Report generation method and device, electronic equipment, storage medium and program product

Abstract

The application discloses a report generation method and device, electronic equipment, a storage medium and a program product, and belongs to the technical field of computers. The method comprises the following steps: based on a received report generation instruction, determining a target report generation requirement for a target object; obtaining a fusion vector corresponding to the target report generation requirement, wherein the fusion vector is used to represent the fusion characteristics of target business data and a target clustering result corresponding to the target business data, the target business data is business data corresponding to the target object and meeting the target report generation requirement, and the target clustering result is obtained by performing multi-layer semantic clustering on the target business data; and based on the fusion vector, an analysis report corresponding to the target object is generated.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a report generation method, apparatus, electronic device, storage medium, and program product. Background Technology

[0002] With the development of the digital economy, the number of dedicated lines opened by government and enterprise customers has surged, and customers are paying increasing attention to business operation service reports. Traditional report generation technologies mainly fall into two categories: one is the manual writing mode, where operations and maintenance personnel manually collect information such as customer dedicated line service lists, traffic data, and fault records by logging into multiple independent systems such as the comprehensive data system, IP workbench, and EOMS, and then compile them into reports according to fixed templates; the other is a simple automated generation technology, which can output reports by calling data from a single system and applying preset templates. Both of these methods suffer from the problem of low accuracy in the generated report content. Summary of the Invention

[0003] This application provides a report generation method, apparatus, electronic device, storage medium, and program product that can improve the accuracy of the generated report content.

[0004] To solve the above-mentioned technical problems, this application is implemented as follows: Firstly, a report generation method is provided, the method comprising: determining a target report generation requirement for a target object based on a received report generation instruction; obtaining a fusion vector corresponding to the target report generation requirement, wherein the fusion vector is used to characterize the fusion features of target business data and the target clustering result corresponding to the target business data, the target business data being the business data corresponding to the target object that meets the target report generation requirement, and the target clustering result being obtained by performing multi-level semantic clustering on the target business data; and generating an analysis report corresponding to the target object based on the fusion vector.

[0005] Secondly, a report generation apparatus is provided, comprising: a determining module, configured to determine a target report generation requirement for a target object based on a received report generation instruction; an acquiring module, configured to acquire a fusion vector corresponding to the target report generation requirement, wherein the fusion vector is used to characterize the fusion features of target business data and the target clustering result corresponding to the target business data, the target business data being business data corresponding to the target object that meets the target report generation requirement, and the target clustering result being obtained by performing multi-level semantic clustering on the target business data; and a generating module, configured to generate an analysis report corresponding to the target object based on the fusion vector.

[0006] Thirdly, an electronic device is provided, including a processor and a memory, wherein the memory stores a program or instructions executable on the processor, the program or instructions, when executed by the processor, perform the steps of the method described in the first aspect.

[0007] Fourthly, a readable storage medium is provided, on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0008] Fifthly, a computer program product is provided, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions that, when executed by a computer, cause the computer to perform the steps of the method described in the first aspect.

[0009] In this embodiment, based on the received report generation instruction, the target report generation requirement for the target object is determined, and the fusion vector corresponding to the target report generation requirement is obtained. The fusion vector is used to characterize the fusion features of the target business data and the target clustering result corresponding to the target business data. The target business data is the business data corresponding to the target object that meets the target report generation requirement. The target clustering result is obtained by performing multi-layer semantic clustering on the target business data. Based on the fusion vector, the report corresponding to the target object is generated. In this way, semantic clustering reduces data noise and can effectively improve the accuracy and relevance of the report content. At the same time, based on the fusion vector, the potential correlation between data such as faults and traffic can be mined, thereby further improving the accuracy of the report content.

