An NLP-based RPA process generation method, device and medium

Abstract

The application provides an NLP-based RPA process generation method, device and medium. The method obtains a corpus data set and processes the corpus data set based on a preprocessing rule. Based on the preprocessed corpus data set, a preset random forest model is trained, and the trained random forest model is used as an NLP classification model. The NLP classification model includes a plurality of classifiers corresponding to the corpus data set. An RPA process description text from a user terminal is obtained, the RPA process description text is input into the NLP classification model to determine a corresponding process description sequence. The RPA process description data is matched with RPA component information. According to the process description sequence, an RPA business process corresponding to the RPA process description text is generated and sent to the user terminal. The method solves the problem that the current RPA business process generation relies on manual work and the process generation efficiency is low, which affects the user experience of the RPA system.

Description

Technical Field

[0001] This application relates to the field of process automation technology, and in particular to an RPA process generation method, device and medium based on Natural Language Processing (NLP). Background Technology

[0002] Robotic Process Automation (RPA) is a business process automation technology based on software robots and artificial intelligence. An RPA system is an application that automates manually operated processes by mimicking the manual actions of end-users on a computer.

[0003] RPA systems, to some extent, liberate labor, enabling repetitive and process-oriented tasks to be automated through established automated workflow systems. However, the prerequisite for an RPA system to automate workflows is that developers identify tasks that can be performed using RPA and, based on standardized conventions for specific tasks, design standardized business processes. Only then can RPA complete the corresponding tasks. In this process, developers need to understand the standardized conventions for specific tasks before they can formulate and set up RPA business processes, wasting human resources. Furthermore, RPA business process generation relies on manual intervention, resulting in low efficiency. Summary of the Invention

[0004] To address the aforementioned issues, embodiments of this application provide an NLP-based RPA process generation method, device, and medium.

[0005] On the one hand, embodiments of this application provide an NLP-based RPA process generation method, which includes:

[0006] Acquire a corpus dataset and process the corpus dataset based on preprocessing rules; the corpus dataset includes at least full-text data and RPA process component description data; the preprocessing rules include at least: data tagging, string parameter tagging, data cleaning, and data segmentation;

[0007] Based on the preprocessed corpus dataset, a pre-set random forest model is trained to serve as an NLP classification model; the NLP classification model includes multiple classifiers corresponding to the corpus dataset.

[0008] Obtain the RPA process description text from the user terminal;

[0009] The RPA process description text is input into the NLP classification model to determine the corresponding process description sequence; wherein, the process description sequence includes several RPA process description data after classifying the RPA process description text; the RPA process description data is matched with RPA component information;

[0010] Based on the process description sequence, an RPA business process corresponding to the RPA process description text is generated and sent to the user terminal.

[0011] In one implementation of this application, the ratio of the full text data to the RPA process component description data in the corpus dataset is 3:7; the RPA process component description data is obtained by crawling the RPA system and preset RPA process component description statements.

[0012] In one implementation of this application, the data tag is used to characterize the permissions of the data in the corpus dataset to be mapped to the RPA business process and the corresponding mapping RPA component type;

[0013] The string parameter marker is used to mark RPA component configuration parameters;

[0014] The data cleaning process includes at least the removal of garbled characters, symbols, and stop words.

[0015] The data segmentation is performed on the corpus dataset using a preset segmentation tool.

[0016] In one implementation of this application, a pre-set random forest model is trained based on the preprocessed corpus dataset, and the trained random forest model is used as an NLP classification model, specifically including:

[0017] The preprocessed corpus dataset is randomly sampled until the number of random samplings reaches a first predetermined value; the number of samples randomly sampled is a second predetermined value.

[0018] The preset random forest model is trained using multiple sets of randomly sampled samples corresponding to the first predetermined value to obtain a classifier of the first predetermined value number.

[0019] The NLP classification model is generated based on a plurality of classifiers and a preset difference classifier of the first predetermined number; the classification result of the difference classifier is different from the classification results of the plurality of classifiers.

[0020] In one implementation of this application, before generating the NLP classification model, the method further includes:

[0021] The model validation metrics of the trained random forest model are determined using a pre-set model test set in the corpus dataset; the model validation metrics include at least: accuracy, recall, false alarm probability, and precision.

