Bidding information correction method, device and storage medium combining RPA and AI

Abstract

The present application relates to the field of computer technology, and in particular to a method, device and storage medium for correcting bidding information in combination with RPA and AI. The bidding information correction method comprises: obtaining at least one bidding information that does not meet the information condition in the bidding information set, and obtaining a bidding information subset, wherein the information condition includes a first natural language processing NLP model and / or a screening condition set; obtaining correction information input for at least one bidding information in the bidding information subset; correcting at least one bidding information based on the correction information, and obtaining a bidding correction set. The present application can improve the accuracy of bidding information correction and improve the user experience.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a bidding information correction method, device and storage medium combining RPA and AI. Background Art

[0002] Robotic Process Automation (RPA), referred to as Robotic Process Automation, uses specific "robot software" to simulate human operations on computers and automatically execute process tasks according to rules.

[0003] Artificial Intelligence (AI) is a technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence.

[0004] Intelligent Document Processing (IDP) is a new generation of automation technology based on artificial intelligence technologies such as Optical Character Recognition (OCR), Computer Vision (CV), Natural Language Processing (NLP), and Knowledge Graph (KG). It can identify, classify, extract, and verify various types of documents, helping enterprises to realize the intelligence and automation of document processing.

[0005] Bidding is an international practice, a product of the highly developed commodity economy, and an organized way of selecting the best for a deal by applying technology, economic methods, and the competitive mechanism of the market economy. When an enterprise conducts bidding, it needs to obtain the required bidding information from the corresponding bidding documents. In order to improve the accuracy of the obtained bidding information, it is necessary to correct the obtained bidding information. However, with the development of science and technology, new formats, new field vocabulary, new expressions, etc. will continue to appear in bidding documents. In this case, the accuracy of the terminal in correcting the bidding information is low, which in turn affects the user experience. Summary of the invention

[0006] The embodiment of the present application provides a bidding information correction method, device and storage medium combining RPA and AI to solve the problems existing in the related technologies. The technical solution is as follows:

[0007] In a first aspect, an embodiment of the present application provides a bidding information correction method combining RPA and AI, including:

[0008] By using a robotic process automation (RPA) system, at least one bidding information that does not meet the information condition in the bidding information set is obtained to obtain a bidding information subset, where the information condition includes a first natural language processing (NLP) model and / or a screening condition set;

[0009] Obtaining correction information input for at least one bidding information in the bidding information subset;

[0010] At least one bidding information is corrected based on the correction information to obtain a bidding correction set.

[0011] In one implementation, obtaining at least one bidding information that does not meet the information condition in the bidding information set to obtain a bidding information subset includes:

[0012] Obtaining the confidence level corresponding to any bidding information in the bidding information set through the first NLP model;

[0013] If the confidence level corresponding to any bidding information is lower than the confidence threshold, the bidding information will be added to the bidding information subset.

[0014] In one implementation, obtaining at least one bidding information that does not meet the information condition in the bidding information set to obtain a bidding information subset includes:

[0015] Get the filter condition set corresponding to the bidding information set;

[0016] At least one bidding information that does not satisfy any screening condition in the screening condition set is obtained from the bidding information set to obtain a bidding information subset.

[0017] In one implementation, obtaining at least one bidding information that does not meet the information condition in the bidding information set to obtain a bidding information subset includes:

[0018] Performing random inspection on the bidding information set, and obtaining the bidding information set after random inspection;

[0019] Obtaining the sampling inspection result information input for any bidding information in the bidding information set after sampling inspection;

[0020] If the sampling result information indicates that any bidding information does not meet the information conditions, then any bidding information will be added to the bidding information subset.

[0021] In one embodiment, the method further comprises:

[0022] The first natural language processing (NLP) model is iterated based on the bidding correction set to obtain a second NLP model.

[0023] In one implementation, the first natural language processing (NLP) model is iterated based on the bidding correction set to obtain a second NLP model, including:

[0024] Obtain the information quantity of at least one bidding information in the bidding correction set;

[0025] If the amount of information is greater than the information amount threshold, incremental training is performed on the first NLP model to obtain a second NLP model, and the bidding correction set is cleared.

[0026] In one embodiment, before obtaining at least one bidding information that does not meet the information condition in the bidding information set to obtain the bidding information subset, the method further includes:

[0027] Obtain a set of unlabeled training samples;

[0028] The third NLP model is trained using the unlabeled training sample set to obtain a fourth NLP model;

[0029] Obtain a set of labeled training samples;

[0030] The model parameters of the fourth NLP model are adjusted using the labeled training sample set to obtain the first NLP model.

[0031] In a second aspect, an embodiment of the present application provides a bidding information correction device combining RPA and AI, including:

[0032] A subset acquisition unit is used to acquire at least one bidding information that does not meet the information condition in the bidding information set through a robotic process automation (RPA) system to obtain a bidding information subset, where the information condition includes a first natural language processing (NLP) model and / or a screening condition set;

[0033] An information acquisition unit, used to acquire correction information input for at least one bidding information in the bidding information subset;

[0034] The information correction unit is used to correct at least one bidding information based on the correction information to obtain a bidding correction set.

[0035] In one embodiment, the subset acquisition unit includes a confidence acquisition subunit and an information adding subunit. The subset acquisition unit is used to acquire at least one bidding information that does not meet the information condition in the bidding information set. When the bidding information subset is obtained:

[0036] A confidence acquisition subunit, used to acquire the confidence corresponding to any bidding information in the bidding information set through the first NLP model;

[0037] The information adding subunit is used to add the bidding information to the bidding information subset if the confidence level corresponding to any bidding information is lower than the confidence level threshold.

[0038] In one embodiment, the subset acquisition unit further includes a set acquisition subunit and an information screening subunit. The subset acquisition unit is used to acquire at least one bidding information that does not meet the information condition in the bidding information set. When the bidding information subset is obtained:

[0039] A collection acquisition subunit is used to acquire a screening condition set corresponding to a bidding information set;

[0040] The information screening subunit is used to obtain at least one bidding information that does not meet any screening condition in the screening condition set from the bidding information set, and obtain a bidding information subset.

