A Question-Answering Method, System, Computing Device and Storage Medium Combining RPA and AI

Abstract

An embodiment of the present application discloses a question-answering method, system, computing device, and storage medium combining RPA and AI, which relates to the technical field of Natural Language Processing (NLP). The question-answering method combining RPA and AI includes: receiving question data of a user, retrieving a first relevant document from a first document library for the question data, generating a reply based on the first relevant document through reading comprehension, and outputting the reply. The method further includes updating the first document library according to a second document library. Since the first document library is automatically updated, the answers given are highly real-time, applicable to fields with strong timeliness, saving a large amount of labor annotation costs, and thus effectively improving the quality of question-answering.

Description

[0001] Cross - reference to related applications

[0002] This application claims the priority of the Chinese patent application with the application number "202010617126.9" and the application name "A Question - Answering Method, System, Computing Device and Storage Medium Based on AI" submitted by Beijing Laiye Network Technology Co., Ltd. and Beijing Benying Network Technology Co., Ltd. on June 30, 2020. Technical field

[0003] This application relates to the field of Natural Language Processing (NLP), and specifically, to a question - answering method, system, computing device and storage medium that combines Robotic Process Automation (RPA) and Artificial Intelligence (AI). Background art

[0004] Robotic Process Automation (RPA) is to simulate human operations on a computer through specific "robot software" and automatically execute process tasks according to rules. Artificial Intelligence (AI) is a technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science. It attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a way similar to human intelligence. The research in this field includes robots, speech recognition, image recognition, Natural Language Processing (NLP) and expert systems, etc. Since its birth, the theory and technology of AI have become increasingly mature, and the application fields have also been continuously expanded. AI can simulate the information process of human consciousness and thinking.

[0005] With the popularization of deep - learning technology, more and more intelligent customer services use deep - learning models to automatically answer users' questions. In the prior art, according to the user's question, a question similar to the user's question is found in the question bank, and then the answer corresponding to the similar question is replied to the user. The disadvantage of this method is that the maintenance cost is relatively high. Especially when applied to fields with strong timeliness such as policy Q&A and commodity Q&A, it is necessary to "manually" update the answers frequently according to the changes of the answers. This is a great workload for those who maintain the answer library. If the database is not updated in time, the replied answers will be out of date, resulting in poor Q&A quality. All existing Q&A models require manual update of answers and are not suitable for fields with strong timeliness.

[0006] Therefore, it is crucial to study a question - answering system that can automatically update answers to improve the response quality of the question - answering system, which has become an urgent problem to be solved. Summary of the Invention

[0007] The embodiments of the present application provide a question - answering method, system, computing device, and storage medium that combine RPA and AI to overcome at least one technical problem existing in the prior art.

[0008] According to the first aspect of the embodiments of the present application, a question - answering method that combines RPA and AI is provided, including:

[0009] Receiving question data from a user;

[0010] Retrieving a first relevant document for the question data in a first document library;

[0011] Based on the first relevant document, generating and outputting a reply through reading comprehension based on Natural Language Processing (NLP) technology.

[0012] According to the second aspect of the embodiments of the present application, a question - answering system that combines RPA and AI is provided, including a question receiving module, a document retrieval module, and a question answering module, where

[0013] The question receiving module is configured to receive question data from a user;

[0014] The document retrieval module is configured to retrieve a first relevant document for the question data in a first document library;

[0015] The question answering module is configured to generate and output a reply through reading comprehension based on Natural Language Processing (NLP) technology according to the first relevant document.

[0016] According to the third aspect of the embodiments of the present application, a computing device is provided, including a storage device and a processor. The storage device is used to store a computer program, and the processor runs the computer program to enable the computing device to execute the steps of the question - answering method that combines RPA and AI.

[0017] According to the fourth aspect of the embodiments of the present application, a storage medium is provided, which stores the computer program used in the computing device. When the computer program is executed by a processor, the steps of the question - answering method that combines RPA and AI are implemented.

[0018] The beneficial effects of the embodiments of the present application are as follows:

[0019] This application provides a question-and-answer method, system, computing device, and storage medium that combine RPA and AI. In the question-and-answer method, after receiving the user's question data, the first relevant document is retrieved from the first document library according to the question data, and reading comprehension is performed on the first relevant document through natural language processing (NLP) technology to generate a reply and send it back to the user. In the question-and-answer method, the first document library is updated according to the second document library. The second relevant document is retrieved from the question library in the second document library, and the second relevant document in the question library is compared with the first relevant document corresponding to the question data. When the preset update rule is satisfied, the first relevant document in the first document library is replaced with the second relevant document, so that the first document library is updated. Since the first document library is kept updated in this question-and-answer method, the answers given are highly real-time, applicable to fields with strong timeliness, and effectively improve the quality of question and answer. In addition, the update of the first document library of the question-and-answer system is automatically completed, thus saving a large amount of laborious annotation costs, solving the problem of poor answer timeliness, and having progressiveness in the improvement of the question-and-answer system.

