Intelligent project management method and system based on pipelined record text

Abstract

The invention belongs to the technical field of project management, and relates to an intelligent project management method and system based on a pipelined record text. The method comprises the steps of extracting task information; determining a task category; attributing the task information to the workflow of the target project, and determining the project progress; determining a current project plan corresponding to the project progress; distributing the task information to the project member with the highest vector similarity; updating the actual progress of the project; and predicting the project completion probability. According to the method, the task information can be extracted according to the pipeline type record text, the project progress is determined by automatically affiliating the task information, the project progress is dynamically and instantly updated, and information lag is avoided; a current project plan corresponding to a project progress can be automatically matched, and project member allocation is performed on task information; the actual progress of the project can be updated, the project completion probability can be predicted, manual intervention is reduced, tasks are automatically extracted, analyzed and managed, and the management efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of project management, and in particular, relates to a project intelligent management method and system based on stream-type record text. Background Art

[0002] The efficiency and accuracy of project management are crucial to the operation of an enterprise. In the traditional project management process, the tracking and management of project progress relies on manual input and maintenance, and project members need to manually update task status, time nodes, and related events. This manual input and maintenance method has significant defects. On the one hand, manual operation is prone to information lag, and it is difficult for project managers to obtain the latest project progress in a timely manner, resulting in insufficient decision-making basis and affecting the timeliness of project advancement. On the other hand, in the manual processing process, the probability of missing key information is high. Once important tasks or time nodes are omitted from the record, it may cause a disconnection in the project links, seriously affecting the continuity of the project. At the same time, manual judgment is also prone to misjudgment, such as inaccurate assessment of the progress of task completion, which will mislead the overall resource allocation and subsequent planning of the project, ultimately leading to low overall management efficiency, increased project costs, and even possible failure risks for the project.

[0003] In addition, in daily work, a large amount of text data in the form of running records is generated, such as meeting minutes, work logs, email exchanges, etc. These text data contain rich project-related information, but the existing technology has failed to fully explore and utilize these valuable data resources in the project management process. The traditional storage method is just to simply record these data, lacking in-depth analysis and effective integration, so that this information cannot be transformed into key elements that help project management, resulting in a waste of data resources. Traditional project management methods have serious deficiencies in the efficiency and accuracy of progress tracking and management, and do not fully utilize the flow-based text data in daily work. Therefore, there is an urgent need for an innovative project management method that can effectively solve the above problems and improve the intelligence level and overall efficiency of project management. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a project intelligent management method and system based on stream-type record text.

[0005] In a first aspect, the present invention provides a project intelligent management method based on a stream-type record text, comprising: Use natural language models to analyze the stream-type record text, extract text data of target entities related to project management, and use named entity recognition methods to extract task information; text data includes task name, task node, time and task content; task information includes time, person, place and event content; Perform semantic analysis on task information, extract task information keywords, match task information keywords with task category keyword library, and determine task category; Based on the task category, combined with the identification of task information dependencies and execution order, the task information is attributed to the workflow of the target project, and the project progress is determined according to the node position corresponding to the workflow of the target project; Match the project progress with each project plan and determine the current project plan corresponding to the project progress; The feature vectors of task keywords and project member labels are extracted through natural language processing algorithms, and task keywords and project member labels are matched by vector similarity, and task information is assigned to the project member with the highest vector similarity. Match the time of task information with the time of task nodes in the current project plan to update the actual progress of the project; Predict the probability of project completion based on the actual progress of the project and the current project plan.