[0041] In one embodiment, the subset acquisition unit further includes a set sampling subunit, a result acquisition subunit and a condition judgment subunit. The subset acquisition unit is used to acquire at least one bidding information that does not meet the information condition in the bidding information set. When the bidding information subset is obtained:

[0042] A collection sampling subunit is used to perform sampling on the bidding information collection and obtain the bidding information collection after sampling;

[0043] A result acquisition subunit is used to acquire the sampling result information input for any bidding information in the bidding information set after sampling;

[0044] The condition judgment subunit is used to add any bidding information to the bidding information subset if the sampling result information indicates that any bidding information does not meet the information condition.

[0045] In one embodiment, the bidding information correction device further includes:

[0046] The model iteration unit is used to iterate the first natural language processing (NLP) model based on the bidding correction set to obtain a second NLP model.

[0047] In one embodiment, the model iteration unit includes a quantity acquisition subunit and a model training subunit, and the model iteration unit is used to iterate the first natural language processing NLP model based on the bidding correction set to obtain the second NLP model:

[0048] A quantity acquisition subunit, used to acquire the information quantity of at least one bidding information in the bidding correction set;

[0049] The model training subunit is used to perform incremental training on the first NLP model to obtain the second NLP model and clear the bidding correction set if the amount of information is greater than the information amount threshold.

[0050] In one embodiment, the bidding information correction device further includes a sample acquisition unit, a model acquisition unit, a marked sample acquisition unit and a model adjustment unit, which are used to acquire at least one bidding information that does not meet the information condition in the bidding information set, before obtaining the bidding information subset:

[0051] A sample acquisition unit, used to acquire a set of unlabeled training samples;

[0052] A model acquisition unit, configured to train the third NLP model using the unlabeled training sample set to obtain a fourth NLP model;

[0053] A labeled sample acquisition unit, used to acquire a set of labeled training samples;

[0054] The model adjustment unit is used to adjust the model parameters of the fourth NLP model by using the labeled training sample set to obtain the first NLP model.

[0055] In a third aspect, an embodiment of the present application provides a terminal combining RPA and AI, the terminal comprising: a memory and a processor. The memory and the processor communicate with each other through an internal connection path, the memory is used to store instructions, the processor is used to execute the instructions stored in the memory, and when the processor executes the instructions stored in the memory, the processor executes the method in any one of the above-mentioned embodiments.

[0056] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a computer, the method in any one of the above-mentioned embodiments is executed.

[0057] The advantages or beneficial effects of the above technical solution include at least:

[0058] In one or related embodiments, at least one bidding information that does not meet the information condition in the bidding information set is obtained to obtain a bidding information subset, wherein the information condition includes a first natural language processing NLP model and / or a screening condition set; correction information input for at least one bidding information in the bidding information subset is obtained; and at least one bidding information is corrected based on the correction information to obtain a bidding correction set. Therefore, by obtaining bidding information that does not meet the information condition and correcting the bidding information, the acquisition of bidding information that does not meet the information condition can be reduced, the inaccurate information obtained based only on the NLP model and / or the screening condition set can be reduced, the information obtained by the NLP model and / or the screening condition set can be adjusted, the accuracy of the correction of the bidding information can be improved, and the user experience can be improved.

[0059] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present application will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the multiple drawings represent the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings only depict some embodiments according to the present application and should not be regarded as limiting the scope of the present application.

[0061] Figure 1 A background schematic diagram showing a bidding information correction method combining RPA and AI according to an embodiment of the present application;

[0062] Figure 2 A background architecture diagram showing a bidding information correction method combining RPA and AI according to an embodiment of the present application;

[0063] Figure 3 A flowchart showing a method for correcting bidding information in combination with RPA and AI according to an embodiment of the present application is shown;

[0064] Figure 4 A flowchart showing a method for correcting bidding information in combination with RPA and AI according to an embodiment of the present application is shown;

[0065] Figure 5 A flowchart of obtaining a first NLP model according to an embodiment of the present application is shown;

[0066] Figure 6 A flowchart of obtaining a bidding information subset according to an embodiment of the present application is shown;

[0067] Figure 7 A schematic diagram showing an information correction interface according to an embodiment of the present application is shown;

[0068] Figure 8 A schematic diagram showing the structure of a bidding information correction device combining RPA and AI according to an embodiment of the present application is shown;

[0069] Fig. 9 A schematic diagram showing the structure of a bidding information correction device combining RPA and AI according to an embodiment of the present application is shown;

[0070] Fig.10 A schematic diagram showing the structure of a bidding information correction device combining RPA and AI according to an embodiment of the present application is shown;

[0071] Fig.11 A schematic diagram showing the structure of a bidding information correction device combining RPA and AI according to an embodiment of the present application is shown;

[0072] Fig.12 A schematic diagram showing the structure of a bidding information correction device combining RPA and AI according to an embodiment of the present application is shown;

[0073] Fig.13 A schematic diagram showing the structure of a bidding information correction device combining RPA and AI according to an embodiment of the present application is shown;

[0074] Fig.14 A schematic diagram showing the structure of a bidding information correction device combining RPA and AI according to an embodiment of the present application is shown;

[0075] Fig.15 A structural block diagram of a terminal according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0076] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be understood as limiting the present application.

[0077] In the description of the present application, the term "plurality" refers to two or more.

[0078] In the description of this application, the term "RPA" refers to the use of specific "robot software" to simulate human operations on computers and automatically execute process tasks according to rules.

[0079] In the description of this application, the term "IDP" refers to Intelligent Document Processing, specifically a new generation of automation technology that uses artificial intelligence technologies such as Optical Character Recognition (OCR), Computer Vision (CV), Natural Language Processing (NLP), and Knowledge Graph (KG) to identify, classify, extract, and verify various types of documents, helping companies to achieve intelligent and automated document processing. The description of this application introduces a typical application scenario of IDP, which is to use intelligent document processing to automatically capture, analyze, and match a large number of bidding documents.

[0080] The intelligent document understanding platform is based on cutting-edge OCR and NLP algorithms, provides a wealth of pre-trained models and self-learning functions, has the ability to process multiple types of complex documents, and realizes end-to-end document processing automation.

