A multi-model fusion development task association method, device, equipment and medium
By collecting and analyzing task report data in the collaborative development community, using multi-model fusion technology to calculate the similarity scores between task reports, automated development task association is achieved, and the problem of time-consuming manual association in the existing technology is solved, improving the problem of improved Task resolution efficiency.
Patent Information
- Application Number
- CN202111542359.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-13
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2041-12-13
AI Technical Summary
The inability of existing technology to achieve automated development task associations leads to the need for a lot of manual time and labor in the collaborative development community to find and associate related task reports, especially in large-scale software projects.
By building an active open source project set in the collaborative development community, collecting development task report data, extracting URL link information using regular expressions, combining structural data analysis model, text semantic representation model and historical correlation model, calculate the similarity between task reports Score, generate multi-model fusion development task correlation model, and realize automated task reporting correlation.
It realizes automated development task correlation, improves task resolution efficiency, reduces manual intervention, and can recommend related task reports for new tasks.
Smart Images

Figure CN114186974B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of software development, and in particular to a method, device, equipment and medium for associating development tasks using multi-model fusion. Background Art
[0002] Social coding was first proposed by the open source community GitHub, aiming to provide a developer-friendly software development environment to help developers connect, collaborate, and develop efficiently. The emergence of social programming has greatly enhanced code reuse and the efficiency of solving development tasks. Developers can participate in reporting and discussing tasks independently. In this way, task reports, as an important type of software development knowledge, are often reported by different developers at different times. In practice, two task reports often contain related information. Developers can link related task reports through URL links during task discussions. In a software project, discovering and linking related task reports can provide developers with more resources and information to solve the target tasks, thereby improving task solving efficiency.
[0003] Currently, in collaborative development communities such as GitHub, the method of associating task reports mainly relies on manual linking by developers. However, the linking process in the real world takes a lot of time and manpower. Especially for large-scale software projects, developers may need to search for a large amount of historical task data before they can locate related tasks through their text description information, and such manual association methods mainly rely on the experience and knowledge of individual developers. Therefore, how to achieve automated development task association has become a technical problem that needs to be solved urgently.
[0004] The above contents are only used to assist in understanding the technical solution of the present invention and do not constitute an admission that the above contents are prior art. Summary of the invention
[0005] The main purpose of the present invention is to provide a multi-model fusion development task association method, device, equipment and medium, aiming to solve the technical problem that the existing technology cannot realize automated development task association.
[0006] To achieve the above object, the present invention provides a multi-model fusion development task association method, the method comprising:
[0007] Build a collection of active open source projects in a collaborative development community based on pre-set metrics;
[0008] In the active open source project set, using an API to collect development task report data of all projects to build a candidate task report database;
[0009] Extracting URL link information in all task reports using regular expressions in the candidate task report database to generate a task report data set;
[0010] Constructing a query task data group and a candidate task data group in the task report data set, and obtaining a similarity score between the query task and each candidate task by using a structured data parsing model, a text semantic representation model, and a historical relevance model respectively;
[0011] The similarity scores between the query task and each candidate task are weighted and summed to obtain a final similarity score between each task report, and a development task association model based on multi-model fusion is constructed according to the final similarity score to generate a task report association tool.
[0012] Optionally, the step of building an active open source project set in the collaborative development community according to preset indicators includes:
[0013] In the collaborative development community GitHub, we use the API to collect basic project information and filter out popular open source projects based on Star, Fork, Delete, and Creation time indicators;
[0014] The selected popular open source projects are used to construct an active open source project set.
[0015] Optionally, the step of collecting development task report data of all projects in the active open source project set using an API to build a candidate task report database includes:
[0016] In the active open source project set, GitHub's Issue API and Pull Request (PR) API are used to collect task report data of all projects. The specific data collected includes: task ID, task processing status, submitter, task title, task description, task comment, submission time, category, label, milestone, etc.
[0017] An alternative task report database is constructed based on the collected report data.
[0018] Optionally, the step of extracting URL link information in all task reports in the candidate task report database using regular expressions to generate a task report data set includes:
[0019] In the candidate task report database, extracting URL link information in all task reports using regular expressions;
[0020] Using GitHub's Cross-referenced API to check URL link information in the task report, filter out actual task report association links, and build an association information benchmark library based on the task report association links;
[0021] The task report data not containing link information is removed from the association information reference library to form a final task report data set.
