Intelligent data acquisition method and device, computer equipment and storage medium

Abstract

The embodiment of the invention discloses an intelligent data collection method and device, computer equipment and a storage medium. The data intelligent collection method is applied to the data intelligent collection system, the data intelligent collection system comprises an RPA assembly and a data processing large model, and the method comprises the steps that the RPA assembly obtains a webpage source code of a target webpage and sends the webpage source code to the data processing large model; the data processing large model determines a collection script of the RPA component according to the webpage source code, and sends the collection script to the RPA component; the RPA component performs multi-modal information collection on the target webpage based on the collection script to obtain multi-modal data corresponding to the target webpage; the large data processing model carries out data fusion and verification processing on the multi-modal data, and target webpage data corresponding to the target webpage are acquired. By implementing the method provided by the embodiment of the invention, the RPA component can dynamically adapt to the webpage structure change to automatically collect the webpage data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet, and particularly relates to a data intelligent collection method and device, a computer device and a storage medium. BACKGROUND

[0002] Robotic Process Automation (RPA) technology can simulate human operations on a computer, automatically perform repetitive and rule-based work, thereby improving work efficiency and reducing human errors. Applying RPA to web page data collection can realize automatic collection of web page data. However, the existing RPA-based automatic collection of web page data relies on fixed rule-based automatic collection scripts for web page data collection. When the web page structure changes, the collection script of the RPA needs to be manually updated to maintain the automatic collection function.

[0003] It can be seen that the existing RPA-based automatic collection of web page data shows obvious vulnerability and inefficiency in the face of dynamic web page environment, and there is an urgent need for a web page data intelligent collection method that can dynamically adapt to changes in web page structure. SUMMARY

[0004] The embodiments of the present application provide a data intelligent collection method and device, a computer device and a storage medium, which can enable the RPA component to dynamically adapt to changes in the web page structure for automatic collection of web page data.

[0005] In a first aspect, the embodiments of the present application provide a data intelligent collection method, which is applied to a data intelligent collection system including an RPA component and a data processing large model. The method comprises the following steps:

[0006] The RPA component acquires the web page source code of a target web page and sends the web page source code to the data processing large model;

[0007] The data processing large model determines the collection script of the RPA component according to the web page source code and sends the collection script to the RPA component;

[0008] The RPA component collects multi-modal information of the target web page based on the collection script to obtain multi-modal data corresponding to the target web page;

[0009] The data processing large model performs data fusion and verification processing on the multi-modal data to collect target web page data corresponding to the target web page.

[0010] In some embodiments, the RPA component performs multi-modal information collection on the target webpage based on the collection script to obtain multi-modal data corresponding to the target webpage, including:

[0011] Determining element positions corresponding to respective element types in the target webpage based on the collection script;

[0012] According to a preset correspondence between element types and collection rules, determining target collection rules corresponding to respective element positions;

[0013] Collecting target element data on respective element positions based on the corresponding target collection rules to obtain the multi-modal data.

[0014] In some embodiments, the multi-modal data includes text data, screenshot data, and API data; and collecting target element data on respective element positions based on the corresponding target collection rules to obtain the multi-modal data includes:

[0015] If the target collection rule is a text copy rule, performing text copy processing on the corresponding element position according to the text copy rule to obtain the text data;

[0016] If the target collection rule is a screenshot rule, performing screenshot processing on the corresponding element position according to the screenshot rule to obtain the screenshot data;

[0017] If the target collection rule is an API data acquisition rule, determining an API identifier of a target API in the corresponding element position, obtaining an API key and request parameters corresponding to the API identifier from a preset API database, and acquiring the API data from the target API according to the API key and the request parameters.

[0018] In some embodiments, the data processing large model performs data fusion and verification processing on the multi-modal data to obtain target webpage data corresponding to the target webpage, including:

[0019] Performing semantic understanding on the text data through NLP technology to obtain text key information in the text data, and mapping the text key information to a preset field sequence to obtain a first field sequence;

[0020] Extracting screenshot text information and image feature information in the screenshot data through OCR technology and CV technology;

[0021] determine screenshot key information according to the screenshot text information and the image feature information, map the screenshot key information to the first field sequence, and verify text key information in the first field sequence based on the screenshot key information to obtain a second field sequence;

[0022] map the API data to the second field sequence according to a data type of the API data, and verify text key information or screenshot key information in the second field sequence based on the API data to obtain a third field sequence;

[0023] determine the target web page data according to the third field sequence.

