Content analysis query method, device, equipment and computer storage medium

Abstract

The present invention relates to the field of computer financial technology, and discloses a content analysis query method, device, equipment and computer storage medium. The method includes: when receiving a content analysis request, obtaining the target content to be analyzed carried by the content analysis request; parsing the target content to obtain the keywords in the target content, and determining the content theme of the target content according to the keywords; querying a preset database to obtain related content similar to the content theme and / or keywords of the target content; comparing the target content with the related content, obtaining similarity information and difference information between the target content and the related content and outputting it. The present invention first determines the keywords of the content, determines the content theme according to the keywords, performs content query according to the content theme, obtains related content similar to the target content, and automatically analyzes the similarity information and difference information between the target content and the related content, thereby improving the efficiency and accuracy of content query and reducing the waste of human resources.

Description

Technical Field

[0001] The present invention relates to the field of computer financial technology (Fintech) technology, and in particular to a content analysis query method, device, equipment and computer storage medium. Background Art

[0002] In recent years, with the rapid development of Internet financial technology (Fintech), more and more technologies (big data, distributed, blockchain, artificial intelligence, etc.) are being applied in the field of computer finance.

[0003] Existing finance involves more and more types of transactions. Financial transactions are not limited to physical objects, but can also be invisible assets, such as patent applications, academic works or literary works. The evaluation of invisible assets in financial transactions is becoming more and more necessary. Currently, the evaluation of invisible assets is usually achieved by humans, that is, manually analyzing the content to be evaluated, finding related content similar to the content to be evaluated, and evaluating it based on the value of the related content. However, the workload of manual content analysis and query is large, which makes the content analysis query inefficient. In addition, due to the different overall levels of manual labor, the accuracy of content analysis queries cannot be guaranteed. Summary of the invention

[0004] The main purpose of the present invention is to propose a content analysis query method, device, equipment and computer storage medium, aiming to solve the technical problems that the current manual content analysis query has a large workload, low content analysis query efficiency and the accuracy of content analysis query cannot be guaranteed.

[0005] To achieve the above object, the present invention provides a content analysis query method, which comprises the following steps:

[0006] Upon receiving a content analysis request, obtaining target content to be analyzed carried in the content analysis request;

[0007] Parsing the target content to obtain keywords in the target content, and determining the content theme of the target content according to the keywords;

[0008] Querying a preset database to obtain related content that is similar to the content theme and / or keywords of the target content;

[0009] The target content is compared with the associated content, and similarity information and difference information between the target content and the associated content are obtained and output.

[0010] Optionally, the step of querying a preset database to obtain related content similar to the content theme and / or keywords of the target content includes:

[0011] Query a preset database to obtain pre-stored content in the preset database in the same technical field as the target content; analyze the pre-stored content through a preset theme extraction model to obtain the content theme of the pre-stored content and the keywords corresponding to the content theme;

[0012] When the target content includes at least two content themes and / or the content themes include at least two keywords, calculating the theme similarity and / or word similarity between the content themes and / or keywords of the target content and the content themes and / or keywords of the pre-stored content;

[0013] The subject similarity and / or the word similarity are weighted to obtain the comprehensive similarity between the target content and the pre-stored content, and the pre-stored content having a comprehensive similarity with the target content higher than a preset threshold is used as the associated content.

[0014] Optionally, the step of comparing the target content with the associated content, obtaining and outputting similarity information and difference information between the target content and the associated content includes:

[0015] Comparing the target content with the associated content, and dividing the content themes of the target content into similar content themes and different content themes;

[0016] The similar content topics and the keywords corresponding to the similar content topics are input as similar information into the similar display area in the preset template, and the different content topics and the keywords corresponding to the different content topics are input as different information into the different display area in the preset template to form a content analysis table.

[0017] Optionally, after the step of comparing the target content with the associated content, obtaining and outputting similarity information and difference information between the target content and the associated content, the method further includes:

[0018] Obtaining the information volume ratio of the similar information in the target content, and when the information volume ratio is less than a preset ratio threshold, determining whether the difference information is disclosed by other content in the pre-stored content database except the associated content;

[0019] If the difference information is not disclosed by other contents in the pre-stored content database except the associated content, analyzing the impact factor of the difference information on the target content and outputting it;

[0020] If the difference information is disclosed by other contents in the pre-stored content database except the associated content, the difference content in the pre-stored content database that discloses the difference information is output.