[0020] The innovation points of the embodiments of this application include:

[0021] 1. In the embodiments of this application, in the question-and-answer method, after receiving the user's question data, the first relevant document is retrieved from the first document library according to the question data, and reading comprehension is performed on the first relevant document through natural language processing (NLP) technology to generate a reply and send it back to the user. In the question-and-answer method, the first document library is updated according to the second document library. The second relevant document is retrieved from the question library in the second document library, and the second relevant document in the question library is compared with the first relevant document corresponding to the question data. When the preset update rule is satisfied, the first relevant document in the first document library is replaced with the second relevant document, so that the first document library is updated. Since the first document library is kept updated in this question-and-answer method, the answers given are highly real-time, applicable to fields with strong timeliness, and effectively improve the quality of question and answer. In addition, the update of the document library of the question-and-answer system is automatically completed, thus saving a large amount of laborious annotation costs, solving the problem of poor answer timeliness, and having progressiveness in the improvement of the question-and-answer system, which is one of the innovation points of the embodiments of this application.

[0022] 2. In the embodiments of the present application, through the first document library, a connection is established between the question and the document containing the answer to the question. Compared with the question-and-answer pairs composed of questions and answers in the prior art, it makes it possible to update the answer to the question by automatically updating the matched document library. One of the advantages of establishing the connection between the question and the document is that, on the one hand, it provides the source of the answer and enhances the interpretability; on the other hand, the update of the document can bring the update of the answer, which is one of the innovative points of the embodiments of the present application.

[0023] 3. In the embodiments of the present application, the first document library is automatically updated according to the second document library. According to the pre-obtained update rules, by comparing the second relevant document with the first relevant document, and the second reply with the first reply, when the update rules are met, the first relevant document in the first document library is replaced with the second relevant document to update the first document library. This way of replacing the associated document, compared with the way of replacing the answer in the question-and-answer pair in the prior art, reduces the necessity of manual participation and ensures the timeliness of the answer of the question-and-answer system through the automatically updated document library, which is one of the innovative points of the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0025] Figure 1 FIG.

[0026] Figure 2 FIG.

[0027] Figure 3 FIG.

[0028] Figure 4 FIG. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0030] It should be noted that the terms "include" and "have" and any variations thereof in the embodiments and accompanying drawings of the present application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally further include steps or units not listed, or may optionally further include other steps or units inherent to these processes, methods, products, or devices.

[0031] In daily life, there is a lot of information that is updated in real time. For example, policy information and commodity information may change over time, and this changed information is published in the public corpus in the form of website announcements or documents. The public corpus includes the data of these recently released updated information. When intelligent customer service is applied to such fields with strong timeliness, there is a problem of delayed answer update. It cannot modify the question answers in a timely manner according to the changed answer-related information, and the Q&A system that requires manual update has a huge workload for the personnel maintaining the answer library. Moreover, the efficiency and effect of manual update often cannot meet the timeliness requirements, resulting in outdated answers being replied, leading to poor Q&A quality. Therefore, a Q&A method and system that can be automatically updated are studied to improve the Q&A quality.

[0032] The embodiments of the present application disclose a Q&A method, system, computing device, and storage medium that combine RPA and AI, which will be described in detail below respectively.

[0033] Embodiment 1

[0034] Figure 1 This is a schematic diagram of the application scenario of a Q&A method that combines RPA and AI provided by the embodiments of the present application. As Figure 1As shown in the figure, there is a question bank Q, a first document library D, and a second document library. Among them, the question bank Q contains various questions asked by users. The second document library is a public corpus, including publicly available information data that is updated in real-time. The documents in the first document library D are generated based on the questions in the question bank Q and the publicly available corpus stored in the second document library. Specifically, for a question q^ in the question bank, the most relevant initial document is retrieved from the second document library, and the retrieved initial document is added to the first document library to form the first document library. When a user asks a question q, the most relevant first relevant document is retrieved from the first document library based on the question q. The retrieved first relevant document is subjected to reading comprehension based on AI. As a possible implementation, based on Natural Language Processing (NLP) technology, the retrieved first relevant document is analyzed to identify the corresponding semantics, and a corresponding reply a is generated based on the identified semantics and sent back to the user, improving the efficiency and reliability of the reply. By regularly and automatically retrieving the most relevant second relevant documents from the second document library based on the questions in the question bank, comparing the second relevant documents with the first relevant documents, and replacing the first relevant documents with the second relevant documents when the preset update rules are met, the first document library is updated, thereby ensuring the timeliness of the reply to the user based on the first document library, reducing the necessity of manual participation, and ensuring the timeliness of the answers of the question-and-answer system through the automatically updated document library.