[0024] In some embodiments, before the RPA component acquires the web page source code of the target web page, the method further comprises:

[0025] The data processing large model determines a plurality of candidate web pages according to a search keyword input by a user;

[0026] The data processing large model prioritizes a plurality of the candidate web pages according to a relevance of each of the candidate web pages to the search keyword and an authority of each of the candidate web pages;

[0027] The data processing large model determines a first preset number of candidate web pages in the priority order as target web pages, the preset number of web pages being an integer greater than or equal to 1;

[0028] The data processing large model sends a corresponding web address of each of the target web pages to the RPA component.

[0029] In some embodiments, the RPA component acquires the web page source code of the target web page, comprising:

[0030] For each of the target web pages, open the target web page according to the corresponding web address;

[0031] Acquire the corresponding web page source code through a preset web page parsing tool.

[0032] In some embodiments, when the preset number of web pages is greater than 1, the target web pages are multiple; the data processing large model performs data fusion and verification processing on the multi-modal data to collect target web page data corresponding to the target web pages, comprising:

[0033] Perform data fusion and verification processing on the multi-modal data corresponding to each of the target web pages to obtain candidate target web page data corresponding to each of the target web pages;

[0034] Correlation analysis is performed on the plurality of candidate target webpage data to obtain the target webpage data.

[0035] In a second aspect, the embodiments of the present application further provide a data intelligent collection system, comprising an RPA component and a data processing large model, wherein:

[0036] The RPA component is configured to obtain webpage source code of a target webpage and send the webpage source code to the data processing large model.

[0037] The data processing large model is configured to determine a collection script of the RPA component according to the webpage source code and send the collection script to the RPA component.

[0038] The RPA component is further configured to collect multi-modal information of the target webpage based on the collection script to obtain multi-modal data corresponding to the target webpage.

[0039] The data processing large model is further configured to perform data fusion and verification processing on the multi-modal data to collect target webpage data corresponding to the target webpage.

[0040] In a third aspect, the embodiments of the present application further provide a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the above method when executing the computer program.

[0041] In a fourth aspect, the embodiments of the present application further provide a computer readable storage medium, wherein the storage medium stores a computer program, and the computer program comprises program instructions, and the program instructions can implement the above method when executed by a processor.

[0042] The embodiment of the present application provides a data intelligent collection method and device, computer equipment and a storage medium. The data intelligent collection method is applied to a data intelligent collection system, the data intelligent collection system comprises an RPA component and a data processing large model, and the method comprises the following steps: the RPA component acquires the web source code of a target web page and sends the web source code to the data processing large model; the data processing large model determines the collection script of the RPA component according to the web source code and sends the collection script to the RPA component; the RPA component collects multi-modal information of the target web page based on the collection script, and obtains multi-modal data corresponding to the target web page; and the data processing large model performs data fusion and verification processing on the multi-modal data, and collects target web page data corresponding to the target web page. The embodiment of the present application can dynamically determine the current collection script of the RPA component through the data processing large model and the currently acquired web source code. Therefore, the RPA component can dynamically adapt to the change of the web page structure and automatically collect web page data. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0044] Figure 1 The flowchart of the data intelligent collection method provided by the embodiment of the present application is shown in the figure.

[0045] Figure 2 Another flowchart of the data intelligent collection method provided by the embodiment of the present application is shown in the figure.

[0046] Figure 3 A sub-flowchart of the data intelligent collection method provided by the embodiment of the present application is shown in the figure.

[0047] Figure 4 Another sub-flowchart of the data intelligent collection method provided by the embodiment of the present application is shown in the figure.

[0048] Figure 5 The schematic block diagram of the data intelligent collection system provided by the embodiment of the present application is shown in the figure.

[0049] Figure 6 The schematic block diagram of the computer equipment provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0050] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are a part rather than all of the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort should fall within the scope of the present application.

[0051] It should be understood that the terms "comprising" and "including" as used in the specification and the appended claims indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0052] It should also be understood that the terms used in the present application specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application specification and the appended claims, the singular forms "a", "an" and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0053] It should be further understood that the term "and / or" used in the present application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations thereof, and includes these combinations.

[0054] The embodiments of the present application provide a data intelligent collection method and device, computer equipment and a storage medium.

[0055] The execution subject of the data intelligent collection method can be a data intelligent collection system provided by the embodiments of the present application, or a computer device integrated with the data intelligent collection system, wherein the data intelligent collection system can be realized in the form of hardware or software, and the computer device can be a terminal or a server.