[0021] Optionally, the step of parsing the target content to obtain keywords in the target content and determining the content theme of the target content according to the keywords includes:

[0022] Perform word segmentation processing on the target content through a preset topic extraction model to obtain a word set corresponding to the target content, remove noise words in the word set, and obtain keywords in the target content;

[0023] The location information and frequency information of the keyword are obtained, and the content theme of the target content is determined according to the location information and frequency information of the keyword through a preset theme extraction model.

[0024] Optionally, before the step of analyzing the pre-stored content by using a preset theme extraction model to obtain the content theme of the pre-stored content and the keywords corresponding to the content theme, the step includes:

[0025] When receiving a model building request, the initial topic extraction model is obtained, and the content in the local storage is marked to form a training sample set;

[0026] Extracting topic training samples from the training sample set, training the initial topic extraction model through the topic training samples, and obtaining a trained topic extraction model and a topic feature vector;

[0027] Processing the topic feature vector by a preset loss function to obtain a loss value, and analyzing the loss value to determine whether the trained topic extraction model has converged;

[0028] When it is detected that the trained topic extraction model converges, the trained topic extraction model is output as a preset topic extraction model.

[0029] Optionally, after the step of obtaining the target content to be analyzed carried in the content analysis request when receiving the content analysis request, the method further includes:

[0030] Acquire the content format of the target content;

[0031] When the content format of the target content is not a text document format, calling a preset OCR model to detect the target content, determine a text area in the target content, and obtain character information contained in the text area, so as to convert the target content into a text document format;

[0032] When the content format of the target content is a text document format, the step of parsing the target content to obtain keywords in the target content and determining the content theme of the target content according to the keywords is performed.

[0033] In addition, to achieve the above-mentioned purpose, the present invention further provides a content analysis query device, the content analysis query device comprising:

[0034] A request receiving module, configured to, upon receiving a content analysis request, obtain the target content to be analyzed carried in the content analysis request;

[0035] A parsing and determining module, configured to parse the target content to obtain keywords in the target content, and determine the content theme of the target content according to the keywords;

[0036] A query acquisition module, used to query a preset database to acquire related content similar to the content theme and / or keywords of the target content;

[0037] The comparison and output module compares the target content with the associated content, obtains similarity information and difference information between the target content and the associated content, and outputs the information.

[0038] In addition, to achieve the above-mentioned purpose, the present invention also provides a content analysis query device, which includes: a memory, a processor, and a computer program corresponding to the content analysis query stored in the memory and executable on the processor, and the computer program corresponding to the content analysis query implements the steps of the content analysis query method described above when executed by the processor.

[0039] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer storage medium, on which a computer program corresponding to the content analysis query is stored, and when the computer program corresponding to the content analysis query is executed by a processor, the steps of the content analysis query method described above are implemented.

[0040] The present invention provides a content analysis query method, device, equipment and computer storage medium. In an embodiment of the present invention, when a content analysis request is received, the target content to be analyzed carried by the content analysis request is obtained; the target content is parsed to obtain keywords in the target content, and the content theme of the target content is determined according to the keywords; a preset database is queried to obtain related content similar to the content theme and / or keywords of the target content; the target content is compared with the related content, and similarity information and difference information between the target content and the related content are obtained and output. In an embodiment of the present invention, the keywords of the target content are determined by automatically parsing the target content, and then the content theme of the target content is determined according to the keywords of the target content, and a content query is performed according to the content theme to obtain related content similar to the target content, and similarity information and difference information of the content are automatically output, so that manual content analysis is not required, which reduces the waste of human resources, improves the efficiency and accuracy of content query, and facilitates users to view content analysis query results. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the device structure of the hardware operating environment involved in the embodiment of the present invention;

[0042] Figure 2 A schematic diagram of a flow chart of building a preset topic extraction model in the first embodiment of the content analysis query method of the present invention;

[0043] Figure 3 This is a flow chart of a first embodiment of the content analysis query method of the present invention;

[0044] Figure 4 It is a schematic diagram of functional modules of an embodiment of a content analysis and query device of the present invention.

[0045] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0046] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0047] like Figure 1 As shown, Figure 1 It is a schematic diagram of the device structure of the hardware operating environment involved in the embodiment of the present invention.

[0048] The content analysis query device of the embodiment of the present invention may be a server device, such as Figure 1 As shown, the content analysis query device may include: a processor 1001, such as a CPU, a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory, or it may be a stable memory (non-volatile memory), such as a disk memory. The memory 1005 may optionally be a storage device independent of the aforementioned processor 1001.