[0006] In a second aspect, the present invention provides a project intelligent management system based on a stream-type record text, including a task information extraction unit, a task classification unit, a project progress analysis unit, a project plan matching unit, a project member allocation unit, an actual progress update unit and a prediction unit; The task information extraction unit is used to analyze the stream-type record text using a natural language model, extract the text data of the target entity associated with project management, and extract the task information using the named entity recognition method; the text data includes the task name, task node, time and task content; the task information includes time, person, place and event content; A task classification unit is used to perform semantic analysis on task information, extract task information keywords, match task information keywords with a task category keyword library, and determine the task category; The project progress analysis unit is used to attribute the task information to the workflow of the target project based on the task category, combined with the identification of the dependency relationship and execution order of the task information, and determine the project progress according to the node position corresponding to the workflow of the target project; The project plan matching unit is used to match the project progress with each project plan and determine the current project plan corresponding to the project progress; A project member assignment unit is used to extract feature vectors of task keywords and project member labels through a natural language processing algorithm, match task keywords and project member labels through vector similarity, and assign task information to the project member with the highest vector similarity; The actual progress update unit is used to match the time of the task information with the time of the task node in the current project plan and update the actual progress of the project; The prediction unit is used to predict the probability of project completion based on the actual progress of the project and the current project plan.

[0007] Based on the above technical solution, the present invention can also be improved as follows.

[0008] Furthermore, the daily transaction record texts of project members are collected through API interfaces or regular data crawling; the daily transaction record texts are cleaned; text cleaning includes data format standardization and missing data filling and interpolation.

[0009] Furthermore, the named entity recognition method is used to extract task information, including: Determine the basic entity information corresponding to the task information; the basic entity information includes the task name, person in charge, deadline and priority; Split the continuous text data into independent word segmentation units and label each independent word segmentation unit; Vectorize the independent word segmentation units to obtain the word segmentation vectors; Build a pre-trained model based on the Transformer model, take the word segmentation vector as input, take the entity label corresponding to the basic entity information as output, add a custom NER layer, configure the loss function, set the learning rate and the number of training rounds, train the pre-trained model and monitor the validation set indicators to obtain the task information extraction model; the validation set indicators include F1 score and accuracy; The word segmentation vector corresponding to the text data is input into the task extraction model to obtain the entity label corresponding to the basic entity information, and the integrity of the entity label in the basic entity information corresponding to the task information is verified.

[0010] Further, semantic analysis is performed on the task information to extract task information keywords, and the task information keywords are matched with the task category keyword library to determine the task category, including: Collect project management task description samples, extract task category keywords, and perform vector representation to build a task category keyword library; task category keywords include meeting, development, testing, and delivery; Perform semantic analysis on task information, extract subject-verb-object structure through dependency syntactic analysis, identify actions and objects through semantic role labeling, and generate word vectors; The cosine similarity between the word vector corresponding to the task information and the word vector of the task category keyword is calculated, and the task category keyword with the largest cosine similarity is used as the task category corresponding to the task information.

[0011] Furthermore, based on the task category, combined with the identification of the dependencies and execution order of the task information, the task information is attributed to the workflow of the target project, including: collecting shared data flows between tasks, the conditions that need to be met for executing tasks and the association between the output of the preceding task and the input of the subsequent task, as well as adjacent tasks of the task information, performing feature matching with each node of the target project, and taking each node of the target project with the highest matching degree as the target node.

[0012] Furthermore, the project progress is matched with each project plan to determine the current project plan corresponding to the project progress, including: organizing all project plans, establishing a hierarchical relationship and association matrix between project plans, and setting a unique identifier for each project plan; setting key milestone nodes in each project plan; quantifying the project progress percentage and establishing a project progress evaluation standard; associating the project progress with the key milestone nodes in the project plan, and determining the project plan with the highest correlation to obtain the current project plan.

[0013] Furthermore, the actual progress of the project is updated by matching the time of the task information with the time of the task node of the current project plan, including: calculating the difference between the time of the task information and the time of the task node of the current project plan, determining the task node within the flexible time window within which the time of the task information is located, and obtaining the actual progress of the project.

