Meteorological disaster text push method and system based on artificial intelligence

Abstract

The artificial intelligence-based meteorological disaster text push method and system provided by the embodiment of the present invention, when receiving a text push processing request for a target user, generates a restatement text for multiple restatement popular science text material units based on the target user's page browsing preference, and obtains the meteorological disaster popular science articles to be pushed. In this way, by analyzing the user's browsing preferences, the meteorological disaster popular science articles can be personalized and customized for push, thereby improving the pertinence of the text push and at the same time improving the richness of the text restatement for the same material.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence and big data push technology, and in particular to a meteorological disaster text push method and system based on artificial intelligence. Background Art

[0002] The significance of popular science on meteorological disasters is very important. First, through popular science propaganda, the public can understand different types of meteorological disasters, such as typhoons, earthquakes, floods, etc., and clarify their potential threats to human life and property. This can help people realize the severity of meteorological disasters, increase vigilance, and take corresponding preventive measures. Secondly, popular science propaganda can not only increase the public's understanding of meteorological disasters, but also teach them the knowledge and skills to deal with disasters. Through education and training, the public can learn how to deal with emergencies, improve their emergency risk avoidance capabilities and awareness of self-rescue and mutual rescue, thereby reducing the losses caused by disasters. Then, popular science propaganda on meteorological disasters can strengthen information exchange and communication between various functional departments and the public. After the public understands the basic knowledge of meteorological disasters, they can better understand the warning information issued by the functional departments and reduce panic and misunderstanding. At the same time, the public can also help the functional departments improve disaster response measures and enhance the ability to cooperate and jointly resist disasters through feedback and suggestions. Further, popular science propaganda on meteorological disasters can increase the public's attention and interest in meteorological scientific research and encourage people to participate in learning and research in related fields. This is conducive to promoting the development of meteorological science, promoting the innovation of disaster prediction, monitoring and response technologies, and enhancing society's ability to prevent and reduce meteorological disasters. In summary, the significance of meteorological disaster science popularization is to enhance public awareness, improve emergency response capabilities, improve communication and cooperation, and promote scientific research and technological innovation. Through these efforts, we can jointly build a safer and more sustainable society.

[0003] One way to popularize science about meteorological disasters is to push popular science articles through PC or mobile phones, but the traditional rigid, mechanical and indiscriminate push mode is difficult to match the actual needs of various users. Summary of the invention

[0004] In order to improve the above problems, the present invention provides a method and system for pushing meteorological disaster texts based on artificial intelligence.

[0005] In a first aspect, a method for pushing meteorological disaster texts based on artificial intelligence is provided, which is applied to an AI personalized customized push processing system, and the method comprises:

[0006] Obtain original popular science text materials and text word frequency distribution relationship network of meteorological disaster events;

[0007] Performing meteorological disaster event attention analysis on the original popular science text material to obtain a meteorological disaster event material set in the original popular science text material;

[0008] Through the text word frequency distribution relationship network and the meteorological disaster event material set, the original popular science text material is subjected to a text block disassembly operation to obtain a plurality of original popular science text material units with different text granularities;

[0009] Parsing out the text retelling professionalism score corresponding to each original popular science text material unit and the text retelling weight factor corresponding to the text retelling professionalism score, wherein the text retelling professionalism score represents the influence coefficient of the text fine-grainedness contained in the original popular science text material unit;

[0010] For each original popular science text material unit, a text retelling weight factor corresponding to the text retelling professionalism score is used to perform a text retelling operation on the original popular science text material unit to obtain a plurality of retelling popular science text material units;

[0011] When a text push processing request for a target user is received, based on the page browsing preference of the target user, a paraphrase text is generated for the multiple paraphrase popular science text material units to obtain a meteorological disaster popular science article to be pushed.

[0012] In some schemes, the text block disassembly operation is performed on the original popular science text material through the text word frequency distribution relationship network and the meteorological disaster event material set to obtain multiple original popular science text material units with different text granularity, including:

[0013] Based on the meteorological disaster event material set, determining a meteorological disaster event word frequency distribution relationship subnet corresponding to the meteorological disaster event in the text word frequency distribution relationship network;

[0014] By setting a word frequency threshold, the word frequency distribution relationship subnet of the meteorological disaster event is subjected to word frequency classification to obtain a word frequency classification result;

[0015] According to the word frequency classification result, the original popular science text material is disassembled into a plurality of original popular science text material units with different text granularities through the meteorological disaster event word frequency distribution relationship subnet.

[0016] In some schemes, the word frequency classification is performed on the meteorological disaster event word frequency distribution relationship subnet by setting a word frequency threshold to obtain a word frequency classification result, including:

[0017] Performing semantic element mining on the meteorological disaster event word frequency distribution relationship subnet to obtain semantic element feature variables corresponding to the meteorological disaster event word frequency distribution relationship subnet;

[0018] Determining a word frequency difference quantitative index of the meteorological disaster event word frequency distribution relationship subnet according to the semantic element characteristic variable;

[0019] The word frequency difference quantification index and the set word frequency threshold are compared and analyzed to obtain the word frequency classification result.

[0020] In some schemes, according to the word frequency classification result, the original popular science text material is disassembled into a plurality of original popular science text material units with different text granularities through the meteorological disaster event word frequency distribution relationship subnet, including:

[0021] According to the word frequency classification result, a feature coding operation is performed through the semantic element feature variables of the meteorological disaster event word frequency distribution relationship subnet and the semantic element feature variables of the original popular science text material to obtain a feature coding result;

[0022] According to the feature coding result, the original popular science text material is disassembled into a plurality of original popular science text material units;

[0023] A fine-grained parsing operation is performed on each original popular science text material unit to obtain multiple original popular science text material units with different text fine-grainedness.

[0024] In some embodiments, the step of disassembling the original popular science text material into a plurality of original popular science text material units according to the feature coding result includes:

[0025] Converting the feature encoding result to obtain a mutually exclusive variable set of local material features;

[0026] Based on the mutually exclusive variable set of local material features, the meteorological disaster event material set is disassembled into a first popular science text material set and a second popular science text material set;

[0027] Based on the first popular science text material set and the second popular science text material set, a third popular science text material set is extracted from the original popular science text material to obtain a plurality of original popular science text material units.

[0028] In some schemes, the determining, based on the meteorological disaster event material set, a meteorological disaster event word frequency distribution relationship subnet corresponding to the meteorological disaster event in the text word frequency distribution relationship network includes identifying material text position data corresponding to the meteorological disaster event material set;

[0029] Determine the basic meteorological disaster event word frequency distribution relationship subnet corresponding to the meteorological disaster event in the text word frequency distribution relationship network through the material text position data;

[0030] The distribution relationship of the basic meteorological disaster event word frequency distribution relationship subnet is updated to obtain the meteorological disaster event word frequency distribution relationship subnet.

[0031] In some schemes, the parsing of the text retelling professionalism score corresponding to each original popular science text material unit and the text retelling weight factor corresponding to the text retelling professionalism score, wherein the text retelling professionalism score represents the influence coefficient of the fine-grained text contained in the original popular science text material unit, includes:

[0032] Compare and analyze the fine-grained text of each original popular science text material unit to obtain contrasting viewpoints;

[0033] According to the comparative viewpoint, the text fine-grainedness of each original popular science text material unit is sequentially sorted to obtain a sequential sorting result;

[0034] According to the results of the order arrangement, the text retelling professionalism score corresponding to each original popular science text material unit and the text retelling weight factor corresponding to the text retelling professionalism score are determined.

