Intelligent operation and maintenance method and device, electronic equipment and storage medium

Abstract

The application discloses an intelligent operation and maintenance method and device, electronic equipment and a storage medium, and relates to the field of artificial intelligence, wherein the method comprises the following steps: receiving an operation and maintenance request initiated by a target chat room, matching a question carried in the operation and maintenance request with a question set in a preset knowledge base to filter out a question in the question set that has the highest similarity with the question carried in the operation and maintenance request, obtaining a target question, obtaining a reply strategy corresponding to the target question based on an index relationship, executing the reply strategy based on pre-statistical multidimensional environment information, obtaining an execution result, and sending the execution result to the target chat room. The application solves the technical problem that environment problems cannot be automatically repaired, user operation and maintenance requirements cannot be solved, and user satisfaction is reduced in the related art.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence, in particular to an intelligent operation and maintenance method and device, an electronic device and a storage medium. BACKGROUND

[0002] In related technologies, there are mainly two kinds of reply operation and maintenance methods for browsing consultation: (1) an intelligent operation and maintenance method based on OCR, which obtains error report pictures sent by a user, recognizes the pictures to obtain text information, directly finds corresponding answers in a knowledge base based on the text information, and if the corresponding answer result is found, an answer page is automatically popped up to notify the user that the answer is complete; (2) an intelligent operation and maintenance robot, or an intelligent operation and maintenance robot based on machine learning, which includes a background processing end, an operation and maintenance database and a plurality of operation and maintenance robots, updates the database by learning strategies or manually specifies rules, and when a fault signal is output to the back end, the back end triggers fault recovery according to the well-maintained rules.

[0003] However, the existing operation and maintenance methods have the following defects: (1) for the intelligent operation and maintenance robot, it is mainly a back-end program, lacks human-computer interaction, and users are basically operation and maintenance experts and operation and maintenance personnel, etc., and users need to have high professional knowledge, so the use threshold is high; (2) for the intelligent operation and maintenance method based on OCR, it needs to rely on fixed rule matching, and can only answer the user's questions, and cannot automatically repair environmental problems, that is, cannot automatically handle and close-loop problems.

[0004] In view of the above problems, no effective solutions have been proposed so far. SUMMARY

[0005] The embodiments of the present application provide an intelligent operation and maintenance method and device, an electronic device and a storage medium, to at least solve the technical problems in related technologies that environmental problems cannot be automatically repaired, user operation and maintenance needs cannot be solved, and user satisfaction is reduced.

[0006] According to an aspect of an embodiment of the present application, an intelligent operation and maintenance method is provided, which includes: receiving an operation and maintenance request initiated by a target chat room; matching a question carried in the operation and maintenance request with a question set in a preset knowledge base to filter out a question in the question set that has the highest similarity with the question carried in the operation and maintenance request, to obtain a target question, wherein the preset knowledge base includes the question set and a strategy set, and each reply strategy in the strategy set is pre-indexed with a question in the question set; obtaining a reply strategy corresponding to the target question based on the index relationship; executing the reply strategy based on pre-statistical multi-dimensional environmental information to obtain an execution result, and sending the execution result to the target chat room.

[0007] Optionally, before receiving the operation and maintenance request initiated by the target chat room, the method further comprises: establishing different types of chat rooms according to application types, wherein the chat rooms are used to display application environment information and operation and maintenance operation dynamics, and initiating the operation and maintenance request according to preset interaction rules; and establishing a chat room group in combination with a plurality of the chat rooms.

[0008] Optionally, the presentation types of the problems carried in the operation and maintenance request include natural language types and / or picture types, and after receiving the operation and maintenance request initiated by the target chat room, the method further comprises: identifying the problems of the natural language types according to preset entity types to obtain identification results; and performing segmentation processing on the identification results to obtain target information.

[0009] Optionally, before matching the problem carried in the operation and maintenance request with the problem set in the preset knowledge base, the method further comprises: extracting keyword words in the preset knowledge base and word positions of each keyword word; forming an inverted index relationship based on the keyword words and the word positions, wherein the inverted index relationship is used to represent position relationships of the keyword words in the reply strategy; and performing aggregation processing on the keyword words to form a semantic index relationship, wherein the semantic index relationship is used to index keyword words with semantic correlation.

[0010] Optionally, before matching the problem carried in the operation and maintenance request with the problem set in the preset knowledge base, the method further comprises: obtaining natural language information, a problem set and a strategy set in a historical time period; and training a semantic representation model and a semantic similarity model based on the natural language information, the problem set and the strategy set.

[0011] Optionally, the step of matching the problem carried in the operation and maintenance request with the problem set in the preset knowledge base comprises: retrieving the target information based on the inverted index relationship, the semantic index relationship and the semantic representation model to obtain a retrieval result, wherein the retrieval result at least includes a matching degree of each problem in the problem set corresponding to the target information; selecting a matching degree greater than a preset threshold based on the retrieval result; representing a problem in the problem set indicated by the selected matching degree as a problem cluster; and selecting the target problem in the problem cluster.

[0012] Optionally, the step of selecting the target problem in the problem cluster comprises: calculating a similarity between each problem in the problem cluster and the target information based on a semantic similarity model; adjusting the similarity based on a preset weight adjustment rule to obtain an adjusted similarity; sorting all the adjusted similarities and selecting a problem indicated by a highest similarity to obtain the target problem.

[0013] Optionally, after receiving the operation and maintenance request initiated by the target chat room, the method further comprises: performing up-sampling processing on the question of the picture type according to a preset up-sampling rule to obtain a feature layer of the question of the picture type; calculating a probability that each pixel in the question of the picture type belongs to text based on the feature layer to obtain a text probability map; forming a dynamic threshold map based on a dynamic threshold of each pixel; obtaining a binary map based on the text probability map and the dynamic threshold map; and forming a text box based on the binary map.

[0014] Optionally, after forming the text box, the method further comprises: performing arctangent calculation based on four points on the text box to obtain an arc; converting the arc into an angle and rotating by the angle; correcting the text box based on a preset text direction classifier and the rotated angle to obtain a horizontal matrix box; and identifying the horizontal matrix box based on a preset text recognizer to obtain a question of a natural language type.

[0015] Optionally, after matching the question carried in the operation and maintenance request with the question set in the preset knowledge base, the method further comprises: in the case that the matching is unsuccessful, initiating a request allocation task to allocate the operation and maintenance request to a responsible terminal, wherein the responsible terminal is configured to process the operation and maintenance request.