[0081] In the description of this application, the term "NLP" refers to Natural Language Processing, which is an important direction in the fields of computer science and artificial intelligence. Specifically, it is a discipline that uses computer technology to analyze, understand and process natural language, using computers as a powerful tool for language research, quantitatively studying language information with the support of computers, and providing language descriptions that can be used by humans and computers. It includes two parts: Natural Language Understanding (NLU) and Natural Language Generation (NLG). NLP is a typical marginal interdisciplinary subject, involving linguistic science, computer science, mathematics, cognition, logic, etc. Its research can realize various theories and methods for effective communication between humans and computers using natural language. The process of processing natural language with computers is called Natural Language Understanding (NLU), Human Language Technology (HLT), Computational Linguistics (HL), Quantitative Linguistics, and Mathematical Linguistics at different times or with different focuses.

[0082] In the description of this application, the term "bidding document" includes bidding documents and tender documents. Among them, the bidding document refers to the invitation document issued by the tenderer to potential bidders and informs them of project requirements, bidding activity rules and contract conditions, etc., which is the main basis for project bidding activities. The tender document refers to the responsive document prepared by the bidder in accordance with the requirements of the bidding document.

[0083] In the description of this application, the term "bidding information" refers to the information obtained by the terminal when the NLP model is used to understand the bidding document. The bidding information includes but is not limited to project name information, company information, funding source information, project scale information, contract estimate price information, project geographical location information, construction period information, bidding scope information, qualification requirements information, deadline information, etc.

[0084] In the description of this application, the term "incremental training" refers to the process in which a model can continuously learn new knowledge from new samples and can preserve most of the previously learned knowledge. Incremental learning involves two points: saving the parameters that need to be persisted at the end of the previous training, and loading the last saved persistent parameters at the beginning of the next training.

[0085] In the description of this application, the term "model parameters" refers to configuration variables within a model. Model parameters can be estimated from data or learned from data. Model parameters are required to make model predictions. Model parameter values ​​can define the model's functionality. Model parameters are often saved as part of a learned model. Model parameters can be estimated using an optimization algorithm, which is an efficient search for possible values ​​for a parameter.

[0086] With the development of science and technology, the terminal technology has become increasingly mature, which has improved the convenience of users' production and life. In the terminal application scenario, users can obtain the bidding information needed in the bidding documents through the information acquisition application.

[0087] According to some embodiments, Figure 1 A background diagram of a bidding information correction method combining RPA and AI according to an embodiment of the present application is shown. Figure 1 As shown, the user can click on the bidding information correction application of the terminal. When the terminal detects that the user clicks on the bidding information correction application, the terminal can display the bidding information correction interface. Based on the bidding information correction interface, the user can input the bidding information set that needs to be corrected and click the "correct" button. When the terminal detects that the user clicks the "correct" button, the terminal can correct the bidding information in the bidding information set.

[0088] According to some embodiments, Figure 2 A background architecture diagram of a bidding information correction method combining RPA and AI according to an embodiment of the present application is shown. Figure 2 As shown, the terminal 11 can upload the correction instruction issued by the user to the server 13 through the network 12. When the server 13 receives the correction instruction, the server 13 can correct the bidding information in the bidding information set through the NLP model. The server 13 can send the corrected bidding information to the terminal 11 through the network 12.

[0089] In some embodiments, since the bidding documents corresponding to the bidding information set that needs to be corrected constantly have new formats, new domain vocabulary, new expressions, etc., the deep learning model is not accurate when correcting the bidding information in the bidding documents, which in turn affects the user experience.

[0090] It is easy to understand that the terminal includes but is not limited to: wearable devices, handheld devices, personal computers, tablet computers, vehicle-mounted devices, smart phones, computing devices or other processing devices connected to wireless modems, etc. In different networks, terminal devices can be called different names, such as: user equipment, access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent or user device, cellular phone, cordless phone, personal digital assistant (PDA), terminal equipment in the fifth generation mobile communication technology (5th generation mobile networks, 5G) network or future evolution network, etc. The terminal can be installed with an operating system, which refers to an operating system that can run in the terminal, a program that manages and controls terminal hardware and terminal applications, and is an indispensable system application in the terminal. The operating system includes but is not limited to Android system, IOS system, Windows phone (WP) system and Ubuntu mobile version operating system, etc.

[0091] According to some embodiments, the terminal 11 may be connected to the server 13 via a network 12. The network 12 is used to provide a communication link between the terminal 11 and the server 13. The network 12 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc. It should be understood that Figure 2 The number of terminals 11, networks 12 and servers 13 is only illustrative. According to actual needs, there can be any number of terminals, networks and servers. For example, the server 13 can be a server cluster composed of multiple servers. The user can use the terminal 11 to interact with the server 13 through the network 12 to perform bidding information correction, etc.

[0092] These and other aspects of the embodiments of the present application will be clear with reference to the following description and drawings. In these descriptions and drawings, some specific implementations in the embodiments of the present application are specifically applied to represent some ways of implementing the principles of the embodiments of the present application, but it should be understood that the scope of the embodiments of the present application is not limited thereto. On the contrary, the embodiments of the present application include all changes, modifications and equivalents that fall within the spirit and connotation of the appended claims.

[0093] The following describes a method for correcting bidding information combining RPA and AI according to an embodiment of the present application in conjunction with the accompanying drawings.

[0094] Figure 3 A flowchart of a bidding information correction method combining RPA and AI according to an embodiment of the present application is shown as follows: Figure 3 As shown, the method may include the following steps:

[0095] Step S101: obtaining at least one bidding information that does not meet the information condition in the bidding information set through the Robotic Process Automation (RPA) system to obtain a bidding information subset;

[0096] According to some embodiments, the bidding information set refers to a set formed by at least one bidding information. The bidding information set does not specifically refer to a fixed set. For example, when the content of the bidding information changes, the bidding information set may also change accordingly. When the number of bidding information changes, the bidding information set may also change accordingly.