[0056] Specifically, the data intelligent collection system provided by the embodiments includes an RPA component and a data processing large model.

[0057] Before executing the data intelligent collection method provided by the present application, the computer equipment needs to be initialized for data collection, specifically, the environment configuration of the RPA component and the access preparation of the data processing large model are needed, as follows:

[0058] Environment configuration of the RPA component: according to the target collection task, the running environment of the RPA component is configured, including setting the browser driver (such as ChromeDriver, etc.), initializing the basic settings of the simulated user behavior (such as mouse moving speed, click delay, etc.).

[0059] Access preparation of the data processing large model: access the selected large model, set up the communication interface, call permission and data transmission format, etc., to ensure that the large model can receive and process the data collected by the RPA component.

[0060] Figure 1 is a flowchart of a data intelligent collection method provided by an embodiment of the present application. As shown in Figure 1 , the method comprises the following steps S110-S140.

[0061] S110, the RPA component acquires the web source code of the target webpage and sends the web source code to the data processing large model.

[0062] In this embodiment, the RPA component can open the target webpage through the browser based on the preset or user input URL, and then acquire the web source code of the target webpage, i.e. the HTML source code, through the preset web parsing tool.

[0063] In other embodiments, the RPA component can also open the target webpage based on the URL sent by the data processing large model. At this time, please refer to Figure 2 , before step S110, the method further comprises:

[0064] S101, the data processing large model determines a plurality of candidate webpages according to the search keywords input by the user;

[0065] S102, the data processing large model prioritizes the plurality of candidate webpages according to the relevance of each candidate webpage to the search keywords and the authority of each candidate webpage;

[0066] S103, the data processing large model determines the candidate webpages in the top preset number of webpages in the priority order as the target webpages, and the preset number of webpages is an integer greater than or equal to 1;

[0067] S104, the data processing large model sends the URLs corresponding to each of the target webpages to the RPA component.

[0068] Specifically, the large model determines the relevance by calculating the semantic similarity between the search keywords and the text content of the candidate web page, and the higher the semantic similarity, the greater the relevance; the authority of the candidate web page can be determined by referring to the authority index of the web page domain name, and then the importance of each candidate web page is determined according to the relevance of each candidate web page, the weight corresponding to the relevance, the authority, and the weight corresponding to the authority, and the candidate web pages are sorted from high to low according to the importance to obtain the priority ranking of the candidate web pages. Based on the priority ranking, N (N is an integer greater than or equal to 1) candidate web pages are selected as target web pages, wherein the target web page is a web page that needs to be collected by the RPA component, and finally the RPA component is sent to the target web page corresponding to the target web page.

[0069] It can be seen that the data processing large model in the embodiment can use its search engine level capability to input keywords related to the collection task, intelligently discover relevant web pages, and prioritize according to factors such as relevance and authority of the web page to determine the target web page. When there are multiple target web pages, subsequent data of multiple related target web pages can be analyzed to mine potential data relationships and values, and improve the breadth and relevance of data search.

[0070] S120, the data processing large model determines the collection script of the RPA component according to the web page source code, and sends the collection script to the RPA component.

[0071] In the embodiment, after the data processing large model obtains the web page source code sent by the RPA component, the semantic information of the web page source code is analyzed by using natural language understanding (NLP) technology, and the key elements and structures such as web page layout and form structure are extracted to determine the collection script, wherein the collection script indicates the element positioning corresponding to each element type (text type, picture type and API type) in the target web page.

[0072] S130, the RPA component collects multi-modal information of the target web page based on the collection script to obtain multi-modal data corresponding to the target web page.

[0073] In some embodiments, as shown in Figure 3 The RPA component collects multi-modal data corresponding to the target web page by the following steps.

[0074] S1301, determining the element positioning corresponding to each element type in the target web page based on the collection script;

[0075] S1302, determine the target acquisition rule corresponding to each element positioning according to the preset correspondence between the element type and the acquisition rule;

[0076] S1303, acquire target element data on each element positioning based on the corresponding target acquisition rule, to obtain the multi-modal data.

[0077] Specifically, the element type includes a text type, a picture type, and an API type, the acquisition rule corresponding to the text type is a text copying rule, the acquisition rule corresponding to the picture type is a screenshot rule, and the acquisition rule corresponding to the API type is an API data acquisition rule. In addition, if an API exists in the target webpage, the embodiment also needs to pre-set an API key and request parameters corresponding to the API.