[0035] In some solutions, for each original popular science text material unit, a text paraphrase weight factor corresponding to the text paraphrase professional score is used to perform a text paraphrase operation on the original popular science text material unit to obtain multiple paraphrase popular science text material units, including:

[0036] Based on the text repetition weight factor, adjusting the set strategy execution variable in the set text material repetition strategy to obtain the target text material repetition strategy;

[0037] According to the target text material retelling strategy, the original popular science text material unit is subjected to a text retelling operation to obtain a plurality of retelling popular science text material units.

[0038] In some embodiments, the text paraphrasing operation is performed on the original popular science text material unit according to the target text material paraphrasing strategy to obtain a plurality of paraphrased popular science text material units, including:

[0039] Performing semantic element mining on the original popular science text material unit to obtain basic semantic element feature variables of the original popular science text material unit;

[0040] According to the target text material repetition strategy, weighted reinforcement is performed through the text repetition weight factor corresponding to the target text material repetition strategy and the basic semantic element feature variable to obtain a semantic element feature weighted reinforcement variable;

[0041] The basic semantic element feature variables of the original popular science text material unit are replaced by the semantic element feature weighted reinforcement variables to obtain a repeated popular science text material unit.

[0042] In some solutions, the obtaining of original popular science text materials and text word frequency distribution relationship network of meteorological disaster events includes:

[0043] Obtain the basic original popular science text materials and the basic text word frequency distribution relationship network of meteorological disaster events;

[0044] Determine a first semantic description vector corresponding to the basic original popular science text material and a second semantic description vector of the basic text word frequency distribution relationship network;

[0045] According to the first semantic description vector and the second semantic description vector, a semantic matching operation is performed on the basic original popular science text material and the basic text word frequency distribution relationship network to obtain the original popular science text material and the text word frequency distribution relationship network.

[0046] In some embodiments, the semantic matching operation is performed on the basic original popular science text material and the basic text word frequency distribution relationship network according to the first semantic description vector and the second semantic description vector to obtain the original popular science text material and the text word frequency distribution relationship network, including:

[0047] Determine a mapping list between the basic original popular science text material and the basic text word frequency distribution relationship network according to the first semantic description vector and the second semantic description vector;

[0048] Performing semantic element mining on the basic original popular science text material and the basic text word frequency distribution relationship network to obtain semantic element feature variables corresponding to the basic original popular science text material and semantic element feature variables corresponding to the basic text word frequency distribution relationship network;

[0049] Based on the mapping list, the semantic element feature variables corresponding to the basic original popular science text material and the semantic element feature variables corresponding to the basic text word frequency distribution relationship network are mapped to obtain the original popular science text material and the text word frequency distribution relationship network.

[0050] In a second aspect, an AI personalized customized push processing system is provided, comprising a processor and a memory communicating with each other, wherein the processor is used to call a computer program from the memory and implement the method described in the first aspect by running the computer program.

[0051] According to a third aspect, a computer-readable storage medium is provided, on which a computer program is stored, and the computer program implements the method described in the first aspect when running.

[0052] The meteorological disaster text push method and system based on artificial intelligence provided by the embodiment of the present invention can obtain the original popular science text material and the text word frequency distribution relationship network of the meteorological disaster event; perform meteorological disaster event attention analysis on the original popular science text material to obtain the meteorological disaster event material set in the original popular science text material; perform text block disassembly operation on the original popular science text material through the text word frequency distribution relationship network and the meteorological disaster event material set to obtain multiple original popular science text material units with different text granularity; parse out the text retelling professional score corresponding to each original popular science text material unit, and the text retelling weight factor corresponding to the text retelling professional score, wherein the text retelling professional score is the weight factor of the text retelling professional score. The influence coefficient of the fine-grained text contained in the original popular science text material unit is represented; for each original popular science text material unit, the text retelling weight factor corresponding to the text retelling professionalism score is used to perform a text retelling operation on the original popular science text material unit to obtain multiple retelling popular science text material units; when a text push processing request for a target user is received, based on the target user's page browsing preference, retelling texts are generated for the multiple retelling popular science text material units to obtain popular science articles on meteorological disasters to be pushed, so that by analyzing the user's browsing preference, personalized popular science articles on meteorological disasters can be pushed, thereby improving the pertinence of text push and at the same time improving the richness of text retelling for the same material. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0054] Figure 1 The present invention provides a flowchart of a method for pushing meteorological disaster text based on artificial intelligence. DETAILED DESCRIPTION

[0055] In order to better understand the above technical scheme, the technical scheme of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical scheme of the present invention, rather than limitations on the technical scheme of the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0056] Figure 1 A meteorological disaster text push method based on artificial intelligence is shown, which is applied to an AI personalized customized push processing system. The method includes the following steps 110-160.

[0057] Step 110: Obtain original popular science text materials and a text word frequency distribution relationship network of meteorological disaster events.

[0058] In an embodiment of the present invention, the meteorological disaster event may be a drought or flood disaster event, the original popular science text material is used to record the relevant popular science content of the drought or flood disaster event, and the text word frequency distribution relationship network is used to characterize the distribution heat of relevant words and sentences in the drought or flood disaster event.

[0059] In this step, it is necessary to obtain the original popular science text materials and the text word frequency distribution relationship network related to the meteorological disaster event. The original popular science text materials are text materials used to record the popular science content related to the drought and flood disaster events. The text word frequency distribution relationship network is used to represent the distribution heat of related words in the drought and flood disaster events. These original popular science text materials include various forms of texts, such as articles, reports, news, etc. These materials can be obtained through various channels, such as searching for relevant materials on the Internet, collecting existing popular science texts, etc. In addition, the text word frequency distribution relationship network is obtained based on the word frequency statistics in the original popular science text materials. It can use charts or other visual methods to show the degree of association and importance between words. This relationship network can help better understand and analyze the use of words in the original popular science text materials. By completing step 110, the original popular science text materials and the text word frequency distribution relationship network related to the meteorological disaster event can be obtained, providing basic data for subsequent processing and analysis.

[0060] Step 120: Perform a meteorological disaster event attention analysis on the original popular science text material to obtain a meteorological disaster event material set in the original popular science text material.

[0061] In an embodiment of the present invention, attention analysis aims to perform targeted material extraction on the original popular science text material, thereby obtaining a meteorological disaster event material set that has a strong correlation with the meteorological disaster event, and the meteorological disaster event material set is a part of the original popular science text material.

[0062] In this step, the original popular science text material is subjected to an attention analysis of meteorological disaster events. This analysis aims to extract content related to meteorological disaster events from the original text in a targeted manner to obtain a meteorological disaster event material set. Attention analysis can be implemented using methods such as natural language processing technology and machine learning algorithms. By performing keyword extraction, semantic analysis and other processing on the original text, the part of the text related to meteorological disaster events can be identified. Attention analysis can help screen and focus on information closely related to meteorological disaster events in popular science texts. The result will be a meteorological disaster event material set, which contains important parts of the original popular science text material related to meteorological disaster events. These materials can be used for subsequent processing and analysis to provide popular science knowledge and information about meteorological disaster events. In step 120, the original popular science text material is subjected to an attention analysis, and content related to meteorological disaster events is extracted from it, and a meteorological disaster event material set is obtained. This step helps to screen and focus on information related to meteorological disaster events in popular science texts, and provides an important data basis for subsequent processing.

[0063] Step 130: Perform text block decomposition operation on the original popular science text material through the text word frequency distribution relationship network and the meteorological disaster event material set to obtain multiple original popular science text material units with different text granularity.

[0064] In an embodiment of the present invention, the text block disassembly operation is used to split the original popular science text material into text material blocks (text material units), and the obtained multiple original popular science text material units have different text granularities (ie, content richness).