[0010] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0011] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0012] Figure 1 This application shows a schematic flowchart of a report generation method provided in an exemplary embodiment. Figure 2 This illustration shows a schematic diagram of the construction process of a training dataset provided in an exemplary embodiment of this application; Figure 3 This invention provides a schematic diagram of the structure of a report generation system according to an exemplary embodiment of the present application. Figure 4 This application shows a schematic diagram of a data processing flow provided by an exemplary embodiment; Figure 5 An interactive schematic diagram of a report generation method provided in an exemplary embodiment of this application is shown; Figure 6 This invention provides a schematic diagram of the structure of a report generation apparatus according to an exemplary embodiment of the present application. Figure 7 A schematic diagram of the structure of an electronic device provided in an exemplary embodiment of this application is shown. Detailed Implementation

[0013] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0014] Figure 1 This illustration shows a flowchart of a report generation method provided in an exemplary embodiment of this application. The method can be executed by an electronic device, which may include a terminal or a network-side device. In other words, the method can be executed by a large language model installed on an electronic device. In some embodiments, the training dataset used to train the large language model can be configured according to... Figure 2 The process shown is used for construction. In some embodiments, the model parameters of the large language model can be optimized using LoRA fine-tuning technology. It should be noted that the training and application processes of this large language model adhere to multiple legal and compliant principles, including legal data sources, compliant data content, compliant data governance, compliant training objectives and schemes, compliant training processes, compliant training environment and tools, and compliant ethical verification of training results, and comply with relevant patent law requirements. Among these: (1) Legality of data sources: The training datasets used in the large language model were all obtained through legal means, covering three categories: publicly authorized data, data authorized by partners, and self-collected compliant data. Among them, the publicly authorized data comes from compliant data sources that comply with relevant open source agreements, and the copyright ownership and scope of authorization have been fully marked, with no unauthorized open source code or data reuse behavior; the data authorized by partners has a formal data use agreement that clearly defines the scope of data use, term and confidentiality obligations, and has a complete authorization chain proof; for the self-collected data, the informed consent procedure has been strictly followed for the parts involving personal information, and the personally identifiable information has been removed through de-identification processing (including but not limited to field masking, feature anonymization, and differential privacy technology application), which fully complies with the requirements of relevant laws and regulations.

[0015] (2) Data Content Compliance: The training dataset used by the large language model undergoes multiple screenings and cleaning processes to remove all non-compliant information and does not involve any illegal acquisition or use of genetic resources. For sensitive data (such as medical and financial data), a privacy computing module (including federated learning and secure multi-party computation technology) is used to ensure that the data is "usable but not visible," avoiding compliance risks during the original data transmission process and ensuring that the data application scenarios and uses comply with public order and good morals and industry regulatory requirements.

[0016] (3) Data Governance Standardization: A complete data traceability system is established during the training process of the large language model to automatically record the source, collection time, annotation process, cleaning rules and permission allocation of the training data, and generate traceable compliance reports to ensure that the data is verifiable throughout its entire lifecycle. The dataset annotation process of the large language model is completed by a professional human R&D team, which clarifies the proportion of human creative contributions, avoids relying on AI-generated data that has not been substantially modified by humans, and complies with the relevant examination requirements for AI patent applications.

[0017] (3) Compliance of training objectives and schemes: The training objective of the large language model is focused on report generation. The training scheme and the final output results do not violate any mandatory provisions of laws and administrative regulations, do not harm the public interest or the legitimate rights and interests of others, and do not pose any potential risks of being used for illegal activities, privacy infringement, or public safety damage. The ethical principle of "intelligent for good" is strictly practiced.

[0018] (4) Compliance of training process: A closed-loop training framework is adopted to achieve compliance and controllability of the training process. The specific process is as follows: First, training samples are obtained through compliant data sources. After the aforementioned data cleaning and desensitization, they are input into the neural network model to generate preliminary training results. Second, an expert system is introduced to verify the preliminary results. Based on preset rules and human expert experience, the feasibility of the results is evaluated, and outputs that may have ethical risks or compliance hazards are corrected (such as removing decision logic that violates public order and good morals and adjusting model parameters that do not comply with safety regulations). Finally, the weight of the loss function is dynamically optimized according to the feedback from the expert system to strengthen the model's learning of compliant results, avoid overfitting errors or non-compliant labels, and form a closed-loop control of "data input - model training - expert verification - parameter optimization - result feedback" to ensure that the entire training process complies with the ethical review requirements of relevant laws.

[0019] (5) Compliance of Training Environment and Tools: The training of the large language model is implemented based on nationally licensed chips and a compliant training platform. All open-source frameworks and components used in the training process have obtained corresponding authorizations, and copyright statements and patent citation information are fully retained, with no infringement or reuse. The training environment is constructed using virtual devices (containers / virtual machines), with fixed random seeds and initial parameter configurations to ensure the reproducibility of the training process. At the same time, through access control, operation log recording, and other methods, risks such as data leakage and parameter tampering during the training process are prevented, ensuring the security and compliance of the training process.

