A method for intelligent recommendation of enterprise activities based on large models

Abstract

The present invention discloses a method for intelligently recommending enterprise activities based on a big model, and the method steps are as follows: step one: data collection; step two: data processing; step three: prompt word construction; step four: recommendation generation, the data collection content includes the enterprise's recent activity history data and feedback data of participating in the activity, the present invention constructs activity recommendation generation prompt words based on three types of data: the enterprise's recent activity history data, the feedback data of participating in the activity, and the novel activity data in the current society, thereby providing rich contextual content for the big model to generate high-quality activity recommendations; the present invention uses the big model to extract key information from the three types of data: the enterprise's recent activity history data, the feedback data of participating in the activity, and the novel activity data in the current society, thereby reducing information redundancy and improving the quality of the input content of the big model under the condition of limited context window.

Description

Technical Field

[0001] The present invention relates to the field of intelligent recommendation technology, and in particular to an enterprise activity intelligent recommendation method based on a large model. Background Art

[0002] Enterprise activities are the product of the combination of ideological and political work with the specific reality of the enterprise, and are an important part of enterprise management.

[0003] In recent years, enterprises at all levels have been constantly exploring new forms of activities, and "smart activities" are a key focus of enterprise work. However, judging from the practice of "smart activities" in enterprises, the current "smart activities" are basically still in the informationization stage of work. The so-called "smart activities" are more about providing more scientific decision-making support for work through the analysis and mining of large amounts of data, or using interactive electronic screens for knowledge promotion. The effect is questionable and there is still a big gap from "smart activities". On the other hand, the increasing formalism of enterprises has seriously affected their development, and enterprise work urgently needs innovation;

[0004] In recent years, large model technologies represented by ChatGPT, Llama, GLM, etc. have flourished, making all walks of life seem to see the dawn of artificial intelligence. Large models, namely large language models (LLM), refer to deep learning models with large-scale parameters and complex computing structures. Many studies have shown that large models improve the expressiveness and predictive performance of models and can handle more complex tasks and data. Large models learn complex patterns and features by training massive data, have stronger generalization capabilities, and can make accurate predictions on unseen data. Currently, large models are widely used in various fields, including natural language processing, computer vision, speech recognition, and recommendation systems.

[0005] At present, enterprises are facing many new problems, new tasks and new requirements. Enterprises should continue to innovate and develop while deeply integrating and helping business operations. The emergence and gradual development of big model technology has brought new directions for innovation and breakthroughs in work. Using big model technology to innovate work, make work more intelligent, and further promote the high-quality development of enterprises is the trend of the times and a development opportunity that enterprises at all levels need to seize. Therefore, it is necessary to design an intelligent recommendation method for enterprise activities based on big models. Summary of the invention

[0006] The purpose of the present invention is to provide a method for intelligently recommending enterprise activities based on a large model to solve the problems raised in the above-mentioned background technology.

[0007] Activities are an important part of corporate work. At present, corporate activities are mainly planned and organized by responsible personnel. With the gradual increase in the number of activities, the activities are becoming more and more conservative and formalistic, and the effect of activities is often greatly reduced. How to carry out more attractive and effective activities has also become a problem faced by event leaders.

[0008] Large models trained with massive amounts of data generally have the capabilities of knowledge integration, content understanding, logical reasoning, problem solving, content generation, etc. Large models that exceed a certain scale will emerge with unexpected complex capabilities and characteristics that small models do not have, demonstrating human-like thinking and intelligence. This "emergence" ability has a certain degree of creativity and has been applied in many fields such as artistic creation, financial risk assessment, and drug research and development.

[0009] There is no doubt that the powerful intelligence of the big model can be used to optimize the organization and management of activities. By integrating and analyzing the company's recent activity history data, feedback from participating in activities, and new activities in the current society, the big model can provide recommendations for corporate activities to enhance the attractiveness and effectiveness of activities, and provide personalized activity recommendations for individuals to enhance participation.

[0010] The main goal of this patent is to build an intelligent recommendation method for corporate activities based on a big model, which aims to integrate three types of data: historical data of corporate activities in recent years, feedback from participants in activities, and novel activities in the current society. It uses the information understanding, logical reasoning, content generation and other capabilities of the big model to intelligently recommend optimized activity plans for the company, reduce the burden on corporate personnel, enhance the attractiveness and effectiveness of activities, promote high-quality work, and help the development of the company.

[0011] To achieve the above object, the present invention provides the following technical solutions:

[0012] A method for intelligent recommendation of enterprise activities based on a large model, the method steps are as follows:

[0013] Step 1: Data collection;

[0014] Step 2: Data processing;

[0015] Step 3: Construction of prompt words;

[0016] Step 4: Recommendation generation.