[0016] According to another aspect of the embodiments of the present application, an intelligent operation and maintenance device is also provided, which comprises: a receiving unit configured to receive an operation and maintenance request initiated by a target chat room; a matching unit configured to match a question carried in the operation and maintenance request with a question set in a preset knowledge base to filter out a question in the question set that has the highest similarity with the question carried in the operation and maintenance request to obtain a target question, wherein the preset knowledge base comprises the question set and a strategy set, and each reply strategy in the strategy set is pre-indexed with a question in the question set; an output unit configured to obtain a reply strategy corresponding to the target question based on the index relationship; and an execution unit configured to execute the reply strategy based on pre-statistical multi-dimensional environment information to obtain an execution result and send the execution result to the target chat room.

[0017] Optionally, the intelligent operation and maintenance device further comprises: a first establishing module configured to establish different types of chat rooms according to application types before receiving the operation and maintenance request initiated by the target chat room, wherein the chat rooms are configured to display application environment information and operation and maintenance operation dynamics and initiate the operation and maintenance request according to a preset interaction rule; and a second establishing module configured to establish a chat room group in combination with a plurality of the chat rooms.

[0018] Optionally, the presentation type of the question carried in the operation and maintenance request comprises a natural language type and / or a picture type, and the intelligent operation and maintenance device further comprises a first identification module, configured to, after receiving the operation and maintenance request initiated by the target chat room, identify the question with the presentation type of the natural language type according to a preset entity type to obtain an identification result; and a first segmentation module, configured to perform segmentation processing on the identification result to obtain target information.

[0019] Optionally, the intelligent operation and maintenance device further comprises a first extraction module, configured to, before matching the question carried in the operation and maintenance request with a question set in a preset knowledge base, extract keyword words in the preset knowledge base and word position of each keyword word; a first formation module, configured to form an inverted index relationship based on the keyword words and the word position, wherein the inverted index relationship is used to represent a position relationship of the keyword words in the reply strategy; and a second formation module, configured to perform aggregation processing on the keyword words to form a semantic index relationship, wherein the semantic index relationship is used to index keyword words with semantic correlation.

[0020] Optionally, the intelligent operation and maintenance device further comprises a first acquisition module, configured to, before matching the question carried in the operation and maintenance request with a question set in a preset knowledge base, acquire natural language information, a question set and a strategy set in a historical time period; and a first training module, configured to train a semantic representation model and a semantic similarity model based on the natural language information, the question set and the strategy set.

[0021] Optionally, the matching unit comprises a first retrieval module, configured to retrieve the target information based on the inverted index relationship, the semantic index relationship and the semantic representation model to obtain a retrieval result, wherein the retrieval result at least comprises a matching degree corresponding to each question in the question set and the target information; a first selection module, configured to select a matching degree greater than a preset threshold based on the retrieval result; a first representation module, configured to represent a question in the question set indicated by the selected matching degree as a question cluster; and a first selection module, configured to select the target question in the question cluster.

[0022] Optionally, the first selection module comprises a first calculation submodule, configured to calculate a similarity between each question in the question cluster and the target information based on a semantic similarity model; a first adjustment submodule, configured to adjust the similarity based on a preset weight adjustment rule to obtain an adjusted similarity; and a first sorting submodule, configured to sort all the adjusted similarities and select a question indicated by a highest similarity to obtain the target question.

[0023] Optionally, the intelligent operation and maintenance apparatus further comprises: a first processing module, configured to, after receiving the operation and maintenance request initiated by the target chat room, perform up-sampling processing on the question of the picture type according to a preset up-sampling rule, to obtain a feature layer of the question of the picture type; a first calculation module, configured to calculate a probability that each pixel in the question of the picture type belongs to text based on the feature layer, to obtain a text probability map; a third forming module, configured to form a dynamic threshold map based on a dynamic threshold of each pixel; a first output module, configured to obtain a binary map based on the text probability map and the dynamic threshold map; and a fourth forming module, configured to form a text box based on the binary map.

[0024] Optionally, the intelligent operation and maintenance apparatus further comprises: a second calculation module, configured to, after forming the text box, perform arctangent calculation based on four points on the text box to obtain an arc; a first conversion module, configured to convert the arc into an angle and rotate the angle; a first correction module, configured to correct the text box based on a preset text direction classifier and the rotated angle, to obtain a horizontal matrix box; and a second recognition module, configured to recognize the horizontal matrix box based on a preset text recognizer, to obtain the question of the natural language type.

[0025] Optionally, the intelligent operation and maintenance apparatus further comprises: a first initiation module, configured to, after matching the question carried in the operation and maintenance request with a question set in a preset knowledge base, initiate a request distribution task to distribute the operation and maintenance request to a responsible terminal in a case where the matching is unsuccessful, wherein the responsible terminal is configured to process the operation and maintenance request.

[0026] According to another aspect of the embodiments of the present application, a computer readable storage medium is also provided, which comprises a stored computer program, wherein the computer readable storage medium controls a device where the computer readable storage medium is located to perform the intelligent operation and maintenance method described above when the computer program runs.

[0027] According to another aspect of the embodiments of the present application, an electronic device is also provided, which comprises one or more processors and a memory, and the memory is configured to store one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the intelligent operation and maintenance method described above.

[0028] In this disclosure, an operation and maintenance (O&M) request initiated by a target chat room is received. The questions carried in the O&M request are matched with a set of questions in a preset knowledge base to filter out the questions in the question set that have the highest similarity to the questions carried in the O&M request, thus obtaining the target question. Based on the index relationship, a response strategy corresponding to the target question is obtained. Based on pre-statistical multi-dimensional environmental information, the response strategy is executed to obtain the execution result, which is then sent to the target chat room. In this application, after receiving an O&M request initiated by a chat room, the question with the highest similarity to the O&M request can be filtered to obtain a corresponding response strategy to resolve the O&M request. After obtaining environmental information, the response strategy is executed to automatically repair relevant environmental problems, thus solving the user's O&M needs, improving the accuracy of error identification, and the accuracy of matching O&M and environmental self-healing strategies. This solves the technical problem in related technologies that cannot automatically repair environmental problems and cannot solve the user's O&M needs, leading to a decline in user satisfaction. Attached Figure Description

[0029] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0030] Figure 1 This is a flowchart of an optional intelligent operation and maintenance method according to an embodiment of the present invention;

[0031] Figure 2 This is a schematic diagram of an optional image message recognition method according to an embodiment of the present invention;

[0032] Figure 3 This is a schematic diagram of an optional selective response strategy method according to an embodiment of the present invention;

[0033] Figure 4 This is a schematic diagram of an optional intelligent operation and maintenance system according to an embodiment of the present invention;

[0034] Figure 5 This is a schematic diagram of an optional intelligent operation and maintenance device according to an embodiment of the present invention;

[0035] Figure 6 This is a hardware structure block diagram of an electronic device (or mobile device) for an intelligent operation and maintenance method according to an embodiment of the present invention. Detailed Implementation

[0036] In the following, the technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application in combination. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work should fall within the protection scope of the present application.