[0097] According to some embodiments, the information condition refers to the condition used by the terminal to determine whether the acquired bidding information is accurate. The information condition does not specifically refer to a fixed condition. The information condition includes but is not limited to a first natural language processing NLP model and / or a set of screening conditions. When the terminal obtains a condition modification instruction for the information condition, the information condition will change accordingly.

[0098] In some embodiments, the first NLP model refers to the NLP model used to determine whether the acquired bidding information is accurate. The first NLP model does not specifically refer to a fixed model. For example, when the terminal obtains a model modification instruction for the first NLP model, the first NLP model will change accordingly. When the information conditions change, the first NLP model will change accordingly.

[0099] In some embodiments, the screening condition refers to the condition preset by the user to determine whether the acquired bidding information is accurate. The screening condition does not specifically refer to a fixed condition. The screening condition includes but is not limited to the existence of bidding amount information, the existence of bidding deadline, etc.

[0100] In some embodiments, the filter condition set refers to a set formed by at least one filter condition. The filter condition set does not specifically refer to a fixed set. For example, when the number of filter conditions changes, the filter condition set may also change accordingly. When the content of the filter condition changes, the filter condition set may also change accordingly.

[0101] According to some embodiments, the bidding information subset refers to a set formed by at least one bidding information in the bidding information set that does not meet the information condition. The bidding information subset does not specifically refer to a fixed subset. For example, when the bidding information set changes, the bidding information subset may also change accordingly. When the information condition changes, the bidding information subset may also change accordingly.

[0102] It is easy to understand that when a user needs to obtain bidding information in a bidding document, the user can obtain a bidding information set corresponding to the bidding document through a terminal. When the terminal obtains the bidding information set, the terminal can obtain at least one bidding information in the bidding information set that does not meet the information condition to obtain a bidding information subset.

[0103] Step S102: obtaining correction information input for at least one bidding information in the bidding information subset;

[0104] According to some embodiments, the correction information refers to information input by a user when correcting at least one bidding information in the bidding information subset when the user obtains the bidding information subset. The correction information does not specifically refer to a fixed information. For example, when the bidding information subset changes, the correction information may also change. When the user who inputs the correction information changes, the correction information may also change.

[0105] For example, when the terminal obtains the bidding information corresponding to the "bidding deadline" as "August", the correction information entered when correcting the bidding information is the specific date "August 24" corresponding to the bidding deadline in the bidding document. Alternatively, when the terminal obtains the bidding information corresponding to the "bidding amount information" as "non-existent", the correction information entered when correcting the bidding information is the specific amount "￥20,000" corresponding to the bidding amount information in the bidding document.

[0106] It is easy to understand that when the terminal obtains the bidding information subset, the terminal can obtain the correction information input for at least one bidding information in the bidding information subset.

[0107] Step S103: Correct at least one bidding information based on the correction information to obtain a bidding correction set.

[0108] According to some embodiments, the bidding correction set refers to a set of at least one corrected bidding information obtained when the terminal corrects at least one bidding information in the bidding information subset based on the correction information. The bidding correction set does not specifically refer to a fixed set. For example, when the bidding information subset changes, the bidding correction set may also change. When the correction information changes, the bidding correction set may also change.

[0109] It is easy to understand that when the terminal obtains the correction information input for at least one bidding information in the bidding information subset, the terminal can correct the at least one bidding information based on the correction information to obtain the bidding correction set.

[0110] In one or related embodiments, at least one bidding information that does not meet the information condition in the bidding information set is obtained to obtain a bidding information subset, wherein the information condition includes a first natural language processing NLP model and / or a screening condition set; correction information input for at least one bidding information in the bidding information subset is obtained; and at least one bidding information is corrected based on the correction information to obtain a bidding correction set. Therefore, by obtaining bidding information that does not meet the information condition and correcting the bidding information, the acquisition of bidding information that does not meet the information condition can be reduced, the inaccurate information obtained based only on the NLP model and / or the screening condition set can be reduced, the information obtained by the NLP model and / or the screening condition set can be adjusted, the accuracy of the correction of the bidding information can be improved, and the user experience can be improved.

[0111] Figure 4 A flowchart of a bidding information correction method combining RPA and AI according to an embodiment of the present application is shown. Figure 4 As shown, the method may include the following steps:

[0112] Step S201: Obtain a set of unlabeled training samples;

[0113] According to some embodiments, the unlabeled training sample set refers to a set formed by at least one unlabeled training sample. The unlabeled training sample set does not specifically refer to a fixed set. For example, when the number of unlabeled training samples changes, the unlabeled training sample set may change. When the content of the unlabeled training samples changes, the unlabeled training sample set may change.

[0114] It is easy to understand that when a user needs to obtain bidding information in a bidding document, the user can correct the bidding information in the bidding document through the terminal. When the terminal corrects the bidding information in the bidding document, the terminal can obtain a set of unlabeled training samples.

[0115] Step S202: training the third NLP model using the unlabeled sample set to obtain a fourth NLP model;

[0116] According to some embodiments, the third NLP model refers to an untrained NLP model. The third NLP model does not specifically refer to a fixed model. For example, when the model parameters of the third NLP model change, the third NLP model may also change. When the model structure of the third NLP model changes, the third NLP model may also change.

[0117] In some embodiments, the fourth NLP model refers to an NLP model obtained by training the third NLP model using an unlabeled sample set. The fourth NLP model does not specifically refer to a fixed model. For example, when the third NLP model changes, the fourth NLP model may also change. When the unlabeled sample set changes, the fourth NLP model may also change.

[0118] It is easy to understand that when the terminal obtains the unlabeled training sample set, the terminal can use the unlabeled training sample set to train the third NLP model to obtain the fourth NLP model.

[0119] Step S203: Obtain a set of labeled training samples;

[0120] According to some embodiments, the labeled training sample set refers to a set formed by at least one labeled training sample. The labeled training sample set does not specifically refer to a fixed set. For example, when the number of labeled training samples changes, the labeled training sample set may change. When the content of the labeled training samples changes, the labeled training sample set may change.

[0121] It is easy to understand that when the terminal obtains the fourth NLP model, the terminal can obtain a set of labeled training samples.