[0049] Those skilled in the art will understand that Figure 1 The device structure shown in the figure does not constitute a limitation of the device, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0050] like Figure 1As shown, the memory 1005 as a computer storage medium may include a computer program corresponding to an operation network communication module, a user interface module, and content analysis query.

[0051] exist Figure 1 In the device shown, the network interface 1004 is mainly used to connect to the background server and communicate data with the background server; the user interface 1003 is mainly used to connect to the client (user end) and communicate data with the client; and the processor 1001 can be used to call the computer program corresponding to the content analysis query stored in the memory 1005, and execute the operations in the following content analysis query method.

[0052] Based on the above hardware structure, an embodiment of the content analysis query method of the present invention is proposed.

[0053] The first embodiment of the content analysis query method of the present invention includes:

[0054] Upon receiving a content analysis request, obtaining target content to be analyzed carried in the content analysis request;

[0055] Parsing the target content to obtain keywords in the target content, and determining the content theme of the target content according to the keywords;

[0056] Querying a preset database to obtain related content that is similar to the content theme and / or keywords of the target content;

[0057] The target content is compared with the associated content, and similarity information and difference information between the target content and the associated content are obtained and output.

[0058] In this embodiment, the content analysis query method is applied to a content query device. The specific type of the content query device is not limited. For example, the content query device can be a terminal or a server. A preset topic extraction model is pre-built in the content query device. The content query device processes the target content through the preset topic extraction model to obtain the content theme of the target content, so as to perform automatic content query analysis according to the content theme of the target content.

[0059] Reference Figure 2 , Figure 2 This is a flow chart of the construction of a preset topic extraction model in the first embodiment of the content analysis query method; the steps of pre-constructing a preset topic extraction model in this embodiment specifically include:

[0060] Step S01, when receiving a model building request, obtain an initial topic extraction model and mark the content in local storage to form a training sample set.

[0061] When receiving the model building request, the content query device obtains the initial topic extraction model, and the content query device marks the content in the local storage to form a training sample set. That is, the content query device outputs the content and prompts the user to mark the content with keywords and topics. The content analysis query device uses the marked content as a topic training sample, and the content query device forms a training sample set with each topic training sample.

[0062] Step S02, extracting topic training samples from the training sample set, training the initial topic extraction model through the topic training samples, and obtaining a trained topic extraction model and a topic feature vector.

[0063] The content query device extracts a preset proportion (the preset proportion can be set according to the specific scenario, for example, the preset proportion is set to 5%) of topic training samples from the training sample set, and the content query device trains an initial topic extraction model through the topic training samples to obtain a trained topic extraction model and a topic feature vector.

[0064] Step S03, processing the topic feature vector through a preset loss function to obtain a loss value, and analyzing the loss value to determine whether the trained topic extraction model has converged.

[0065] A loss function is preset in the content query device. The preset loss function can be constructed according to the model. The content query device inputs the topic feature vector formed by training into the preset loss function, processes the topic feature vector through the preset loss function, and obtains a loss value. The content query device compares the loss value with a preset loss value (the preset loss value can be set according to the gradient) to judge the topic extraction model obtained by training; if the loss value is less than the preset loss value, it is determined that the topic extraction model has converged; if the loss value is greater than or equal to the preset loss value, it is determined that the topic extraction model has not converged.

[0066] Step S04, when it is detected that the trained topic extraction model converges, the trained topic extraction model is output as a preset topic extraction model.

[0067] When the content query device detects that the loss value is less than the preset loss value, it determines that the topic extraction model has converged and the content query device outputs the trained topic extraction model as the preset topic extraction model. In this embodiment, the content query device fully learns the model so that the topic extracted by the preset topic extraction model is more accurate.

[0068] Reference Figure 3 , Figure 3 This is a flow chart of a first embodiment of a content analysis query method of the present invention. In this embodiment, the content analysis query method includes:

[0069] Step S10: upon receiving a content analysis request, obtaining the target content to be analyzed carried in the content analysis request.

[0070] The content query device receives the content query request, and the triggering method of the content query request is not specifically limited, that is, the content query request can be actively triggered by the user, for example, the user clicks the "content query" button on the content query device display page to actively trigger the content query request; in addition, the content query request can also be automatically triggered, for example, the content query device is pre-set: when receiving new search content, the content query request is automatically triggered.

[0071] The content query device receives the content query request, and obtains the target content associated with the content query request. The specific content and format of the target content are not limited. For example, the target content can be a text document format, PDF (short for Portable Document Format, meaning "portable document format") or image format, etc.

