structured information.

[0101] Figure 3 It is a schematic diagram of the process of the retrieval method provided by the present invention, such as Figure 3 As shown, the method includes:

[0102] Step 310, receiving a target support direction sent by a user terminal;

[0103] Step 330, from the structured information of each support policy text, determine the support policy text whose support direction matches the target support direction as the target support policy text, and return the structured information of the target support policy text to the user terminal, the structured information is determined based on the information structuring method.

[0104] Specifically, the support policy texts from various sources can be structured through the above embodiments, thereby realizing the transformation of the support policy texts into structured information. The structured information of the support policy texts provides convenience for information search and positioning and project application matching. Moreover, on this basis, a retrieval system can be constructed to facilitate the rapid search and matching of target information.

[0105] The user can input the target support direction through a user terminal in the form of a smart phone, computer, tablet computer, etc., and send the target support direction to the server for retrieval. The target support direction here is the support direction for which relevant information is expected to be found from the support policy text.

[0106] After receiving the target support direction, the support policy text corresponding to the target support direction can be located in the structured information of each support policy text as the target support policy text, and the target support policy text is returned to the user terminal for viewing by the user terminal.

[0107] As a preference, matching-related information may also be provided for the target support policy text.

[0108] The method provided by the embodiment of the present invention realizes rapid retrieval of support policy texts in target support directions through structured information, helps to improve information query efficiency and project application matching efficiency, and can assist project application companies to quickly and accurately query support policy text information.

[0109] Based on the above embodiment, in step 320, determining the support policy text whose support direction matches the target support direction from the structured information of each support policy text includes:

[0110] Step 321, semantically encode the target support direction based on the semantic model to obtain target support direction features;

[0111] Step 322, matching the target support direction feature with the support direction feature of each support policy text, and obtaining a support policy text whose support direction feature matches the target support direction feature;

[0112] The support direction characteristics of each support policy text are obtained by semantically encoding the paragraphs with the content type of support direction in each support policy text based on the semantic model.

[0113] Specifically, the semantic model is used to extract the semantic features of the paragraphs with the content type of the support direction in each support policy text. The semantic model may include an encoding layer for semantically encoding each paragraph, so as to obtain a support direction representation vector that can represent the paragraph, that is, the support direction feature of each support policy text. The support direction feature can be output as a result in the form of a vector representation.

[0114] The received target support direction is input into the semantic model, and the target support direction is semantically encoded through the encoding layer of the semantic model, thereby obtaining a target support direction feature that can characterize the semantics of the target support direction. Next, the target support direction feature is matched with the support direction feature of each support policy text, thereby obtaining a support policy text whose support direction feature matches the target support direction feature.

[0115] Furthermore, the matching degree of the support policy text can be obtained, and the support policy text can be filtered according to a preset matching degree threshold, and the policy text with a matching degree higher than the matching degree threshold can be sent to the user terminal, and the policy text with a matching degree lower than the matching degree threshold can be filtered.

[0116] The method provided in the embodiment of the present invention performs semantic encoding on the target support direction based on the semantic model, matches the encoded target support direction characteristics with the support direction characteristics of each support policy text, and obtains the support direction characteristics matching the target support direction characteristics, thereby further improving the efficiency and accuracy of target information query.

[0117] Based on any of the above embodiments, the semantic model is obtained by performing unsupervised training on the initial semantic model based on positive samples and negative samples;

[0118] The positive sample includes the first supporting direction segment and the second supporting direction segment, and the second supporting direction segment is obtained by fine-tuning the first supporting direction segment;

[0119] The counterexample sample includes a first support direction segment and a third support direction segment, and the third support direction segment and the first support direction segment represent different support directions.

[0120] Specifically, before executing step 321, the initial semantic model may be pre-trained in an unsupervised manner. The unsupervised training may reduce the annotation cost, and the trained initial semantic model may be used as the semantic model. Specifically, the model training may be performed in the following manner:

[0121] First, a large number of sample segments containing content types of support directions are collected, and any first support direction segment is selected as a reference segment. On this basis, the first support direction segment is fine-tuned to obtain a second support direction segment, and a positive sample is constructed based on the first support direction segment and the second support direction segment. In addition, a segment representing a different support direction from the first support direction segment is used as a third support direction segment, and a negative sample is constructed based on the first support direction segment and the third support direction segment. Among them, the number of samples of the positive sample and the negative sample can be the same, or there can be a certain proportional relationship, which is not specifically limited in the embodiment of the present invention.