[0014] Furthermore, based on the actual progress of the project and the current project plan, the probability of project completion is predicted, including: collecting historical data, calculating the time deviation coefficient, resource consumption rate and critical path deviation to construct a sample data set; constructing a probability prediction model, using a Monte Carlo simulation engine to train the probability prediction model, and inputting the data of the sample data set into the probability prediction model to obtain the probability of project completion; Assume the time deviation coefficient is The actual completion time is The planned completion time is , the total critical path duration is ,but: ; Assume that the resource consumption rate is , the actual resource usage is The total planned resources are , the process weight factor is ,but: ; Assume the critical path deviation is , the number of key processes completed is The total number of key processes is ,but: ; Assume that the probability of project completion is , , , and is the model parameter of the probability prediction model, then the probability prediction model is expressed as: .

[0015] Furthermore, after predicting the probability of project completion, task reminders are pushed to project members.

[0016] The beneficial effects of the present invention are as follows: the present invention can extract task information based on flow-type record text, determine project progress by automatically attributing task information, dynamically and instantly update project progress, and avoid information lag; it can automatically match the current project plan corresponding to the project progress and assign project members to task information; it can update the actual progress of the project and predict the probability of project completion, reduce manual intervention, realize automatic extraction, analysis and management of tasks, and improve management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 A schematic diagram of a project intelligent management method based on a stream-type record text provided in Example 1 of the present invention; Figure 2 A system block diagram of a project intelligent management system based on stream-type record text provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0018] Example 1 As an example, Figure 1 As shown, in order to solve the above technical problems, this embodiment provides a project intelligent management method based on a stream-type record text, including: Use natural language models to analyze the stream-type record text, extract text data of target entities related to project management, and use named entity recognition methods to extract task information; text data includes task name, task node, time and task content; task information includes time, person, place and event content; Perform semantic analysis on task information, extract task information keywords, match task information keywords with task category keyword library, and determine task category; Based on the task category, combined with the identification of task information dependencies and execution order, the task information is attributed to the workflow of the target project, and the project progress is determined according to the node position corresponding to the workflow of the target project; Match the project progress with each project plan and determine the current project plan corresponding to the project progress; The feature vectors of task keywords and project member labels are extracted through natural language processing algorithms, and task keywords and project member labels are matched by vector similarity, and task information is assigned to the project member with the highest vector similarity. Match the time of task information with the time of task nodes in the current project plan to update the actual progress of the project; Predict the probability of project completion based on the actual progress of the project and the current project plan.

[0019] The present invention can extract task information based on flow-type record text, determine project progress by automatically attributing task information, dynamically and instantly update project progress, and avoid information lag; it can automatically match the current project plan corresponding to the project progress and assign project members to task information; it can update the actual progress of the project and predict the probability of project completion, reduce manual intervention, realize automatic extraction, analysis and management of tasks, and improve management efficiency. Optionally, collect the daily transaction record text of project members through the API interface (Application Programming Interface) or regular data crawling; perform text cleaning on the daily transaction record text; text cleaning includes data format standardization and missing data filling and interpolation.

[0020] Specifically, data is collected through the API interface, such as: constructing an API request based on the HTTP request library, and after sending the request, receiving the response data returned by the API containing the text of daily transaction records.

[0021] The data format is standardized to JSON or XML format, etc. For missing values ​​of numeric fields, the mean, median, and mode are selected to fill them. If the data has time series characteristics, linear interpolation, polynomial interpolation, and other methods are used to estimate missing values.

[0022] Optionally, use named entity recognition to extract task information, including: Determine the basic entity information corresponding to the task information; the basic entity information includes the task name, person in charge, deadline and priority; Split the continuous text data into independent word segmentation units and label each independent word segmentation unit; Vectorize the independent word segmentation units to obtain the word segmentation vectors; Build a pre-trained model based on the Transformer model, take the word vector as input, take the entity label corresponding to the basic entity information as output, add a custom NER (Named Entity Recognition) layer, configure the loss function, set the learning rate and the number of training rounds, train the pre-trained model and monitor the validation set indicators to obtain the task information extraction model; the validation set indicators include F1 score and accuracy; The word segmentation vector corresponding to the text data is input into the task extraction model to obtain the entity label corresponding to the basic entity information, and the integrity of the entity label in the basic entity information corresponding to the task information is verified.