[0022] Optionally, before the step of constructing a query task data group and a candidate task data group in the task report data set, and respectively using a structured data parsing model, a text semantic representation model, and a historical relevance model to obtain a similarity score between the query task and each candidate task, the step further includes:
[0023] Extracting text data from each task report data in the task report dataset, including task report title, description and comments;
[0024] removing stop words, numbers, punctuation marks, and other non-alphabetic characters from the text data;
[0025] The remaining words are transformed into root forms using the Snowball Stemmer technique in NLTK to reduce the feature dimension and unify similar words into a common representation to obtain the preprocessed task report data.
[0026] Optionally, the step of constructing a query task data group and a candidate task data group in the task report data set, and obtaining a similarity score between the query task and each candidate task by using a structured data parsing model, a text semantic representation model, and a task association network model, respectively, comprises:
[0027] In the task report data set, the latest 40% samples are selected as the query task data set according to the creation time of the task report, and the task report data is used as the candidate task data set;
[0028] Use the structural data parsing model to calculate the similarity score of the structural information between the query task and each candidate task. S ;
[0029] Use the text semantic representation model to calculate the textual information similarity score between the query task and each candidate task T ;
[0030] Use the historical relevance model to calculate the historical information similarity score between the query task and each candidate task H .
[0031] Optionally, the step of performing weighted summation on the similarity scores between the query task and each candidate task to obtain a final similarity score between each task report, and constructing a development task association model based on multi-model fusion according to the similarity score to generate a task report association tool includes:
[0032] Performing weighted summation of similarity scores between the query task and each candidate task to obtain a final similarity score, and constructing a development task association model based on multi-model fusion according to the final similarity score;
[0033] The model is evaluated using the Top-k recall evaluation metric and the task report dataset.
[0034] According to the evaluation results, the optimal sub-model weight combination is selected to form a task report association tool.
[0035] In addition, to achieve the above purpose, the present invention also proposes a multi-model fusion development task association device, the device comprising:
[0036] A project building module, which is used to build a set of active open source projects in a collaborative development community based on preset indicators;
[0037] A data construction module, used for collecting development task report data of all projects in the active open source project set by using an API to construct a candidate task report database;
[0038] A link acquisition module, used to extract URL link information in all task reports in the candidate task report database using regular expressions to generate a task report data set;
[0039] A task calculation module is used to construct a query task data group and a candidate task data group in the task report data set, and respectively use a structured data parsing model, a text semantic representation model, and a historical relevance model to obtain a similarity score between the query task and each candidate task;
[0040] The tool generation module is used to perform weighted summation of the similarity scores between the query task and each candidate task to obtain a final similarity score between each task report, and to construct a development task association model based on multi-model fusion according to the final similarity score to generate a task report association tool.
[0041] In addition, to achieve the above-mentioned purpose, the present invention also proposes a computer device, which includes: a memory, a processor, and a multi-model fusion development task association program stored in the memory and executable on the processor, wherein the multi-model fusion development task association program is configured to implement the multi-model fusion development task association method as described above.
[0042] In addition, to achieve the above-mentioned purpose, the present invention also proposes a medium, on which a multi-model fusion development task association program is stored, and when the multi-model fusion development task association program is executed by a processor, the steps of the multi-model fusion development task association method described above are implemented.
[0043] The present invention constructs an active open source project set in a collaborative development community according to preset indicators; in the active open source project set, API is used to collect development task report data of all projects to construct an alternative task report database; regular expressions are used to extract URL link information in all task reports in the alternative task report database to generate a task report data set; a query task data group and a candidate task data group are constructed in the task report data set, and a similarity score between a query task and each candidate task is calculated by using a structural data parsing model, a text semantic representation model and a historical association model respectively; the similarity score between the query task and each candidate task is weighted and summed to obtain a final similarity score between each task report, and a development task association model based on multi-model fusion is constructed according to the final similarity score to generate a task report association tool, and recommendation of related task reports for new tasks is realized by combining with deep learning technology, and automatic development task association is realized by weighted summing of similarity scores to screen out the optimal weight and finally constructing a task report association tool. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 It is a structural diagram of a development task-associated device for multi-model fusion of a hardware operating environment involved in an embodiment of the present invention;
[0045] Figure 2 It is a flowchart of the first embodiment of the development task association method of multi-model fusion of the present invention. DETAILED DESCRIPTION
[0046] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0047] Reference Figure 1 , Figure 1 A schematic diagram of the structure of development task-associated equipment for multi-model fusion of a hardware operating environment involved in an embodiment of the present invention.