[0020] (6) Ethical verification and compliance of training results: After the model is trained, it is further verified through third-party ethical compliance assessment and algorithm filing review to ensure that the model output results do not violate social morality or harm public interests. For sensitive scenarios that may be involved, a special result verification mechanism is established to ensure that the model always complies with relevant laws and regulations in practical applications.

[0021] In summary, the data and training process used in the large language model in this application strictly comply with relevant regulations and do not violate laws, social ethics, public interests, or the use of genetic resources. Therefore, it fully meets the compliance requirements for patent authorization.

[0022] like Figure 1 As shown, the report generation method may include the following steps: S110: Based on the received report generation instructions, determine the target report generation requirements for the target object.

[0023] The target task instruction can be sent based on user interaction (such as clicking, swiping, voice input, etc.) or automatically generated based on preset operation and maintenance strategies or condition-triggered strategies (such as a certain indicator exceeding a set threshold). After receiving the report generation instruction, the large language model will respond to and parse the instruction, extracting the operated object, i.e., the target object, and the target report generation requirement from the instruction.

[0024] S120: Obtain the fusion vector corresponding to the target report generation requirement.

[0025] The fusion vector is used to characterize the fusion features of the target business data and the target clustering result corresponding to the target business data. The target business data is the business data corresponding to the target object that meets the requirements for generating the target report. The target clustering result is obtained by performing multi-level semantic clustering on the target business data.

[0026] In this step, multi-layer semantic clustering effectively reduces noise in the target business data, thereby minimizing the impact of data noise on subsequent processing. Based on this, the analysis report generated from the denoised target business data more accurately reflects the true situation and key features, effectively improving the accuracy and reliability of the report results.

[0027] In some embodiments, the target business data includes structured and unstructured data. In other embodiments, the structured data is collected by calling system API interfaces such as government and enterprise operation and maintenance systems, integrated resource management systems, IP workbenches, and EOMS, and may include dedicated line service lists, service traffic, fault information, critical protection records, etc. In still other embodiments, the unstructured data may include text content such as operation and maintenance work orders and fault case libraries extracted through OCR technology.

[0028] S130: Based on the fusion vector, generate an analysis report corresponding to the target object.

[0029] Understandably, by deeply integrating target business data with its corresponding target clustering structure, it is possible to uncover potential correlations between data such as faults and traffic, thereby significantly improving the accuracy of report content.

[0030] In this embodiment, based on the received report generation instruction, the target report generation requirement for the target object is determined, and the fusion vector corresponding to the target report generation requirement is obtained. The fusion vector is used to characterize the fusion features of the target business data and the target clustering result corresponding to the target business data. The target business data is the business data corresponding to the target object that meets the target report generation requirement. The target clustering result is obtained by performing multi-layer semantic clustering on the target business data. Based on the fusion vector, the report corresponding to the target object is generated. In this way, semantic clustering reduces data noise and can effectively improve the accuracy and relevance of the report content. At the same time, based on the fusion vector, the potential correlation between data such as faults and traffic can be mined, thereby further improving the accuracy of the report content.

[0031] In some embodiments, S120 may include the following steps: S122: Determine the retrieval information corresponding to the target report generation requirement, wherein the retrieval information includes business type, data dimension, and semantic keywords.

[0032] S124: If a fusion vector corresponding to the target report generation requirement is matched from the database based on the retrieval information, the fusion vector is extracted; or, if no fusion vector corresponding to the target report generation requirement is matched from the database based on the retrieval information, multi-level semantic clustering is performed on the collected target business data to obtain the target clustering result, and feature fusion is performed on the target business data and the target clustering result to obtain the fusion vector.

[0033] Understandably, based on this retrieval information, a search and matching process is performed in the database. If a fusion vector corresponding to the target report generation requirement can be successfully retrieved, then the fusion vector is directly extracted; if no corresponding fusion vector is found in the database, then the fusion vector needs to be determined. In this embodiment, by reusing the existing processing results in the database, redundant calculations and processing can be effectively avoided, thereby effectively improving the efficiency of business report generation; at the same time, the fusion vector is determined only when the database lacks data, ensuring the integrity and reliability of the data required for report generation.