[0023] Build pre-trained models based on Transformer models, such as the BERT (Bidirectional Encoder Representations from Transformers) model and the RoBERTa (Robustly Optimized BERT Approach) model. These models are pre-trained on large-scale corpora and have language understanding capabilities. Add a custom named entity recognition layer, which is used to process the input word segmentation vectors and predict the entity label corresponding to each word segmentation. The learning rate is used to control the step size of the model parameter update during the training process.

[0024] Configure a loss function such as the cross entropy loss function to measure the difference between the model prediction result and the true label. The accuracy is the proportion of samples predicted correctly by the model to the total samples.

[0025] Optionally, semantic analysis is performed on the task information to extract keywords of the task information, and the keywords of the task information are matched with a task category keyword library to determine the task category, including: Collect project management task description samples, extract task category keywords, and perform vector representation to build a task category keyword library; task category keywords include meeting, development, testing, and delivery; Perform semantic analysis on task information, extract subject-verb-object structure through dependency syntactic analysis, identify actions and objects through semantic role labeling, and generate word vectors; The cosine similarity between the word vector corresponding to the task information and the word vector of the task category keyword is calculated, and the task category keyword with the largest cosine similarity is used as the task category corresponding to the task information.

[0026] In actual application, project management task description samples can come from different sources such as project documents, task assignment records, and work reports. Use word embedding models (such as Word2Vec (Word to Vector) model or GloVe (Global Vectors for Word Representation) or Transformer-based word vector model) to vectorize the extracted task category keywords. Word vectors map words to numerical vectors in a low-dimensional space, so that words with similar semantics are closer in the vector space, so that the semantic similarity between words can be measured by calculating the distance between vectors. Convert each task category keyword into a corresponding word vector and store it in the task category keyword library. Through semantic role labeling, the relationship between actions and objects in task information can be more accurately grasped, thereby gaining a deeper understanding of the semantics of the task.

[0027] The similarity between the two vectors is determined by calculating the cosine value of the angle between them, with a value range of -1 to 1. The closer the value is to 1, the more similar the two vectors are. After calculating all the cosine similarities, find the task category keyword with the largest cosine similarity to the task information word vector, and determine the task category corresponding to this task category keyword as the task category of the task information, so as to achieve the classification of the task information.

[0028] Optionally, based on the task category, combined with identifying the dependencies and execution order of the task information, the task information is attributed to the workflow of the target project, including: collecting shared data flows between tasks, the conditions that need to be met for executing tasks and the association between the output of the preceding task and the input of the subsequent task, as well as adjacent tasks of the task information, performing feature matching with each node of the target project, and taking each node of the target project with the highest matching degree as the target node.

[0029] Specifically, we analyze the task information of each task category in detail and sort out the data input and output involved in the execution of the task. For example, a data processing task may receive data from another data collection task as input. We record the flow and transmission relationship of these data to form a shared data flow between tasks.

[0030] According to the dependency and execution order of task information, find the adjacent tasks of each task information. Adjacent tasks can be predecessor tasks and successor tasks, and there are logical connections between adjacent tasks. For example, in a project, the predecessor task of the task "design product prototype" is "clarify product requirements", and the successor task is "conduct product testing".

[0031] The workflow of the target project includes the various stages of the project, the task nodes contained in each stage, and the relationship between the task nodes. Clarify the function, input requirements, output results, and execution conditions of each node. For example, the workflow of a software development project may include stages such as requirements analysis, design, coding, testing, and deployment, and each stage contains multiple specific nodes.