[0037] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0038] In order for those skilled in the art to better understand the present application, the following explains some terms or names involved in the embodiments of the present application:

[0039] Optical Character Recognition (OCR) is a process of scanning text materials and then analyzing image files to obtain text and layout information.

[0040] Configuration Management Database (CMDB) is a logical warehouse containing information of the whole life cycle of configuration items and the relationship between configuration items.

[0041] Natural Language Processing (NLP) is used to study language problems in human-computer interaction.

[0042] It should be noted that the intelligent operation and maintenance method and device in the present disclosure can be used in the field of artificial intelligence to solve user operation and maintenance requests, and can also be used in any field other than the field of artificial intelligence to solve user operation and maintenance requests. The application field of the intelligent operation and maintenance method and device in the present disclosure is not limited.

[0043] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for display, analyzed data, etc.) involved in the present disclosure are all information and data authorized by the user or authorized by all parties.

[0044] The following embodiments of the present application can be applied to various systems / applications / devices that solve user operation and maintenance requests. The present application can establish a user-oriented chat room group, so that users can initiate operation and maintenance requests in the chat room. The presentation type of the operation and maintenance request can be in natural language form or in picture form. The specific content of the operation and maintenance request can be various, for example, explicit operation and maintenance operation, environment abnormal error code, error screenshot when the platform is abnormal, etc. The chat room background will send the operation and maintenance request initiated by the user to the micro-service cluster according to the preset mode (for example, the mode of http request) for processing (the micro-service cluster includes NLP and OCR services). After analyzing and understanding the natural language or picture information of the user, the execution strategy is selected, and the necessary environment information is extracted from the CMDB system. The central control module is returned to call the corresponding server to execute the strategy. For operation and maintenance requests that cannot be matched and scenarios that cannot be implemented for environment self-healing, task allocation can be performed to allocate the operation and maintenance task to the responsible person in the form of messages, emails, etc. The closed-loop processing of the operation and maintenance task can be effectively realized.

[0045] The present application will be described in detail below in conjunction with various embodiments.

[0046] Embodiment one

[0047] According to the embodiments of the present application, an intelligent operation and maintenance method embodiment is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in a different order.

[0048] Figure 1 is a flowchart of an optional intelligent operation and maintenance method according to an embodiment of the present application, as shown in Figure 1 The method comprises the following steps:

[0049] Step S101, receiving an operation and maintenance request initiated by a target chat room.

[0050] Step S102, matching the problem carried in the operation and maintenance request with the problem set in the preset knowledge base to filter out the problem in the problem set that has the highest similarity with the problem carried in the operation and maintenance request, and obtaining a target problem, wherein the preset knowledge base comprises: a problem set and a strategy set. Each reply strategy in the strategy set is pre-indexed with the problems in the problem set.

[0051] Step S103, based on the index relationship, obtaining the reply strategy corresponding to the target problem.

[0052] Step S104, based on the pre-statistical multi-dimensional environment information, executing the reply strategy to obtain the execution result, and sending the execution result to the target chat room.

[0053] Through the above steps, the operation and maintenance request initiated by the target chat room can be received, the problem carried in the operation and maintenance request is matched with the problem set in the preset knowledge base to filter out the problem in the problem set with the highest similarity with the problem carried in the operation and maintenance request, the target problem is obtained, based on the index relationship, the reply strategy corresponding to the target problem is obtained, based on the pre-statistical multi-dimensional environment information, the reply strategy is executed to obtain the execution result, and the execution result is sent to the target chat room. In the embodiment of the application, after receiving the operation and maintenance request initiated by the chat room, the problem with the highest similarity with the operation and maintenance request can be filtered out to obtain the corresponding reply strategy to solve the operation and maintenance request, and after obtaining the environment information, the reply strategy is executed to automatically repair the related environment problem, thereby solving the operation and maintenance demand of the user, improving the error recognition accuracy and the accuracy of the operation and maintenance and environment self-healing strategy matching, and further solving the technical problems in the related art that the environment problem cannot be automatically repaired, the operation and maintenance demand of the user cannot be solved, and the user satisfaction is reduced.

[0054] The embodiment of the application will be described in detail below in combination with the above steps.

[0055] In the embodiment of the application, before receiving the operation and maintenance request initiated by the target chat room, the method further comprises: establishing different types of chat rooms according to application types, wherein the chat room is used to display application environment information and operation and maintenance operation dynamics, and initiates the operation and maintenance request according to a preset interaction rule; and establishing a chat room group in combination with a plurality of chat rooms.

[0056] In the embodiment, in the chat room group, each application has its own independent chat room (i.e., different types of chat rooms can be established according to application types), and user groups can also be distinguished according to applications, and users in the same group can master the latest application environment information and operation and maintenance operation dynamics in the chat room (i.e., the chat room is used to display application environment information and operation and maintenance operation dynamics). In the embodiment, the chat room can be implemented using node.js coding, or can be implemented using other coding, which is not limited herein.

[0057] In this embodiment, the operation and maintenance assistant robot can be set in the chat room, and the user can initiate an operation and maintenance request according to a preset interaction rule (the specific interaction rule can be set by the user, for example, mail interaction, instant message interaction, for example, triggering interaction with the operation and maintenance assistant robot in the chat room by mentioning the operation and maintenance assistant robot), wherein the type of the operation and maintenance request can be natural language or picture form, and the specific content of the operation and maintenance request can be various, for example, explicit operation and maintenance operation, environment exception error code, error screenshot when the platform is abnormal, and the like.

[0058] Optionally, different permission roles can be set for the users in each group, so that the users initiate different operation and maintenance requests according to different permissions.