[0034] In some embodiments, the service type may include data transmission leased line, Internet leased line, etc., the data dimension may include fault data, traffic data, complaint data, etc., and the semantic keywords may include other keywords such as fault cause, optimization suggestion type, etc.

[0035] In some embodiments, fusing features of the target business data and the target clustering result to obtain the fusion vector may include: mapping the feature vectors corresponding to the structured data and the feature vectors corresponding to the unstructured data included in the target business data to a multi-dimensional space to obtain a target feature representation; assigning weights to the target feature representation and the target clustering result using a self-attention mechanism; and performing weighted fusion of the target feature representation and the target clustering result to obtain a fusion vector. For example, the fusion vector can be represented by the following formula: Where Vstruct represents the target feature representation, and Vsemantic represents the target clustering result. The weights represent the target feature representations. The weights representing the target clustering results, for example, , .

[0036] In some embodiments, performing multi-level semantic clustering on the collected target business data to obtain the target clustering result includes: clustering the target business data according to the business type, the data dimension, and semantic similarity respectively; and taking the clustering result with a similarity greater than a preset threshold to the target report generation requirement as the target clustering result.

[0037] It is understandable that this multi-layer semantic clustering includes three layers: (1) First-level clustering: Divided into data transmission leased lines and Internet leased lines according to business type, with the number of clusters set to 2.

[0038] (2) Second-level clustering: Divided into fault data, traffic data and complaint data according to data type, with the number of clusters set to 3.

[0039] (3) Third-level clustering: Classified by semantic similarity, such as multiple semantic clusters such as fault cause and optimization suggestion type.

[0040] In some embodiments, after collecting target business data, the structured data included in the target business data undergoes a first preprocessing step, and the unstructured data included in the target business data undergoes a second preprocessing step. The first preprocessing includes, but is not limited to, at least one of the following: inspection, cleaning (e.g., Z-Score method for outlier removal), standardization, deduplication, and feature engineering. The second preprocessing includes, but is not limited to, at least one of the following: text segmentation, word segmentation, entity recognition (e.g., extracting key entities such as object names and fault types), and converting image data to text via image OCR.

[0041] In other embodiments, the target business data is clustered according to the business type, the data dimension, and semantic similarity, respectively, including: generating feature vectors corresponding to preprocessed structured data and unstructured data, respectively; and clustering the feature vectors according to the business type, the data dimension, and semantic similarity, respectively.

[0042] In this embodiment, after clustering is completed, redundant data can be filtered out by calculating the similarity between each clustering result and the target report generation requirement, thereby determining the most relevant target clustering result. In some embodiments, this similarity can be cosine similarity, which can be expressed by the following formula: , where a represents the target report generation requirement and b is the center vector of the cluster.

[0043] In some embodiments, after taking the clustering results with a similarity greater than a preset threshold to the target report generation requirement as the target clustering result, the method further includes: constructing a three-level index based on the business type, the data dimension, and semantic keywords, wherein the semantic keywords are determined by clustering according to the semantic similarity; and associating the fusion vector with the index information corresponding to the three-level index and storing it in the database.

[0044] It is understood that this application embodiment constructs a three-level index of "business type - data dimension - semantic keywords" and stores the fused vector and index information in the database. This allows for the retrieval of relevant data from the database upon receiving a report generation instruction. In some embodiments, the retrieval can be performed using the F query function: Where T represents the target report generation requirement, K represents the index keywords, and the F function returns the relevant fusion vector through cosine similarity matching.

[0045] In some embodiments, generating an analysis report corresponding to the target object based on the fusion vector includes: parsing the target report generation requirements to generate at least one subtask; matching each subtask in an instruction set and a prompt word set to obtain the execution instructions and prompt words corresponding to the subtask; and generating an analysis report corresponding to the target object according to a target template corresponding to the target report generation requirements based on the execution instructions and prompt words corresponding to each subtask.

[0046] Understandably, the target report generation requirements are structured and broken down into multiple independently executable subtasks. For each subtask, a precise match is performed against a pre-defined set of instructions and prompts to obtain the corresponding execution instructions and prompts. Then, based on these instructions and prompts, each subtask is completed sequentially or in parallel, thereby generating the complete target report. In this embodiment, the task decomposition and instruction / prompt matching mechanism significantly improves the automation and flexibility of report generation, reduces reliance on manually written instructions, and effectively enhances the accuracy of the large language model's understanding of industry requirements. Simultaneously, since the instructions and prompts are pre-defined and standardized, consistency between different subtasks is ensured, reducing logical conflicts or content duplication, thereby improving the efficiency and quality of report generation.