[0072] Step S20: parsing the target content to obtain keywords in the target content, and determining the content theme of the target content according to the keywords.

[0073] The content query device parses the target content to obtain keywords in the target content, and determines the content theme of the target content according to the keywords, specifically including:

[0074] Step a1, performing word segmentation processing on the target content through a preset topic extraction model to obtain a word set corresponding to the target content, removing noise words in the word set, and obtaining keywords in the target content;

[0075] Step a2, obtaining the position information and frequency information of the keyword, and determining the content theme of the target content according to the position information and frequency information of the keyword through a preset theme extraction model.

[0076] The content query device performs word segmentation processing on the target content through a preset topic extraction model to obtain a word set corresponding to the target content. The content query device removes noise words in the word set to obtain keywords in the target content; the content query device extracts the location information and frequency information of the keywords. The content query device processes the location information and frequency information of the keywords through a preset topic extraction model to determine the content theme of the target content.

[0077] That is, in this embodiment, the content query device uses a preset topic extraction model to perform semantic analysis on the content of the target content and count the topic of each sentence. The topic statistics are based on the position, number, and frequency of a series of keywords with the same semantics. To determine whether this sentence or paragraph conforms to one or more topics. If there are multiple topics, the weight of each topic can also be counted. For example: the keyword A1 of topic A appears 2 times, and the keyword B1 of topic B appears 1 time, then it can be roughly calculated that the weight of topic A is 2 / 3, and the weight of topic B is 1 / 3.

[0078] Step S30: query a preset database to obtain related content that is similar to the content theme and / or keywords of the target content.

[0079] The content query device extracts the category information, technical field, keywords, application scenario and other information of the target content according to the NLP (Neuro-Linguistic Programming, natural language processing) technology, and clusters and classifies the pre-stored content database according to the category information, technical field, keywords, application scenario and other information; the content query device obtains the pre-stored content of the same cluster as the target content (the same cluster can be the same technical field information, the same scenario or the same keyword), the content query device obtains the content theme of the pre-stored content, the content query device compares the content theme and / or keywords of the target content with the content theme and / or keywords of the pre-stored content, and obtains the related content similar to the content theme and / or keywords of the target content. Specifically, the content query device calculates the keyword similarity and / or theme similarity between the target content and the pre-stored content in the content database, the content query device performs weight analysis on the keyword similarity and / or theme similarity to determine the comprehensive similarity, and the content query device uses the pre-stored content with the highest comprehensive similarity as the related content.

[0080] That is, the content query device performs semantic analysis on the content of the pre-stored content and counts the subject of each sentence. Subject statistics are based on the position, number, and frequency of a series of keywords with the same semantics to determine whether the sentence or paragraph conforms to one or more subjects. If there are multiple subjects, the weights of each subject can also be counted. For example: the keyword A1 of subject A of the pre-stored content appears 2 times, and the keyword B1 of subject B appears 1 time. It can be roughly calculated that the weight of subject A is 2 / 3, and the weight of subject B is 1 / 3. When the content query device compares the content subject of the target content with the content subject of the pre-stored content for similarity, it determines the subject similarity of each content. The content query device multiplies the similarity of each subject by the weight corresponding to the content subject to obtain the comprehensive similarity. The content query device uses the pre-stored content with the highest comprehensive similarity as the associated content.

[0081] Step S40: compare the target content with the associated content, obtain similarity information and difference information between the target content and the associated content, and output the information.

[0082] The content query device compares the target content with the associated content, obtains and outputs similarity information and difference information between the target content and the associated content, specifically including:

[0083] Step b1, comparing the target content with the associated content, and dividing the content themes of the target content into similar content themes and different content themes;

[0084] Step b2, input the similar content topics and the keywords corresponding to the similar content topics as similar information into the similar display area in the preset template, and input the different content topics and the keywords corresponding to the different content topics as different information into the different display area in the preset template to form a content analysis table.

[0085] The content query device compares the target content with the associated content, and divides the content topics of the target content into similar content topics (similar content topics refer to content topics included in both the target content and the associated content) and different content topics (different content topics refer to content topics included in the target content but not included in the associated content); the content query device obtains similar content topics and keywords corresponding to the similar content topics, and inputs the similar content topics and keywords corresponding to the similar content topics as similar information into a similar display area in a preset template; the content query device obtains different content topics and keywords corresponding to the different content topics, and inputs the different content topics and keywords corresponding to the different content topics as different information into a different display area in a preset template, thereby forming a content analysis table to facilitate users to view content query results.