[0122] Step S204: using the labeled training sample set, adjusting the model parameters of the fourth NLP model to obtain the first NLP model;

[0123] The specific process is as described above and will not be repeated here.

[0124] According to some embodiments, Figure 5 FIG. 1 is a flowchart of obtaining a first NLP model according to an embodiment of the present application. Figure 5 As shown, by training the third NLP model with the unlabeled sample set, the third NLP model can learn the commonalities in the unlabeled sample set to obtain the fourth NLP model. Afterwards, the model parameters of the fourth NLP model are adjusted using the labeled training sample set. In this case, the model only needs to learn the labeled training sample set based on the commonalities in the unlabeled sample set. Therefore, the efficiency of model training can be improved, and the accuracy of the first NLP model in correcting bidding information can be improved.

[0125] It is easy to understand that when the terminal obtains the annotated training sample set, the terminal can use the annotated training sample set to adjust the model parameters of the fourth NLP model to obtain the first NLP model.

[0126] Step S205: obtaining at least one bidding information that does not meet the information condition in the bidding information set through the Robotic Process Automation (RPA) system to obtain a bidding information subset;

[0127] The specific process is as described above and will not be repeated here.

[0128] According to some embodiments, when the terminal obtains a subset of bidding information, the terminal obtains at least one bidding information in the bidding information set that does not meet the information condition, for example, through a first NLP model, through a screening condition set, or through random sampling.

[0129] In some embodiments, when the terminal obtains the bidding information subset by means of the first NLP model, the terminal can obtain the confidence corresponding to any bidding information in the bidding information set by means of the first NLP model; if the confidence corresponding to any bidding information is lower than the confidence threshold, the bidding information is added to the bidding information subset. Therefore, the terminal can determine whether the bidding information meets the information condition by determining the size of the confidence and the confidence threshold, thereby improving the accuracy of obtaining the bidding information subset.

[0130] In some embodiments, the confidence level refers to the degree of trustworthiness corresponding to the extracted information. The confidence level does not specifically refer to a fixed confidence level. For example, when the bidding information changes, the confidence level may also change. When the first NLP model changes, the confidence level may also change.

[0131] In some embodiments, the terminal may obtain the confidence level corresponding to any bidding information according to the following formula:

[0132]

[0133] Among them, S i is the confidence level corresponding to the bidding information i, e i is the actual score corresponding to the bidding information i, ∑ j e j is the standard score corresponding to the bidding information i.

[0134] For example, when the terminal obtains a bidding information subset through the first NLP model, the terminal can obtain the confidence level of bidding information A in the bidding information set as A1. If A1 is less than the confidence threshold, the bidding information A is added to the bidding information subset; the terminal can also obtain the confidence level of bidding information B in the bidding information set as B1. If B1 is not less than the confidence threshold, the bidding information B is not added to the bidding information subset.

[0135] In some embodiments, when the terminal obtains a bidding information subset by means of a screening condition set, the terminal may obtain a screening condition set corresponding to the bidding information set; obtain at least one bidding information that does not satisfy any screening condition in the screening condition set from the bidding information set to obtain the bidding information subset. Therefore, the accuracy of obtaining the bidding information subset may be improved.

[0136] For example, when the screening condition set includes the existence of bidding amount information and the existence of bidding deadline, if the terminal obtains bidding information corresponding to the "bidding amount information" in the bidding information set, the terminal can add the "bidding amount information" to the bidding information subset. Alternatively, when the terminal obtains that the bidding information corresponding to the "bidding deadline" in the bidding information set is incomplete, for example, only date information, only minute information, only hour information, etc., the terminal can add the bidding information to the bidding information subset. In some embodiments, the terminal can simultaneously obtain the bidding information subset by means of the first NLP model and the screening condition set, such as Figure 6 Only when the confidence corresponding to any bidding information in the bidding information set is not less than the confidence threshold and all the screening conditions in the screening condition set are met at the same time, the terminal will not add any bidding information to the bidding information subset. Therefore, the accuracy of obtaining the bidding information subset can be improved.

[0137] For example, if the terminal obtains the confidence level of bidding information C in the bidding information set as C1, C1 is not less than the confidence threshold, but bidding information C does not meet the screening condition C2 in the screening condition set, then the terminal adds bidding information C to the bidding information subset. If the terminal obtains bidding information D in the bidding information set that meets all the screening conditions in the screening condition set, but the confidence level of bidding information D is D1, which is less than the confidence threshold, then the terminal adds bidding information D to the bidding information subset.

[0138] In some embodiments, when the terminal obtains a bidding information subset by sampling, the terminal can sample the bidding information set to obtain the sampled bidding information set; obtain sampling result information input for any bidding information in the sampled bidding information set; if the sampling result information indicates that any bidding information does not meet the information condition, then add any bidding information to the bidding information subset. Therefore, the accuracy of obtaining the bidding information subset can be improved.

[0139] In some embodiments, when the terminal performs random inspection on the bidding information set, the random inspection method includes but is not limited to periodic inspection, irregular periodic inspection, fixed quantity inspection, etc.

[0140] For example, the terminal may perform random sampling on the bidding information set every three days to obtain the bidding information set after random sampling; the terminal may also perform random sampling on the bidding information set every random number of days to obtain the bidding information set after random sampling; the terminal may also perform random sampling on the bidding information set whenever the number of bidding information in the bidding information set reaches a preset quantity threshold to obtain the bidding information set after random sampling.

[0141] It is easy to understand that when a user needs to obtain bidding information in a bidding document, the user can obtain a bidding information set corresponding to the bidding document through a terminal. When the terminal obtains the bidding information set, the terminal can obtain at least one bidding information in the bidding information set that does not meet the information condition to obtain a bidding information subset.

[0142] Step S206: obtaining correction information input for at least one bidding information in the bidding information subset;

[0143] The specific process is as described above and will not be repeated here.

[0144] According to some embodiments, when the terminal obtains the bidding information subset, the terminal can display at least one bidding information in the bidding information subset in the information correction interface. When the user opens the correction interface, the user can enter the correction information corresponding to any bidding information in the correction window corresponding to the bidding information, such as Figure 7 shown.