[0122] It can be understood that the higher the similarity of the support direction features between the positive example samples, the more the support direction features corresponding to the positive example samples can reflect the common features of the support direction features of the first support direction segment and the second support direction segment.

[0123] The lower the similarity of the support direction features between the counterexample samples, the more the support direction features corresponding to the counterexample samples can reflect the difference characteristics of the support direction features of the first support direction segment and the third support direction segment.

[0124] Then, the positive sample and the negative sample are input into the initial semantic model for training. During the training process, the initial semantic model can amplify and learn the common features of the positive sample, that is, the support direction features of the first support direction segment and the second support direction segment; and at the same time amplify and learn the difference features of the support direction features of the negative sample, that is, the first support direction segment and the third support direction segment. The semantic model trained in this way can better distinguish the differences between the support direction features.

[0125] The method provided in the embodiment of the present invention performs unsupervised training on the initial semantic model based on constructed positive samples and negative samples, so that the semantic model obtained by training can better distinguish the differences between supporting direction features, thereby improving the efficiency and accuracy of target information query.

[0126] Based on any of the above embodiments, step 310 specifically includes:

[0127] Receive target support directions and target declaration conditions sent by user terminals;

[0128] Accordingly, step 320 specifically includes:

[0129] Determine the matching degree of the target support policy text based on the target application conditions, the application condition formula of the target support policy text and the application condition restriction type;

[0130] Return the target support policy text and matching degree to the user terminal;

[0131] The application condition formula is obtained by dividing the paragraphs in the target support policy text whose content type is application condition into sentences, and converting the clauses obtained by the sentence division into formulas based on the preset template;

[0132] The type of restriction on application conditions is determined based on the keywords in the paragraphs of the target support policy text whose content type is application conditions.

[0133] Specifically, in addition to the case where the user only inputs the target support direction, the target support direction and the target declaration condition can also be input simultaneously through the user terminal, and the target declaration condition here can be one or more.

[0134] After receiving the target support direction and target application conditions input by the user, the support policy text can be determined according to the method of the above embodiment. On this basis, the matching degree of the target support policy text is determined according to the target application conditions.

[0135] First, the paragraphs in the target support policy text with the content type of application conditions can be divided into sentences, and the clauses obtained by the sentence division can be converted into formulas based on the preset template to obtain the application condition formula. The preset template can be in the form of X opY, where X matches the items in the financial statements through keywords and common aliases; the op template contains common operators such as greater than, less than, not more than, not less than, etc.; Y is a specific number, which can be a specific amount, a ratio, or an interval.

[0136] For example, if the user enters the economic indicators in the declaration conditions, the sub-items in the policy-declaration conditions-economic indicators in the structured information can be pre-sentenced, and then the clauses can be converted into formulas through the template. If the user enters the financial report item information, the declaration condition formula can be used to automatically calculate whether the conditions are met. When the user enters one or more target declaration conditions, it can be calculated whether each target declaration condition is met.

[0137] Furthermore, the application condition restriction type refers to whether the application condition in the target support policy text belongs to one of (one of the following is met, all are met, and the following cannot be applied for). Since the target support policy text usually contains multiple application conditions, and the application condition restriction type is clearly stated in the paragraph, the application condition restriction type can be determined based on the keywords in the paragraph whose content type is application conditions.

[0138] Then, according to the matching result of each target declaration condition and the declaration condition restriction type, all target declaration conditions can be subjected to logical operations such as AND, OR, and COMPLEMENT to obtain the matching result of the target support policy text. The matching degree of the target support policy text can be obtained by dividing the matching result by the number of declaration conditions of the target support policy. The target support policy text and the matching degree can be returned to the user terminal in the order of high to low matching degree.

[0139] The method provided by the embodiment of the present invention enables the user to quickly obtain the matching degree of the target support policy text, and at the same time can more conveniently find the target support policy text with a higher matching degree.