[0086] In this embodiment, the target content is automatically parsed to determine the keywords of the target content, and then the content theme of the target content is determined based on the keywords of the target content. A content query is performed based on the content theme to obtain related content similar to the target content, and similarity information and difference information of the content are automatically output. Manual content analysis is not required, which reduces the waste of human resources, improves the efficiency and accuracy of content query, and makes it convenient for users to view content analysis query results.

[0087] Based on the first embodiment of the content analysis query method of the present invention, a second embodiment of the content analysis query method of the present invention is proposed.

[0088] This embodiment is the step after step S10 in the first embodiment. The difference between this embodiment and the above embodiment is that:

[0089] Acquire the content format of the target content;

[0090] When the content format of the target content is not a text document format, calling a preset OCR model to detect the target content, determine a text area in the target content, and obtain character information contained in the text area, so as to convert the target content into a text document format;

[0091] When the content format of the target content is a text document format, the step of parsing the target content to obtain keywords in the target content and determining the content theme of the target content according to the keywords is performed.

[0092] The content query device obtains the content format of the target content; when the content format of the target content is not a text document format (such as a text document format of doc or txt), the content query device calls a preset OCR model (the preset OCR (Optical Character Recognition) model refers to a pre-set character recognition model) to detect the target content, the content query device determines the text area in the target content, the content query device obtains the character information contained in the text area, and the text document format converts the target content into a text document format to facilitate content query and mining; when the content format of the target content is a text document format, the content query device executes step S20 in the first embodiment: parsing the target content to obtain keywords in the target content, and determining the content theme of the target content according to the keywords. In this implementation, the content query device can perform similar queries on target contents of different formats, mine the differences in content, and make the target content analysis more comprehensive and convenient.

[0093] Based on the above-mentioned embodiments of the content analysis query method of the present invention, a third embodiment of the content analysis query method of the present invention is proposed.

[0094] This embodiment is a refinement of step S30 in the third embodiment. The difference between this embodiment and the above embodiment is that:

[0095] Querying a preset database to obtain pre-stored content in the preset database in the same technical field as the target content;

[0096] Analyze the pre-stored content by using a preset theme extraction model to obtain content themes of the pre-stored content and keywords corresponding to the content themes;

[0097] When the target content includes at least two content themes and / or the content themes include at least two keywords, calculating the theme similarity and / or word similarity between the content themes and / or keywords of the target content and the content themes and / or keywords of the pre-stored content;

[0098] The subject similarity and / or the word similarity are weighted to obtain the comprehensive similarity between the target content and the pre-stored content, and the pre-stored content having a comprehensive similarity with the target content higher than a preset threshold is used as the associated content.

[0099] The content query device obtains the technical field information of the target content, and the content query device queries the pre-stored content database to obtain the pre-stored content with the same technical field information in the pre-stored content database. In this embodiment, only the pre-stored content with the same technical field information as the target content is analyzed, which reduces the content analysis query amount and effectively ensures the accuracy of the content analysis query. The content query device analyzes the pre-stored content through a preset topic extraction model to obtain the content topic of the set content; when the target content contains at least two content topics and / or the content topic contains two keywords, the content query device calculates the topic similarity between each content topic of the target content and each content topic of the content, and / or the content query device calculates the word similarity between the keywords in each content topic of the target content and the keywords in each content topic of the content; the content query device performs weighted calculation on each topic similarity and / or word similarity to obtain the comprehensive similarity between the target content and the pre-stored content, and the content query device uses the pre-stored content with a comprehensive similarity with the target content higher than a preset threshold (the preset threshold can be flexibly set according to the query accuracy, for example, set to 80) as the associated content.

[0100] The content query device uses NLP (natural language processing) technology to analyze and extract the subject and keywords of the target content. The content query device compares the subject and keywords of the target content, as well as the weights of each subject, with the subject, keywords and weights of the documents in the library, and mines out the related content with similar subjects to the target content, and calculates the comprehensive similarity between the contents through the subject, keywords and weights of each subject. The content query device uses the pre-stored content with a comprehensive similarity with the target content higher than a preset threshold (the preset threshold can be flexibly set according to the query accuracy, for example, set to 80%) as the related content. In this embodiment, the comprehensive similarity is determined according to the content subject of each content, thereby determining the related content, so that the query of the related content is more accurate.