[0048] like Figure 1 As shown, the development task-associated device of the multi-model fusion may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (Wireless-Fidelity, WI-FI) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM) memory, or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.
[0049] Those skilled in the art will understand that Figure 1 The structure shown in does not constitute a limitation on the development task-associated device of multi-model fusion, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.
[0050] like Figure 1 As shown, the memory 1005 as a storage medium may include an operating system, a data storage module, a network communication module, a user interface module, and a multi-model fusion development task association program.
[0051] exist Figure 1 In the multi-model fusion development task association device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the multi-model fusion development task association device of the present invention can be set in the multi-model fusion development task association device, and the multi-model fusion development task association device calls the multi-model fusion development task association program stored in the memory 1005 through the processor 1001, and executes the multi-model fusion development task association method provided by the embodiment of the present invention.
[0052] The embodiment of the present invention provides a multi-model fusion development task association method, referring to Figure 2 , Figure 2 It is a flowchart of the first embodiment of the development task association method of multi-model fusion of the present invention.
[0053] In this embodiment, the multi-model fusion development task association method includes the following steps:
[0054] Step S10: Build an active open source project set in the collaborative development community according to preset indicators.
[0055] In the specific implementation, GitHub's API is used to collect popular open source projects that meet the following conditions: at least 10 Stars; forked at least once; not deleted and not a Fork project; the project was created after 2010 and before 2021; at the same time, active open source projects that meet the following conditions are screened out: at least 100 Issues or Pull Requests; at least 3 code contributors; development activities such as code submission, development task processing, contribution mergers, and submitted comments in the past 3 months.
[0056] Furthermore, the step of constructing an active open source project set in the collaborative development community according to preset indicators includes: in the collaborative development community GitHub, using the API to collect basic information data of the project, and screening out popular open source projects according to the Star, Fork, Delete and Creation time indicators; and constructing an active open source project set with the screened popular open source projects.
[0057] Step S20: In the active open source project set, the development task report data of all projects are collected by using API to build a candidate task report database.
[0058] In the specific implementation, based on the set of active open source projects, GitHub's Issue API is used to collect Issue report data of all projects. The specific data collection content includes: task ID, task processing status (open, closed), submitter, task title, task description, task comments, submission time, category ("0" represents Issue), label, milestone, etc.; GitHub's Pull request (PR) API is used to collect PR report data of all projects. The specific data collection content includes: task ID, task processing status (open, closed), submitter, task title, task description, task comments, submission time, category ("1" represents PR), label, milestone, etc.; the collection data results of Issue report data of all projects and PR report data of all projects are combined to build an alternative task report database.
[0059] Furthermore, the step of using the API to collect development task report data of all projects in the active open source project set to build an alternative task report database includes: using GitHub's Issue API and Pull Request (PR) API to collect task report data of all projects in the active open source project set, and the specific data collected is: task ID, task processing status, submitter, task title, task description, task comments, submission time, category, label, milestone, etc.; building an alternative task report database based on the collected report data.
[0060] Step S30: extracting URL link information in all task reports in the candidate task report database using regular expressions to generate a task report data set.
[0061] In the specific implementation, for the candidate task report dataset, the regular expression (github.com / [a-zA-Z0-9-_.] / [a-zA-Z0-9-_.] / issues|pull / [0-9]+)) is used to extract the URL link information in all task reports; GitHub's Cross-referenced API is used to check the link information in the task report, filter out the actual task report related links, and build a task related information benchmark library. In the specific work, only internal links of the project are considered, and cross-project task links are not considered for the time being. According to the task related information benchmark library, the task report data that does not contain link information is removed to form the final task report dataset;
[0062] Furthermore, the step of extracting URL link information from all task reports in the alternative task report database using regular expressions to generate a task report data set includes: extracting URL link information from all task reports in the alternative task report database using regular expressions; checking the URL link information in the task reports using GitHub's Cross-referenced API, screening out actual task report association connections, and building an association information benchmark library based on the task report association connections; removing task report data that does not contain link information from the association information benchmark library to form a final task report data set.