[0140] Based on any of the above embodiments, when the support policy text is structured, the application time can also be converted into monthly data, so that when the user searches, the retrieved target support policy text can be sorted according to the application month.

[0141] Based on any of the above embodiments, Figure 4is a flow chart of the support policy text retrieval method provided by the present invention, such as Figure 4 As shown, the method includes:

[0142] (1) Obtain the supporting policy texts for enterprises. First, obtain the information of the released policy documents from various data sources, and then filter out the supporting policy texts for enterprises.

[0143] (2) Structuring the information of the enterprise support policy text to obtain structured information.

[0144] Information can be structured according to five dimensions: support direction, application conditions, support methods, policy parties and application time.

[0145] (3) Based on structured information, a search platform is constructed for information search and location and project application matching.

[0146] After the user inputs the target support direction and target application conditions, the target support policy text and the matching degree of the target support policy text that match the target support direction and target application conditions are automatically matched, and the target support policy text and the matching degree are returned to the user terminal.

[0147] The information structuring device provided by the present invention is described below. The information structuring device described below and the information structuring method described above can be referenced to each other.

[0148] Figure 5 is a schematic diagram of the structure of the information structuring device provided by the present invention, such as Figure 5 As shown, the device comprises:

[0149] A text acquisition unit 510 is used to acquire a support policy text for an enterprise;

[0150] A structuring unit 520, configured to structure the support policy text to obtain a structural type of each paragraph in the support policy text;

[0151] A classification unit 530, configured to classify the content of each segment to obtain a content type of each segment, wherein the content type includes at least one of a support method, a support direction, an application condition, a policy party, and an application time;

[0152] The information determination unit 540 is used to determine the structured information of the support policy text based on the structure type and content type of each paragraph.

[0153] The information structuring device provided by the embodiment of the present invention structures the support policy text, and classifies the content from multiple dimensions such as support method, support direction, application conditions, policy party and application time, and determines the structured information of the support policy text based on the structure type and content type obtained by classification. Thus, the information structuring of the support policy text is reliably and accurately realized, and the structured information obtained thereby facilitates information search, positioning and information matching.

[0154] Based on any of the above embodiments, the device further includes an object classification module, which is used to:

[0155] When the content type of any paragraph is a declaration condition, the paragraph keywords of the any paragraph are extracted, and based on the paragraph keywords, the restriction objects of the declaration condition indicated by the any paragraph are classified to obtain the restriction object type of the any paragraph, and the restriction object type is added to the content type of the any paragraph.

[0156] Based on any of the above embodiments, the text acquisition unit 510 is further used to:

[0157] Obtain candidate policy texts;

[0158] Based on the text classifier, the candidate policy text is classified based on the content related to application conditions and application procedures in the support policy text;

[0159] The candidate policy texts whose text classification results are support policies are determined as support policy texts for enterprises.

[0160] Based on any of the above embodiments, Figure 6 is a schematic diagram of the structure of the retrieval device provided by the present invention, such as Figure 6 As shown, the device comprises:

[0161] A direction receiving unit 610 is used to receive a target support direction sent by a user terminal;

[0162] The text determination unit 620 is used to determine the support policy text whose support direction matches the target support direction from the structured information of each support policy text, as the target support policy text, and return the structured information of the target support policy text to the user terminal, wherein the structured information is determined based on the information structuring method as described above.

[0163] The device provided by the embodiment of the present invention realizes rapid retrieval of support policy texts in target support directions through structured information, which helps to improve information query efficiency and project application matching efficiency, and can assist project application companies to quickly and accurately query support policy text information.

[0164] Based on any of the above embodiments, the text determination unit 620 is further configured to:

[0165] Based on the semantic model, semantic encoding is performed on the target support direction to obtain target support direction features;

[0166] Matching the target support direction feature with the support direction features of each support policy text to obtain a support policy text whose support direction feature matches the target support direction feature;

[0167] The support direction features of each support policy text are obtained by semantically encoding the paragraphs with the content type of support direction in each support policy text based on the semantic model.