[0065] In this step, the original popular science text material is decomposed into text blocks using the text word frequency distribution relationship network and the meteorological disaster event material set. The purpose is to split the original text material into multiple units with different text granularity. First, the text word frequency distribution relationship network is used to analyze the word usage in the original text. This can help understand the importance and correlation of each word in the text. Based on this information, it can be determined which words or phrases are more important and common in describing meteorological disaster events. Then, combined with the meteorological disaster event material set, the original text is decomposed into text blocks. This means that the original text is divided into multiple smaller text blocks or units, each of which contains specific content related to meteorological disaster events. This division can be carried out according to the logical structure of the text, thematic distribution, etc., so as to better organize and manage text information. Through the text block decomposition operation, multiple original popular science text material units with different text granularity are obtained. These units can be paragraphs, sentences or even smaller text fragments. They each contain specific information related to meteorological disaster events, providing more fine-grained text data for subsequent processing and analysis. In step 130, the original popular science text material is decomposed into text blocks using the text word frequency distribution relationship network and the meteorological disaster event material set, thereby obtaining a plurality of original popular science text material units with different text granularities. Such decomposition operation helps to better organize and manage text information, and provides more specific and detailed text units for subsequent processing and analysis.

[0066] Step 140: parse out the text retelling professionalism score corresponding to each original popular science text material unit and the text retelling weight factor corresponding to the text retelling professionalism score, wherein the text retelling professionalism score represents the influence coefficient of the text fine-grainedness contained in the original popular science text material unit.

[0067] In the embodiment of the present invention, the text retelling professional score is used to characterize the importance of the text fine-grained content contained in the original popular science text material unit, and is also used to characterize the professionalism of the material content of the original popular science text material unit. The higher the text retelling professional score, the stronger the professionalism of the material content of the original popular science text material unit. Therefore, when performing text retelling later, it is necessary to consider the matching of the professionalism of the material content of the original popular science text material unit and the professionalism of the reader.

[0068] In this step, each original popular science text material unit is parsed to obtain the text retelling professionalism score and the corresponding text retelling weight factor. The text retelling professionalism score indicates the importance of the text granularity contained in the text unit, while the text retelling weight factor guides the subsequent text retelling operation. First, each original popular science text material unit is evaluated and given a text retelling professionalism score. This score can be determined based on a variety of factors, such as the professionalism, accuracy, and depth of the text. A higher text retelling professionalism score indicates that the text granularity contained in the text unit has higher importance and professionalism. Then, a text retelling weight factor is assigned to each text retelling professionalism score. This weight factor is used to guide subsequent text retelling operations and determine the direction and degree of retelling. According to specific needs, different weight factors can be assigned to make the retelling results more in line with the expected popular science or professionalism. By parsing each original popular science text material unit and calculating the text retelling professionalism score and text retelling weight factor, the importance of each text unit and its influence in the retelling process can be understood. This provides a basis for subsequent text paraphrasing operations, ensuring that the paraphrasing results are consistent with the target requirements. In step 140, each original popular science text material unit is parsed to obtain a text paraphrasing professionalism score and a corresponding text paraphrasing weight factor. The use of these scores and factors can guide subsequent text paraphrasing operations, ensuring that the paraphrasing results achieve the expected results in terms of popular science and professionalism.

[0069] Step 150: for each original popular science text material unit, a text retelling weight factor corresponding to the text retelling professionalism score is used to perform a text retelling operation on the original popular science text material unit to obtain a plurality of retelling popular science text material units.

[0070] In the embodiment of the present invention, the text paraphrase weight factor can guide the text paraphrase operation of the original popular science text material unit, for example, it can guide the text paraphrase of popular science or the text paraphrase of professional research. In this way, the obtained paraphrase popular science text material unit can also be the paraphrase result of popular science or the paraphrase result of professional research.

[0071] In this step, the text retelling operation is performed according to the previously obtained text retelling weight factor and the original popular science text material unit to generate a new popular science text with target popular science. By combining the text retelling weight factor, the degree and direction of retelling of each text unit can be adjusted. A higher weight factor means that the content of the text unit is more important and critical, and it is necessary to maintain greater accuracy and professionalism in the retelling process. On the contrary, a lower weight factor means that the content of the text unit can be slightly changed or simplified in the retelling. According to the guidance of the text retelling weight factor, the original popular science text material unit is retelled. This may include rewriting sentences, modifying words, adding explanations or examples, etc., to achieve better popular science effects. During the retelling process, it is necessary to pay attention to maintaining the accuracy, clarity and comprehensibility of the text, and ensure that the retelling text is consistent with the target popular science. Through such a text retelling operation, a new popular science text with target popular science is generated. These new texts are designed to better convey and explain knowledge related to meteorological disaster events, and provide readers or audiences with popular science information that is easier to understand and accept. In step 150, a text paraphrase operation is performed according to the text paraphrase weight factor and the original popular science text material unit to generate a new popular science text with the target popular science. This process adjusts the degree and direction of the text paraphrase to achieve a better popular science effect, making the final popular science text easier to understand and accept.

[0072] Step 160: upon receiving a text push processing request for a target user, based on the page browsing preference of the target user, generate a paraphrase text for the multiple paraphrase popular science text material units to obtain a meteorological disaster popular science article to be pushed.

[0073] In an embodiment of the present invention, when a text push processing request is received, the page browsing preferences of the target user can be analyzed to determine the target user's level of expertise with respect to meteorological disaster events, and personalized retelling text customization can be generated based on the level of expertise to obtain popular science articles on meteorological disasters to be pushed.

[0074] In step 160, when a text push processing request for a target user is received, the processing is performed according to the target user's page browsing preference. By analyzing the target user's page browsing preference, the target user's professional level for meteorological disaster events can be understood, and a personalized paraphrase text is generated based on this level, and finally a meteorological disaster popular science article to be pushed is obtained. For example, the target user is assumed to be a student majoring in meteorology, and he often browses web pages and materials related to meteorology. Based on his page browsing preference, it can be determined that he has a high level of professionalism in meteorological disaster events. According to this professional level, personalized paraphrase text customization can be performed. More professional terms, in-depth explanations and specific cases may be used to present the popular science knowledge of meteorological disaster events. Such a paraphrase text will meet the target user's needs for professional information and provide more in-depth and comprehensive popular science content. Finally, by executing step 160, a meteorological disaster popular science article to be pushed is generated, and the article is personalized according to the target user's page browsing preference. Customization of the paraphrase text is performed to meet the target user's professional needs for meteorological disaster events. In step 160, according to the received text push processing request for the target user, a plurality of paraphrasing popular science text material units are generated based on the target user's page browsing preference. By analyzing the target user's page browsing preference, the professional level is determined, and a personalized paraphrasing text is generated accordingly, and finally the meteorological disaster popular science article to be pushed is obtained.

[0075] In other possible application scenarios, it is assumed that the target user is an ordinary person who has no professional background or in-depth understanding of meteorological disaster events. According to his page browsing preferences, it can be inferred that he pays more attention to general news, lifestyle or entertainment content. Based on this situation, corresponding paraphrase text generation can be performed to convey popular science information to the target user and help them understand and respond to meteorological disaster events. For example, the generated articles to be pushed may use more concise and easy-to-understand language to avoid excessive use of professional terms and technical details. Actual cases, charts or animations may be provided to illustrate different types of meteorological disasters, preventive measures or behavioral suggestions. In addition, the opinions and suggestions of authoritative organizations may be cited to increase credibility. Through step 160, paraphrase text generation will be performed according to the page browsing preferences of the target users of the general public. The generated text will try to avoid being too professional and complex so that the general public can easily understand and accept it. The focus is on providing practical and easy-to-digest popular science information to help them enhance their cognition and prevention awareness of meteorological disaster events. In step 160, when the target user is an ordinary person, paraphrase text generation is performed according to his page browsing preferences. By using simple and easy-to-understand language, real-life cases and authoritative opinions, useful and easy-to-understand popular science information is provided to meet the needs of the general public on meteorological disaster events.