[0017] As a further solution of the present invention: the data collection content includes the company's recent activity history data, feedback data from participating in activities, and novel activity data in the current society.

[0018] As a further solution of the present invention: the collection of the enterprise's recent activity history data refers to the collection of activity plans, activity records and news reports of the enterprise's recent activities.

[0019] As a further solution of the present invention: the feedback data collection of participation activities refers to the collection of feedback information after participating in corporate activities, such as collected activity evaluation surveys, written personal experiences, and proposed activity suggestions.

[0020] As a further solution of the present invention: the collection of data on novel activities in the current society refers to collecting information on activities that are relatively novel and popular in the society in recent years from the Internet.

[0021] As a further solution of the present invention: the data processing is used to shorten the length of each piece of data so that more pieces of data can be added to the prompt word, allowing the large model to obtain more contextual knowledge, thereby facilitating the large model to generate activity recommendations that better meet user expectations.

[0022] As a further solution of the present invention: the prompt word construction refers to the activity recommendation prompt word construction, and the activity recommendation prompt word construction method steps are as follows:

[0023] S1: Select the historical data of the enterprise's activities within N years, assuming N is 1, and record the historical data as D;

[0024] S2: Select the feedback data of the enterprise participating in the activities within N years, assuming that N is also 1, and record the feedback data as F;

[0025] S3: Select Y novel activity data in society within X years, assuming X is 1, Y is 10, and record the activity data as N;

[0026] S4: The three types of data D, F and N are integrated to construct prompt words for activity recommendations according to user expectations.

[0027] As a further solution of the present invention: when constructing the activity recommendation prompt words, it is necessary to control the selection of three types of data D, F and N to avoid exceeding the input context window size of the large model.

[0028] As a further solution of the present invention: the step 4 is used to pass the constructed activity recommendation prompt words into the large model to obtain the generated activity recommendation.

[0029] As a further solution of the present invention: in the step 1, the enterprise's recent activity history data, feedback data of participating in activities and novel activity data in the current society all need to be deduplicated when collected.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] 1. The present invention constructs activity recommendation generation prompt words based on three types of data: recent activity history data of enterprises, feedback data of participating activities, and novel activity data in the current society, providing rich contextual content for large models to generate high-quality activity recommendations;

[0032] 2. The present invention uses a large model to extract key information from three types of data: the company's recent activity history data, feedback data from participating in activities, and novel activity data in the current society, thereby reducing information redundancy and improving the quality of the input content of the large model when the context window is limited. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 Flowchart of the steps for the intelligent recommendation method for enterprise activities based on big models. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] See also Figure 1 In an embodiment of the present invention, a method for intelligently recommending enterprise activities based on a large model has the following steps:

[0036] Step 1: Data collection;

[0037] The data collection content includes the company's recent activity history data, feedback data from participating in activities, and new activity data in the current society. The company's recent activity history data collection refers to the collection of activity plans, activity records, news reports, etc. of the company's recent activities (such as within half a year or within a year). For companies with a high degree of work informatization, these activity data generally exist in the work system, and the work system can be connected to collect relevant data. For companies with a low degree of work informatization, it is necessary to build a collection system to collect relevant data through manual entry or web crawlers. For example, many corporate departments will conduct news reports on the company's official website, WeChat public account, internal corporate community and other channels after the event. The data source is relatively fixed, and full-time personnel can be arranged to collect data irregularly or set up web crawlers to regularly capture relevant data. The role of the company's recent activity history data is to let the big model know the company's recent activities and avoid recommending duplicate activity plans;

[0038] The collection of feedback data from participating in activities refers to the collection of feedback information after participating in corporate activities, such as collected activity evaluation surveys, written personal experiences, proposed activity suggestions, etc. The collection of feedback data from participating in activities can be done in the same way as the collection of historical data of recent corporate activities. The role of feedback data from participating in activities is to provide optimization direction for the large model to generate activity recommendations;

[0039] The collection of novel social activity data refers to collecting information about relatively novel and popular social activities in recent years from the Internet. This type of data can be collected by searching general search engines such as Baidu, Bing, and Google, or crawled by web crawlers. The role of novel social activity data is to provide a reference for the large model to generate innovative activity recommendations;

[0040] Step 2: Data processing;

[0041] The data processing is used to shorten the length of each piece of data so that more pieces of data can be added to the prompt word, allowing the large model to obtain more contextual knowledge, thereby facilitating the large model to generate activity recommendations that better meet user expectations.