[0047] In some embodiments, before generating the analysis report corresponding to the target object based on the fusion vector, the method further includes: constructing a prompt word set. Constructing the prompt word set may include: determining multiple initial prompt words according to the business scenario, wherein the multiple initial prompt words may include fault analysis prompt words (Gprompt), traffic monitoring prompt words (Lprompt), complaint-related prompt words (Zprompt), general logic prompt words (Pprompt), etc.; optimizing the initial prompt words based on historical operation and maintenance Q&A pairs and related text blocks to ensure the professionalism of the prompt words.

[0048] In other embodiments, before generating the analysis report corresponding to the target object based on the fusion vector, the method further includes: constructing an instruction set. The instruction set may include: data query instructions, analysis and reasoning instructions, report generation instructions, etc.

[0049] In some embodiments, execution instructions and prompts can be matched using cosine similarity to select those with high relevance.

[0050] In some embodiments, an instruction parsing function can be used to match the execution instruction with the prompt word. This instruction parsing function can be: Where M is the subtask to be broken down, E is the instruction set or prompt word set, and the F function returns the list of matching instructions or prompt words.

[0051] In some embodiments, before generating an analysis report corresponding to the target object based on the execution instructions and prompts for each subtask and according to the target template corresponding to the target report generation requirement, the method may further include: obtaining a target template corresponding to the target report generation requirement. That is, selecting a template corresponding to the target report generation requirement from a template library, wherein the template library includes standardized templates such as basic information, network health assessments, and optimization suggestions. In some embodiments, the target template may include one or more standardized modules.

[0052] In some embodiments, generating an analysis report corresponding to the target object based on the execution instructions and prompts for each subtask and according to a target template corresponding to the target report generation requirements may include: filling the target template with analysis data corresponding to the fusion vector through a large model processing function. Where D is the analysis result corresponding to the fusion vector, Mb is the target template, and the LLM function generates the standardized report content.

[0053] In some embodiments, the analysis results include the execution results of each subtask. That is, for each subtask, its corresponding execution instructions, prompt words, and fusion vectors are input into the large language model, which then executes the task to obtain the execution result corresponding to that subtask.

[0054] In some embodiments, determining the target report generation requirement for a target object based on the received report generation instruction includes: in response to receiving an initial report generation instruction input by a user, determining a fuzzy object and a fuzzy report generation requirement; providing the user with a list of candidate objects based on the fuzzy object, and providing the user with a requirement explanation question based on the fuzzy report generation requirement; receiving the target object selected by the user based on the list of candidate objects; receiving the content of the user's response to the requirement explanation question, and determining the target report generation requirement based on the content of the response.

[0055] Understandably, the user establishes a session with the large language model through a client dialog box, inputting an initial report generation command. This command carries a fuzzy object and a fuzzy report generation requirement. The large language model calls the precise customer query API, returning a list of customers (including customer IDs) corresponding to the fuzzy object—the candidate object list. The user then selects a precise customer, which becomes the target object. Simultaneously, since the user's report generation requirement is also fuzzy, the large language model generates related requirement explanation questions based on this fuzzy requirement and sends them back to the user for answering. Based on the user's response, the precise target report generation requirement is then determined.

[0056] In this embodiment, relevant information is confirmed through multi-round question-and-answer sessions, thereby achieving dynamic capture of personalized needs and solving the problem of insufficient adaptability of traditional solutions.

[0057] Based on the above embodiments, this application also provides a report generation system, such as... Figure 3 As shown, it may include the following modules: data acquisition module 310, semantic clustering and noise reduction module 320, data fusion module 330, index construction module 340, scenario-based prompt word library module 350, dedicated line instruction set module 360, large language model interactive control module 370, and report generation and optimization module 380.

[0058] The data acquisition module 310 is used to perform the following steps: Step 1: Collect business data. By calling the API interfaces of systems such as government and enterprise operation and maintenance systems, integrated resource management systems, IP workbench, and EOMS, collect structured data such as dedicated line business lists, business traffic, fault conditions, and critical protection records of key customers; and extract unstructured data text content such as front-line operation and maintenance work orders and fault case libraries through OCR technology.