[0032] Match each feature of the task information (task category, input and output data, execution conditions and adjacent tasks, etc.) with the features of each node of the target project one by one. For task category, directly compare whether the task category to which the task information belongs is consistent or similar with the task category of the node; for input and output data, check whether the input data requirements and output data characteristics of the task information match the input and output of the node; for execution conditions, check whether the execution conditions of the task information match the execution conditions of the node; for adjacent tasks, analyze whether the adjacent task relationship of the task information is consistent with the previous and next task relationship of the node in the workflow.

[0033] Set a corresponding weight for each feature matching item, and calculate the matching score between the task information and each node according to the matching degree. For example, the weight of task category matching can be set higher because it is an important attribute of the task; the weight of input and output data matching is second; the weight of execution conditions and adjacent task matching is also set to appropriate values ​​according to their importance. By comprehensively calculating the matching scores of various features, the total matching degree between the task information and each node is obtained.

[0034] Optionally, match the project progress with each project plan to determine the current project plan corresponding to the project progress, including: organizing all project plans, establishing a hierarchical relationship and association matrix between project plans, and setting a unique identifier for each project plan; setting key milestone nodes in each project plan; quantifying the project progress percentage and establishing project progress evaluation standards; associating the project progress with the key milestone nodes in the project plan, and determining the project plan with the highest correlation to obtain the current project plan.

[0035] Optionally, the actual progress of the project is updated by matching the time of the task information with the time of the task node of the current project plan, including: calculating the difference between the time of the task information and the time of the task node of the current project plan, determining the task node within the flexible time window within which the time of the task information falls, and obtaining the actual progress of the project.

[0036] Optionally, based on the actual progress of the project and the current project plan, the probability of project completion is predicted, including: collecting historical data, calculating the time deviation coefficient, resource consumption rate and critical path deviation to construct a sample data set; constructing a probability prediction model, using a Monte Carlo simulation engine to train the probability prediction model, and inputting the data of the sample data set into the probability prediction model to obtain the probability of project completion; Assume the time deviation coefficient is The actual completion time is The planned completion time is , the total critical path duration is ,but: ; Assume that the resource consumption rate is , the actual resource usage is The total planned resources are , the process weight factor is ,but: ; Assume the critical path deviation is , the number of key processes completed is The total number of key processes is ,but: ; Assume that the probability of project completion is , , , and is the model parameter of the probability prediction model, then the probability prediction model is expressed as: .

[0037] Optionally, after predicting the probability of project completion, push task reminders to project members.

[0038] Notification systems (such as email, instant messaging apps, and calendar apps) push task reminders in real time.

[0039] Example 2 Based on the same principle as the method shown in Example 1 of the present invention, as shown in the attached Figure 2 As shown, the embodiment of the present invention also provides a project intelligent management system based on a stream-type record text, including a task information extraction unit, a task classification unit, a project progress analysis unit, a project plan matching unit, a project member allocation unit, an actual progress update unit and a prediction unit; The task information extraction unit is used to analyze the stream-type record text using a natural language model, extract the text data of the target entity associated with project management, and extract the task information using the named entity recognition method; the text data includes the task name, task node, time and task content; the task information includes time, person, place and event content; A task classification unit is used to perform semantic analysis on task information, extract task information keywords, match task information keywords with a task category keyword library, and determine the task category; The project progress analysis unit is used to attribute the task information to the workflow of the target project based on the task category, combined with the identification of the dependency relationship and execution order of the task information, and determine the project progress according to the node position corresponding to the workflow of the target project; The project plan matching unit is used to match the project progress with each project plan and determine the current project plan corresponding to the project progress; A project member assignment unit is used to extract feature vectors of task keywords and project member labels through a natural language processing algorithm, match task keywords and project member labels through vector similarity, and assign task information to the project member with the highest vector similarity; The actual progress update unit is used to match the time of the task information with the time of the task node in the current project plan and update the actual progress of the project; The prediction unit is used to predict the probability of project completion based on the actual progress of the project and the current project plan.

[0040] Optionally, collect the daily transaction record text of project members through API interface or regular data crawling; perform text cleaning on the daily transaction record text; text cleaning includes data format standardization and missing data filling and interpolation.