[0022] Based on the comparison results between the model validation index value and the corresponding preset threshold, it is determined whether the trained random forest model has completed training.

[0023] In one implementation of this application, the RPA process description text is input into the NLP classification model to determine the corresponding process description sequence, specifically including:

[0024] After inputting the RPA process description text into the NLP classification model, the classification results of each classifier and the preset differential classifier for each word segment in the RPA process description text are determined.

[0025] According to the word order of the RPA process description text, the classification results of each segmentation are added to the process description sequence in sequence.

[0026] In one implementation of this application, generating an RPA business process corresponding to the RPA process description text based on the process description sequence specifically includes:

[0027] Based on the process description sequence, determine the RPA component information corresponding to each RPA process description data; the RPA component information includes at least: RPA component type, RPA component configuration parameters, and RPA component tag.

[0028] Based on the RPA component information corresponding to each of the process description sequences, generate a visual RPA component corresponding to each of the RPA process description data;

[0029] Based on the temporal relationship of the process description sequence, the interaction attributes of each of the visualized RPA components are determined to generate the RPA business process.

[0030] In one implementation of this application, after acquiring the corpus dataset and processing the corpus dataset based on preprocessing rules, the method further includes:

[0031] The corpus dataset is divided into several data vectors according to a preset partitioning rule, and corresponding data vector information is generated. The preset partitioning rule is based on sentences. The data vector information includes the number of words contained in the data vector, the frequency of occurrence, and the tag. The frequency of occurrence is the number of times the sentence appears in the corpus dataset.

[0032] The data vector information is then updated to the preprocessed corpus dataset.

[0033] On the other hand, embodiments of this application also provide an NLP-based RPA process generation device, the device comprising:

[0034] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0035] Acquire a corpus dataset and process the corpus dataset based on preprocessing rules; the corpus dataset includes at least full-text data and RPA process component description data; the preprocessing rules include at least: data tagging, string parameter tagging, data cleaning, and data segmentation;

[0036] Based on the preprocessed corpus dataset, a pre-set random forest model is trained to serve as an NLP classification model; the NLP classification model includes multiple classifiers corresponding to the corpus dataset.

[0037] Obtain the RPA process description text from the user terminal;

[0038] The RPA process description text is input into the NLP classification model to determine the corresponding process description sequence; wherein, the process description sequence includes several RPA process description data after classifying the RPA process description text; the RPA process description data is matched with RPA component information;

[0039] Based on the process description sequence, an RPA business process corresponding to the RPA process description text is generated and sent to the user terminal.

[0040] This application also provides an NLP-based RPA process generation traceability non-volatile computer storage medium storing computer-executable instructions, which are configured as follows:

[0041] Acquire a corpus dataset and process the corpus dataset based on preprocessing rules; the corpus dataset includes at least full-text data and RPA process component description data; the preprocessing rules include at least: data tagging, string parameter tagging, data cleaning, and data segmentation;

[0042] Based on the preprocessed corpus dataset, a pre-set random forest model is trained to serve as an NLP classification model; the NLP classification model includes multiple classifiers corresponding to the corpus dataset.

[0043] Obtain the RPA process description text from the user terminal;

[0044] The RPA process description text is input into the NLP classification model to determine the corresponding process description sequence; wherein, the process description sequence includes several RPA process description data after classifying the RPA process description text; the RPA process description data is matched with RPA component information;

[0045] Based on the process description sequence, an RPA business process corresponding to the RPA process description text is generated and sent to the user terminal.

[0046] The above technical solution can generate standardized RPA business components and processes based on the user's natural language, reducing reliance on manual labor and minimizing human resource investment. Automated process generation ensures high efficiency. This solves the current problems of manual process generation, low efficiency, and negative impact on user experience with RPA systems. Attached Figure Description

[0047] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0048] Figure 1 This is a flowchart illustrating an NLP-based RPA process generation method in an embodiment of this application.

[0049] Figure 2 This is a schematic diagram of another RPA process generation method based on NLP in the embodiments of this application;

[0050] Figure 3 This is a schematic diagram illustrating the generation of an NLP classification model for an NLP-based RPA process generation method in an embodiment of this application.