[0059] Step 2: Data Preprocessing. For structured data, after loading, inspection, cleaning (Z-Score method to remove outliers), standardization, deduplication, and feature engineering, a standardized structured dataset is output. For unstructured data, text segmentation, word segmentation, entity recognition (extracting key entities such as customer names and fault types) are performed, and image data is converted into text using image OCR to generate an initial text dataset. For example, such as... Figure 4 The diagram shown is a schematic of the data processing flow.

[0060] The semantic clustering noise reduction module 320 is used to perform the following steps: Step 1: Text Vector Generation. The initial text dataset is converted into high-dimensional semantic vectors using a BERT pre-trained model, and the structured data is used to generate feature vectors through an embedding layer.

[0061] Step 2: Multi-level semantic clustering. Based on the improved semantic clustering algorithm, the text vectors are clustered hierarchically: (1) First-level clustering: Divided into data transmission leased lines and Internet leased lines according to business type, with the number of clusters set to 2.

[0062] (2) Second-level clustering: Divided into fault data, traffic data and complaint data according to data type, with the number of clusters set to 3.

[0063] (3) Third-level clustering: Classify according to semantic similarity, such as fault cause, optimization suggestion type, etc., generate hundreds of semantic clusters, and output the core semantic dataset after noise reduction.

[0064] Step 3: Clustering result filtering. Semantic clusters with a relevance of ≥0.7 to the task instructions are retained by calculating cosine similarity, and redundant data is filtered out.

[0065] Data fusion module 330 is used to perform the following steps: Step 1: Vector Alignment. This involves mapping the feature vectors of structured and unstructured data to a unified multidimensional space.

[0066] Step 2: Semantic fusion. A self-attention mechanism is used to assign weights to data from different sources, generating a fusion vector.

[0067] The index building module 340 is used to: build a three-level index of "business type-data dimension-semantic keyword", store the fused vector and index information in the vector database; and retrieve relevant fused vectors through the F query function after receiving task instructions.

[0068] The scenario-based prompt word library module 350 is used to: determine prompt words according to the dedicated line business scenario. These prompt words can include fault analysis prompt words (Gprompt), traffic monitoring prompt words (Lprompt), complaint-related prompt words (Zprompt), and general logic prompt words (Pprompt); then, optimize the prompt words by combining the operation and maintenance Q&A pairs summarized in actual production with relevant text blocks to ensure the professionalism of the prompt words; finally, match the task instructions and prompt words through cosine similarity, giving priority to the prompt words with the highest relevance, and calling Pprompt by default when no match is found.

[0069] The dedicated line instruction set module 360 ​​is used to generate multiple execution instructions, including data query instructions, analysis and reasoning instructions, report generation instructions, etc., and then match tasks and instructions through the instruction parsing function.

[0070] The large model interactive control module 370 is used to: respond to the user establishing a session with the large language model through a client dialog box and inputting fuzzy task commands; the large language model calls the precise customer query API and returns a list of customers (including customer IDs) corresponding to the fuzzy names; the user then selects and confirms the precise customers.

[0071] The report generation and optimization module 380 is used to: select the corresponding template based on business type and reporting requirements. The template library includes standardized modules such as basic information, network health assessment, and optimization suggestions, supporting flexible combinations. Then, data is populated into the template through a large language model processing function, and finally, an analysis report is generated and pushed to the client, allowing users to download it directly.

[0072] In this embodiment, a semantic clustering algorithm is used to process multi-source heterogeneous operation and maintenance data in the communications industry, reducing data noise interference and improving the accuracy and relevance of report content. Then, a prompt word library and instruction set adapted to the communication leased line business scenario are constructed, and customer needs are dynamically obtained by combining a streaming question-and-answer interaction mechanism. Furthermore, the semantic features of data from multiple systems are integrated, and the efficiency of report generation is improved and the cost of manual intervention is reduced through model fine-tuning and process optimization, while enhancing the accuracy and professionalism of report generation.

[0073] The following is through Figure 5 Regarding the above Figure 3 The report generation system shown is illustrated as an example. For example, it is assumed that the client is a government or enterprise operations and maintenance system. Figure 5 As shown, the following steps may be included: S510: Users input commands on the government and enterprise operation and maintenance platform to establish a session with the large language model.