[0168] Based on any of the above embodiments, the device further includes a model training unit, which is used to perform unsupervised training on the initial semantic model based on the positive sample and the negative sample;

[0169] The positive example sample includes a first supporting direction segment and a second supporting direction segment, wherein the second supporting direction segment is obtained by fine-tuning the first supporting direction segment;

[0170] The counterexample sample includes the first support direction segment and a third support direction segment, and the third support direction segment and the first support direction segment represent different support directions.

[0171] Based on any of the foregoing embodiments, the direction receiving unit 610 is further configured to:

[0172] Receive target support directions and target declaration conditions sent by user terminals;

[0173] Accordingly, the text determination unit 620 is further configured to:

[0174] Determining the matching degree of the target support policy text based on the target application condition, the application condition formula of the target support policy text and the application condition restriction type;

[0175] Returning the target support policy text and the matching degree to the user terminal;

[0176] The application condition formula is obtained by dividing the paragraphs in the target support policy text whose content type is application condition into sentences, and converting the clauses obtained by the sentence division into formulas based on a preset template;

[0177] The application condition restriction type is determined based on keywords in a paragraph whose content type is application conditions in the target support policy text.

[0178] Figure 7 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7 As shown, the storage medium

[0179] It may include: a processor 710, a communication interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communication interface 720 and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the information structuring method, which includes: obtaining the support policy text for the enterprise; structuring the support policy text to obtain the structure type of each paragraph in the support policy text; classifying the paragraph content of each paragraph to obtain the content type of each paragraph, and the content type includes at least one of the support method, support direction, application conditions, policy party and application time; based on the structure type and content type of each paragraph, determining the structured information of the support policy text.

[0180] The processor 710 can also call the logical instructions in the memory 730 to execute the retrieval method, which includes: receiving the target support direction sent by the user terminal; determining, from the structured information of each support policy text, a support policy text whose support direction matches the target support direction as the target support policy text, and returning the structured information of the target support policy text to the user terminal, wherein the structured information is determined based on the information structuring method.

[0181] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0182] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the information structuring method provided by the above methods, which includes: obtaining a support policy text for an enterprise; structuring the support policy text to obtain the structural type of each paragraph in the support policy text; classifying the paragraph content of each paragraph to obtain the content type of each paragraph, and the content type includes at least one of support method, support direction, application conditions, policy party and application time; based on the structural type and content type of each paragraph, determining the structured information of the support policy text.

[0183] In addition, when the computer program is executed by the processor, the computer can execute the retrieval method provided by the above-mentioned methods, which method includes: receiving the target support direction sent by the user terminal; determining, from the structured information of each support policy text, a support policy text whose support direction matches the target support direction as the target support policy text, and returning the structured information of the target support policy text to the user terminal, wherein the structured information is determined based on the information structuring method.

[0184] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the information structuring method provided by the above-mentioned methods, the method comprising: obtaining a support policy text for an enterprise; structuring the support policy text to obtain a structural type of each paragraph in the support policy text; classifying the paragraph content of each paragraph to obtain a content type of each paragraph, the content type including at least one of support method, support direction, application conditions, policy party and application time; determining the structured information of the support policy text based on the structural type and content type of each paragraph.

[0185] In addition, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it is implemented to execute the retrieval method provided by the above-mentioned methods. The method includes: receiving a target support direction sent by a user terminal; determining, from the structured information of each support policy text, a support policy text whose support direction matches the target support direction as the target support policy text, and returning the structured information of the target support policy text to the user terminal, wherein the structured information is determined based on the information structuring method.

[0186] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0187] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0188] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A search method, characterized in that: include: Receiving a target support direction sent by a user terminal; Determining, from the structured information of each support policy text, a support policy text whose support direction matches the target support direction as the target support policy text, and returning the structured information of the target support policy text to the user terminal; The structured information is determined based on the following steps: Acquire the support policy text for the enterprise; structure the support policy text to obtain the structural type of each paragraph in the support policy text; select the paragraph whose structural type is the main title from the paragraphs, and classify the paragraph contents of the selected paragraphs to obtain the content type of each paragraph, wherein the content type includes support method, support direction, application conditions, policy party and application time; Determining the structural information of the support policy text based on the structural type and content type of each paragraph; The structured information is a tree structure, wherein each node is text content under the structure type, and the parent node of the text content includes the content type.