[0051] Figure 4 This is a schematic diagram of the structure of an NLP-based RPA process generation device in an embodiment of this application. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0053] Natural Language Processing (NLP) is an artificial intelligence technology designed to enable computers to understand and generate human language. In text classification applications, NLP technology categorizes texts based on their content, assigning them to different categories. It has wide applications in information filtering, sentiment analysis, and spam filtering.

[0054] The application of NLP technology in text classification mainly involves the following aspects: First, NLP technology can extract text features, which can be used to train machine learning models to help machines understand and classify text. Second, NLP technology can be applied to sentiment analysis, using methods such as vocabulary corpora, rule matching, and machine learning to classify text according to sentiment, thereby achieving sentiment analysis. Third, NLP technology can perform topic classification. Topic classification aims to divide text into different topic categories. NLP technology can perform topic classification of text through methods such as word frequency statistics, keyword extraction, and topic models.

[0055] If NLP technology can be used in conjunction with an RPA system to build business processes, the efficiency of process generation will be further improved, business processes will be generated more conveniently, and manpower input will be reduced.

[0056] Based on this, this application provides an NLP-based RPA process generation method, device, and medium to solve the problem that current methods for generating RPA business processes rely on manual labor, resulting in low process generation efficiency and affecting the user experience of the RPA system.

[0057] The various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0058] This application provides an NLP-based RPA process generation method, such as... Figure 1 As shown, the method may include steps S101-S105:

[0059] S101, the server obtains the corpus dataset and processes the corpus dataset based on preprocessing rules.

[0060] The corpus dataset should include at least full-text data and RPA process component description data. Preprocessing rules should include at least: data tagging, string parameter tagging, data cleaning, and data segmentation.

[0061] It should be noted that the server, as the execution subject of the NLP-based RPA process generation method, exists only as an example, and the execution subject is not limited to the server. This application does not make any specific limitation in this regard.

[0062] The corpus dataset was constructed using two datasets to ensure coverage of the Chinese language and RPA business processes. The datasets include a publicly available Chinese text classification dataset containing full-text data, and a Chinese business process component description dataset containing RPA process component descriptions. The Chinese text classification dataset is primarily used for data augmentation to ensure coverage of user input text. The Chinese business process component description dataset is derived from web scraping of RPA systems and the collection and expansion of RPA business process component description statements to emphasize RPA business processes. The data provides augmentation for classification labels; to ensure the accuracy of subsequent classifiers, the ratio of the publicly available Chinese text classification dataset to the Chinese business process component description dataset is adjusted to 3:7.

[0063] In other words, the ratio of full-text data to RPA process component description data in the corpus dataset is 3:7. The RPA process component description data was obtained by crawling RPA systems and pre-defined RPA process component description statements.

[0064] In this embodiment, data tags are used to characterize the permissions and corresponding RPA component types mapped from data in the corpus dataset to RPA business processes. String parameter tags are used to mark RPA component configuration parameters. Data cleaning includes at least the removal of garbled characters, symbols, and stop words. Data segmentation is performed on the corpus dataset using a preset segmentation tool.

[0065] String parameter tags are primarily used to extract English string parameters. Typically, the user-input process description text contains consecutive English strings representing component attributes to be entered into a specified RPA component. This consecutive English string is extracted, tagged with a data tag, and stored. The data tag is named `str`. a 'a' is the sequence number of the data tag, which is used for attribute input in the subsequent RPA component generation process.

[0066] Data cleaning primarily targets the Chinese business process description dataset, including removing potential garbled characters and symbols, HTML tags in the crawler text, and symbols from business process description statements. After data cleaning, the text data is separated into training and test sets at a 9:1 ratio for algorithm training and accuracy verification. Tokenization is used to delineate word boundaries within sentences, mainly involving splitting and labeling the training set from sentences to words and phrases. Tokenization can be implemented using Chinese lexical analysis toolkits such as the Language Technology Platform (LTP) or the Tsinghua University Lexical Analyzer for Chinese (THULAC).