[0074] S520: The large language model platform returns a precise customer list to users through streaming question answering.

[0075] S530: Users select target customers based on a customer list.

[0076] S540: Push the selected target customers to the large language model platform via the flow rate question and answer.

[0077] S550: The large language model platform selects the corresponding report template directory based on the business type of the target customer.

[0078] S560: The large language model platform pushes the corresponding report catalog information to users through streaming question answering.

[0079] S570: Users select the target report directory based on the report directory.

[0080] S580: Pushes the selected target report directory to the large language model platform via flow rate question answering.

[0081] S590: The large language model platform calls the government and enterprise report generation API interface to push report generation information.

[0082] S595: Generate and output reports and provide downloads.

[0083] Figure 6 This illustration shows a schematic diagram of the structure of a report generation apparatus provided in one embodiment of this application, as shown below. Figure 6 As shown, the report generation device 600 may include: a determining module 610, an acquiring module 620, and a generating module 630.

[0084] In this embodiment, the determining module 610 is used to determine the target report generation requirement for the target object based on the received report generation instruction; the obtaining module 620 is used to obtain the fusion vector corresponding to the target report generation requirement, wherein the fusion vector is used to characterize the fusion feature of the target business data and the target clustering result corresponding to the target business data, the target business data is the business data corresponding to the target object that meets the target report generation requirement, and the target clustering result is obtained by performing multi-layer semantic clustering on the target business data; and the generating module 630 is used to generate an analysis report corresponding to the target object based on the fusion vector.

[0085] In some embodiments, the acquisition module 620 is specifically used to: determine the retrieval information corresponding to the target report generation requirement, wherein the retrieval information includes business type, data dimension, and semantic keywords; extract the fusion vector if a fusion vector corresponding to the target report generation requirement is matched from the database based on the retrieval information; or, if a fusion vector corresponding to the target report generation requirement is not matched from the database based on the retrieval information, perform multi-level semantic clustering on the collected target business data to obtain the target clustering result, and perform feature fusion on the target business data and the target clustering result to obtain the fusion vector.

[0086] In some embodiments, the acquisition module 620 is specifically used to: cluster the target business data according to the business type, the data dimension, and the semantic similarity, respectively; and take the clustering result with a similarity greater than a preset threshold to the target report generation requirement as the target clustering result.

[0087] In some embodiments, the device further includes a storage module for: constructing a three-level index based on the business type, the data dimension, and semantic keywords, wherein the semantic keywords are determined by clustering according to the semantic similarity; and associating the fusion vector with the index information corresponding to the three-level index and storing it in the database.

[0088] In some embodiments, the generation module 630 is specifically used to: parse the target report generation requirement and generate at least one subtask; match each subtask in the instruction set and the prompt word set respectively to obtain the execution instruction and prompt word corresponding to the subtask; and generate an analysis report corresponding to the target object according to the target template corresponding to the target report generation requirement based on the execution instruction and the prompt word corresponding to each subtask.

[0089] In some embodiments, the acquisition module 610 is specifically configured to: in response to receiving an initial report generation instruction input by a user, determine a fuzzy object and a fuzzy report generation requirement; provide the user with a list of candidate objects based on the fuzzy object, and provide the user with an initial requirement explanation question based on the fuzzy report generation requirement; receive the target object selected by the user based on the list of candidate objects; receive the content of the user's reply to the requirement explanation question, and determine the target report generation requirement based on the content of the reply.

[0090] The report generation apparatus provided in this application embodiment can implement the various processes implemented in the method embodiment of the above-described report generation method. To avoid repetition, it will not be described again here.

[0091] The report generation device in this application embodiment can be a device, or it can be a component, integrated circuit, or chip in an electronic device. This application embodiment does not impose specific limitations.

[0092] One report generation device in this application embodiment can be a device with an operating system. The operating system can be Android, iOS, or other possible operating systems; this application embodiment does not specifically limit the specific operating system.

[0093] like Figure 7 As shown in the figure, this application embodiment also provides an electronic device 700, including a processor 710 and a memory 720. The memory 720 stores a program or instructions that can run on the processor 710. When the program or instructions are executed by the processor 710, they implement the various processes of the above-described report generation method embodiment and can achieve the same technical effect. To avoid repetition, they will not be described again here.

[0094] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described report generation method embodiments and achieve the same technical effects. To avoid repetition, they will not be described again here.