[0101] For example: the target content has 6 topics, each topic has 5 keywords, topic A has a weight of 30%, topic B has a weight of 20%, topic C has a weight of 15%, topic D has a weight of 15%, topic E has a weight of 10%, and topic F has a weight of 10%. After comparing with the documents in the library, it is found that there are 0 documents similar to all 6 topics, and there are 5 documents similar to 5 topics, with comprehensive similarities of 84%, 75%, 90%, 68% and 77% respectively. The preset threshold in the content query device is 80%, and the content query device uses the pre-stored content with comprehensive similarities of 84% and 90% as related content.

[0102] Based on the above-mentioned embodiments of the content analysis query method of the present invention, a fourth embodiment of the content analysis query method of the present invention is proposed.

[0103] This embodiment is the step after step S40 in the fourth embodiment. The difference between this embodiment and the above embodiment is that:

[0104] Obtaining the information volume ratio of the similar information in the target content, and when the information volume ratio is less than a preset ratio threshold, determining whether the difference information is disclosed by other content in the pre-stored content database except the associated content;

[0105] If the difference information is not disclosed by other contents in the pre-stored content database except the associated content, analyzing the impact factor of the difference information on the target content and outputting it;

[0106] If the difference information is disclosed by other contents in the pre-stored content database except the associated content, the difference content in the pre-stored content database that discloses the difference information is output.

[0107] In the present implementation, the content query device obtains the information volume ratio of similar information in the target content. When the information volume ratio is less than a preset ratio threshold (the preset ratio threshold can be set according to a specific scenario, for example, set to 90%), the content query device determines whether the difference information is disclosed by other contents except the associated content in the pre-stored content database; if the difference information is not disclosed by other contents except the associated content in the pre-stored content database, the content query device analyzes the influence factor of the difference information on the target content (the influence factor can be flexibly set according to different factors) and outputs it; if the difference information is disclosed by other contents except the associated content in the pre-stored content database, the content query device outputs the difference content in the pre-stored content database that discloses the difference information.

[0108] In this embodiment, the difference information is queried for the second time and the difference content is displayed, which reduces the number of steps for the user to make a second query and makes the difference mining more effective. In addition, in this embodiment, when the difference information is not disclosed, the impact factor of the difference information can be determined to determine the value of the target content. This eliminates the need for manual information analysis and makes content mining more effective.

[0109] Reference Figure 4 , Figure 4 This is a functional module diagram of an embodiment of a content analysis query device of the present invention; the present invention also provides a content analysis query device, the content analysis query device comprising:

[0110] The request receiving module 10 is used to obtain the target content to be analyzed carried in the content analysis request when receiving the content analysis request;

[0111] A parsing and determining module 20, configured to parse the target content to obtain keywords in the target content, and determine the content theme of the target content according to the keywords;

[0112] A query acquisition module 30 is used to query a preset database to acquire related content similar to the content theme and / or keywords of the target content;

[0113] The comparison and output module 40 compares the target content with the associated content, obtains similarity information and difference information between the target content and the associated content, and outputs the information.

[0114] In one embodiment, the query acquisition module includes:

[0115] A query acquisition unit is used to query a preset database to obtain pre-stored content in the preset database in the same technical field as the target content; analyze the pre-stored content through a preset theme extraction model to obtain the content theme of the pre-stored content and the keywords corresponding to the content theme;

[0116] a similarity calculation unit, configured to calculate the subject similarity and / or word similarity of the content subject and / or keyword of the target content and the content subject and / or keyword of the pre-stored content when the target content contains at least two content subjects and / or the content subject contains at least two keywords;

[0117] The content determination unit is used to perform weighted calculation on the subject similarity and / or the word similarity to obtain the comprehensive similarity between the target content and the pre-stored content, and to use the pre-stored content whose comprehensive similarity with the target content is higher than a preset threshold as the associated content.

[0118] In one embodiment, the comparison output module 40 includes:

[0119] The topic division is used to compare the target content with the associated content, and to divide the content topics of the target content into similar content topics and different content topics;

[0120] An input generation unit is used to input the similar content topics and the keywords corresponding to the similar content topics as similar information into the similar display area in the preset template, and input the different content topics and the keywords corresponding to the different content topics as difference information into the different display area in the preset template to form a content analysis table.