[0076] It can be understood that by applying steps 110 to 160, the original popular science text material and the text word frequency distribution relationship network of the meteorological disaster event can be obtained; the original popular science text material is subjected to a meteorological disaster event attention analysis to obtain a meteorological disaster event material set in the original popular science text material; the original popular science text material is subjected to a text block disassembly operation through the text word frequency distribution relationship network and the meteorological disaster event material set to obtain a plurality of original popular science text material units with different text granularities; the text retelling professionalism score corresponding to each original popular science text material unit and the text retelling weight factor corresponding to the text retelling professionalism score are parsed out, wherein the text retelling professionalism score represents the original science The influence coefficient of the fine-grained text contained in the popular text material unit; for each original popular science text material unit, the text retelling weight factor corresponding to the text retelling professionalism score is used to perform text retelling operation on the original popular science text material unit to obtain multiple retelling popular science text material units; when a text push processing request for a target user is received, based on the target user's page browsing preference, retelling text is generated for multiple retelling popular science text material units to obtain popular science articles on meteorological disasters to be pushed, so that by analyzing the user's browsing preferences, personalized popular science articles on meteorological disasters can be pushed, thereby improving the pertinence of text push and at the same time improving the richness of text retelling for the same material.

[0077] In order to better understand the above technical solution, an original popular science text material involving drought and flood disasters is used as an example for explanation below.

[0078] Drought and Flood Disasters: The Extreme Mystery of Water Resources

[0079] Drought and flood are common and disturbing natural disasters that have a profound impact on our lives, agriculture and the environment. When we face drought or flood, water supply and irrigation systems encounter problems, food production decreases, ecosystems are damaged, and even larger-scale social and economic problems may occur. Understanding the causes and countermeasures of drought and flood disasters is crucial for us to better adapt to and mitigate their impacts.

[0080] The causes of drought and flood disasters are mainly related to abnormal climate. Drought is usually caused by long-term low rainfall, high temperature and excessive evaporation. Climate change has aggravated the frequency and intensity of drought, making drought an important problem worldwide. In contrast, floods are caused by extreme climate events such as heavy rainfall, rainstorms, and typhoons. Floods not only cause urban waterlogging and farmland flooding, but may also trigger secondary disasters such as mountain torrents and mudslides.

[0081] We can take various forms of prevention and control measures to deal with drought and flood disasters. When facing drought, water conservation is the most critical link. We can reduce water consumption by improving irrigation technology, promoting water-saving equipment and rationally managing water resources. At the same time, we should strengthen drought monitoring and early warning systems and take timely response measures, such as changing crops and adjusting planting structures, to reduce the losses caused by drought.

[0082] When facing floods, we need to focus on flood control project construction and water resource dispatching. Measures such as building dams, improving drainage systems, and strengthening river dredging can effectively reduce the impact of floods on human society and farmland. In addition, improving the public's self-rescue ability and safety awareness, and strengthening the community's flood prevention preparations and emergency response are also important means to reduce the impact of flood disasters.

[0083] In addition to response measures, we should also strengthen scientific research and monitoring and early warning capabilities for drought and flood disasters. Through in-depth research in areas such as climate change and water cycles, we can improve our understanding of drought and flood disasters and provide a scientific basis for disaster management. At the same time, establishing a sound meteorological monitoring and early warning system and issuing drought and flood warning information in a timely manner will help people make corresponding preparations and take actions.

[0084] In short, drought and flood disasters are natural disasters caused by abnormal climate, which have serious impacts on human society, agriculture and ecological environment. Understanding their causes and countermeasures is of great significance for reducing the impact of disasters and mitigating their effects. The prevention and control of drought and flood disasters requires the joint efforts of the whole society, including the participation of various functional departments, scientific research institutions, farmers and the public.

[0085] We can also prevent drought and flood disasters by improving the efficiency of water resource management. Reasonable planning of the balance between water supply and demand, strengthening sewage treatment and recycling water resources are key measures to ensure the sustainable use of water resources. In addition, establishing a mutually beneficial and win-win regional water resource cooperation mechanism and promoting cross-border cooperation and exchanges will help solve the problem of uneven distribution of water resources.

[0086] At the same time, it is also crucial to educate the public and raise society's awareness of meteorological disaster risks. Organizing relevant science popularization activities, conducting disaster response drills, and formulating disaster emergency plans can help the public better understand the nature and hazards of drought and flood disasters, and learn the corresponding self-rescue and prevention knowledge.

[0087] In short, drought and flood disasters are complex and serious problems that require multi-party cooperation and comprehensive response. Through scientific research, strengthening monitoring and early warning, strengthening the implementation of prevention and control measures, and public participation, we can reduce the impact of drought and flood disasters, improve society's adaptability, protect people's lives and property, and achieve sustainable development.

[0088] In some application scenarios, examples of popular science articles on meteorological disasters to be pushed corresponding to the above articles are generated by retelling them from a professional perspective as follows.

[0089] Drought and Flood Disasters: The Extreme Mystery of Water Resources

[0090] Introduction: Drought and flood are common and disturbing natural disasters that have profound impacts on life, agriculture and the environment. It is crucial to understand their causes and countermeasures in order to adapt to and mitigate their impacts.

[0091] text:

[0092] 1. Causes of drought and flood disasters: closely related to abnormal climate. Drought is caused by long-term low rainfall, high temperature and evaporation, and climate change has aggravated the frequency and intensity of drought. Floods are caused by extreme climate events such as heavy rainfall, rainstorms, and typhoons.

[0093] 2. Countermeasures:

[0094] Drought prevention and control: save water, improve irrigation technology, improve water resource management efficiency, strengthen drought monitoring and early warning systems, and take timely measures such as crop transformation and adjustment of planting structure;

[0095] Flood response: construct flood control projects, improve drainage systems, strengthen river dredging, enhance the public's self-rescue capabilities and safety awareness, and strengthen community flood preparedness and emergency response.

[0096] 3. Scientific research and monitoring and early warning:

[0097] Strengthen scientific research on drought and flood disasters, gain a deeper understanding of climate change and water cycles, and provide a scientific basis for disaster management;

[0098] Establish a complete meteorological monitoring and early warning system, and issue drought and flood warning information in a timely manner.

[0099] 4. Water resources management and cooperation:

[0100] Rationally plan the balance between water supply and demand, strengthen sewage treatment and water recycling to ensure sustainable use;

[0101] Establish a mutually beneficial and win-win regional water resources cooperation mechanism, promote cross-border cooperation and exchanges, and solve the problem of uneven distribution of water resources.

[0102] 5. Raise public awareness:

[0103] Organize science popularization activities, carry out disaster response drills, and formulate disaster emergency plans to help the public understand the nature and hazards of drought and flood disasters, and learn self-rescue and prevention knowledge.

[0104] Conclusion: Drought and flood disasters are complex and serious problems that require multi-party cooperation and comprehensive response. Through scientific research, strengthening monitoring and early warning, strengthening the implementation of prevention and control measures and public participation, we can reduce the impact of drought and flood disasters, improve social adaptability and achieve sustainable development.