[0041] Optionally, use named entity recognition to extract task information, including: Determine the basic entity information corresponding to the task information; the basic entity information includes the task name, person in charge, deadline and priority; Split the continuous text data into independent word segmentation units and label each independent word segmentation unit; Vectorize the independent word segmentation units to obtain the word segmentation vectors; Build a pre-trained model based on the Transformer model, take the word segmentation vector as input, take the entity label corresponding to the basic entity information as output, add a custom NER layer, configure the loss function, set the learning rate and the number of training rounds, train the pre-trained model and monitor the validation set indicators to obtain the task information extraction model; the validation set indicators include F1 score and accuracy; The word segmentation vector corresponding to the text data is input into the task extraction model to obtain the entity label corresponding to the basic entity information, and the integrity of the entity label in the basic entity information corresponding to the task information is verified.

[0042] Optionally, semantic analysis is performed on the task information to extract keywords of the task information, and the keywords of the task information are matched with a task category keyword library to determine the task category, including: Collect project management task description samples, extract task category keywords, and perform vector representation to build a task category keyword library; task category keywords include meeting, development, testing, and delivery; Perform semantic analysis on task information, extract subject-verb-object structure through dependency syntactic analysis, identify actions and objects through semantic role labeling, and generate word vectors; The cosine similarity between the word vector corresponding to the task information and the word vector of the task category keyword is calculated, and the task category keyword with the largest cosine similarity is used as the task category corresponding to the task information.

[0043] Optionally, based on the task category, combined with identifying the dependencies and execution order of the task information, the task information is attributed to the workflow of the target project, including: collecting shared data flows between tasks, the conditions that need to be met for executing tasks and the association between the output of the preceding task and the input of the subsequent task, as well as adjacent tasks of the task information, performing feature matching with each node of the target project, and taking each node of the target project with the highest matching degree as the target node.

[0044] Compare the matching scores of the task information and each node of the target project to find the node with the highest matching degree. This node with the highest matching degree is used as the target node of the task information in the target project workflow, which means that the task information should be assigned to the location of this node and become part of the target project workflow.

[0045] Optionally, match the project progress with each project plan to determine the current project plan corresponding to the project progress, including: organizing all project plans, establishing a hierarchical relationship and association matrix between project plans, and setting a unique identifier for each project plan; setting key milestone nodes in each project plan; quantifying the project progress percentage and establishing project progress evaluation standards; associating the project progress with the key milestone nodes in the project plan, and determining the project plan with the highest correlation to obtain the current project plan.

[0046] Specifically, collect all ongoing or developed project plan documents. These project plan documents come from various project leaders and cover various projects, such as software development project plans, engineering project construction plans, and marketing activity plans, etc. Summarize the collected project plan documents and remove duplicate or invalid project plan documents.

[0047] Specifically, a correlation matrix is ​​created with project plans as rows and columns. The elements in the matrix are used to represent the degree of correlation between different project plans. The correlation relationship may include resource sharing, task dependency, and data interaction.

[0048] According to the task flow and time schedule of each project plan, find out the key milestone nodes that have a key impact on the success of the project and mark them in the corresponding project plan document, such as assigning a unique number to each milestone node so that it can be accurately identified in project progress tracking and correlation analysis. Choose appropriate indicators to measure project progress. Common indicators include the number of tasks completed, the proportion of workload completed, the proportion of project cost expenditure and the delivery of key results. Divide the project progress into different key milestone nodes, set the corresponding progress percentage range for each key milestone node, and calculate the current project progress percentage.

[0049] Optionally, the actual progress of the project is updated by matching the time of the task information with the time of the task node of the current project plan, including: calculating the difference between the time of the task information and the time of the task node of the current project plan, determining the task node within the flexible time window within which the time of the task information falls, and obtaining the actual progress of the project.