[0067] Furthermore, in this embodiment of the application, after obtaining the corpus dataset and processing it based on preprocessing rules, the method further includes:

[0068] The server divides the corpus dataset into several data vectors according to a preset partitioning rule and generates corresponding data vector information. The preset partitioning rule divides the dataset by sentence. The data vector information includes the number of words contained in the data vector, their frequency of occurrence, and a tag. The frequency of occurrence refers to the number of times the sentence appears in the corpus dataset. The data vector information is then updated to the preprocessed corpus dataset.

[0069] In other words, the server can extract features from the corpus dataset and convert the text in the corpus dataset into a unified computer language without losing information. This conversion process converts the text information in the corresponding text dataset into sentence vector information, i.e., data vector information, based on word segmentation information. The sentence vector information contains information such as the vocabulary and frequency of occurrence of each sentence.

[0070] S102, the server trains a pre-set random forest model based on the pre-processed corpus dataset, so that the trained random forest model can be used as an NLP classification model.

[0071] NLP classification models include multiple classifiers corresponding to the corpus dataset.

[0072] In this embodiment of the application, a pre-set random forest model is trained based on the pre-processed corpus dataset, so that the trained random forest model can be used as an NLP classification model, specifically including:

[0073] The server randomly samples the preprocessed corpus dataset until the number of random samples reaches a first predetermined value. The number of randomly sampled samples is a second predetermined value. Next, the server trains a pre-defined random forest model using multiple sets of randomly sampled samples corresponding to the first predetermined value, to obtain a first predetermined number of classifiers. Then, the server generates an NLP classification model based on the first predetermined number of classifiers and a pre-defined discriminant classifier. The classification result of the discriminant classifier differs from the classification results of the multiple classifiers.

[0074] The server can train the model for classification using a preprocessed corpus dataset. The server can use a Random Forest (RF) model for classification training. During training, random sampling refers to sampling a second predetermined value (e.g., m) of samples from the corpus dataset, and then randomly sampling a first predetermined value (e.g., n times). The n randomly sampled samples are then used to generate n random sample sets D. niWhere i is the sample index, greater than 0 and less than m. Models are trained using a set of n randomly sampled samples to construct n classifiers C. ni The final text classification result is determined by voting from n classifiers. The NLP classification model produces N+1 classification results, where N corresponds to the number of RPA business process components in different categories, and 1 corresponds to other ignored categories labeled as R. e .

[0075] The voting method for obtaining the classification results is represented as follows:

[0076]

[0077] Where H(x) represents the final classification result obtained by the NLP classification model after classifying the RPA process description text x, and R j This indicates the use of n classifiers C ni More than half of the obtained classification results are identified as belonging to the same category. j represents the RPA business process component classification, T represents the total number of classifier classification results, k represents the classifier classification result sequence number, and other represents other categories besides the classifier classification results.

[0078] For example, if an RPA process description text states that data is written from a certain row and column of table A, and more than half of the classifiers in an NLP classification model classify the data as "cell write", then the final output classification result will be "cell write".

[0079] In this embodiment of the application, before generating the NLP classification model, the method further includes:

[0080] The server uses a pre-defined model test set in the corpus dataset to determine the model validation metrics for the trained random forest model. These metrics include at least: accuracy, recall, false alarm probability, and precision. Based on the comparison of these validation metrics with corresponding pre-defined thresholds, the server determines whether the trained random forest model has completed its training.

[0081] In other words, the server can validate the trained model using the aforementioned model test set. If the accuracy, recall, false alarm probability, and precision all match their respective preset thresholds, the random forest model is considered successfully trained. If the accuracy is greater than the first preset threshold, the recall is less than the second preset threshold, the false alarm probability is less than the third preset threshold, and the precision is greater than the fourth preset threshold, the model is considered validated. The first, second, third, and fourth preset thresholds can be set by the user; this application does not impose specific limitations on them.

[0082] After obtaining the NLP classification model, it can be deployed on the server as an independent API interface to provide call support for business process text classification.

[0083] S103, the server obtains the RPA process description text from the user terminal.