[0042] After the three types of data are collected, they need to be passed to the big model for content summary, extracting the main information from the data, and filtering out invalid information. Since the big model has a length limit for input, when constructing the prompt word, it is necessary to provide the big model with as much useful information as possible within a certain length range;

[0043] Step 3: Construction of prompt words;

[0044] Prompt construction refers to the construction of activity recommendation prompts. Prompts refer to input text used to guide the large model to generate outputs of a specific type or style. Prompts directly affect the model's understanding and generated content, so the design of prompts is crucial to the output quality and relevance of the model. The present invention constructs activity recommendation prompts based on three types of data: recent activity history data of the enterprise, feedback data from participating in activities, and novel activity data in the current society. The steps of the activity recommendation prompt construction method are as follows:

[0045] S1: Select the historical data of the enterprise's activities within N years, assuming N is 1, and record the historical data as D;

[0046] S2: Select the feedback data of the enterprise participating in the activities within N years, assuming that N is also 1, and record the feedback data as F;

[0047] S3: Select Y novel activity data in society within X years, assuming X is 1, Y is 10, and record the activity data as N;

[0048] S4: The three types of data D, F and N are integrated to construct prompt words for activity recommendations according to user expectations.

[0049] When constructing the activity recommendation prompt words, it is necessary to control the selection of three types of data D, F and N to avoid exceeding the input context window size of the large model. If the prompt word length exceeds the input context window size of the large model, it is necessary to appropriately reduce the use of the three types of data.

[0050] Step 4: Recommendation generation.

[0051] The constructed activity recommendation prompt words are passed into the big model to obtain the generated activity recommendations.

[0052] In the step 1, the company's recent activity history data, feedback data from participating in activities, and new activity data in the current society all need to be deduplicated when collected to avoid data redundancy.

[0053] References

[0054] [1] Zhang Qi, Gui Tao, Zheng Rui, Huang Xuanjing. Introduction and Practice of Large Language Models[M]. Shanghai: Fudan University Press, 2023.

[0055] [2] You Yang. Practical AI Big Model[M]. Beijing: Publishing House of Electronics Industry, 2023.

[0056] [3] Wang Xiaohua. Large model development and fine-tuning from scratch: based on PyTorch and ChatGLM[M]. Beijing: Publishing House of Electronics Industry, 2023.

[0057] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for intelligent recommendation of enterprise activities based on a large model, characterized by: The method steps are as follows: Step 1: Data collection; Step 2: Data processing; Step 3: Construction of prompt words; Step 4: Recommendation generation.

2. The method for intelligent recommendation of enterprise activities based on a large model according to claim 1, characterized in that: The data collection content includes the company's recent activity history data, feedback data from participating in activities, and novel activity data in the current society.

3. The method for intelligent recommendation of enterprise activities based on a large model according to claim 2, characterized in that: The collection of the enterprise's recent activity history data refers to the collection of activity plans, activity records and news reports of the enterprise's recent activities.

4. The method for intelligent recommendation of enterprise activities based on a large model according to claim 2, characterized in that: The feedback data collection of participation activities refers to the collection of feedback information after participating in corporate activities, such as collected activity evaluation surveys, written personal experiences, and proposed activity suggestions.

5. The method for intelligent recommendation of enterprise activities based on a large model according to claim 2, characterized in that: The novel activity data collection in the current society refers to collecting the novel and popular activity information in the society in recent years from the Internet.

6. The method for intelligent recommendation of enterprise activities based on a large model according to claim 1, characterized in that: The data processing is used to shorten the length of each piece of data so that more pieces of data can be added to the prompt word, allowing the large model to obtain more contextual knowledge, thereby facilitating the large model to generate activity recommendations that better meet user expectations.

7. The method for intelligently recommending enterprise activities based on a large model according to claim 1, characterized in that: The prompt word construction refers to the activity recommendation prompt word construction, and the activity recommendation prompt word construction method steps are as follows: S1: Select the historical data of the enterprise's activities within N years, assuming N is 1, and record the historical data as D; S2: Select the feedback data of the enterprise participating in the activities within N years, assuming that N is also 1, and record the feedback data as F; S3: Select Y novel activity data in society within X years, assuming X is 1, Y is 10, and record the activity data as N; S4: The three types of data D, F and N are integrated to construct prompt words for activity recommendations according to user expectations.

8. The method for intelligent recommendation of enterprise activities based on a large model according to claim 7, characterized in that: When constructing the activity recommendation prompt words, it is necessary to control the selection of three types of data D, F and N to avoid exceeding the input context window size of the large model.

9. The method for intelligent recommendation of enterprise activities based on a large model according to claim 1, characterized in that: The step 4 is used to pass the constructed activity recommendation prompt words into the large model to obtain the generated activity recommendation.

10. The method for intelligent recommendation of enterprise activities based on a large model according to claim 1, characterized in that: In the step 1, the enterprise's recent activity history data, feedback data on participation in activities, and novel activity data in the current society all need to be deduplicated when collected.

Images