[0145] It is easy to understand that when the terminal obtains the bidding information subset, the terminal can obtain the correction information input for at least one bidding information in the bidding information subset.

[0146] Step S207: Correct at least one bidding information based on the correction information to obtain a bidding correction set;

[0147] The specific process is as described above and will not be repeated here.

[0148] For example, when the terminal obtains the correction information E2 corresponding to the bidding information E1, the terminal can correct the bidding information E1 based on the correction information E2, and put the corrected bidding information E2 into the bidding correction set. When the terminal obtains the correction information F2 corresponding to the bidding information F1, the terminal can correct the bidding information F1 based on the correction information F2, and put the corrected bidding information F2 into the bidding correction set.

[0149] It is easy to understand that when the terminal obtains the correction information input for at least one bidding information in the bidding information subset, the terminal can correct the at least one bidding information based on the correction information to obtain the bidding correction set.

[0150] Step S208: iterating the first natural language processing (NLP) model based on the bidding correction set to obtain a second NLP model;

[0151] According to some embodiments, the second NLP model refers to an NLP model obtained by iterating the first NLP model. The second NLP model does not specifically refer to a fixed model. For example, when the first NLP model changes, the second NLP model may also change. When the iteration method of the first NLP model changes, the second NLP model may also change.

[0152] According to some embodiments, when the terminal obtains the bidding correction set, the terminal can obtain the information quantity of at least one bidding information in the bidding correction set. If the information quantity obtained by the terminal is greater than the information quantity threshold, the first NLP model is incrementally trained to obtain the second NLP model, and the bidding correction set is cleared. Therefore, the terminal can use the second NLP model to perform bidding information correction, which can improve the accuracy of bidding information correction.

[0153] For example, when the information quantity threshold set by the terminal is 50, if the amount of information in the bidding correction set obtained by the terminal is 50, the terminal can perform incremental training on the first NLP model to obtain the second NLP model and clear the bidding correction set.

[0154] In some embodiments, when the terminal performs incremental training on the first NLP model, the terminal may use the bidding correction set as the training sample set; based on the training sample set, the model parameters of the first NLP model are adjusted to obtain the second NLP model. Therefore, the first NLP model may be incrementally trained by adjusting the parameters, which may improve the accuracy of obtaining the second NLP model.

[0155] It is easy to understand that when the terminal obtains the bidding correction set, the terminal can iterate the first natural language processing NLP model based on the bidding correction set to obtain the second NLP model.

[0156] In an embodiment of the present application, by obtaining an unlabeled training sample set; using the unlabeled training sample set to train the third NLP model to obtain a fourth NLP model; obtaining a labeled training sample set; using the labeled training sample set to adjust the model parameters of the fourth NLP model to obtain the first NLP model; thus, the efficiency of model training can be improved, and the accuracy of the first NLP model in correcting bidding information can be improved. At least one bidding information that does not meet the information condition in the bidding information set is obtained to obtain a bidding information subset; correction information input for at least one bidding information in the bidding information subset is obtained; at least one bidding information is corrected based on the correction information to obtain a bidding correction set; thus, the accuracy of the bidding information correction can be improved. The first natural language processing NLP model is iterated based on the bidding correction set to obtain the second NLP model; therefore, the terminal can correct the bidding information through the second NLP model, which can reduce the acquisition of bidding information that does not meet the information conditions, reduce the inaccurate information acquired based only on the NLP model and / or the screening condition set, and adjust the information acquired by the NLP model and / or the screening condition set, which can improve the accuracy of the information acquired when correcting the bidding information for information resources with new formats, new domain vocabulary, and new expressions, thereby improving the user experience.

[0157] The following are device embodiments of the present application, which can be used to execute the method embodiments of the present application. For details not disclosed in the device embodiments of the present application, please refer to the method embodiments of the present application.

[0158] See also Figure 8 , which is a structural diagram of a bidding information correction device combining RPA and AI in one embodiment of the present application. The bidding information correction device combining RPA and AI can be implemented as all or part of the device through software, hardware, or a combination of both. The bidding information correction device 800 combining RPA and AI includes a subset acquisition unit 801, a bidding information correction unit 802, and an information correction unit 803, wherein:

[0159] The subset acquisition unit 801 is used to acquire at least one bidding information that does not meet the information condition in the bidding information set through the robotic process automation RPA system to obtain a bidding information subset, where the information condition includes a first natural language processing NLP model and / or a screening condition set;

[0160] The bidding information correction unit 802 is used to obtain correction information input for at least one bidding information in the bidding information subset;

[0161] The information correction unit 803 is used to correct at least one bidding information based on the correction information to obtain a bidding correction set.

[0162] According to some embodiments, Fig. 9 The following is a schematic diagram showing the structure of a bidding information correction device combining RPA and AI according to an embodiment of the present application. Fig. 9 As shown, the subset acquisition unit 801 includes a confidence acquisition subunit 811 and an information adding subunit 821. The subset acquisition unit 801 is used to acquire at least one bidding information that does not meet the information condition in the bidding information set. When the bidding information subset is obtained:

[0163] The confidence acquisition subunit 811 is used to acquire the confidence corresponding to any bidding information in the bidding information set through the first NLP model;

[0164] The information adding subunit 821 is used to add the bidding information to the bidding information subset if the confidence level corresponding to any bidding information is lower than the confidence level threshold.

[0165] According to some embodiments, Fig.10 The following is a schematic diagram showing the structure of a bidding information correction device combining RPA and AI according to an embodiment of the present application. Fig.10 As shown, the subset acquisition unit 801 further includes a set acquisition subunit 831 and an information screening subunit 841. The subset acquisition unit 801 is used to acquire at least one bidding information that does not meet the information condition in the bidding information set. When the bidding information subset is obtained:

[0166] The set acquisition subunit 831 is used to acquire the screening condition set corresponding to the bidding information set;

[0167] The information screening subunit 841 is used to obtain at least one bidding information that does not meet any screening condition in the screening condition set from the bidding information set, and obtain a bidding information subset.