[0084] like Figure 2 As shown, this application includes an RPA process generator 210 and an NLP text classifier 220. Users can input RPA process description text (business process component description text) through a text input interaction window 211, and execute the text classification service 230 of the NLP text classifier. The NLP text classifier is built after steps such as acquiring the corpus dataset 221, data preprocessing 222 (data tagging, string parameter tagging, data cleaning, data segmentation), feature extraction 223, and NLP classification model construction 224. String parameter tagging can perform classification services on the RPA process description text immediately after tagging, without needing to complete the NLP classification model for string classification. Then, after the text classification service is completed, RPA process description data 240 is obtained. The server then performs RPA process drawing 250 and generates a process panel 260 of the RPA business process for the user.

[0085] The structural diagram of the NLP text classifier during its construction process is shown below. Figure 3 As shown, the process includes an ensemble learning phase 301 for acquiring the corpus dataset, random sampling 302 to obtain n sample sets 303, sample classification training to obtain n classifiers 304, and further obtaining a voting classifier 305.

[0086] S104, the server inputs the RPA process description text into the NLP classification model to determine the corresponding process description sequence.

[0087] The process description sequence includes several RPA process description data sets after categorizing the RPA process description text. The RPA process description data is matched with RPA component information.

[0088] In this embodiment of the application, the server inputs the RPA process description text into the NLP classification model to determine the corresponding process description sequence, specifically including:

[0089] After the server inputs the RPA process description text into the NLP classification model, it determines the classification results of each classifier and the preset differential classifier for each word segment in the RPA process description text. Following the word order of the RPA process description text, the classification results of each word segment are added sequentially to the process description sequence.

[0090] A differential classifier refers to a classifier whose label classification result is R. eThe classifier, because the RPA process description text contains several words, each word may correspond to a process component of a business process. The server can generate the classification results of each word and the process description sequence of the classification results by the NLP classification model according to the word order. For example, [cell writing, formula switching, cell merging], etc.

[0091] The RPA process generator provides an interactive window for inputting descriptive text for business process components. This window receives the user's input and passes it to the NLP classification API service interface. The classification model provides a set of time-series descriptive data. After removing any potentially ignored classifications, this descriptive data, combined with English string parameters extracted during data preprocessing, is transformed into component descriptive data that can directly generate RPA processes.

[0092] S105, the server generates an RPA business process corresponding to the RPA process description text based on the process description sequence, and sends it to the user terminal.

[0093] In this embodiment of the application, an RPA business process corresponding to the RPA process description text is generated based on the process description sequence, specifically including:

[0094] Based on the sequence of process descriptions, the server determines the RPA component information corresponding to each RPA process description data. The RPA component information includes at least: RPA component type, RPA component configuration parameters, and RPA component tags.

[0095] In other words, the server can convert each RPA process description data into standard component data of the RPA process generator according to the process description sequence. The standard component data of the RPA process generator can be pre-set with information such as RPA component type, RPA component configuration parameters, and RPA component tags, and then use the converted RPA process description data as RPA component information.

[0096] Among them, English string parameters are filled into the corresponding configuration parameter entries. When there is no corresponding string parameter in the business process component description text, the configuration parameter input is ignored. For components with multiple configuration parameters, the characteristics and context of the English string parameters are analyzed, the key of the component configuration parameter is mapped and matched with the context of the string parameter, and the configuration is automatically filled in.

[0097] Next, the server generates visual RPA components corresponding to each RPA process description data based on the RPA component information corresponding to the process description sequence. Then, based on the temporal relationship of the process description sequence, it determines the interaction attributes of each visual RPA component to generate the RPA business process.

[0098] In other words, the server takes RPA component information as input, transforms it into visual components, and draws them onto the workflow panel. The resulting workflow panel consists of component design, component encapsulation, and visualization. Component design involves filling in the attributes, descriptions, and parameter configurations of each component. Component encapsulation adds visual parameters, sets appearance, node interaction attributes, and connects the various independent components. The visualization displays the encapsulated workflow components on the workflow panel, allowing for appropriate visual configuration and adjustments.

[0099] Furthermore, this application can also optimize the corpus dataset, feature extraction steps, and NLP classification model.