[0059] In this embodiment, through the chat room group, the user groups of different applications and different environments can be fully isolated, the single communication channel of the same user group is locked, the information difference between users is reduced, and the communication efficiency and processing efficiency of operation and maintenance matters are improved. At the same time, an independent environment exception error sample set can be maintained according to the application, which can highlight the exception features of the application, improve the error recognition accuracy and the matching accuracy of the operation and maintenance and environment self-healing strategies, and through the concept of combining operation and maintenance and chat, various daily operation and maintenance work, including environment monitoring, alarm and early warning, demand realization, resource coordination and the like, can be completed in a natural language manner, thereby reducing the use threshold of the user.

[0060] In step S101, an operation and maintenance request initiated by a target chat room is received.

[0061] Optionally, the presentation type of the problem carried in the operation and maintenance request includes a natural language type and / or a picture type, and after receiving the operation and maintenance request initiated by the target chat room, the method further includes: identifying the problem with the presentation type of the natural language type according to a preset entity type to obtain an identification result; and performing segmentation processing on the identification result to obtain target information.

[0062] In the embodiment of the application, the user can initiate an operation and maintenance request in a natural language type and / or a picture type (that is, the presentation type of the problem carried in the operation and maintenance request includes a natural language type and / or a picture type), and after receiving the operation and maintenance request initiated by the user, the natural language can be identified according to a preset entity type (for example, entity class, time class and number class) named entity and cut to obtain target information.

[0063] Optionally, after receiving the operation and maintenance request initiated by the target chat room, the method further comprises: performing up-sampling processing on the question of the picture type according to a preset up-sampling rule to obtain a feature layer of the question of the picture type; calculating a probability that each pixel in the question of the picture type belongs to text based on the feature layer to obtain a text probability map; forming a dynamic threshold map based on a dynamic threshold of each pixel; obtaining a binary map based on the text probability map and the dynamic threshold map; and forming a text box based on the binary map.

[0064] Optionally, after forming the text box, the method further comprises: performing arctangent calculation based on four points on the text box to obtain an arc; converting the arc into an angle and rotating the angle; correcting the text box based on a preset text direction classifier and the rotated angle to obtain a horizontal matrix box; and identifying the horizontal matrix box based on a preset text recognizer to obtain the question of the natural language type.

[0065] In the embodiment of the application, for the operation and maintenance request of the picture type, the picture can be recognized as a natural language for processing, Figure 2 is a schematic diagram of an optional picture message recognition method according to an embodiment of the application, as Figure 2 shown, a user picture message is processed through a differentiable binarization text detection (i.e., DB text detection), a detection box correction, a convolutional recurrent neural network (CRNN) text recognition and the like to obtain output content (i.e., a message in the form of a natural language), which is as follows:

[0066] The DB text detection module in the embodiment can be packaged using Python, can first generate a feature layer of the picture through up-sampling (i.e., performing up-sampling processing on the question of the picture type according to a preset up-sampling rule to obtain a feature layer of the question of the picture type), then calculate a probability that each pixel in the question of the picture type belongs to text based on the feature layer to form a text probability map, then form a dynamic threshold map according to a dynamic threshold of each pixel, and then generate a DB binary map through the text probability map and the dynamic threshold map, and expand a label according to the DB binary map to form a text box.

[0067] The detection box correction module in the embodiment can be implemented by Python, can calculate an arc by taking an arctangent of four points of the text box, convert the arc into an angle, rotate the image by the angle, correct the text box through a preset text direction classifier to obtain a correct horizontal matrix box.

[0068] The CRNN in the embodiment is a widely used text recognizer (i.e., a preset text recognizer), which can be packaged using Python, and the CRNN integrates feature extraction and sequence modeling and is an end-to-end text recognition network, and the CRNN can be used to identify the horizontal matrix box in the embodiment to convert the question of the picture type into a question of the natural language type.

[0069] Optionally, before matching the problem carried in the operation and maintenance request with the problem set in the preset knowledge base, further comprising: extracting the keyword and the word position of each keyword in the preset knowledge base; forming an inverted index relationship based on the keyword and the word position, wherein the inverted index relationship is used to represent the position relationship of the keyword in the reply strategy; and performing aggregation processing on the keyword to form a semantic index relationship, wherein the semantic index relationship is used to index the keywords with semantic correlation.

[0070] In the embodiment of the application, a knowledge base in natural language form (i.e., a preset knowledge base) can be maintained, which includes a problem set (e.g., operation and maintenance requests, environment error problems, etc.) and a strategy set, and then an inverted index relationship and a semantic index relationship are respectively established for the demand set (i.e., the problem set and the strategy set), wherein the inverted index is responsible for extracting the keyword and the word position in the demand set (i.e., extracting the keyword and the word position of each keyword in the preset knowledge base, and forming an inverted index relationship based on the keyword and the word position), forming a “keyword-strategy” matrix, i.e., the inverted index relationship is used to represent the position relationship of the keyword in the reply strategy. The semantic index can perform aggregation processing on the keyword in the strategy and index the keywords with semantic correlation. In the embodiment, the ANN (approximate nearest neighbor search) implemented by Python can be run for processing.

[0071] Optionally, the embodiment can make a proper noun dictionary for a certain operation and maintenance field (financial software operation and maintenance field), and by using the proper noun dictionary, the index can be established in a targeted manner to avoid excessive cutting of sentences when generating the index.

[0072] Optionally, before matching the problem carried in the operation and maintenance request with the problem set in the preset knowledge base, further comprising: obtaining natural language information, a problem set and a strategy set in a historical time period; and training a semantic representation model and a semantic similarity model based on the natural language information, the problem set and the strategy set.

[0073] In the embodiment of the application, the semantic representation model and the semantic similarity model can be trained by the obtained natural language information, the problem set and the strategy set in the historical time period, wherein the semantic representation model can use BOW (bag of words), CNN (convolutional neural network) and other algorithms, the semantic similarity model can use MatchPyramid (construct a matching matrix) and other algorithms, and both can be implemented by Python. The embodiment does not limit the specific algorithm for building the model, and the algorithm can be selected according to the actual situation.

[0074] In step S102, the question carried in the operation and maintenance request is matched with the question set in the preset knowledge base to screen out a question in the question set that has the highest similarity with the question carried in the operation and maintenance request, and a target question is obtained. The preset knowledge base includes the question set and a strategy set, and each reply strategy in the strategy set is pre-indexed with a question in the question set.