[0078] At this time, the acquisition of target element data on each element positioning based on the corresponding target acquisition rule to obtain the multi-modal data includes:

[0079] If the target acquisition rule is a text copying rule, the corresponding element positioning is subjected to text copying processing according to the text copying rule to obtain the text data.

[0080] If the target acquisition rule is a screenshot rule, the corresponding element positioning is subjected to screenshot processing according to the screenshot rule to obtain the screenshot data.

[0081] If the target acquisition rule is an API data acquisition rule, the API identifier of the target API in the corresponding element positioning is determined, the API key and the request parameters corresponding to the API identifier are acquired from a preset API database, and the API data is acquired from the target API according to the API key and the request parameters.

[0082] Specifically, in the webpage browsing process, the RPA component recognizes and extracts unstructured text information in the page, such as article content, product description, etc., based on the current acquisition script. For some information in the webpage that is difficult to obtain directly in the form of text (such as pictures, charts, verification code pictures, etc.), the RPA performs a screenshot operation based on the acquisition script, saves the screenshot as a picture file format, and provides materials for subsequent multi-modal analysis. If the target webpage has a corresponding API, the RPA component sends a request to the API according to the pre-configured API key and request parameters, acquires related data, for example, for the inventory API, the API key is used to access the API, the inventory acquisition request carrying the corresponding request parameter (such as product 123) is sent to the database recording the real-time inventory through the API, and the real-time inventory data returned by the database is acquired.

[0083] It can be seen that the RPA component in the embodiment can select different data extraction methods according to different data types, instead of relying on a single data extraction tool, thereby improving the data extraction capability.

[0084] Further, the RPA component interacts with the data processing large model in real time. When the data processing large model determines that the current web page needs to be verified based on the screenshot data collected by the RPA component, the data processing large model generates a verification script in real time and sends it to the RPA component, so that the RPA component performs verification operations on the current page based on the real-time updated verification script. For example, the data processing large model identifies that the current web page needs to perform a moving operation on a verification picture. At this time, the data processing large model determines the current position and target position of the picture that needs to be moved by analyzing the screenshot data, and generates a verification script according to the current position and target position of the picture and sends it to the RPA component.

[0085] S140, the data processing large model performs data fusion and verification processing on the multi-modal data to collect target web page data corresponding to the target web page.

[0086] In some embodiments, as shown in Figure 4 The data processing large model specifically performs data fusion and verification processing on the multi-modal data to collect target web page data corresponding to the target web page by the following steps:

[0087] S1401, performing semantic understanding on the text data by NLP technology to obtain text key information in the text data, and mapping the text key information to a preset field sequence to obtain a first field sequence;

[0088] Specifically, the data processing large model performs semantic understanding on the text type data by natural language processing (NLP) technology, identifies the key information in the text, and then maps it to a pre-set target field sequence, wherein the target field sequence includes column fields of different attributes. In this embodiment, the identified text key information is mapped to the column field corresponding to the attribute. For example, the text data is a product introduction article, and the data processing large model identifies the product name, specification, price and other key information from the article, and then maps the specific product name, specification, price and other key information to the column field corresponding to the attribute in the target field sequence, for example, maps the specific product name to the column field corresponding to the product name attribute, and maps the specific specification parameter to the column field corresponding to the specification attribute.

[0089] In the embodiment, the preset field sequence is dynamically generated based on the search keyword input by the user. For example, in the e-commerce data collection, the user inputs the keywords: collect the pictures, rules, price, inventory, and comments of the product A, and the field sequence with the attributes of picture, rule, price, and inventory is generated.

[0090] In the embodiment, the user can also set the multi-round extraction logic, for example, first extract the basic information of the product, and then extract the comment information.

[0091] S1402, extract the screenshot text information and image feature information in the screenshot data by the OCR technology and the CV technology;

[0092] Specifically, the screenshot data is analyzed by the computer vision (CV), the screenshot text information in the screenshot data is extracted by the optical character recognition (OCR) based on the CV analysis result, and the image feature information of the screenshot data is extracted by the image recognition processing based on the CV analysis result.

[0093] S1403, determine the screenshot key information according to the screenshot text information and the image feature information, map the screenshot key information to the first field sequence, and verify the text key information in the first field sequence based on the screenshot key information to obtain the second field sequence;

[0094] Specifically, after the screenshot text information and the image feature information are extracted, the screenshot text information and the image feature information are associated, for example, the size information of the product is determined by the screenshot text information, the color and shape information of the product are determined by the image feature information, and the product description information containing the size, color, and shape of the product is determined by the association of the screenshot text information and the image feature information as the screenshot key information.