[0121] In one embodiment, the content analysis query device includes:

[0122] an acquisition judgment module, configured to acquire the information volume ratio of the similar information in the target content, and when the information volume ratio is less than a preset ratio threshold, determine whether the difference information is disclosed by other content in the pre-stored content database except the associated content;

[0123] The first output is used to analyze and output the influence factor of the difference information on the target content if the difference information is not disclosed by other contents other than the associated content in the pre-stored content database;

[0124] The second output module is configured to output the difference content in the pre-stored content database that discloses the difference information if the difference information is disclosed by other content in the pre-stored content database except the associated content.

[0125] In one embodiment, the analysis and determination module includes:

[0126] A keyword acquisition unit, configured to perform word segmentation processing on the target content through a preset topic extraction model, obtain a word set corresponding to the target content, remove noise words from the word set, and obtain keywords in the target content;

[0127] The theme determination unit is used to obtain the position information and frequency information of the keyword, and determine the content theme of the target content according to the position information and frequency information of the keyword through a preset theme extraction model.

[0128] In one embodiment, the content analysis query device includes:

[0129] The topic extraction module is used for sample tagging. When receiving a model building request, it obtains the initial topic extraction model and tags the content in the local storage to form a training sample set;

[0130] A model training module, used to extract topic training samples from the training sample set, train the initial topic extraction model through the topic training samples, and obtain a trained topic extraction model and a topic feature vector;

[0131] A loss value determination module, used to process the topic feature vector through a preset loss function to obtain a loss value, and analyze the loss value to determine whether the trained topic extraction model converges;

[0132] The model output module is used to output the trained topic extraction model as a preset topic extraction model when it is detected that the trained topic extraction model converges.

[0133] In one embodiment, the content analysis query device includes:

[0134] A format acquisition module, used to acquire the content format of the target content;

[0135] a format conversion module, for, when the content format of the target content is not a text document format, calling a preset OCR model to detect the target content, determine the text area in the target content, and obtain character information contained in the text area, so as to convert the target content into a text document format;

[0136] When the content format of the target content is a text document format, the parsing and determining module 20 is executed to parse the target content to obtain keywords in the target content, and determine the content theme of the target content according to the keywords.

[0137] Among them, the method implemented when the content analysis query device is executed can refer to the various embodiments of the content analysis query method of the present invention, and will not be repeated here.

[0138] In this embodiment, the target content is automatically parsed to determine the keywords of the target content, and then the content theme of the target content is determined based on the keywords of the target content. A content query is performed based on the content theme to obtain related content similar to the target content, and similarity information and difference information of the content are automatically output. Manual content analysis is not required, which reduces the waste of human resources, improves the efficiency and accuracy of content query, and facilitates users to view content analysis query results. The present invention also provides a computer storage medium.

[0139] The computer storage medium of the present invention stores a computer program corresponding to the content analysis query, and when the computer program corresponding to the content analysis query is executed by the processor, the steps of the content analysis query method described above are implemented.

[0140] Among them, the method implemented when the computer program corresponding to the content analysis query running on the processor is executed can refer to the various embodiments of the content analysis query method of the present invention, and will not be repeated here.

[0141] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0142] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0143] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0144] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A content analysis query method, characterized in that: The content analysis query method comprises the following steps: Upon receiving a content analysis request, obtaining target content to be analyzed carried in the content analysis request; Parsing the target content to obtain keywords in the target content, and determining the content theme of the target content according to the keywords; The step of parsing the target content to obtain keywords in the target content and determining the content theme of the target content according to the keywords includes: Performing word segmentation processing on the target content to obtain a word set corresponding to the target content; Removing noise words from the word set to obtain keywords in the target content; Acquire the position information and frequency information of the keyword, and determine the content theme of the target content according to the position information and frequency information of the keyword through a preset theme extraction model; Querying a preset database to obtain related content that is similar to the content theme and / or keywords of the target content; The step of querying a preset database to obtain related content similar to the content theme and / or keywords of the target content includes: Querying a preset database to obtain pre-stored content in the preset database in the same technical field as the target content; Analyze the pre-stored content by using a preset theme extraction model to obtain content themes of the pre-stored content and keywords corresponding to the content themes; When the target content includes at least two content themes and / or the content themes include at least two keywords, calculating the theme similarity and / or word similarity between the content themes and / or keywords of the target content and the content themes and / or keywords of the pre-stored content; Performing a weighted operation on the topic similarity and / or the word similarity to obtain a comprehensive similarity between the target content and the pre-stored content, and taking the pre-stored content whose comprehensive similarity with the target content is higher than a preset threshold as the associated content, wherein the preset threshold is flexibly set according to the query accuracy; The target content is compared with the associated content, and similarity information and difference information between the target content and the associated content are obtained and output.