[0075] Optionally, the step of matching the question carried in the operation and maintenance request with the question set in the preset knowledge base includes: searching target information based on an inverted index relationship, a semantic index relationship and a semantic representation model to obtain a search result, wherein the search result at least includes a matching degree corresponding to each question in the question set and the target information; selecting a matching degree greater than a preset threshold based on the search result; representing the question in the question set indicated by the selected matching degree as a question cluster; and selecting a target question in the question cluster.

[0076] In the embodiment of the present application, after the inverted index relationship, the semantic index relationship and the semantic representation model are obtained, the processed natural language operation and maintenance request (i.e., the target information) can be searched to search all questions (i.e., through searching, the matching degree corresponding to each question in the question set and the target information is obtained, the matching degree greater than the preset threshold is selected, and the question in the question set indicated by the selected matching degree is represented as a question cluster) same as and similar to the target information. Then, the target question can be selected in the question cluster.

[0077] Optionally, the step of selecting the target question in the question cluster includes: calculating the similarity between each question in the question cluster and the target information based on a semantic similarity model; adjusting the similarity based on a preset weight adjustment rule to obtain an adjusted similarity; sorting all the adjusted similarities and selecting a question indicated by the highest similarity to obtain the target question.

[0078] In the embodiment of the present application, the similarity between each question in the question cluster and the target information can be evaluated and scored. Then, based on the preset weight adjustment rule, the similarity is adjusted (for example, the linear regression model and GBDT (gradient boosting) packaged by Python can be used to do regression analysis on the question cluster after similarity calculation, and the similarity scores of the questions are adjusted according to the preset weight (which can be set according to the inverted index and the semantic index)), all the adjusted similarities are sorted, and the question indicated by the highest similarity is selected to obtain the target question.

[0079] Optionally, the user can maintain a proper noun dictionary, which can include proper nouns in a certain operation and maintenance field and the user's own diction habits, and can be used for similarity evaluation and scoring to improve the score of specific words.

[0080] In this embodiment, Figure 3 This is a schematic diagram of an optional selective response strategy method according to an embodiment of the present invention, such as... Figure 3 As shown, it includes: a natural language analysis module, an index generation module, a model training module, a retrieval module, a similarity calculation module, a re-ranking module, and a domain-specific terminology dictionary. In this embodiment, the user can maintain a knowledge base and a policy set. Through this knowledge base and policy set, an inverted index and a semantic index can be generated. Through a large set of natural language data, as well as the knowledge base and policy set, a semantic representation model and a semantic similarity model can be trained. After the user initiates a request, the natural language request is processed (including: Named Entity Recognition (NER), recombination, tagging, etc.) to obtain the processed language information. The system uses the obtained inverted index, semantic index, and semantic representation model to perform semantic and term retrieval on the processed language information, obtaining all documents that are the same as or similar to the language information. These documents are then passed to the similarity calculation module in list form for semantic and term similarity calculation. Subsequently, the linear model and GBDT model (gradient boosting) in the reordering module are used to adjust the obtained similarity, thereby selecting the strategy or environment exception error that is closest to the user's request (i.e., hitting the knowledge base or strategy). At the same time, users can also maintain a domain-specific terminology dictionary, which can be used in modules such as index generation, retrieval, and similarity calculation.

[0081] Optionally, after matching the issues carried in the maintenance request with the set of issues in the preset knowledge base, the method further includes: if the matching fails, initiating a request to assign a task, and assigning the maintenance request to the responsible terminal, wherein the responsible terminal is used to process the maintenance request.

[0082] In this embodiment of the invention, for unmatched maintenance requests and scenarios where environmental self-healing cannot be achieved, task allocation can be performed to assign the maintenance request to the responsible person in the form of IM messages, emails, etc. (that is, in the case of unsuccessful matching, a request to allocate tasks is initiated to assign the maintenance request to the responsible terminal to handle the maintenance request), which can realize effective processing and closed-loop of maintenance tasks.

[0083] Step S103: Based on the index relationship, obtain the response strategy corresponding to the target question.

[0084] In this embodiment of the invention, after obtaining the target question, the corresponding response strategy can be obtained based on the index relationship (i.e., the inverted index relationship and the semantic index relationship) pre-established between the question and the response strategy.

[0085] Step S104: Based on the pre-statistical multi-dimensional environmental information, execute the response strategy, obtain the execution result, and send the execution result to the target chat room.

[0086] In the embodiment of the application, the environment information of multiple statistical dimensions can be aggregated by the CMDB module, after obtaining the reply strategy, the necessary environment information can be extracted from the CMDB module, then the reply strategy is executed, and the execution result is sent to the target chat room to notify the related user that the operation and maintenance request is completed, thereby realizing automatic matching and execution of the strategy to automatically heal the environment problem of the application.

[0087] In the embodiment of the application, a low-threshold intelligent operation and maintenance method is provided, which can accurately identify user operation and maintenance requirements and environment abnormalities through various ways, and automatically match and execute strategies to automatically heal the environment problem of the application, which can achieve the following beneficial effects:

[0088] (1) The chat room group can fully isolate different application and different environment user groups, lock the single communication channel of the same user group, reduce the information difference between users, and improve the communication efficiency and processing efficiency of operation and maintenance matters. At the same time, according to the application, an independent environment abnormal error sample set is maintained, which can highlight the abnormal features of the application and improve the accuracy of error identification and operation and maintenance, and environment self-healing strategy matching.

[0089] (2) The concept of combining operation and maintenance and chat is adopted, and various daily operation and maintenance work, including environment monitoring, alarm warning, demand realization, resource coordination, etc., is completed in a natural language manner, which can reduce the use threshold of users.

[0090] (3) The OCR technology is used to identify pictures to extract the content therein, which can increase the diversity of operation and maintenance entrances, further reduce the use threshold of users, and improve the efficiency.

[0091] (4) By maintaining a domain-specific name dictionary (which can include financial software operation and maintenance domain-specific words and user habit words), the system's ability to recognize user intent in a specific scenario is significantly improved. Users and operation and maintenance experts specify a certain number of "demand-strategy" and "error-strategy" sets for model generation, which improves the hit rate of accurate operation and maintenance strategies.

[0092] (5) The operation and maintenance strategy and the self-healing strategy are maintained by users, and the operation and maintenance experts further abstract, optimize, and comply with the regulations. The intelligent robot schedules the microservices to replace the manual operation of users such as operation and maintenance, testing, and development, which improves the work efficiency, ensures the operation safety and compliance, and saves the labor cost.