[0095] Because text on a webpage may be locked, directly copying the text may result in garbled characters or incorrect text data. In such cases, it's necessary to validate the text key information of the corresponding attributes by taking a screenshot. If the validation passes (i.e., the data for the corresponding attribute is the same), the text key information for that attribute is retained. If the validation fails, the screenshot key information replaces the text key information for that attribute. For example, if the price field for product A in the first field sequence corresponds to a price of 100 yuan, but the screenshot key information indicates that the price of product A should be 150 yuan, then the price of product A in the corresponding column for the price attribute in the first field sequence needs to be replaced with 150 yuan. That is, when the text key information and the screenshot key information differ, the screenshot key information takes precedence. Furthermore, for attribute information not extracted from the text key information, the specific data under the first field sequence is further supplemented based on the screenshot key information.

[0096] S1404. Map the API data to the second field sequence according to the data type of the API data, and perform verification processing on the text key information or screenshot key information in the second field sequence based on the API data to obtain the third field sequence.

[0097] In this embodiment, when mapping the API data to the second field sequence according to the data type of the API data, if there is no specific attribute data under the corresponding attribute field, the corresponding API data is directly mapped to the attribute field of the corresponding column. If there is specific attribute data under the corresponding attribute field, the API data is used to overwrite the previously mapped data. That is, if there is text key information or screenshot key information under the corresponding attribute field, and it is different from the API data, the text key information or screenshot key information under the corresponding attribute field is replaced with the API data, and the API data shall prevail in the end.

[0098] For example, if the inventory quantity in the second field sequence is 120, while the inventory quantity obtained through the API is 110, then the inventory quantity is replaced with 110. If the inventory quantity obtained through the API is the same as the inventory quantity that already exists in the second field sequence, then there is no need to modify the inventory quantity.

[0099] S1405. Determine the target webpage data based on the third field sequence.

[0100] In this embodiment, after obtaining the third field sequence, the large language model further refines the data in the third field sequence and uses the refined data as the final extracted target webpage data.

[0101] It can be seen that the data joint verification processing can be performed on the multi-modal data, and the accuracy of webpage data extraction is improved.

[0102] Further, in some embodiments, when the preset number of webpages is greater than 1, the target webpages are multiple; the data processing large model performs data fusion and verification processing on the multi-modal data to collect target webpage data corresponding to the target webpages, including: performing data fusion and verification processing on the multi-modal data corresponding to each target webpage respectively to obtain candidate target webpage data corresponding to each target webpage respectively; performing correlation analysis processing on multiple candidate target webpage data to obtain the target webpage data.

[0103] Specifically, for each target webpage, steps S110-S130 are performed respectively, so that the data processing large model obtains multi-modal data corresponding to each target webpage respectively, and then determines third field sequences (candidate target webpage data) corresponding to each target webpage respectively based on steps S1401-S1404, and finally performs correlation analysis processing on multiple third field sequences to obtain fourth field sequences, and then obtains the target webpage data based on the fourth field sequences.

[0104] Among them, when performing correlation analysis processing on multiple third field sequences, the attributes data of each third field sequence are fused for correlation supplement, and for the case that multiple different attribute data correspond to the same attribute field, the attribute data of the target webpage with the greatest importance is determined as the final attribute data under the attribute field, and finally the fourth field sequence is obtained, and then the data in the fourth field sequence is polished in language, and the polished data is taken as the final extracted target webpage data.

[0105] It can be seen that the embodiment can perform correlation analysis on the data of multiple related webpages, and the breadth and correlation of the data search range are improved.

[0106] In addition, the data intelligent collection system provided by the present application is also configured with an incremental learning mechanism, and the data continuously collected by the RPA component is fed back to the large model, and the large model performs incremental learning on these new data. Through learning, the large model can quickly adapt to new types of webpages, improve the recognition accuracy of data in specific fields, and continuously optimize the collection effect.

[0107] To sum up, the data intelligent collection method provided by the embodiment is applied to a data intelligent collection system, the data intelligent collection system comprises an RPA component and a data processing large model, and the method comprises the following steps: the RPA component acquires the web source code of a target web page and sends the web source code to the data processing large model; the data processing large model determines a collection script of the RPA component according to the web source code and sends the collection script to the RPA component; the RPA component collects multi-modal information of the target web page based on the collection script and obtains multi-modal data corresponding to the target web page; and the data processing large model performs data fusion and verification processing on the multi-modal data and collects target web page data corresponding to the target web page. The embodiment of the application can dynamically determine the current collection script of the RPA component through the data processing large model and the currently acquired web source code. It can be seen that the scheme can enable the RPA component to dynamically adapt to changes in the web page structure and automatically collect web page data.