[0050] Specifically, if the actual start time or actual completion time of the task information is within the flexible time window of the task node, the task information is considered to match the task node. For example, if the flexible time window of the task node is between 9 and 11 days of the planned completion time, and the difference between the actual completion time and the planned completion time of the task information is within this range, then the task information is considered to match the task node. In actual application, the task node with the smallest time difference is preferentially selected as the matching object, or the matching task node is determined as the actual progress of the project based on the logical order and dependency of the tasks.

[0051] Optionally, based on the actual progress of the project and the current project plan, the probability of project completion is predicted, including: collecting historical data, calculating the time deviation coefficient, resource consumption rate and critical path deviation to construct a sample data set; constructing a probability prediction model, using a Monte Carlo simulation engine to train the probability prediction model, and inputting the data of the sample data set into the probability prediction model to obtain the probability of project completion; Assume the time deviation coefficient is The actual completion time is The planned completion time is , the total critical path duration is ,but: ; Assume that the resource consumption rate is , the actual resource usage is The total planned resources are , the process weight factor is ,but: ; Assume the critical path deviation is , the number of key processes completed is The total number of key processes is ,but: ; Assume that the probability of project completion is , , , and is the model parameter of the probability prediction model, then the probability prediction model is expressed as: .

[0052] Through multiple random sampling to simulate various possible project execution situations, the sample data set is input into the probability prediction model, and the model parameters are continuously adjusted. , , and , so that the project completion probability predicted by the probability prediction model is as close as possible to the actual completion status of historical projects, thereby determining the optimal model parameters.

[0053] Optionally, after predicting the probability of project completion, push task reminders to project members.

[0054] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. The project intelligent management method based on the stream-type record text is characterized by: include: Use natural language models to analyze the stream-type record text, extract text data of target entities related to project management, and use named entity recognition methods to extract task information; text data includes task name, task node, time and task content; task information includes time, person, place and event content; Perform semantic analysis on task information, extract task information keywords, match task information keywords with task category keyword library, and determine task category; Based on the task category, combined with the identification of task information dependencies and execution order, the task information is attributed to the workflow of the target project, and the project progress is determined according to the node position corresponding to the workflow of the target project; Match the project progress with each project plan and determine the current project plan corresponding to the project progress; The feature vectors of task keywords and project member labels are extracted through natural language processing algorithms, and task keywords and project member labels are matched by vector similarity, and task information is assigned to the project member with the highest vector similarity. Match the time of task information with the time of task nodes in the current project plan to update the actual progress of the project; Predict the probability of project completion based on the actual progress of the project and the current project plan.

2. According to claim 1, the project intelligent management method based on the running record text is characterized in that: Collect the daily transaction record text of project members through API interface or regular data crawling; perform text cleaning on the daily transaction record text; text cleaning includes data format standardization and missing data filling and interpolation.

3. The project intelligent management method based on the running record text according to claim 1 is characterized in that: Using named entity recognition methods, we extract task information, including: Determine the basic entity information corresponding to the task information; the basic entity information includes the task name, person in charge, deadline and priority; Split the continuous text data into independent word segmentation units and label each independent word segmentation unit; Vectorize the independent word segmentation units to obtain the word segmentation vectors; Build a pre-trained model based on the Transformer model, take the word segmentation vector as input, take the entity label corresponding to the basic entity information as output, add a custom NER layer, configure the loss function, set the learning rate and the number of training rounds, train the pre-trained model and monitor the validation set indicators to obtain the task information extraction model; the validation set indicators include F1 score and accuracy; The word segmentation vector corresponding to the text data is input into the task extraction model to obtain the entity label corresponding to the basic entity information, and the integrity of the entity label in the basic entity information corresponding to the task information is verified.