[0093] (6) For cases that the platform cannot answer or cannot match the environment self-healing strategy, the operation and maintenance task can be distributed to specific responsible persons in the form of real-time messages and emails according to the CMDB account, to realize effective processing and closed loop of the operation and maintenance task.

[0094] Embodiment two

[0095] Figure 4 This is a schematic diagram of an optional intelligent operation and maintenance system according to an embodiment of the present invention, such as... Figure 4 As shown, this includes: chat room groups directly facing users (including chat rooms for application 1, application 2, ..., chat rooms for application N, etc.), which automatically capture and passively receive maintenance requests, environment error messages, error images, etc. sent by users (i.e., requests initiated by maintenance, development, testing, etc.), and submit requirements (including natural language and image recognition formats) to the central control (i.e., the central control module responsible for coordinating the functions of various modules, which can be maintained by maintenance, development, testing, etc.). After the microservice cluster (including NLP and OCR services and various maintenance policies, which can be maintained by maintenance, development, testing, etc.) parses the natural language and image recognition formats, it selects the execution policy, returns the policy scheduling result to the central control, and then the central control retrieves the ledger information (including environment information, etc.) from the CMDB module and returns it to the central control module to invoke the corresponding microservice execution policy (i.e., execute the policy through the server in the test environment server group and return the result). The microservice cluster can then feed back the result to the corresponding chat room. For user requests that cannot be matched or scenarios where self-healing of the environment cannot be achieved, messages can be pushed to the WEB push server in the form of real-time messages and emails based on the CMDB module. The WEB push server then pushes the messages to the chat room, and the chat room provides feedback to maintenance, development, testing and other personnel for processing.

[0096] In this embodiment, the CMDB module aggregates environmental information and responsible parties across multiple statistical dimensions. In a large-scale, multi-type test environment, code from multiple applications at different deployment times is deployed simultaneously. The CMDB can register each application's environment in chronological order of deployment time. Each environment is further divided into multiple nodes at the smallest service granularity, including application servers, Oracle databases, MySQL databases, batch servers, PaaS containers, DSF services, etc. Each service node contains information such as its type, operating system, server IP address, server environment user information, and other various data elements.

[0097] In this embodiment of the invention, after the microservice cluster receives the specific execution strategy and environment information sent by the central control module, it executes the strategy on the correct server with the corresponding microservice and waits for the processing result. Then, it feeds the result back to the central control module and notifies the message push server, and finally pushes it to the chat room to be displayed to the user. This realizes automatic matching and execution of strategies to self-heal and solve the application's environmental problems.

[0098] Example 3

[0099] The intelligent operation and maintenance device provided in this embodiment includes multiple implementation units, each of which corresponds to a specific implementation step in Embodiment 1 above.

[0100] Figure 5 This is a schematic diagram of an optional intelligent operation and maintenance device according to an embodiment of the present invention, such as... Figure 5 As shown, the device may include: a receiving unit 50, a matching unit 51, an output unit 52, and an execution unit 53, wherein...

[0101] The receiving unit 50 is used to receive maintenance requests initiated by the target chat room;

[0102] The matching unit 51 is used to match the questions carried in the operation and maintenance request with the question set in the preset knowledge base, so as to filter out the questions in the question set that have the highest similarity with the questions carried in the operation and maintenance request, and obtain the target question. The preset knowledge base includes: a question set and a strategy set. Each response strategy in the strategy set and the questions in the question set are pre-indexed.

[0103] Output unit 52 is used to obtain the response strategy corresponding to the target question based on the index relationship;

[0104] The execution unit 53 is used to execute a response strategy based on pre-statistical multi-dimensional environmental information, obtain the execution result, and send the execution result to the target chat room.

[0105] The aforementioned device can receive maintenance requests initiated by a target chat room via a receiving unit 50, match the issues carried in the maintenance request with a set of issues in a preset knowledge base via a matching unit 51 to filter out the issues in the issue set that have the highest similarity to the issues carried in the maintenance request, thus obtaining the target issue. An output unit 52 obtains a response strategy corresponding to the target issue based on an index relationship, and an execution unit 53 executes the response strategy based on pre-statistical multi-dimensional environmental information to obtain the execution result, which is then sent to the target chat room. In this embodiment of the invention, after receiving a maintenance request initiated by a chat room, the issue with the highest similarity to the maintenance request can be filtered out to obtain a corresponding response strategy to resolve the maintenance request. After obtaining environmental information, the response strategy is executed to automatically repair related environmental issues, thus addressing the user's maintenance needs, improving the accuracy of error identification, and the accuracy of matching maintenance and environmental self-healing strategies. This solves the technical problem in related technologies where environmental issues cannot be automatically repaired, user maintenance needs cannot be met, leading to decreased user satisfaction.

[0106] Optionally, the intelligent operation and maintenance device further comprises a first establishing module configured to establish different types of chat rooms according to application types before receiving the operation and maintenance request initiated by the target chat room, wherein the chat room is used to display application environment information and operation and maintenance operation dynamics, and initiates the operation and maintenance request according to a preset interaction rule; and a second establishing module configured to establish a chat room group in combination with a plurality of chat rooms.

[0107] Optionally, the presentation type of the question carried in the operation and maintenance request comprises a natural language type and / or a picture type, and the intelligent operation and maintenance device further comprises a first identifying module configured to identify the question with the natural language type according to a preset entity type after receiving the operation and maintenance request initiated by the target chat room, to obtain an identification result; and a first segmenting module configured to perform segmentation processing on the identification result to obtain target information.

[0108] Optionally, the intelligent operation and maintenance device further comprises a first extracting module configured to extract a keyword and a word position of each keyword in the preset knowledge base before matching the question carried in the operation and maintenance request with a question set in the preset knowledge base; a first forming module configured to form an inverted index relationship based on the keyword and the word position, wherein the inverted index relationship is used to represent a position relationship of the keyword in a reply strategy; and a second forming module configured to perform aggregation processing on the keyword to form a semantic index relationship, wherein the semantic index relationship is used to index the keywords with semantic correlation.

[0109] Optionally, the intelligent operation and maintenance device further comprises a first obtaining module configured to obtain natural language information, a question set and a strategy set in a historical time period before matching the question carried in the operation and maintenance request with the question set in the preset knowledge base; and a first training module configured to train a semantic representation model and a semantic similarity model based on the natural language information, the question set and the strategy set.