[0063] Step S40: constructing a query task data group and a candidate task data group in the task report data set, and respectively using a structural data parsing model, a text semantic representation model and a historical association model to obtain a similarity score between the query task and each candidate task.
[0064] Furthermore, before the step of constructing a query task data group and a candidate task data group in the task report data set, and respectively using a structural data parsing model, a text semantic representation model, and a historical association model to obtain a similarity score between the query task and each candidate task, the method further includes: extracting text data from each task report data in the task report data set, including the title, description, and comments of the task report; deleting stop words, numbers, punctuation marks, and other non-alphabetic characters from the text data; and converting the remaining words into root forms using the Snowball Stemmer technology in NLTK to reduce feature dimensions and unify similar words into a common representation to obtain preprocessed task report data.
[0065] In the specific implementation, the remaining words are converted to their root forms using the Snowball Stemmer technique in NLTK to reduce the feature dimension and unify similar words into a common representation.
[0066] Furthermore, the step of constructing a query task data group and a candidate task data group in the task report data set, and respectively using a structural data parsing model, a text semantic representation model, and a historical association model to obtain a similarity score between the query task and each candidate task includes: in the task report data set, selecting the latest 40% samples as the query task data group according to the creation time of the task report, and using the task report data as the candidate task data group; using the structural data parsing model to calculate the structural information similarity Score between the query task and each candidate task S ; Use the text semantic representation model to calculate the textual information similarity score between the query task and each candidate task T ; Use the historical relevance model to calculate the similarity score of the historical information between the query task and each candidate task H .
[0067] In the specific implementation, for each project, according to the creation time of the task report, the latest 40% samples are selected as the query task data group, and all task report data are candidate task data groups; the structured data of the task report is extracted and analyzed, and the structural data parsing model is used to calculate the structural information similarity score between the query task and each candidate task (the task creation time must be earlier than the query task). S. The structured data variables of the task report include: task processing state state (Boolean type, "0" represents open, "1" represents closed), submitter submitter (text type), category label type (Boolean type, "0" represents Issue, "1" represents PR), label label (text type), milestone milestone (text type), description complexity complexity (numeric type, the total number of words in the task report title and description), comment comment (numeric type, the number of comments), etc. Among them, for a specific text type variable X, it is necessary to summarize and remove duplicates of all X variable values in the task report data to obtain N different X values, and then encode different X values with natural numbers {1,…,N} in turn, and establish a one-to-one mapping between numerical values and X text values; for task reports with multiple labels, only the first label is selected as the analysis object. Then, for each task report, a feature vector {state,submitter,type,label,milestone,complexity,comment} is constructed to represent its structural information. For the structural information feature vector V of given two task reports s1 and V s2 , use cosine similarity to calculate their structural information similarity score Score S , the calculation method is as follows:
[0068]
[0069] Extract and analyze the text data of the task report, and use the text semantic representation model to calculate the text information similarity score between the query task and each candidate task (the task creation time must be earlier than the query task) T Based on the preprocessed task report document data (task title, task description and task comment), the Bert text representation model is used to calculate the text similarity score between the query task and each candidate task (the task creation time must be earlier than the query task). B . Call the BERTClient function in the BERT_serving.client library to extract features from each sentence in the task report. The representation vector dimension threshold can be set to 100, 200, 500, 1000, etc. For the text information feature vector V of the given two task reports t1 and V t2 , use cosine similarity to calculate their text information similarity score Score T , the calculation method is shown in the formula:
[0070]
[0071] According to the task report submitter information, extract all the task report data submitted by the submitter or participated in the comments, and use the historical relevance model to calculate the historical information similarity score between the query task and each candidate task (the task creation time must be earlier than the query task). H According to the submitter information of the task report, the submitter's historical participation (submission or participation in comment) task report ID is extracted from the task report dataset, and the task reports are arranged in reverse order according to the submission time. For each task report, a feature vector {id1, id2, ..., id n}. The dimension threshold of the representation vector can be set to 100, 200, 500, 1000, etc. If the submitter has not yet generated historical participation information, or the existing dimension is lower than the dimension threshold, the feature vector is filled with "0". Therefore, for the historical information feature vector V of the given two task reports h1 and V h2 , use cosine similarity to calculate their historical information similarity score Score H , the calculation method is shown in the formula:
[0072]
[0073] Step S50: performing weighted summation on the similarity scores between the query task and each candidate task to obtain a final similarity score between each task report, and constructing a development task association model based on multi-model fusion according to the final similarity score to generate a task report association tool.