[0095] The processor mentioned above is the processor in the terminal described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk. In some examples, the readable storage medium may be a non-transient readable storage medium.

[0096] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described report generation method embodiment and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0097] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0098] This application also provides a computer program / program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, which, when executed by a computer, cause the computer to perform the various processes of the above-described report generation method embodiment and achieve the same technical effect. To avoid repetition, these will not be described again here.

[0099] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0100] From the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of computer software products plus necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware. The computer software product is stored in a storage medium (such as ROM, RAM, magnetic disk, optical disk, etc.) and includes several instructions to cause the terminal or network-side device to execute the methods described in the various embodiments of this application.

[0101] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other implementations under the guidance of this application without departing from the spirit and scope of the claims. All of these implementations are within the protection scope of this application.

Claims

1. A report generation method, characterized in that, The method includes: Based on the received report generation instructions, determine the target report generation requirements for the target object; Obtain the fusion vector corresponding to the target report generation requirement, wherein the fusion vector is used to characterize the fusion feature of the target business data and the target clustering result corresponding to the target business data, the target business data is the business data corresponding to the target object that meets the target report generation requirement, and the target clustering result is obtained by performing multi-level semantic clustering on the target business data; Based on the fusion vector, an analysis report corresponding to the target object is generated.

2. The method according to claim 1, characterized in that, The step of obtaining the fusion vector corresponding to the target report generation requirement includes: Determine the retrieval information corresponding to the target report generation requirement, wherein the retrieval information includes business type, data dimension, and semantic keywords; If a fusion vector corresponding to the target report generation requirement is matched from the database based on the retrieval information, the fusion vector is extracted; or, if no fusion vector corresponding to the target report generation requirement is matched from the database based on the retrieval information, multi-level semantic clustering is performed on the collected target business data to obtain the target clustering result, and feature fusion is performed on the target business data and the target clustering result to obtain the fusion vector.

3. The method according to claim 2, characterized in that, The step of performing multi-level semantic clustering on the collected target business data to obtain the target clustering result includes: The target business data is clustered according to the business type, the data dimension, and the semantic similarity, respectively. Clustering results whose similarity to the target report generation requirements is greater than a preset threshold are taken as the target clustering results.

4. The method according to claim 3, characterized in that, After selecting clustering results with a similarity greater than a preset threshold to the target report generation requirement as the target clustering result, the method further includes: Based on the business type, the data dimension, and the semantic keywords, a three-level index is constructed, wherein the semantic keywords are determined by clustering according to the semantic similarity. The fusion vector is associated with the index information corresponding to the third-level index and stored in the database.

5. The method according to claim 1, characterized in that, The step of generating an analysis report corresponding to the target object based on the fusion vector includes: The target report generation requirements are parsed to generate at least one sub-task; Each subtask is matched against the instruction set and the prompt word set to obtain the execution instruction and prompt word corresponding to the subtask; Based on the execution instructions and prompts corresponding to each subtask, an analysis report corresponding to the target object is generated according to the target template corresponding to the target report generation requirements.

6. The method according to claim 1, characterized in that, The process of determining the target report generation requirements for the target object based on the received report generation instructions includes: In response to receiving the initial report generation instruction from the user, determine the fuzzy object and the fuzzy report generation requirement; Based on the fuzzy object, a list of candidate objects is fed back to the user, and based on the fuzzy report generation requirement, an explanation of the initial requirement is fed back to the user. Receive the target object selected by the user based on the candidate object list; Receive the user's response to the question regarding the requirement, and determine the target report generation requirement based on the response.

7. A task processing device, characterized in that, include: The determination module is used to determine the target report generation requirements for the target object based on the received report generation instructions; The acquisition module is used to acquire the fusion vector corresponding to the target report generation requirement, wherein the fusion vector is used to characterize the fusion feature of the target business data and the target clustering result corresponding to the target business data, the target business data is the business data corresponding to the target object that meets the target report generation requirement, and the target clustering result is obtained by performing multi-level semantic clustering on the target business data; The generation module is used to generate an analysis report corresponding to the target object based on the fusion vector.

8. An electronic device, characterized in that, It includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the method as described in any one of claims 1-6.

9. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions that, when executed by a computer, cause the computer to perform the steps of the method as described in any one of claims 1 to 6.

Images