[0110] Optionally, the matching unit comprises a first searching module configured to search the target information based on the inverted index relationship, the semantic index relationship and the semantic representation model to obtain a search result, wherein the search result at least comprises a matching degree of each question in the question set corresponding to the target information; a first selecting module configured to select the matching degree greater than a preset threshold based on the search result; a first representing module configured to represent the question in the question set indicated by the selected matching degree as a question cluster; and a first selecting module configured to select a target question in the question cluster.

[0111] Optionally, the first selecting module comprises: a first calculating submodule, configured to calculate similarity between each question in the question cluster and the target information based on a semantic similarity model; a first adjusting submodule, configured to adjust the similarity based on a preset weight adjustment rule to obtain adjusted similarity; and a first sorting submodule, configured to sort all the adjusted similarity and select a question indicated by the highest similarity to obtain the target question.

[0112] Optionally, the intelligent operation and maintenance apparatus further comprises: a first processing module, configured to perform up-sampling processing on the question of the picture type by a preset up-sampling rule to obtain a feature layer of the question of the picture type after receiving the operation and maintenance request initiated by the target chat room; a first calculating module, configured to calculate a probability that each pixel in the question of the picture type belongs to text based on the feature layer to obtain a text probability map; a third forming module, configured to form a dynamic threshold map based on a dynamic threshold of each pixel; a first output module, configured to obtain a binary map based on the text probability map and the dynamic threshold map; and a fourth forming module, configured to form a text box based on the binary map.

[0113] Optionally, the intelligent operation and maintenance apparatus further comprises: a second calculating module, configured to perform arctangent calculation based on four points on the text box to obtain an arc after forming the text box; a first converting module, configured to convert the arc into an angle and rotate the angle; a first correcting module, configured to correct the text box based on a preset text direction classifier and the rotated angle to obtain a horizontal matrix box; and a second recognizing module, configured to recognize the horizontal matrix box based on a preset text recognizer to obtain the question of the natural language type.

[0114] Optionally, the intelligent operation and maintenance apparatus further comprises a first initiating module, configured to initiate a request distribution task to distribute the operation and maintenance request to a responsible terminal in a case where the matching is unsuccessful after matching the question carried in the operation and maintenance request with the question set in the preset knowledge base, wherein the responsible terminal is configured to process the operation and maintenance request.

[0115] The apparatus described above can further comprise a processor and a memory, and the receiving unit 50, the matching unit 51, the output unit 52, the execution unit 53, etc. are stored in the memory as program units, and the corresponding functions are realized by the processor executing the program units stored in the memory.

[0116] The processor described above comprises a core, and the core retrieves the corresponding program units from the memory. The core can be set to one or more, and the execution result is obtained by adjusting the core parameters to execute the reply strategy, and the execution result is sent to the target chat room.

[0117] The above-mentioned memory can include a non-permanent memory in a computer readable medium, a random access memory (RAM) and / or a non-volatile memory such as a read-only memory (ROM) or a flash memory (flash RAM), and the memory includes at least one memory chip.

[0118] The application further provides a computer program product adapted to execute the program of the following method steps when executed on a data processing device: receiving an operation and maintenance request initiated by a target chat room, matching a question carried in the operation and maintenance request with a question set in a preset knowledge base to screen out a question in the question set that has the highest similarity with the question carried in the operation and maintenance request, obtaining a target question, obtaining a reply strategy corresponding to the target question based on an index relationship, executing the reply strategy based on pre-statistical multi-dimensional environment information, obtaining an execution result, and sending the execution result to the target chat room.

[0119] According to another aspect of the embodiments of the application, a computer readable storage medium is also provided, which includes a stored computer program, wherein the computer readable storage medium controls a device where the computer readable storage medium is located to execute the intelligent operation and maintenance method described above when the computer program runs.

[0120] According to another aspect of the embodiments of the application, an electronic device is also provided, which includes one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the intelligent operation and maintenance method described above.

[0121] Figure 6 is a hardware structure block diagram of an electronic device (or mobile device) for an intelligent operation and maintenance method according to an embodiment of the application. As shown in Figure 6 , the electronic device can include one or more (in the figure, 102a, 102b, …, 102n are used to show) processors 102 (the processor 102 can include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. In addition, it can also include a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which can be included as one of the ports of the I / O interface), a network interface, a keyboard, a power supply and / or a camera. Those skilled in the art can understand that Figure 6 The structure shown is only schematic, which does not limit the structure of the above-mentioned electronic device. For example, the electronic device can include more or fewer components than Figure 6 shown, or have a different configuration from Figure 6 shown.

[0122] The above-mentioned serial numbers of the embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0123] In the above-mentioned embodiments of the present application, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0124] In several embodiments provided in the present application, it should be understood that the disclosed technical contents can be implemented by other ways. Among them, the above-mentioned device embodiments are only schematic, for example, the division of the units can be a logical function division, and actual implementation can have another division way, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units or modules shown or discussed can be indirect coupling or communication connection through some interfaces, units or modules, which can be electrical or other forms.

[0125] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place or distributed to multiple units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0126] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be realized in the form of hardware or in the form of software functional unit.

[0127] The integrated unit, if realized in the form of software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the whole or part of the technical solutions which essentially contribute to the prior art can be embodied in the form of software product, which is stored in a storage medium and includes a plurality of instructions for making a computer device (which can be a personal computer, a server or a network device, etc.) execute all or part of the steps of the method described in each embodiment of the present application. The above-mentioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and various program code storage media.

[0128] The above merely describes the preferred embodiments of the present application, and it should be pointed out that, for those skilled in the art, several improvements and refinements can be made without departing from the principles of the present application, and these improvements and refinements should also be considered as falling within the protection scope of the present application.