[0074] Furthermore, the steps of weighted summing the similarity scores between the query task and each candidate task to obtain a final similarity score between each task report, and constructing a development task association model based on multi-model fusion according to the similarity score to generate a task report association tool include: weighted summing the similarity scores between the query task and each candidate task to obtain a final similarity score, and constructing a development task association model based on multi-model fusion according to the final similarity score; evaluating the model using a Top-k recall level evaluation indicator and the task report data set; and selecting the optimal sub-model weight combination according to the evaluation result to form a task report association tool.
[0075] In the specific implementation, the three sub-model similarity scores obtained above are weighted and summed to build a development task association model based on multi-model fusion, and a variety of evaluation indicators are used to evaluate the model, and the optimal sub-model weight combination is selected to form the final task report association tool. The specific implementation steps are as follows: The three similarity scores obtained in S5 are weighted and summed, and the three score weights are A, B, and C, respectively, so as to calculate the final similarity score Score between each task report pair. The calculation method is shown in the formula:
[0076] Score=A.Score s +B.Score T +C.Score H
[0077] The model is evaluated using the Top-k recall rate (R@k) evaluation index and the task-related information benchmark library. The purpose of R@k is to check whether the top-k recommendation results are correct. For the task report i to be queried, its R@k can be calculated as follows. In actual evaluation, k can be selected from 1 to 10:
[0078]
[0079] It can be understood that different weights (summed to be 1) are assigned to the above three scoring weights A, B, and C, respectively, and the query task report association results of all projects are evaluated according to the model evaluation indicators mentioned above, and the average values of R@1, R@5, and R@10 of all projects are calculated, and the three indicators are added together to form the final evaluation indicator. According to the final evaluation indicator, the optimal sub-model weight combination is selected, and on this basis, the three sub-models are combined to form the final task report association tool.
[0080] This embodiment builds an active open source project set in a collaborative development community according to preset indicators; in the active open source project set, the development task report data of all projects are collected using an API to build an alternative task report database; in the alternative task report database, the URL link information in all task reports is extracted using regular expressions to generate a task report data set; a query task data group and a candidate task data group are constructed in the task report data set, and a similarity score between the query task and each candidate task is calculated using a structural data parsing model, a text semantic representation model, and a historical association model respectively; the similarity scores between the query task and each candidate task are weighted and summed to obtain a final similarity score between each task report, and a development task association model based on multi-model fusion is constructed according to the final similarity score to generate a task report association tool, and the recommendation of related task reports for new tasks is realized by combining with deep learning technology, and the automated development task association is realized by weighted summing of similarity scores to screen out the optimal weight and finally constructing a task report association tool.
[0081] In addition, an embodiment of the present invention further proposes a medium on which a multi-model fusion development task association program is stored. When the multi-model fusion development task association program is executed by a processor, the steps of the multi-model fusion development task association method described above are implemented.
[0082] The embodiments or specific implementation methods of the multi-model fusion development task association device of the present invention can refer to the above-mentioned method embodiments, which will not be repeated here.
[0083] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.
[0084] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.
[0085] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory / random access memory, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.
[0086] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. A method for developing task association by multi-model fusion, characterized in that: The method comprises: Build a collection of active open source projects in a collaborative development community based on pre-set metrics; In the active open source project set, using an API to collect development task report data of all projects to build a candidate task report database; Extracting URL link information in all task reports in the candidate task report database using regular expressions to generate a task report data set; Constructing a query task data group and a candidate task data group in the task report data set, and obtaining a similarity score between the query task and each candidate task by using a structured data parsing model, a text semantic representation model, and a historical relevance model respectively; Performing weighted summation on the similarity scores between the query task and each candidate task to obtain a final similarity score between each task report, and constructing a development task association model based on multi-model fusion according to the final similarity score to generate a task report association tool; The step of constructing a query task data group and a candidate task data group in the task report data set, and obtaining a similarity score between the query task and each candidate task by using a structured data parsing model, a text semantic representation model, and a historical relevance model, respectively, comprises: In the task report data set, the latest 40% samples are selected as the query task data set according to the creation time of the task report, and the task report data is used as the candidate task data set; Use the structural data parsing model to calculate the structural information similarity score between the query task and each candidate task S ; Use the text semantic representation model to calculate the text information similarity Score between the query task and each candidate task T ; Use the historical relevance model to calculate the historical information similarity Score between the query task and each candidate task H .