[0035] Embodiment 2

[0036] Figure 2 This is a schematic flowchart of a question-and-answer method combining RPA and AI provided by an embodiment of the present application. As Figure 2 shown, a question-and-answer method combining RPA and AI includes:

[0037] 110. Receive the question data of the user.

[0038] Among them, the question data is the question raised by the user. For example, if the question raised by the user is "Is an integrated dryer good?", the question data can be a single question or multiple questions, which is not limited in this embodiment.

[0039] 120. Retrieve the first relevant document from the first document library for the question data.

[0040] In this embodiment, for each piece of question data according to the question data raised by the user, keyword extraction is performed, and keyword matching is performed in the first document library based on the extracted keywords to retrieve the corresponding relevant document. The relevant document contains the answer to the corresponding question data. For the sake of distinction, it is called the first relevant document.

[0041] 130. Based on the first relevant document, generate a response through reading comprehension based on Natural Language Processing (NLP) technology and output it. Optionally, the question-and-answer method further includes:

[0042] 140. Update the first document library according to the second document library. The second document library is a public corpus, including instantaneously updated public information data.

[0043] Establish the correspondence between the question data and the most relevant document, so as to realize the timely update of the question answer by updating the first document library.

[0044] Optionally, the step of updating the first document library according to the second document library includes:

[0045] 142. Retrieve the most relevant second relevant document in the second document library according to the question data in the pre-initialized question library.

[0046] In this embodiment, multiple question data proposed by users are stored in the question library, and instantaneously updated public information data are stored in the second document library. Each question is retrieved in the second document library to obtain the relevant document most relevant to the corresponding question data, which is called the second relevant document for the sake of distinction. That is to say, based on the pre-set question library and the second document library, the correspondence between each question and the most relevant second document is established.

[0047] Obtain the first word vector according to the corpus information of the question data in the question library, convert the multiple documents in the second document library into corresponding second word vectors, then calculate the similarity between the first word vector and the multiple second word vectors, and use the document with the highest similarity calculation result value as the second relevant document for this question, and update the first document library through the second relevant document.

[0048] 144. Update the first document library according to the second relevant document and the first relevant document.

[0049] Optionally, the step of updating the first document library according to the second relevant document and the first relevant document includes:

[0050] 1442. If the second relevant document corresponding to the question data is different from the first relevant document, generate a first response through reading comprehension according to the first relevant document, and generate a second response through reading comprehension according to the second relevant document.

[0051] To effectively update the first document library, not only do we need to compare relevant documents, but also compare the responses generated from relevant documents, set corresponding update rules according to user requirements, and replace documents when the update rules are met.

[0052] 1444. According to the pre-obtained update rules, compare the first relevant document with the second relevant document, and compare the first response with the second response. If the second relevant document or the second response meets the pre-obtained update rules, then replace the first relevant document corresponding to the question data in the first document library with the second relevant document to form a new first document library.

[0053] Optionally, the pre-obtained update rules include:

[0054] The second response contains the first response;

[0055] The score of the second response in reading comprehension is higher than that of the first response;

[0056] The second relevant document was uploaded more recently than the first relevant document;

[0057] The reading volume of the second relevant document is greater than that of the first relevant document;

[0058] The website where the second relevant document was published is cited by the website where the first relevant document was published.

[0059] Optionally, the question-answering method further includes: 100. Retrieve the first relevant document in the second document library according to the question data in the pre-initialized question library to form the first document library.

[0060] In this embodiment, the question-answering method can update the reply to the question by updating the document in the question-document pair by establishing the connection between the question and the document. Since the questions in the question library are regularly and automatically retrieved to match the second relevant document, and the first document library is updated according to the second relevant document, the first document library is automatically updated, thereby realizing the automatic update of the answers to the questions by this question-answering method, ensuring the timeliness of the answers, not requiring a large amount of manual annotation to update the document library, and using an automated method to retrieve documents and add them to the answer library, improving the quality of question answering.

[0061] Embodiment Three

[0062] Figure 3 It is a schematic structural diagram of a question-answering system combining RPA and AI provided by an embodiment of the present application. As Figure 3As shown in the figure, a question-answering system 300 combining RPA and AI is provided, including a question receiving module 310, a document retrieval module 320, and a question answering module 330, where

[0063] The question receiving module 310 is configured to receive question data from a user.

[0064] The document retrieval module 320 is configured to retrieve a first relevant document from a first document library based on the question data.

[0065] The question answering module 330 is configured to generate and output a reply based on the first relevant document through reading comprehension using natural language processing (NLP) technology.