[0100] Corpus optimization primarily enhances the coverage of text related to business process component descriptions, resulting in more stable component generation. Corpus optimization expands the Chinese business process component description text, increasing the proportion of specialized descriptive vocabulary to ensure the accuracy of data segmentation within the business processes. Furthermore, during data segmentation, multiple Chinese lexical analysis toolkits are compared to obtain a better segmentation scheme for classifying RPA business processes.

[0101] After the feature extraction process vectorizes the text, the resulting vectors have a large dimension, which affects the training speed and classification performance of the model. Before inputting the data into the classification model for training, data dimensionality reduction methods are used to reduce the vector dimension.

[0102] NLP model optimization occurs during model building and its main purpose is to improve the classification performance of the classification model. Model optimization involves three steps: feature selection, hyperparameter tuning, and sample balance adjustment.

[0103] Feature selection: This is performed using a feature importance assessment method. By considering mutual information, correlation, and redundancy among features, an importance score is calculated for each feature to achieve feature selection. Feature selection can identify and remove irrelevant features, reduce interfering features in the classification process, and thus improve the accuracy of the model.

[0104] Hyperparameter tuning: Hyperparameters tuned during model optimization include the number of decision trees, the maximum depth of each decision tree, and the minimum number of samples required for node splits. Adjusting these hyperparameters aims to improve the model's recognition capabilities and accuracy.

[0105] Sample balance adjustment: This mainly involves increasing the amount of minority class text in the component description classification through data augmentation methods, enabling the model to better adapt to the features of minority class text. To maintain sentence integrity, synonym replacement and replacement of words with close proximity in the vector space are used to achieve the purpose of sample balance adjustment.

[0106] The above technical solution provides a technical approach for generating RPA workflows that integrates NLP technology. This includes processes such as corpus collection, data preprocessing, feature extraction, NLP model construction, component description data generation, RPA workflow data transformation, RPA workflow visualization, corpus optimization, feature extraction optimization, and NLP model optimization. The method combines RPA business process component design rules with user-friendliness considerations, designing a business process generation method that directly converts natural language text into business process components, providing users with a better RPA workflow design experience.

[0107] The above technical solutions can generate standardized RPA business components and processes based on the user's natural language, without relying too much on manual labor, reducing the investment of human resources. Automated business process generation can ensure process generation efficiency.

[0108] Figure 4 A schematic diagram of the structure of an NLP-based RPA process generation device provided in this application embodiment is shown below. Figure 4 As shown, the device includes:

[0109] At least one processor; and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to:

[0110] Acquire the corpus dataset and process it based on preprocessing rules. The corpus dataset includes at least full-text data and RPA process component description data. Preprocessing rules include at least: data tagging, string parameter tagging, data cleaning, and data segmentation. Based on the preprocessed corpus dataset, train a pre-defined random forest model to serve as an NLP classification model. The NLP classification model includes multiple classifiers corresponding to the corpus dataset. Acquire the RPA process description text from the user terminal. Input the RPA process description text into the NLP classification model to determine the corresponding process description sequence. The process description sequence includes several RPA process description data after classifying the RPA process description text. Match the RPA process description data with RPA component information. Based on the process description sequence, generate the RPA business process corresponding to the RPA process description text and send it to the user terminal.

[0111] This application also provides an NLP-based RPA process generation traceability non-volatile computer storage medium storing computer-executable instructions, which are configured as follows:

[0112] Acquire the corpus dataset and process it based on preprocessing rules. The corpus dataset includes at least full-text data and RPA process component description data. Preprocessing rules include at least: data tagging, string parameter tagging, data cleaning, and data segmentation. Based on the preprocessed corpus dataset, train a pre-defined random forest model to serve as an NLP classification model. The NLP classification model includes multiple classifiers corresponding to the corpus dataset. Acquire the RPA process description text from the user terminal. Input the RPA process description text into the NLP classification model to determine the corresponding process description sequence. The process description sequence includes several RPA process description data after classifying the RPA process description text. Match the RPA process description data with RPA component information. Based on the process description sequence, generate the RPA business process corresponding to the RPA process description text and send it to the user terminal.