[0168] According to some embodiments, Fig.11 The following is a schematic diagram showing the structure of a bidding information correction device combining RPA and AI according to an embodiment of the present application. Fig.11 As shown, the subset acquisition unit 801 also includes a set sampling subunit 851, a result acquisition subunit 861 and a condition judgment subunit 871. The subset acquisition unit 801 is used to obtain at least one bidding information that does not meet the information condition in the bidding information set. When the bidding information subset is obtained:

[0169] The collection sampling subunit 851 is used to perform sampling on the bidding information collection and obtain the bidding information collection after sampling;

[0170] The result acquisition subunit 861 is used to acquire the sampling result information input for any bidding information in the bidding information set after sampling;

[0171] The condition judgment subunit 871 is used to add any bidding information to the bidding information subset if the sampling result information indicates that any bidding information does not meet the information condition.

[0172] According to some embodiments, Fig.12 The following is a schematic diagram showing the structure of a bidding information correction device combining RPA and AI according to an embodiment of the present application. Fig.12 As shown, the bidding information correction device 800 also includes:

[0173] The model iteration unit 804 is used to iterate the first natural language processing (NLP) model based on the bidding correction set to obtain a second NLP model.

[0174] According to some embodiments, Fig.13 The following is a schematic diagram showing the structure of a bidding information correction device combining RPA and AI according to an embodiment of the present application. Fig.13 As shown, the model iteration unit 804 includes a quantity acquisition subunit 814 and a model training subunit 824. The model iteration unit 804 is used to iterate the first natural language processing NLP model based on the bidding correction set to obtain the second NLP model:

[0175] The quantity acquisition subunit 814 is used to acquire the information quantity of at least one bidding information in the bidding correction set;

[0176] The model training subunit 824 is used to perform incremental training on the first NLP model to obtain a second NLP model and clear the bidding correction set if the amount of information is greater than the information amount threshold.

[0177] According to some embodiments, Fig.14 The following is a schematic diagram showing the structure of a bidding information correction device combining RPA and AI according to an embodiment of the present application. Fig.14 As shown, the bidding information correction device 800 also includes a sample acquisition unit 805, a model acquisition unit 806, a marked sample acquisition unit 807 and a model adjustment unit 808, which are used to obtain at least one bidding information that does not meet the information condition in the bidding information set, before obtaining the bidding information subset:

[0178] A sample acquisition unit 805 is used to acquire a set of unlabeled training samples;

[0179] A model acquisition unit 806 is used to train the third NLP model using the unlabeled training sample set to obtain a fourth NLP model;

[0180] The labeled sample acquisition unit 807 is used to acquire a set of labeled training samples;

[0181] The model adjustment unit 808 is used to adjust the model parameters of the fourth NLP model using the labeled training sample set to obtain the first NLP model.

[0182] The functions of each module in each device in the embodiments of the present application can be found in the corresponding description in the above method, and will not be repeated here.

[0183] In an embodiment of the present application, a subset acquisition unit is used to acquire at least one bidding information that does not meet the information condition in the bidding information set to obtain a bidding information subset, and the information condition includes a first natural language processing NLP model and / or a screening condition set; an information acquisition unit is used to acquire correction information input for at least one bidding information in the bidding information subset; an information correction unit is used to correct at least one bidding information based on the correction information to acquire a bidding correction set. Therefore, by acquiring bidding information that does not meet the information condition and correcting the bidding information, the accuracy of the information acquired when correcting the bidding information for information resources with new formats, new domain vocabulary, and new expressions can be improved, thereby improving the user experience.

[0184] Fig.15 FIG. 2 shows a block diagram of a terminal according to an embodiment of the present application. Fig.15 As shown, the terminal includes: a memory 1510 and a processor 1520, and the memory 1510 stores a computer program that can be run on the processor 1520. When the processor 1520 executes the computer program, the bidding information correction method combining RPA and AI in the above embodiment is implemented. The number of the memory 1510 and the processor 1520 can be one or more.

[0185] The terminal also includes:

[0186] The communication interface 1530 is used to communicate with external devices and perform data exchange transmission.

[0187] If the memory 1510, the processor 1520 and the communication interface 1530 are implemented independently, the memory 1510, the processor 1520 and the communication interface 1530 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.15 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0188] Optionally, in a specific implementation, if the memory 1510, the processor 1520 and the communication interface 1530 are integrated on a chip, the memory 1510, the processor 1520 and the communication interface 1530 can communicate with each other through an internal interface.

[0189] An embodiment of the present application provides a computer-readable storage medium storing a computer program, which implements the method provided in the embodiment of the present application when the program is executed by a processor.

[0190] An embodiment of the present application also provides a chip, which includes a processor for calling and executing instructions stored in the memory from the memory, so that a communication device equipped with the chip executes the method provided by the embodiment of the present application.

[0191] An embodiment of the present application also provides a chip, including: an input interface, an output interface, a processor and a memory, wherein the input interface, the output interface, the processor and the memory are connected via an internal connection path, and the processor is used to execute the code in the memory. When the code is executed, the processor is used to execute the method provided in the embodiment of the application.

[0192] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. It is worth noting that the processor may be a processor supporting the advanced RISC machines (ARM) architecture.

[0193] Further, optionally, the above-mentioned memory may include a read-only memory and a random access memory, and may also include a non-volatile random access memory. The memory may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may include a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may include a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available. For example, static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct rambus RAM (DR RAM).

[0194] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.

[0195] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0196] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0197] Any process or method description in the flow chart or otherwise described herein can be understood to represent a module, fragment or portion of a code including one or more executable instructions for implementing the steps of a specific logical function or process. And the scope of the preferred embodiment of the present application includes other implementations, in which the functions may not be performed in the order shown or discussed, including in a substantially simultaneous manner or in a reverse order according to the functions involved.

[0198] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, which can be embodied in any computer-readable medium for use by an instruction execution system, apparatus or device (such as a computer-based system, a system including a processor or other system that can fetch instructions from an instruction execution system, apparatus or device and execute instructions), or used in combination with these instruction execution systems, apparatuses or devices.