[0105] In some application scenarios, examples of popular science articles on meteorological disasters to be pushed corresponding to the above articles are generated by retelling them from a popular science perspective as follows:

[0106] Drought and Flood Disasters: The Extreme Mystery of Water Resources

[0107] Introduction: Drought and flood disasters have a profound impact on our lives, agriculture and environment. Through two drought and flood disaster events, we can understand their causes and countermeasures, which will help us better adapt to and mitigate their impacts.

[0108] text:

[0109] 1. Popularization of drought events:

[0110] Drought in the Midwest of the United States in 2012: Drought caused by long-term low rainfall, high temperatures and evaporation led to dry farmland and depletion of water sources, causing serious impacts on the economy and ecosystems. Climate change has increased the frequency and intensity of droughts.

[0111] Popular science on countermeasures: Water conservation is the key. We should improve irrigation technology, promote water-saving equipment, rationally manage water resources, conduct drought monitoring and early warning, change crops, adjust planting structure, etc. to reduce losses caused by drought.

[0112] 2. Popular science on flood events:

[0113] Floods in the Yangtze River Basin, China, in 2019: Heavy rainfall caused floods that flooded cities and farmland, threatening human life and property. Inadequate river dredging and flood control projects exacerbated the floods.

[0114] Popularization of response measures: Strengthening flood control project construction, improving drainage systems, and clearing river silt can reduce the impact of floods. It is also very important to improve the public's self-rescue ability and safety awareness, and strengthen community flood preparedness and emergency response.

[0115] Through the popularization of these two drought and flood disasters, we can have a more specific understanding of the causes and countermeasures of drought and flood disasters. Saving water, improving irrigation technology, drought monitoring and early warning are the keys to dealing with droughts; strengthening flood control projects, improving drainage systems, and enhancing the public's self-rescue ability and safety awareness are important means to mitigate the impact of floods.

[0116] Conclusion: Drought and flood disasters are natural disasters caused by abnormal climate, which have serious impacts on human society, agriculture and the ecological environment. Understanding their causes and countermeasures, strengthening scientific research and monitoring and early warning, and promoting water resources management and public participation will help mitigate the impact of drought and flood disasters and achieve sustainable development.

[0117] In some optional embodiments, step 130 performs a text block decomposition operation on the original popular science text material through the text word frequency distribution relationship network and the meteorological disaster event material set to obtain multiple original popular science text material units with different text granularity, including steps 131 to 133.

[0118] Step 131: Based on the meteorological disaster event material set, determine the meteorological disaster event word frequency distribution relationship subnetwork corresponding to the meteorological disaster event in the text word frequency distribution relationship network.

[0119] In step 131, it is assumed that there is a meteorological disaster event material set, which includes descriptions and related information of multiple drought and flood disaster events. These materials can be used to construct a text word frequency distribution relationship network, which will show the frequency distribution and relationship between words. Then, the subnetwork corresponding to each meteorological disaster event can be determined in this word frequency distribution relationship network.

[0120] For example, if a flood event in southern China is selected as a meteorological disaster event material, text data related to the event can be extracted from the material set, such as news reports, research articles, etc. Then, by processing and analyzing these texts, a text word frequency distribution relationship network with keywords as nodes can be constructed to show the frequency and correlation between different words. In this relationship network, words related to flood events (such as flood, waterlogging, drainage, flood control, etc.) will form a subnetwork, showing the frequency distribution and relationship between these words.

[0121] Step 131 is to help better understand the characteristics and related terms of meteorological disaster events. Through the word frequency distribution relationship subnet, we can see which words are used more frequently in a specific event, which explains the key characteristics and influencing factors of the event. This helps to gain a deeper understanding of the differences and commonalities between different meteorological disaster events.

[0122] Step 132: by setting a word frequency threshold, the word frequency distribution relationship subnet of the meteorological disaster event word frequency classification is performed to obtain a word frequency classification result.

[0123] In step 132, the meteorological disaster event word frequency distribution relationship subnet is subjected to word frequency classification according to the set word frequency threshold, thereby obtaining a word frequency classification result.

[0124] Taking drought and flood disasters as an example, suppose a word frequency threshold is set to 5, which means that a word is considered a high-frequency word only when its frequency is greater than or equal to 5. In the word frequency distribution relationship subnetwork, if the frequency of a word (such as drought) exceeds 5 times, it is classified as a high-frequency word. On the contrary, if the frequency of a word (such as drainage) is less than 5 times, it is classified as a low-frequency word.

[0125] Through the word frequency classification in step 132, it is possible to better understand which terms and concepts are more common and important in meteorological disaster events. The high-frequency word classification provides a set of key words that can accurately describe and explain the core content of meteorological disaster events. The low-frequency word classification includes some less common but still important words, which can provide clues for further research and exploration.

[0126] Step 133: Based on the word frequency classification result, the original popular science text material is disassembled into a plurality of original popular science text material units with different text granularities through the meteorological disaster event word frequency distribution relationship subnet.

[0127] In step 133, based on the word frequency classification result and the meteorological disaster event word frequency distribution relationship subnet, the original popular science text material is disassembled into a plurality of original popular science text material units with different text granularities.

[0128] Assume that there is an original popular science article about drought and flood disasters. According to the word frequency classification result of step 132, this article can be disassembled into multiple text fine-grained material units.

[0129] For example, words in the high-frequency word category (such as drought, irrigation, water conservation, etc.) can form a fine-grained unit to specifically introduce the causes, impacts, and countermeasures of drought and flood disasters. Words in the low-frequency word category (such as drainage facilities, soil moisture content, rainfall intensity, etc.) can form another fine-grained unit to discuss in depth the relevant scientific knowledge and technical details of drought and flood disasters.

[0130] By breaking down the original popular science text material into multiple material units with different text granularity, more accurate and targeted information can be provided according to the needs and interests of readers. Such a breaking down operation helps to meet the needs of different reader groups and provide more operable and practical popular science content.

[0131] The beneficial effect of step 131 to step 133 is that by using the text word frequency distribution relationship network and the meteorological disaster event material set, the original popular science text material is fine-grainedly disassembled and classified. This can make the popular science content more accurate, systematic and easy to understand, improve the reader's cognition and understanding of meteorological disaster events, and meet the needs and interests of different readers.

[0132] In some optional embodiments, step 132 sets a word frequency threshold to perform word frequency classification on the meteorological disaster event word frequency distribution relationship subnet to obtain a word frequency classification result, including steps 1321 to 1323.

[0133] Step 1321: perform semantic element mining on the meteorological disaster event word frequency distribution relationship subnetwork to obtain semantic element feature variables corresponding to the meteorological disaster event word frequency distribution relationship subnetwork.

[0134] For drought and flood disaster events, semantic element mining is performed to obtain semantic element feature variables related to these events. For example, in the text materials related to drought and flood disasters, some keywords closely related to drought and flood events can be found, such as drought, water resources, drainage facilities, rainfall intensity, etc. By analyzing the semantic relationship and co-occurrence pattern between these keywords, the semantic element feature variables of drought and flood disaster events can be obtained. These variables may include weather conditions (such as rainfall, evaporation rate), environmental factors (such as soil moisture content, topography), and impact range (such as cities, farmland), etc.

[0135] Step 1322: Determine the word frequency difference quantitative index of the meteorological disaster event word frequency distribution relationship subnet based on the semantic element feature variable.

[0136] Based on the semantic element feature variables, the word frequency difference quantitative indicators of the drought and flood disaster event word frequency distribution relationship subnetwork are determined. For example, the word frequency difference of keywords in different drought and flood disaster events can be calculated, such as the frequency difference between drought and water resources. These quantitative indicators can help compare the differences and commonalities in keyword frequencies between different events.