[0066] Optionally, the question-answering system further includes an update module 340:

[0067] The update module 340 is configured to update the first document library according to a second document library.

[0068] In this embodiment, a question-answering system 300 combining RPA and AI is provided, which can implement the functions of the question-answering method combining RPA and AI. The corresponding implementation steps and effects can be referred to in the method section. [[ID=2,0]]

[0069] Embodiment 4

[0070] Figure 4 The following is a schematic structural diagram of a computing device provided by an embodiment of the present application. As Figure 4 shown, a computing device 400 is provided, including a storage device 410 and a processor 420. The storage device 410 is used to store a computer program, and the processor 420 runs the computer program to enable the computing device 400 to execute the steps of the question-answering method combining RPA and AI.

[0071] In this embodiment, a storage medium is provided, which stores the computer program used in the computing device. When the computer program is executed by a processor, it implements the steps of the question-answering method combining RPA and AI.

[0072] In summary, the embodiments of the present application provide a question-answering method and system, a computing device, and a storage medium combining RPA and AI. Through a periodically updated automatically document library, the question-answering system can reply to user questions based on the latest relevant documents. The process of document update is automatically completed, saving a large amount of laborious annotation costs, facilitating the maintenance of the question-answering system, meeting the timeliness requirements of user questions, and being applicable to fields with strong real-time requirements.

[0073] Those of ordinary skill in the art can understand that the accompanying drawings are only schematic diagrams of an embodiment, and the modules or processes in the drawings are not necessarily essential for implementing the present application.

[0074] Those of ordinary skill in the art can understand that the modules in the device in the embodiment can be distributed in the device of the embodiment according to the description of the embodiment, or can be correspondingly changed to be located in one or more devices different from this embodiment. The modules of the above embodiments can be combined into one module, or further split into multiple sub-modules.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A question-and-answer method combining RPA and AI, characterized in that, including: Receiving the problem data of the user; Retrieving the first relevant document in the first document library according to the problem data; Generating a reply based on reading comprehension according to the first relevant document based on natural language processing (NLP) technology and outputting it; It also includes: Updating the first document library according to the second document library, where the first document library is to retrieve the initial document in the second document library according to the problem data in the pre-initialized problem library, and add the retrieved initial document to the first document library to form the first document library; The step of updating the first document library according to the second document library includes: Retrieving the second relevant document in the second document library according to the problem data in the pre-initialized problem library; If the second relevant document corresponding to the problem data is different from the first relevant document, generating the first reply based on reading comprehension according to the first relevant document, and generating the second reply based on reading comprehension according to the second relevant document; According to the pre-obtained update rule, comparing the first relevant document with the second relevant document, and comparing the first reply with the second reply. If the second relevant document or the second reply meets the pre-obtained update rule, replacing the first relevant document corresponding to the problem data in the first document library with the second relevant document to form a new first document library.

2. The method according to claim 1, wherein The pre-obtained update rule includes: The second reply contains the first reply; The score of the reading comprehension of the second reply is higher than that of the first reply; The second relevant document is uploaded more recently than the first relevant document; The reading volume of the second relevant document is greater than that of the first relevant document; The publishing website of the second relevant document is cited by the publishing website of the first relevant document.

3. A question-answering system combining RPA and AI, characterized in that, Including a problem receiving module, a document retrieval module, and a problem reply module, where The problem receiving module is configured to receive the problem data of the user; The document retrieval module is configured to retrieve the first relevant document from the problem data in the first document library; The problem reply module is configured to generate a reply based on reading comprehension according to the first relevant document based on natural language processing (Natural Language Processing, NLP) technology and output it; It also includes an update module: The update module is configured to update the first document library according to the second document library, where the first document library is to retrieve the initial document in the second document library according to the problem data in the pre-initialized problem library, and add the retrieved initial document to the first document library to form the first document library; The step of updating the first document library according to the second document library includes: Retrieving the second relevant document in the second document library according to the problem data in the pre-initialized problem library; If the second relevant document corresponding to the problem data is different from the first relevant document, generate a first response through reading comprehension based on the first relevant document, and generate a second response through reading comprehension based on the second relevant document; According to the pre-obtained update rule, compare the first relevant document with the second relevant document, and compare the first response with the second response. If the second relevant document or the second response meets the pre-obtained update rule, replace the first relevant document corresponding to the problem data in the first document library with the second relevant document to form a new first document library.

4. A computing device, characterized in that, It includes a storage device and a processor. The storage device is used to store a computer program, and the processor runs the computer program to enable the computing device to execute the steps of the method according to any one of claims 1-2.

5. A storage medium, characterized in that, It stores the computer program used in the computing device according to claim 4, and when the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1-2.

Images