[0113] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0114] The devices, media, and methods provided in this application are one-to-one correspondences. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0115] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0116] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for generating RPA processes based on NLP, characterized in that, The method includes: Acquire a corpus dataset and process the corpus dataset based on preprocessing rules; the corpus dataset includes at least full-text data and RPA process component description data; the preprocessing rules include at least: data tagging, string parameter tagging, data cleaning, and data segmentation; Based on the preprocessed corpus dataset, a pre-set random forest model is trained to serve as an NLP classification model. The NLP classification model includes multiple classifiers corresponding to the corpus dataset. Specifically, this includes: randomly sampling the preprocessed corpus dataset until the number of random samples reaches a first predetermined value; the number of randomly sampled samples is a second predetermined value; training the pre-set random forest model using multiple sets of randomly sampled samples corresponding to the first predetermined value to obtain the first predetermined number of classifiers; generating the NLP classification model based on the first predetermined number of classifiers and a pre-set differential classifier; the classification result of the differential classifier differs from the classification results of the multiple classifiers. Obtain the RPA process description text from the user terminal; The RPA process description text is input into the NLP classification model to determine the corresponding process description sequence; wherein, the process description sequence includes several RPA process description data after classifying the RPA process description text; the RPA process description data is matched with RPA component information; Based on the process description sequence, an RPA business process corresponding to the RPA process description text is generated and sent to the user terminal.

2. The method for generating RPA processes based on NLP according to claim 1, characterized in that, The ratio of the full-text data to the RPA process component description data in the corpus dataset is 3:7; the RPA process component description data is obtained by crawling the RPA system and preset RPA process component description statements.

3. The method for generating RPA processes based on NLP according to claim 1, characterized in that, The data tags are used to characterize the permissions and corresponding RPA component types for mapping data in the corpus dataset to RPA business processes; The string parameter marker is used to mark RPA component configuration parameters; The data cleaning process includes at least the removal of garbled characters, symbols, and stop words. The data segmentation is performed on the corpus dataset using a preset segmentation tool.

4. The NLP-based RPA process generation method according to claim 1, characterized in that, Before generating the NLP classification model, the method further includes: The model validation metrics of the trained random forest model are determined using a pre-set model test set in the corpus dataset; the model validation metrics include at least: accuracy, recall, false alarm probability, and precision. Based on the comparison results between the model validation index value and the corresponding preset threshold, it is determined whether the trained random forest model has completed training.

5. The method for generating RPA processes based on NLP according to claim 1, characterized in that, The RPA process description text is input into the NLP classification model to determine the corresponding process description sequence, specifically including: After inputting the RPA process description text into the NLP classification model, the classification results of each classifier and the preset differential classifier for each word segment in the RPA process description text are determined. According to the word order of the RPA process description text, the classification results of each segmentation are added to the process description sequence in sequence.

6. The RPA process generation method based on NLP according to claim 1, characterized in that, Based on the process description sequence, an RPA business process corresponding to the RPA process description text is generated, specifically including: Based on the process description sequence, determine the RPA component information corresponding to each RPA process description data; the RPA component information includes at least: RPA component type, RPA component configuration parameters, and RPA component tag. Based on the RPA component information corresponding to each of the process description sequences, generate a visual RPA component corresponding to each of the RPA process description data; Based on the temporal relationship of the process description sequence, the interaction attributes of each of the visualized RPA components are determined to generate the RPA business process.

7. The method for generating RPA processes based on NLP according to claim 1, characterized in that, After acquiring the corpus dataset and processing it based on preprocessing rules, the method further includes: The corpus dataset is divided into several data vectors according to a preset partitioning rule, and corresponding data vector information is generated. The preset partitioning rule is based on sentences. The data vector information includes the number of words contained in the data vector, the frequency of occurrence, and the tag. The frequency of occurrence is the number of times the sentence appears in the corpus dataset. The data vector information is then updated to the preprocessed corpus dataset.

8. An RPA process generation device based on NLP, characterized in that, The device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform an NLP-based RPA process generation method as described in any one of claims 1-7.

9. A non-volatile computer storage medium for RPA process generation traceability based on NLP, storing computer-executable instructions, characterized in that, The computer-executable instructions are capable of executing an NLP-based RPA process generation method as described in any one of claims 1-7.

Images