2. The content analysis query method according to claim 1, characterized in that: The step of comparing the target content with the associated content, obtaining similarity information and difference information between the target content and the associated content, and outputting the information includes: Comparing the target content with the associated content, and dividing the content themes of the target content into similar content themes and different content themes; The similar content topics and the keywords corresponding to the similar content topics are input as similar information into the similar display area in the preset template, and the different content topics and the keywords corresponding to the different content topics are input as different information into the different display area in the preset template to form a content analysis table.

3. The content analysis query method according to claim 1, characterized in that: After the step of comparing the target content with the associated content, obtaining and outputting similarity information and difference information between the target content and the associated content, the method further includes: Obtaining the information volume ratio of the similar information in the target content, and when the information volume ratio is less than a preset ratio threshold, determining whether the difference information is disclosed by other content in the pre-stored content database except the associated content; If the difference information is not disclosed by other contents in the pre-stored content database except the associated content, analyzing the impact factor of the difference information on the target content and outputting it; If the difference information is disclosed by other contents in the pre-stored content database except the associated content, the difference content in the pre-stored content database that discloses the difference information is output.

4. The content analysis query method according to claim 3, characterized in that: Before the step of analyzing the pre-stored content by using a preset theme extraction model to obtain the content theme of the pre-stored content and the keywords corresponding to the content theme, the method includes: When receiving a model building request, the initial topic extraction model is obtained, and the content in the local storage is marked to form a training sample set; Extracting topic training samples from the training sample set, training the initial topic extraction model through the topic training samples, and obtaining a trained topic extraction model and a topic feature vector; Processing the topic feature vector by a preset loss function to obtain a loss value, and analyzing the loss value to determine whether the trained topic extraction model has converged; When it is detected that the trained topic extraction model converges, the trained topic extraction model is output as a preset topic extraction model.

5. The content analysis query method according to any one of claims 1 to 4, characterized in that: After the step of obtaining the target content to be analyzed carried in the content analysis request upon receiving the content analysis request, the method further includes: Acquire the content format of the target content; When the content format of the target content is not a text document format, calling a preset OCR model to detect the target content, determine a text area in the target content, and obtain character information contained in the text area, so as to convert the target content into a text document format; When the content format of the target content is a text document format, the step of parsing the target content to obtain keywords in the target content and determining the content theme of the target content according to the keywords is performed.

6. A content analysis query device, characterized in that: The content analysis query device comprises: A request receiving module, configured to, upon receiving a content analysis request, obtain the target content to be analyzed carried in the content analysis request; A parsing and determining module is used to parse the target content to obtain keywords in the target content, and determine the content theme of the target content according to the keywords, wherein the parsing and determining module is specifically used to perform word segmentation processing on the target content to obtain a word set corresponding to the target content; remove noise words in the word set to obtain keywords in the target content; obtain position information and frequency information of the keywords, and determine the content theme of the target content according to the position information and frequency information of the keywords through a preset theme extraction model; A query acquisition module is used to query a preset database to obtain related content similar to the content theme and / or keywords of the target content, wherein the query acquisition module is specifically used to query the preset database to obtain pre-stored content in the preset database in the same technical field as the target content; analyze the pre-stored content through a preset theme extraction model to obtain the content theme of the pre-stored content and the keywords corresponding to the content theme; when the target content contains at least two content themes and / or the content theme contains at least two keywords, calculate the theme similarity and / or word similarity between the content theme and / or keywords of the target content and the content theme and / or keywords of the pre-stored content; perform a weighted operation on the theme similarity and / or the word similarity to obtain the comprehensive similarity between the target content and the pre-stored content, and use the pre-stored content whose comprehensive similarity with the target content is higher than a preset threshold as the related content, wherein the preset threshold is flexibly set according to the query accuracy; The comparison and output module compares the target content with the associated content, obtains similarity information and difference information between the target content and the associated content, and outputs the information.

7. A content analysis query device, characterized in that: The content analysis query device includes: a memory, a processor, and a computer program corresponding to the content analysis query stored in the memory and executable on the processor. When the computer program corresponding to the content analysis query is executed by the processor, the steps of the content analysis query method as described in any one of claims 1 to 5 are implemented.

8. A computer storage medium, characterized in that: The computer storage medium stores a computer program corresponding to the content analysis query, and when the computer program corresponding to the content analysis query is executed by the processor, the steps of the content analysis query method according to any one of claims 1 to 5 are implemented.

Images