[0137] Step 1323: Compare and analyze the word frequency difference quantification index and the set word frequency threshold to obtain the word frequency classification result.

[0138] The word frequency difference quantitative index is compared and analyzed with the set word frequency threshold to obtain the word frequency classification results of drought and flood disaster events. For example, if the word frequency threshold is set to 10, it means that only when the frequency of a word is greater than or equal to 10, it is classified as a high-frequency word. If the frequency of a keyword exceeds the threshold, it is classified as a high-frequency word; otherwise, it is classified as a low-frequency word.

[0139] Through the implementation of steps 1321 to 1323, the text material of drought and flood disaster events can be disassembled into the results of different word frequency classifications. This fine-grained classification helps to deeply understand the characteristics and influencing factors of drought and flood disasters, and provide more targeted popular science content. Readers can obtain information related to their areas of interest from the word frequency classification results, and understand different types of drought and flood disaster events, response strategies, and scientific background knowledge. Overall, this method improves the accuracy and operability of popular science content, and promotes the public's level of awareness of drought and flood disasters and their ability to respond.

[0140] In some optional embodiments, the step 133 decomposes the original popular science text material into multiple original popular science text material units with different text granularity through the meteorological disaster event word frequency distribution relationship subnet based on the word frequency classification result, including steps 1331-1333.

[0141] Step 1331: Based on the word frequency classification result, a feature coding operation (logical operation) is performed through the semantic element feature variables of the meteorological disaster event word frequency distribution relationship subnet and the semantic element feature variables of the original popular science text material to obtain a feature coding result.

[0142] Based on the word frequency classification results and the semantic element feature variables of the meteorological disaster event word frequency distribution relationship subnetwork, as well as the semantic element feature variables of the original popular science text material, feature encoding operations are performed. Feature encoding here refers to expressing the relationship between different features through logical operations, such as using Boolean operators for combination, comparison or screening. Through feature encoding, feature encoding results that describe fine-grained features of different texts are obtained.

[0143] Step 1332: according to the feature coding result, the original popular science text material is disassembled into a plurality of original popular science text material units.

[0144] According to the feature encoding results, the original popular science text material is disassembled into multiple original popular science text material units. These units can be sentences, paragraphs or other smaller text units in the original text. The disassembled units retain part or all of the content of the original text material and are reassembled according to the feature encoding results so that each unit has different fine-grained text features.

[0145] Step 1333: Perform a fine-grained analysis operation on each original popular science text material unit to obtain multiple original popular science text material units with different text fine-grainedness.

[0146] Fine-grained parsing is performed on each original popular science text material unit to obtain original popular science text material units with different text fine-grainedness. Fine-grained parsing can include analysis and processing of the structure, grammar, semantics, etc. of the text. Through fine-grained parsing, the information conveyed in each unit can be further understood, and more specific and accurate popular science content can be extracted.

[0147] Through the implementation of steps 1331 to 1333, the original popular science text material can be disassembled into multiple original popular science text material units with different text granularity. This disassembly and analysis process helps to deeply understand the popular science content related to drought and flood disaster events, and provides richer and more detailed information. Readers can select text units of interest according to their own needs and obtain specific fine-grained popular science knowledge. On the whole, this method improves the customizability and personalization of popular science content, and promotes the public's level of awareness and response capabilities for drought and flood disasters.

[0148] Under some preferred design ideas, step 1332 decomposes the original popular science text material into multiple original popular science text material units based on the feature encoding result, including steps 13321 to 13323.

[0149] Step 13321: convert the feature encoding result to obtain a mutually exclusive variable set of local material features.

[0150] According to the feature encoding results, the conversion operation is performed to obtain the local material feature mutually exclusive variable set. Here, the local material feature mutually exclusive variable set refers to distinguishing and extracting the features with mutually exclusive relationships in the feature encoding results. For example, assuming that the feature encoding results contain three features: drought, rainfall, and water resources. Through the conversion operation, they can be divided into different mutually exclusive variable sets to represent their independence.

[0151] Step 13322: Based on the mutually exclusive variable set of local material features, the meteorological disaster event material set is decomposed into a first popular science text material set and a second popular science text material set.

[0152] Based on the mutually exclusive variable set of local material features, the meteorological disaster event material set is disassembled into the first popular science text material set and the second popular science text material set. The meteorological disaster event material set here refers to the part of the original popular science text material related to drought and flood disaster events. According to the classification of the mutually exclusive variable set of local material features, the content with the same characteristics in the original material can be divided into the first and second popular science text material sets.

[0153] Step 13323: extract a third popular science text material set from the original popular science text material based on the first popular science text material set and the second popular science text material set to obtain a plurality of original popular science text material units.

[0154] According to the first and second popular science text material sets, a third popular science text material set is extracted from the original popular science text material to obtain a plurality of original popular science text material units. The third popular science text material set here is a subset extracted according to a specific combination of local features. For example, if the first popular science text material set contains drought and water resource features, and the second popular science text material set contains rainfall features, then the content related to drought, water resources and rainfall can be extracted from the original text as the third popular science text material set.

[0155] Through the implementation of step 13321-step 13323, the original popular science text material can be disassembled into multiple original popular science text material units with different text granularity. This disassembly process is based on the feature encoding results and the mutually exclusive variable set of local material features, so that each text unit has a specific popular science nature under a specific feature combination. Doing so helps to provide more personalized and targeted popular science information to meet readers' needs for different features and topics. On the whole, the implementation of step 13321-step 13323 improves the customizability and accuracy of popular science content, and promotes the public's in-depth understanding of drought and flood disasters and the improvement of their ability to respond.

[0156] In some optional embodiments, the step 131 of determining the meteorological disaster event word frequency distribution relationship subnetwork corresponding to the meteorological disaster event in the text word frequency distribution relationship network based on the meteorological disaster event material set includes steps 1311-1313.

[0157] Step 1311: Identify the material text location data corresponding to the meteorological disaster event material set.

[0158] In the drought and flood disaster event scenario, there may be a meteorological disaster event material set, which includes relevant news reports, expert comments, social media posts, etc. In this step, it is necessary to identify the location data of these material texts, that is, to determine the location of each text in the material set.

[0159] Step 1312: Determine the basic meteorological disaster event word frequency distribution relationship subnet corresponding to the meteorological disaster event in the text word frequency distribution relationship network through the material text position data.

[0160] By using the material text position data obtained in step 1311, a text set related to drought and flood disaster events can be extracted from the text word frequency distribution relationship network. Then, a basic meteorological disaster event word frequency distribution relationship subnet is constructed based on the word frequency distribution relationship of these texts. This subnet will contain important words and phrases related to drought and flood disasters and the association information between them.

[0161] Step 1313: update the distribution relationship of the basic meteorological disaster event word frequency distribution relationship subnet to obtain the meteorological disaster event word frequency distribution relationship subnet.

[0162] In this step, the distribution relationship of the basic meteorological disaster event word frequency distribution relationship subnetwork is updated. This means that the word frequency distribution relationship may be updated based on new material text or other information sources to more accurately reflect the characteristics and trends of drought and flood disaster events. Through the updated distribution relationship, a more complete and accurate meteorological disaster event word frequency distribution relationship subnetwork can be obtained.

[0163] By executing steps 1311 to 1313, relevant material texts can be extracted from the meteorological disaster event material set, and a basic meteorological disaster event word frequency distribution relationship subnet can be constructed through the word frequency distribution relationship network. Then, by updating the distribution relationship, the subnet is continuously optimized to more accurately reflect the characteristics and trends of drought and flood disaster events. Such a process can help to deeply understand drought and flood disaster events, thereby providing more useful information and insights in terms of prediction, early warning and response strategy formulation. Overall, steps 1311 to 1313 can improve the cognition and response capabilities of drought and flood disaster events, thereby having a positive effect on reducing the impact of disasters and protecting people's lives and property.