[0108] Figure 5 is a schematic block diagram of a data intelligent collection system provided by the embodiment of the application. As shown in Figure 5 Corresponding to the above data intelligent collection method, the application further provides a data intelligent collection system 500. The data intelligent collection system 500 comprises units for executing the above data intelligent collection method, and the data intelligent collection system 500 can be configured in a terminal or a server. Specifically, please refer to Figure 5 The data intelligent collection system 500 comprises an RPA component 501 and a data processing large model 502, wherein:

[0109] The RPA component 501 is configured to acquire the web source code of a target web page and send the web source code to the data processing large model 502.

[0110] The data processing large model 502 is configured to determine a collection script of the RPA component 501 according to the web source code and send the collection script to the RPA component 501.

[0111] The RPA component 501 is further configured to collect multi-modal information of the target web page based on the collection script and obtain multi-modal data corresponding to the target web page.

[0112] The data processing large model 502 is further configured to perform data fusion and verification processing on the multi-modal data and collect target web page data corresponding to the target web page.

[0113] In some embodiments, the RPA component 501, when performing the step of performing multi-modal information collection on the target webpage based on the collection script to obtain multi-modal data corresponding to the target webpage, is specifically configured to:

[0114] determining element positions corresponding to each element type in the target webpage based on the collection script;

[0115] determining target collection rules corresponding to each element position based on a preset correspondence between element types and collection rules;

[0116] collecting target element data on each element position based on the corresponding target collection rules to obtain the multi-modal data.

[0117] In some embodiments, the multi-modal data includes text data, screenshot data, and API data; and the RPA component 501, when performing the step of collecting target element data on each element position based on the corresponding target collection rules to obtain the multi-modal data, is specifically configured to:

[0118] if the target collection rule is a text copy rule, performing text copy processing on the corresponding element position based on the text copy rule to obtain the text data;

[0119] if the target collection rule is a screenshot rule, performing screenshot processing on the corresponding element position based on the screenshot rule to obtain the screenshot data;

[0120] if the target collection rule is an API data acquisition rule, determining an API identifier of a target API in the corresponding element position, obtaining an API key and request parameters corresponding to the API identifier from a preset API database, and acquiring the API data from the target API based on the API key and the request parameters.

[0121] In some embodiments, the data processing large model 502, when performing the step of performing data fusion and verification processing on the multi-modal data to collect target webpage data corresponding to the target webpage, is specifically configured to:

[0122] performing semantic understanding on the text data through NLP technology to obtain text key information in the text data and mapping the text key information to a preset field sequence to obtain a first field sequence;

[0123] extracting screenshot text information and image feature information in the screenshot data through OCR technology and CV technology;

[0124] determine screenshot key information according to the screenshot text information and the image feature information, map the screenshot key information to the first field sequence, and perform verification processing on text key information in the first field sequence based on the screenshot key information to obtain a second field sequence;

[0125] map the API data to the second field sequence according to a data type of the API data, and perform verification processing on text key information or screenshot key information in the second field sequence based on the API data to obtain a third field sequence;

[0126] determine the target web page data according to the third field sequence.

[0127] In some embodiments, before the RPA component 501 performs the step of obtaining the web page source code of the target web page, the data processing large model 502 is further configured to:

[0128] determine a plurality of candidate web pages according to a search keyword input by a user;

[0129] perform priority ranking on the plurality of candidate web pages according to a relevance of each candidate web page to the search keyword and an authority of each candidate web page;

[0130] determine a candidate web page in a front preset number of candidate web pages in the priority ranking as a target web page, the preset number of web pages being an integer greater than or equal to 1;

[0131] send a corresponding web address of each target web page to the RPA component 501.

[0132] In some embodiments, when the RPA component 501 performs the step of obtaining the web page source code of the target web page, it is specifically configured to:

[0133] for each target web page, open the target web page according to the corresponding web address;

[0134] obtain the corresponding web page source code through a preset web page analysis tool.

[0135] In some embodiments, when the preset number of web pages is greater than 1, there are multiple target web pages; when the data processing large model 502 performs the step of performing data fusion and verification processing on the multi-modal data to obtain target web page data corresponding to the target web page, it is specifically configured to:

[0136] perform data fusion and verification processing on the multi-modal data corresponding to each target web page to obtain candidate target web page data corresponding to each target web page;

[0137] Correlation analysis is performed on the plurality of candidate target webpage data to obtain the target webpage data.