4. The project intelligent management method based on the streamlining record text according to claim 1 is characterized in that: Perform semantic analysis on task information, extract task information keywords, match task information keywords with task category keyword library, and determine task category, including: Collect project management task description samples, extract task category keywords, and perform vector representation to build a task category keyword library; task category keywords include meeting, development, testing, and delivery; Perform semantic analysis on task information, extract subject-verb-object structure through dependency syntactic analysis, identify actions and objects through semantic role labeling, and generate word vectors; The cosine similarity between the word vector corresponding to the task information and the word vector of the task category keyword is calculated, and the task category keyword with the largest cosine similarity is used as the task category corresponding to the task information.

5. The project intelligent management method based on the streamlining record text according to claim 1 is characterized in that: Based on the task category, combined with the identification of the dependencies and execution order of task information, the task information is attributed to the workflow of the target project, including: collecting shared data flows between tasks, the conditions that need to be met for executing tasks and the association between the output of the previous task and the input of the subsequent task, as well as the adjacent tasks of the task information, and performing feature matching with each node of the target project, and taking each node of the target project with the highest matching degree as the target node.

6. The project intelligent management method based on the running record text according to claim 1 is characterized in that: Match the project progress with each project plan to determine the current project plan corresponding to the project progress, including: organizing all project plans, establishing hierarchical relationships and association matrices between project plans, and setting a unique identifier for each project plan; setting key milestone nodes in each project plan; quantifying the project progress percentage and establishing project progress evaluation standards; associating the project progress with the key milestone nodes in the project plan, and determining the project plan with the highest degree of association to obtain the current project plan.

7. The project intelligent management method based on the running record text according to claim 1 is characterized in that: According to the matching of the time of the task information with the time of the task node of the current project plan, the actual progress of the project is updated, including: calculating the difference between the time of the task information and the time of the task node of the current project plan, determining the task node within the flexible time window in which the time of the task information is located, and obtaining the actual progress of the project.

8. The project intelligent management method based on the running record text according to claim 1 is characterized in that: According to the actual progress of the project and the current project plan, the probability of project completion is predicted, including: collecting historical data, calculating the time deviation coefficient, resource consumption rate and critical path deviation to build a sample data set; building a probability prediction model, using the Monte Carlo simulation engine to train the probability prediction model, and inputting the data of the sample data set into the probability prediction model to obtain the probability of project completion; Assume the time deviation coefficient is The actual completion time is The planned completion time is , the total critical path duration is ,but: ； Assume that the resource consumption rate is , the actual resource usage is The total planned resources are , the process weight factor is ,but: ； Assume the critical path deviation is , the number of key processes completed is The total number of key processes is ,but: ； Assume that the probability of project completion is , , , and is the model parameter of the probability prediction model, then the probability prediction model is expressed as: 。 9. The project intelligent management method based on the running record text according to claim 1 is characterized in that: After predicting the probability of project completion, push task reminders to project members.

10. The project intelligent management system based on the running record text is characterized by: It includes task information extraction unit, task classification unit, project progress analysis unit, project plan matching unit, project member allocation unit, actual progress update unit and prediction unit; The task information extraction unit is used to analyze the stream-type record text using a natural language model, extract the text data of the target entity associated with project management, and extract the task information using the named entity recognition method; the text data includes the task name, task node, time and task content; the task information includes time, person, place and event content; A task classification unit is used to perform semantic analysis on task information, extract task information keywords, match task information keywords with a task category keyword library, and determine the task category; The project progress analysis unit is used to attribute the task information to the workflow of the target project based on the task category, combined with the identification of the dependency relationship and execution order of the task information, and determine the project progress according to the node position corresponding to the workflow of the target project; The project plan matching unit is used to match the project progress with each project plan and determine the current project plan corresponding to the project progress; A project member assignment unit is used to extract feature vectors of task keywords and project member labels through a natural language processing algorithm, match task keywords and project member labels through vector similarity, and assign task information to the project member with the highest vector similarity; The actual progress update unit is used to match the time of the task information with the time of the task node in the current project plan and update the actual progress of the project; The prediction unit is used to predict the probability of project completion based on the actual progress of the project and the current project plan.

Images