[0164] In some optional embodiments, step 140 parses out the text retelling professionalism score corresponding to each original popular science text material unit, and the text retelling weight factor corresponding to the text retelling professionalism score, wherein the text retelling professionalism score represents the influence coefficient of the fine-grained text contained in the original popular science text material unit, including steps 141 to 143.

[0165] Step 141: Conduct a comparative analysis on the text granularity of each original popular science text material unit to obtain comparative viewpoints.

[0166] Suppose there is a set of original popular science text material units covering different aspects of drought and flood disasters. In this step, the textual fine-grained analysis of these text units will be compared to find differences in viewpoints. For example, one text unit may emphasize the impact of drought on crop yields, while another text unit focuses on the impact of drought on water supply.

[0167] Step 142: Based on the comparison viewpoint, the text fine-grained text of each original popular science text material unit is sequentially sorted to obtain a sequential sorting result.

[0168] Based on the contrasting viewpoints in step 141, the text granularity of each original popular science text material unit can be sequentially arranged to accurately reflect the content hierarchy and relevance of each text unit. For example, if a text unit focuses on the impact of drought on crop yields, while another text unit focuses on the impact of drought on water supply, the text units can be sequentially arranged based on these viewpoints to ensure the logic and coherence of the content.

[0169] Step 143: Determine the text retelling professionalism score corresponding to each original popular science text material unit and the text retelling weight factor corresponding to the text retelling professionalism score based on the sequence sorting result.

[0170] According to the results of the order arrangement in step 142, a text retelling professionalism score can be determined for each original popular science text material unit. The text retelling professionalism score characterizes the influence coefficient of the text fine-grainedness contained in the original popular science text material unit. This score can be evaluated based on the accuracy, professionalism and information richness of the text.

[0171] At the same time, the corresponding text retelling weight factor can also be determined for the text retelling professionalism score. The text retelling weight factor is used to measure the importance of different text retelling professionalism scores to the overall text retelling result. By determining the weight factor, the contribution of the original popular science text material unit in the overall text retelling can be more accurately reflected.

[0172] By executing steps 141 to 143, the original popular science text material units in the drought and flood disaster event scenario can be analyzed, sorted and evaluated. By comparing viewpoints and sorting in order, the key viewpoints and content levels of the text can be accurately grasped, thereby obtaining a clearer and more organized text retelling result. By determining the professional score and weight factor of the text retelling, the quality and importance of the text retelling can be quantified, and useful guidance can be provided for subsequent processing. Overall, steps 141 to 143 help to improve the quality and effect of the text retelling of drought and flood disaster events, and further enhance the ability to understand and apply relevant information.

[0173] In some optional embodiments, for each original popular science text material unit in step 150, a text paraphrase operation is performed on the original popular science text material unit using a text paraphrase weight factor corresponding to the text paraphrase professionalism score to obtain multiple paraphrase popular science text material units, including steps 151 and 152.

[0174] Step 151: Based on the text repetition weight factor, adjust the set strategy execution variable in the set text material repetition strategy to obtain the target text material repetition strategy.

[0175] Assume that there is a set of original popular science text material units obtained through text retelling evaluation, and the text retelling professionalism score and the corresponding text retelling weight factor are determined for each unit. In this step, the execution variables of the text material retelling strategy are adjusted according to the text retelling weight factor. For example, if a certain original popular science text material unit has a higher text retelling professionalism score and weight factor, it can be given a higher importance or priority in the target text material retelling strategy.

[0176] Step 152: According to the target text material paraphrase strategy, the original popular science text material unit is subjected to a text paraphrase operation to obtain a plurality of paraphrase popular science text material units.

[0177] According to the target text material retelling strategy obtained in step 151, a text retelling operation is performed on each original popular science text material unit. This means that the variables will be executed according to the set strategy, and the text will be modified, reorganized or supplemented to achieve the target retelling effect. For example, for a certain original popular science text material unit, if the target text material retelling strategy sets the requirements of improving readability and emphasizing key points, the text may be simplified, the paragraph structure may be reorganized, or charts or images that highlight key points may be added.

[0178] When responding to drought and flood disaster events, as far as step 151 is concerned, it is assumed that there is a set of original popular science text material units, which contain different aspects of information about drought and flood disasters. The text retelling professionalism score and the corresponding text retelling weight factor have been determined for each unit. Now, it is necessary to adjust the execution variables of the text material retelling strategy according to these weight factors. In terms of drought, if a certain original text unit has a higher text retelling professionalism score and weight factor, it can be given a higher importance in the target text material retelling strategy. Therefore, in the target text material retelling strategy, strategy execution variables will be set, such as emphasizing the key points of the text unit, adding relevant examples or data support, and ensuring accurate communication of the impact of drought on crop yields. In terms of waterlogging, if another original text unit has a higher text retelling professionalism score and weight factor, it can be given a higher importance in the target text material retelling strategy. Therefore, in the target text material retelling strategy, strategy execution variables may be set, such as emphasizing the challenges of waterlogging in the text unit to the urban drainage system, providing solutions or suggestions, and ensuring accurate communication of the impact of waterlogging on people's lives.

[0179] As for step 151, based on the target text material retelling strategy determined in step 151, the original popular science text material unit is retelled. The following are two examples: Retelling of drought-related text units: According to the target text material retelling strategy, the impact of drought on crop yields in this text unit will be emphasized. The unit may be modified and supplemented to highlight key points and data support. For example, actual cases can be added to illustrate crop failures caused by droughts and adaptive agricultural measures for droughts. In this way, the retelling text unit will more accurately convey the impact of drought on crop yields and related solutions; Retelling of waterlogging-related text units: According to the target text material retelling strategy, the challenges of waterlogging to urban drainage systems in this text unit will be emphasized. The unit may be modified and supplemented to highlight key points and provide solutions. For example, actual cases can be added to illustrate the importance of urban waterlogging and flood risk management caused by waterlogging. At the same time, technical measures for improving urban drainage systems and emergency warning mechanisms for waterlogging can be introduced.

[0180] By executing steps 151 to 152, the original popular science text material unit can be operated according to the text retelling weight factor and the target text material retelling strategy to obtain multiple retelling popular science text material units. These retelling text units have better professionalism, accuracy and readability, and meet specific retelling requirements and target strategies. Overall, steps 151 to 152 can improve the quality and effective communication capabilities of popular science texts in drought and flood disaster event scenarios, making relevant information easier to understand and absorb. At the same time, based on the adjustment and operation of text retelling scores and weight factors, the retelling results can also be optimized to make them more in line with the requirements of the target retelling strategy. This helps to improve the public's level of awareness of drought and flood disaster events, enhance the effectiveness of scientific communication, and promote the improvement of society's ability to respond to disasters.

[0181] Under some possible design ideas, in step 152, based on the target text material retelling strategy, the original popular science text material unit is subjected to a text retelling operation to obtain multiple retelling popular science text material units, including steps 1521 to 1523.

[0182] Step 1521: perform semantic element mining on the original popular science text material unit to obtain basic semantic element feature variables of the original popular science text material unit.

[0183] Assume that there is a set of original popular science text material units, which contain different aspects of information about drought and flood disasters. In this step, semantic element mining is performed on each text unit to identify and extract its basic semantic element feature variables. For example, for a text unit describing drought, the basic semantic element features that may be identified include the cause of the drought, the scope of impact, and the damage to crops.