[0199] It should be understood that the various parts of the present application can be implemented with hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented with software or firmware stored in a memory and executed by a suitable instruction execution system. All or part of the steps of the above embodiment method can be completed by instructing the relevant hardware through a program, which can be stored in a computer-readable storage medium, and when the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0200] In addition, each functional unit in each embodiment of the present application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into one module. The above-mentioned integrated module can be implemented in the form of hardware or in the form of a software functional module. If the above-mentioned integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. The storage medium can be a read-only memory, a disk or an optical disk, etc.

[0201] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A bidding information correction method combining robotic process automation (RPA) and artificial intelligence (AI), characterized in that: include: By means of a robotic process automation (RPA) system, at least one bidding information that does not satisfy an information condition in a bidding information set is obtained to obtain a bidding information subset, wherein the information condition includes a first natural language processing (NLP) model and a screening condition set; the bidding information set refers to a set formed by gathering at least one bidding information; Obtaining correction information input for at least one bidding information in the bidding information subset; Correcting the at least one bidding information based on the correction information to obtain a bidding correction set; Wherein, before obtaining at least one bidding information that does not meet the information condition in the bidding information set to obtain the bidding information subset, the method further includes: Obtain a set of unlabeled training samples; Using the unlabeled training sample set to train the third NLP model to obtain a fourth NLP model; Obtain a set of labeled training samples; Using the labeled training sample set, adjusting the model parameters of the fourth NLP model to obtain the first NLP model; Wherein, the method further comprises: Obtaining the information quantity of at least one bidding information in the bidding correction set; If the amount of information is greater than the information amount threshold, the bidding correction set is used as a training sample set, and based on the training sample set, the model parameters of the first NLP model are adjusted to obtain a second NLP model, and the bidding correction set is cleared, and the second NLP model is used to correct the bidding information.

2. The method according to claim 1, characterized in that The step of obtaining at least one bidding information that does not meet the information condition in the bidding information set to obtain a bidding information subset includes: Obtaining the confidence level corresponding to any bidding information in the bidding information set through the first NLP model; If the confidence level corresponding to any bidding information is lower than the confidence level threshold, the bidding information is added to the bidding information subset.

3. The method according to claim 1, characterized in that The step of obtaining at least one bidding information that does not meet the information condition in the bidding information set to obtain a bidding information subset includes: Obtaining a screening condition set corresponding to the bidding information set; At least one bidding information that does not satisfy any screening condition in the screening condition set is obtained from the bidding information set to obtain a bidding information subset.

4. The method according to claim 1, characterized in that: The step of obtaining at least one bidding information that does not meet the information condition in the bidding information set to obtain a bidding information subset includes: Performing a random inspection on the bidding information set to obtain a bidding information set after random inspection; Obtaining sampling result information input for any bidding information in the sampling bidding information set; If the sampling result information indicates that any of the bidding information does not meet the information condition, then the any of the bidding information is added to the bidding information subset.

5. A bidding information correction device combining robotic process automation (RPA) and artificial intelligence (AI), characterized in that: include: A subset acquisition unit is used to acquire, through a robotic process automation (RPA) system, at least one bidding information that does not meet an information condition in a bidding information set, and obtain a bidding information subset, wherein the information condition includes a first natural language processing (NLP) model and / or a screening condition set; the bidding information set refers to a set formed by gathering at least one bidding information; An information acquisition unit, used to acquire correction information input for at least one bidding information in the bidding information subset; An information correction unit, configured to correct the at least one bidding information based on the correction information to obtain a bidding correction set; The bidding information correction device also includes: A model iteration unit, the model iteration unit comprising a quantity acquisition subunit and a model training subunit, the model iteration unit is used to iterate the first natural language processing NLP model based on the bidding correction set, and when a second NLP model is obtained: The quantity acquisition subunit is used to acquire the information quantity of at least one bidding information in the bidding correction set; The model training subunit is used for, if the amount of information is greater than the information amount threshold, using the bidding correction set as a training sample set, adjusting the model parameters of the first NLP model based on the training sample set to obtain a second NLP model, and clearing the bidding correction set, and the second NLP model performs bidding information correction; The device is also used to: obtain an unlabeled training sample set; use the unlabeled training sample set to train the third NLP model to obtain a fourth NLP model; obtain a labeled training sample set; use the labeled training sample set to adjust the model parameters of the fourth NLP model to obtain the first NLP model.

6. The device according to claim 5, characterized in that The subset acquisition unit includes a confidence acquisition subunit and an information adding subunit, and the subset acquisition unit is used to acquire at least one bidding information that does not meet the information condition in the bidding information set, and when the bidding information subset is obtained: The confidence acquisition subunit is used to acquire the confidence corresponding to any bidding information in the bidding information set through the first NLP model; The information adding subunit is used to add the bidding information to the bidding information subset if the confidence corresponding to any bidding information is lower than the confidence threshold.

7. The device according to claim 5, characterized in that The subset acquisition unit further includes a set acquisition subunit and an information screening subunit. The subset acquisition unit is used to acquire at least one bidding information that does not meet the information condition in the bidding information set. When the bidding information subset is obtained: The set acquisition subunit is used to acquire the screening condition set corresponding to the bidding information set; The information screening subunit is used to obtain at least one bidding information that does not meet any screening condition in the screening condition set from the bidding information set, so as to obtain a bidding information subset.

8. The device according to claim 5, characterized in that The subset acquisition unit further includes a set sampling subunit, a result acquisition subunit and a condition judgment subunit. The subset acquisition unit is used to acquire at least one bidding information that does not meet the information condition in the bidding information set. When the bidding information subset is obtained: The set sampling subunit is used to perform sampling on the bidding information set to obtain the bidding information set after sampling; The result acquisition subunit is used to obtain the sampling result information input for any bidding information in the bidding information set after the sampling inspection; The condition judgment subunit is used to add any bidding information to the bidding information subset if the sampling result information indicates that any bidding information does not meet the information condition.

9. A terminal combining RPA and AI, characterized in that: include: A processor and a memory, wherein the memory stores instructions, and the instructions are loaded and executed by the processor to implement the method according to any one of claims 1 to 4.

10. A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images