2. The search method according to claim 1, characterized in that: The step of determining, from the structured information of each support policy text, a support policy text whose support direction matches the target support direction includes: Based on the semantic model, the target support direction is semantically encoded to obtain the target support direction feature; the target support direction feature is matched with the support direction feature of each support policy text to obtain a support policy text whose support direction feature matches the target support direction feature; the support direction feature of each support policy text is obtained by semantically encoding the paragraphs with the content type of support direction in each support policy text based on the semantic model.

3. The search method according to claim 2, characterized in that: The semantic model is obtained by unsupervised training of the initial semantic model based on positive samples and negative samples; The positive example sample includes a first supporting direction segment and a second supporting direction segment, wherein the second supporting direction segment is obtained by fine-tuning the first supporting direction segment; The counterexample sample includes the first support direction segment and a third support direction segment, and the third support direction segment and the first support direction segment represent different support directions.

4. The search method according to any one of claims 1 to 3, characterized in that: The receiving the target support direction sent by the user terminal includes: Receive target support directions and target declaration conditions sent by user terminals; The step of determining, from the structured information of each support policy text, a support policy text whose support direction matches the target support direction as the target support policy text, and returning the structured information of the target support policy text to the user terminal includes: Determining the matching degree of the target support policy text based on the target application condition, the application condition formula of the target support policy text and the application condition restriction type; Returning the target support policy text and the matching degree to the user terminal; The application condition formula is obtained by dividing the paragraphs in the target support policy text whose content type is application condition into sentences, and converting the clauses obtained by the sentence division into formulas based on a preset template; The application condition restriction type is determined based on keywords in a paragraph whose content type is application conditions in the target support policy text.

5. The search method according to claim 1, characterized in that: The process of classifying the content of each paragraph to obtain the content type of each paragraph further includes: When the content type of any paragraph is a declaration condition, the paragraph keywords of the any paragraph are extracted, and based on the paragraph keywords, the restriction objects of the declaration condition indicated by the any paragraph are classified to obtain the restriction object type of the any paragraph, and the restriction object type is added to the content type of the any paragraph.

6. The search method according to claim 1, characterized in that: The obtaining of the supporting policy text for the enterprise includes: Obtain candidate policy texts; Based on the text classifier, the candidate policy text is classified based on the content related to application conditions and application procedures in the support policy text; The candidate policy texts whose text classification results are support policies are determined as support policy texts for enterprises.

7. A search device, characterized in that: include: A direction receiving unit, used for receiving a target support direction sent by a user terminal; a text determination unit, configured to determine, from the structured information of each support policy text, a support policy text whose support direction matches the target support direction as the target support policy text, and return the structured information of the target support policy text to the user terminal; An information structuring unit is used to obtain a support policy text for an enterprise; structure the support policy text to obtain a structural type of each paragraph in the support policy text; filter out paragraphs whose structural type is a main title from each paragraph, and classify the paragraph contents of each filtered paragraph to obtain a content type of each paragraph, wherein the content type includes support method, support direction, application conditions, policy party and application time; Determining the structural information of the support policy text based on the structural type and content type of each paragraph; The structured information is a tree structure, wherein each node is text content under the structure type, and the parent node of the text content includes the content type.

8. The search device according to claim 7, characterized in that: The direction receiving unit includes a conditional receiving subunit, which is used for: Receive target support directions and target declaration conditions sent by user terminals; Accordingly, the text determination unit includes a matching degree determination subunit, which is used to: Determining the matching degree of the target support policy text based on the target application condition, the application condition formula of the target support policy text and the application condition restriction type; Returning the target support policy text and the matching degree to the user terminal; The application condition formula is obtained by dividing the paragraphs in the target support policy text whose content type is application condition into sentences, and converting the clauses obtained by the sentence division into formulas based on a preset template; The application condition restriction type is determined based on keywords in a paragraph whose content type is application conditions in the target support policy text.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the search method according to any one of claims 1 to 6 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the search method according to any one of claims 1 to 6 is implemented.

Images