[0138] In summary, the embodiment of the application can dynamically determine the current collection script of the RPA component 501 through the data processing large model 502 and the currently obtained webpage source code. It can be seen that the present scheme can enable the RPA component 501 to dynamically adapt to changes in webpage structure for webpage data automatic collection.

[0139] It should be noted that those skilled in the art can clearly understand the specific implementation process of the above data intelligent collection system and each unit, which can refer to the corresponding description in the foregoing method embodiments. For the convenience and brevity of description, it will not be repeated here.

[0140] The above data intelligent collection system can be implemented in the form of a computer program, which can run on a computer device as shown in the computer device. Figure 6 .

[0141] Please refer to Figure 6 , Figure 6 is a schematic block diagram of a computer device provided by an embodiment of the application. The computer device 600 can be a terminal or a server, and the data intelligent collection system is deployed on the computer device 600, which includes an RPA component and a data processing large model.

[0142] Referring to Figure 6 , the computer device 600 includes a processor 602, a memory, and a network interface 605 connected through a system bus 601, wherein the memory can include a non-volatile storage medium 603 and an internal memory 604.

[0143] The non-volatile storage medium 603 can store an operating system 6031 and a computer program 6032. The computer program 6032 includes program instructions, which when executed, can enable the processor 602 to perform a data intelligent collection method.

[0144] The processor 602 is configured to provide computing and control capabilities to support the operation of the entire computer device 600.

[0145] The internal memory 604 provides an environment for the running of the computer program 6032 in the non-volatile storage medium 603, and the computer program 6032, when executed by the processor 602, can enable the processor 602 to perform a data intelligent collection method.

[0146] The network interface 605 is configured to perform network communication with other devices. Those skilled in the art can understand that Figure 6The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device 600 to which the scheme of the present application is applied. Specifically, the computer device 600 can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0147] The processor 602 is configured to run the computer program 6032 stored in the memory to implement the following steps:

[0148] The RPA component is configured to acquire the web source code of the target webpage and send the web source code to the data processing large model;

[0149] The data processing large model is configured to determine the collection script of the RPA component according to the web source code and send the collection script to the RPA component;

[0150] The RPA component is configured to perform multi-modal information collection on the target webpage based on the collection script to obtain multi-modal data corresponding to the target webpage;

[0151] The data processing large model is configured to perform data fusion and verification processing on the multi-modal data to collect target webpage data corresponding to the target webpage.

[0152] It should be understood that, in the embodiments of the present application, the processor 602 can be a central processing unit (CPU), and the processor 602 can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0153] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiments of the method can be completed by a computer program instructing relevant hardware. The computer program includes program instructions, and the computer program can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the above-mentioned embodiments of the method.

[0154] Therefore, the application further provides a storage medium. The storage medium can be a computer-readable storage medium. The storage medium stores a computer program, where the computer program includes program instructions. The program instructions are executed by a processor to enable the processor to perform the following steps:

[0155] The RPA component is used to obtain web source code of the target webpage, and the web source code is sent to the data processing large model;

[0156] The data processing large model is used to determine a collection script of the RPA component according to the web source code, and the collection script is sent to the RPA component;

[0157] The RPA component is used to perform multi-modal information collection on the target webpage based on the collection script, to obtain multi-modal data corresponding to the target webpage;

[0158] The data processing large model is used to perform data fusion and verification processing on the multi-modal data, to collect target webpage data corresponding to the target webpage.

[0159] The storage medium can be a U disk, a mobile hard disk, a read-only memory (ROM), a magnetic disk or an optical disk, and various computer-readable storage media that can store program codes.

[0160] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been described in the above description in general terms. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the application.

[0161] In several embodiments provided in the application, it should be understood that the disclosed apparatus and method can be implemented in other ways. For example, the apparatus embodiments described above are only schematic. For example, the division of each unit is only a logical function division, and actual implementation can have another division manner. For example, a plurality of units or components can be combined or integrated into another system, or some features can be omitted or not executed.

[0162] The steps in the method embodiments of the present application can be adjusted in sequence, combined and reduced according to actual needs. The units in the device embodiments of the present application can be combined, divided and reduced according to actual needs. In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0163] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a storage medium. Based on such understanding, the technical solutions of the present application essentially or say the part of the prior art that contributes to the technical solutions, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application.