Claims

1. An intelligent operation and maintenance method, characterized in that, The application comprises the following steps: According to the application type, different types of chat rooms are established, which are used to display application environment information and operation dynamic, and initiate operation requests according to preset interaction rules; different permission roles are set for users of each chat room to initiate different operation requests; A chat room group is established by combining multiple chat rooms; An operation request initiated by a target chat room is received; The problem carried in the operation request is matched with a problem set in a preset knowledge base to filter out a problem in the problem set that has the highest similarity with the problem carried in the operation request, and a target problem is obtained, wherein the preset knowledge base comprises the problem set and a strategy set, each reply strategy in the strategy set is pre-indexed with a problem in the problem set; the index relationship is established by making a special term dictionary in the operation field and establishing the index relationship by using the special term dictionary; Based on the index relationship, a reply strategy corresponding to the target problem is obtained; Based on the pre-statistical multi-dimensional environment information, the reply strategy is executed to obtain an execution result, and the execution result is sent to the target chat room; Before matching the problem carried in the operation request with the problem set in the preset knowledge base, the following steps are further included: Extracting key words and word positions of each key word in the preset knowledge base; Based on the key words and the word positions, an inverted index relationship is formed, wherein the inverted index relationship is used to represent the position relationship of the key words in the reply strategy; The key words are aggregated to form a semantic index relationship, wherein the semantic index relationship is used to index the key words with semantic correlation; The presentation type of the problem carried in the operation request includes a natural language type and / or a picture type, and after receiving the operation request initiated by the target chat room, the following steps are further included: According to a preset entity type, the problem with the natural language type is identified to obtain an identification result; The identification result is segmented to obtain target information; Before matching the problem carried in the operation request with the problem set in the preset knowledge base, the following steps are further included: Obtaining natural language information, a problem set and a strategy set in a historical time period; Based on the natural language information, the problem set and the strategy set, a semantic representation model and a semantic similarity model are trained; The step of matching the problem carried in the operation request with the problem set in the preset knowledge base comprises the following steps: Based on the inverted index relationship, the semantic index relationship and the semantic representation model, the target information is retrieved to obtain a retrieval result, wherein the retrieval result at least includes a matching degree of each problem in the problem set corresponding to the target information; Based on the retrieval result, a matching degree greater than a preset threshold is selected; The selected problem in the problem set indicated by the matching degree is represented as a problem cluster; The target problem is selected in the problem cluster.

2. The method of claim 1, wherein, The step of selecting the target problem in the problem cluster comprises the following steps: Based on a semantic similarity model, a similarity between each question in the question cluster and the target information is calculated; Based on a preset weight adjustment rule, the similarity is adjusted to obtain an adjusted similarity; All the adjusted similarities are sorted, and a question indicated by the highest similarity is selected to obtain the target question.

3. The method of claim 1, wherein, After receiving the operation and maintenance request initiated by the target chat room, further comprising: Through a preset upsampling rule, the questions of the picture type are upsampling processed to obtain a feature layer of the questions of the picture type; Based on the feature layer, a probability that each pixel in the questions of the picture type belongs to text is calculated to obtain a text probability map; Based on the dynamic threshold value of each pixel, a dynamic threshold value map is formed; Based on the text probability map and the dynamic threshold value map, a binary map is obtained; Based on the binary map, a text box is formed.

4. The method of claim 3, wherein, After forming the text box, further comprising: Based on four points on the text box, an inverse tangent calculation is performed to obtain an radian; The radian is converted into an angle, and the angle is rotated; Based on a preset text direction classifier and the rotated angle, the text box is corrected to obtain a horizontal matrix box; Based on a preset text recognizer, the horizontal matrix box is recognized to obtain a natural language type question.

5. The method of claim 1, wherein, After matching the question carried in the operation and maintenance request with the question set in the preset knowledge base, further comprising: In the case of unsuccessful matching, a request allocation task is initiated to allocate the operation and maintenance request to a responsible terminal, wherein the responsible terminal is used to process the operation and maintenance request.

6. An intelligent operation and maintenance device, characterized in that, Comprising: A first establishing module is configured to establish different types of chat rooms according to application types before receiving an operation and maintenance request initiated by a target chat room, wherein the chat room is used to display application environment information and operation and maintenance operation dynamics, and initiates an operation and maintenance request according to a preset interaction rule; a second establishing module is configured to establish a chat room group in combination with a plurality of chat rooms; different permission roles are set for users of each chat room to initiate different operation and maintenance requests; A receiving unit is configured to receive an operation and maintenance request initiated by a target chat room; A matching unit is configured to match a question carried in the operation and maintenance request with a question set in a preset knowledge base to filter out a question with the highest similarity between the question carried in the operation and maintenance request and the question set in the operation and maintenance request, and obtain a target question, wherein the preset knowledge base includes the question set and a strategy set, each reply strategy in the strategy set and a question in the question set have a pre-established index relationship; the index relationship includes making a special term dictionary of the operation and maintenance field, and establishing an index relationship using the special term dictionary; An output unit is configured to obtain a reply strategy corresponding to the target question based on the index relationship; An execution unit is configured to execute the reply strategy based on pre-statistical multidimensional environment information to obtain an execution result, and send the execution result to the target chat room. The apparatus further comprises a first extraction module configured to extract keyword and word position of each keyword in the preset knowledge base before matching the question carried in the operation and maintenance request with the question set in the preset knowledge base; a first formation module configured to form an inverted index relationship based on the keyword and the word position, wherein the inverted index relationship is used to represent the position relationship of the keyword in the reply strategy; and a second formation module configured to aggregate the keyword to form a semantic index relationship, wherein the semantic index relationship is used to index the keyword with semantic correlation. The presentation type of the question carried in the operation and maintenance request comprises a natural language type and / or a picture type, and the intelligent operation and maintenance apparatus further comprises a first identification module configured to identify the question with the natural language type according to a preset entity type after receiving the operation and maintenance request initiated by the target chat room to obtain an identification result; and a first segmentation module configured to segment the identification result to obtain target information. The apparatus further comprises a first acquisition module configured to acquire natural language information, a question set and a strategy set in a historical time period before matching the question carried in the operation and maintenance request with the question set in the preset knowledge base; and a first training module configured to train a semantic representation model and a semantic similarity model based on the natural language information, the question set and the strategy set. The matching unit comprises a first retrieval module configured to retrieve the target information based on the inverted index relationship, the semantic index relationship and the semantic representation model to obtain a retrieval result, wherein the retrieval result at least comprises a matching degree of each question in the question set corresponding to the target information; a first selection module configured to select the matching degree greater than a preset threshold based on the retrieval result; a first representation module configured to represent the question in the question set indicated by the selected matching degree as a question cluster; and a first selection module configured to select a target question in the question cluster.

7. A computer readable storage medium characterized in that, The computer readable storage medium comprises a stored computer program, wherein the computer program controls the device where the computer readable storage medium is located to perform the intelligent operation and maintenance method in any one of claims 1 to 5 when the computer program is running.

8. An electronic device, comprising: The apparatus comprises one or more processors and a memory configured to store one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the intelligent operation and maintenance method in any one of claims 1 to 5.

Images