2. The method according to claim 1, characterized in that The steps of building an active open source project set in a collaborative development community according to preset indicators include: In the collaborative development community GitHub, we use the API to collect basic project information and filter out popular open source projects based on Star, Fork, Delete, and Creation time indicators; The selected popular open source projects are used to construct an active open source project set.
3. The method according to claim 1, characterized in that The step of collecting development task report data of all projects in the active open source project set using an API to build a candidate task report database includes: In the active open source project set, GitHub's Issue API and Pull Request API are used to collect task report data of all projects. The specific data collected includes: task ID, task processing status, submitter, task title, task description, task comment, submission time, category, label, milestone, etc. An alternative task report database is constructed based on the collected report data.
4. The method according to claim 1, characterized in that The step of extracting URL link information in all task reports using regular expressions in the candidate task report database to generate a task report data set includes: In the candidate task report database, extracting URL link information in all task reports using regular expressions; Using GitHub's Cross-referenced API to check URL link information in the task report, filter out actual task report association links, and build an association information benchmark library based on the task report association links; The task report data not containing link information is removed from the association information reference library to form a final task report data set.
5. The method according to claim 1, characterized in that Before the step of constructing a query task data group and a candidate task data group in the task report data set, and obtaining a similarity score between the query task and each candidate task by using a structured data parsing model, a text semantic representation model, and a historical relevance model, the method further includes: Extracting text data from each task report data in the task report dataset, including task report title, description and comments; removing stop words, numbers, punctuation marks, and other non-alphabetic characters from the text data; The remaining words are transformed into root forms using the Snowball Stemmer technique in NLTK to reduce the feature dimension and unify similar words into a common representation to obtain the preprocessed task report data.
6. The method according to claim 1, characterized in that The step of performing weighted summation of the similarity scores between the query task and each candidate task to obtain a final similarity score between each task report, and constructing a development task association model based on multi-model fusion according to the similarity score to generate a task report association tool includes: Performing weighted summation of similarity scores between the query task and each candidate task to obtain a final similarity score, and constructing a development task association model based on multi-model fusion according to the final similarity score; The model is evaluated using the Top-k recall evaluation metric and the task report dataset. According to the evaluation results, the optimal sub-model weight combination is selected to form a task report association tool.
7. A multi-model fusion development task association device, used to implement the multi-model fusion development task association method according to any one of claims 1 to 6, characterized in that: The device comprises: A project building module, which is used to build a set of active open source projects in a collaborative development community based on preset indicators; A data construction module, used for collecting development task report data of all projects in the active open source project set by using an API to construct a candidate task report database; A link acquisition module, used to extract URL link information in all task reports in the candidate task report database using regular expressions to generate a task report data set; A task calculation module is used to construct a query task data group and a candidate task data group in the task report data set, and respectively use a structured data parsing model, a text semantic representation model, and a historical relevance model to obtain a similarity score between the query task and each candidate task; The tool generation module is used to perform weighted summation of the similarity scores between the query task and each candidate task to obtain a final similarity score between each task report, and to construct a development task association model based on multi-model fusion according to the final similarity score to generate a task report association tool.
8. A multi-model fusion development task association device, characterized in that: The device includes: a memory, a processor, and a multi-model fusion development task association program stored in the memory and executable on the processor, wherein the multi-model fusion development task association program is configured to implement the steps of the multi-model fusion development task association method as described in any one of claims 1 to 6.
9. A medium, characterized in that The medium stores a multi-model fusion development task association program, which, when executed by a processor, implements the steps of the multi-model fusion development task association method as described in any one of claims 1 to 6.
Citation Information
Patent Citations
Event matching by analysis of text characteristics (e-match)
US10108697B1
Service for providing item recommendations
WO2008121737A1
Digital information capture and retrieval
WO2019008394A1