[0164] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A data intelligent acquisition method, characterized in that, The intelligent data acquisition method is applied to an intelligent data acquisition system, which includes an RPA component and a large data processing model. The method includes: The RPA component obtains the source code of the target webpage and sends the source code to the large data processing model. The large data processing model determines the collection script of the RPA component based on the webpage source code and sends the collection script to the RPA component; The RPA component collects multimodal information from the target webpage based on the collection script, and obtains multimodal data corresponding to the target webpage. The large data processing model performs data fusion and verification on the multimodal data to collect the target webpage data corresponding to the target webpage.

2. The method according to claim 1, characterized in that, The RPA component collects multimodal information from the target webpage based on the collection script, obtaining multimodal data corresponding to the target webpage, including: Based on the collection script, determine the element location corresponding to each element type in the target webpage; Based on the preset correspondence between element types and acquisition rules, determine the target acquisition rules corresponding to the location of each element. Based on the corresponding target acquisition rules, target element data at each element location is acquired to obtain the multimodal data.

3. The method according to claim 2, characterized in that, The multimodal data includes text data screenshots and API data; the multimodal data is obtained by collecting target element data for each element location based on the corresponding target acquisition rules, including: If the target acquisition rule is a text copying rule, then the corresponding element is located and text copying is performed according to the text copying rule to obtain the text data; If the target acquisition rule is a screenshot rule, then the corresponding element location is screenshotted according to the screenshot rule to obtain the screenshot data; If the target acquisition rule is an API data acquisition rule, then determine the API identifier of the target API in the corresponding element location, obtain the API key and request parameters corresponding to the API identifier from the preset API database, and obtain the API data from the target API according to the API key and the request parameters.

4. The method according to claim 3, characterized in that, The large-scale data processing model performs data fusion and verification on the multimodal data to collect target webpage data corresponding to the target webpage, including: The text data is semantically understood using NLP technology to obtain key text information, and the key text information is mapped to a preset field sequence to obtain a first field sequence. The screenshot text information and image feature information are extracted from the screenshot data using OCR and CV technologies. Based on the screenshot text information and the image feature information, key screenshot information is determined, and the key screenshot information is mapped to the first field sequence. Based on the key screenshot information, the key text information in the first field sequence is verified to obtain the second field sequence. The API data is mapped to the second field sequence according to the data type of the API data, and the key text information or screenshot information in the second field sequence is verified based on the API data to obtain the third field sequence; The target webpage data is determined based on the sequence of the third field.

5. The method according to claim 1, characterized in that, Before the RPA component obtains the source code of the target webpage, the method further includes: The large-scale data processing model determines multiple candidate web pages based on the user's input search keywords; The data processing model prioritizes multiple candidate web pages based on their relevance to the search keywords and their authority. The data processing model determines the candidate web pages with the first preset number of web pages in the priority ranking as the target web pages, where the preset number of web pages is an integer greater than or equal to 1. The large data processing model sends the URLs corresponding to each of the target web pages to the RPA component.

6. The method according to claim 5, characterized in that, The RPA component obtains the source code of the target webpage, including: For each of the target web pages, open the target web page according to the corresponding URL; The corresponding webpage source code is obtained using a preset webpage parsing tool.

7. The method according to claim 5, characterized in that, When the number of preset web pages is greater than 1, there are multiple target web pages; the large data processing model performs data fusion and verification processing on the multimodal data to collect target web page data corresponding to the target web pages, including: Data fusion and verification processes are performed on the multimodal data corresponding to each target webpage to obtain candidate target webpage data corresponding to each target webpage. The target webpage data is obtained by performing correlation analysis on multiple candidate target webpage data.

8. A data intelligent acquisition system, characterized in that, The intelligent data acquisition system includes RPA components and a large data processing model, wherein: The RPA component is used to obtain the source code of the target webpage and send the source code to the data processing big model. The large data processing model is used to determine the collection script of the RPA component based on the webpage source code, and send the collection script to the RPA component; The RPA component is also used to collect multimodal information from the target webpage based on the collection script, and obtain multimodal data corresponding to the target webpage; The large data processing model is also used to perform data fusion and verification processing on the multimodal data, and to collect the target webpage data corresponding to the target webpage.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The computer device is equipped with the intelligent data acquisition system, which includes an RPA component and a large data processing model. When the processor executes the computer program, it implements the intelligent data acquisition method as described in any one of claims 1-7.

10. A storage medium, characterized in that, The storage medium stores a computer program, which includes program instructions that, when executed by a processor, cause the processor to perform the intelligent data acquisition method as described in any one of claims 1-7.

Images