[0184] Step 1522: Based on the target text material retelling strategy, weighted reinforcement is performed using the text retelling weight factor corresponding to the target text material retelling strategy and the basic semantic element feature variable to obtain a semantic element feature weighted reinforcement variable.

[0185] According to the target text material retelling strategy, the text retelling weight factor and the basic semantic element feature variable will be used for weighted reinforcement. For example, if the importance of drought to crops is emphasized in the target text material retelling strategy, a higher weight factor may be given to the semantic element features related to crop damage. Through weighted reinforcement, the semantic element feature weighted reinforcement variable is obtained.

[0186] Step 1523: replace the basic semantic element feature variables of the original popular science text material unit with the semantic element feature weighted reinforcement variables to obtain a repeated popular science text material unit.

[0187] The basic semantic element feature variables of the original popular science text material unit are iterated through the weighted reinforcement variables of the semantic element features obtained in step 1522. Specifically, the weight or value of the semantic element features of each text unit is adjusted according to the weighted reinforcement variables to achieve the goal of retelling the text. For example, in a text unit describing waterlogging, if the target text material retelling strategy requires emphasizing the challenges of the urban drainage system, the weight values ​​related to the semantic element features such as drainage system load and rainwater runoff may be increased. In this way, a popular science text material unit after retelling is obtained, and its basic semantic element features are iteratively adjusted to meet the requirements of the target retelling strategy.

[0188] By executing steps 1521 to 1523, the original popular science text material unit is retelled using semantic element mining and weighted reinforcement. The advantage of this is that key semantic elements can be extracted and weighted reinforcement can be performed according to the target text material retelling strategy, making the retelling text more accurate and focused. Through the iterative adjustment in step 1523, the text content can be further optimized according to the retelling requirements while maintaining the basic semantic elements.

[0189] Under some optional design ideas, the step 110 of obtaining the original popular science text material and the text word frequency distribution relationship network of the meteorological disaster event includes steps 111 to 113.

[0190] Step 111: Obtain the basic original popular science text materials and the basic text word frequency distribution relationship network of meteorological disaster events.

[0191] In this step, basic original popular science text materials related to drought and flood disasters are collected. These text materials can include text data from various sources such as professional reports, academic papers, and news reports. At the same time, a basic text word frequency distribution relationship network is also constructed, which records the frequency of occurrence of different words in the text and the correlation between them.

[0192] Step 112: Determine a first semantic description vector corresponding to the basic original popular science text material and a second semantic description vector of the basic text word frequency distribution relationship network.

[0193] In this step, a semantic description vector is determined for the basic original popular science text material and the basic text word frequency distribution relationship network. For example, for the basic original popular science text material describing the drought, key words related to the drought may be extracted and combined into a first semantic description vector. At the same time, a second semantic description vector is constructed based on the word frequency and association information in the basic text word frequency distribution relationship network.

[0194] Step 113: perform a semantic matching operation on the basic original popular science text material and the basic text word frequency distribution relationship network based on the first semantic description vector and the second semantic description vector to obtain the original popular science text material and the text word frequency distribution relationship network.

[0195] By performing a semantic matching operation, the first semantic description vector is compared with the second semantic description vector. For example, if the first semantic description vector contains key words describing the drought, and the second semantic description vector of the basic text word frequency distribution relationship network has a high frequency and correlation with the key words, then it can be considered that these basic original popular science text materials and the text word frequency distribution relationship network are related to the drought.

[0196] By executing steps 111 to 113, the basic original popular science text materials and text word frequency distribution relationship network of drought and flood disaster events can be obtained, and semantic matching operations can be performed. The advantage of doing so is that key information related to drought and flood disaster events can be extracted, and the degree of relevance can be determined through semantic matching. By constructing a text word frequency distribution relationship network, the frequency of occurrence and relevance of relevant words in drought and flood disaster events can be further analyzed and understood. Overall, steps 111 to 113 help to deeply understand the popular science text information of drought and flood disaster events, and provide basic data support for subsequent processing and response work.

[0197] In some exemplary embodiments, step 113 performs a semantic matching operation on the basic original popular science text material and the basic text word frequency distribution relationship network based on the first semantic description vector and the second semantic description vector to obtain the original popular science text material and the text word frequency distribution relationship network, including steps 1131 to 1133.

[0198] Step 1131: Determine a mapping list between the basic original popular science text material and the basic text word frequency distribution relationship network based on the first semantic description vector and the second semantic description vector.

[0199] In this step, the first semantic description vector and the second semantic description vector are used for matching to find out the elements that are associated between the basic original popular science text material and the basic text word frequency distribution relationship network. For example, if the first semantic description vector includes key words describing drought conditions, and the second semantic description vector contains words related to these key words with high frequency and association, then they can be recorded in the mapping list to indicate that there is a mapping relationship between them.

[0200] Step 1132: perform semantic element mining on the basic original popular science text material and the basic text word frequency distribution relationship network to obtain semantic element feature variables corresponding to the basic original popular science text material and semantic element feature variables corresponding to the basic text word frequency distribution relationship network.

[0201] In this step, semantic element mining is performed on the basic original popular science text materials and the basic text word frequency distribution relationship network. By analyzing the relationship between text content and vocabulary, the semantic element feature variables corresponding to the basic original popular science text materials can be extracted, including important features such as descriptions of the causes of drought and the scope of impact. At the same time, the semantic element feature variables corresponding to the basic text word frequency distribution relationship network can also be obtained, which records the frequency of different words and the correlation between them.

[0202] Step 1133: Based on the mapping list, the semantic element feature variables corresponding to the basic original popular science text material and the semantic element feature variables corresponding to the basic text word frequency distribution relationship network are mapped to obtain the original popular science text material and the text word frequency distribution relationship network.

[0203] The semantic element feature variables corresponding to the basic original popular science text material and the semantic element feature variables corresponding to the basic text word frequency distribution relationship network are mapped through the mapping list obtained in step 1131. Specifically, the corresponding semantic element feature variables are matched and integrated according to the mapping relationship to obtain the original popular science text material and the text word frequency distribution relationship network related to the drought and flood disaster event.

[0204] By executing steps 1131 to 1133, the basic original popular science text material and the basic text word frequency distribution relationship network can be mapped according to the matching results of the semantic description vector to obtain the original popular science text material and the text word frequency distribution relationship network. The advantage of this is that information from different sources can be integrated and associated, semantic element features related to drought and flood disaster events can be extracted, and the corresponding text word frequency distribution relationship network can be constructed.

[0205] In some independent embodiments, when a text push processing request for a target user is received, as described in step 160, based on the page browsing preference of the target user, a paraphrase text is generated for the multiple paraphrase popular science text material units to obtain the popular science articles on meteorological disasters to be pushed, which at least includes: determining the page browsing preference viewpoint of the target user through the page browsing behavior data of the target user; determining the professional evaluation label of the target user by using the page browsing preference viewpoint; and generating paraphrase text for the multiple paraphrase popular science text material units in combination with the professional evaluation label to obtain the popular science articles on meteorological disasters to be pushed.

[0206] In the drought and flood disaster event scenario, step 160 describes how to generate popular science articles on meteorological disasters to be pushed based on the page browsing preferences of the target user. The following is an example explanation: Assume that there is a weather warning application that provides users with popular science information about drought and flood disasters. The application collects users' page browsing behavior data, including the different pages and content they visit in the application. When a drought and flood disaster event occurs, the application receives a text push processing request for the target user. According to step 160, it analyzes the page browsing preference views of the target user, that is, the browsing behavior data of the user in the application, to understand their preferences for different types of content. For example, if the application finds that the target user often views information related to drought, it can be concluded that the user is more interested in drought issues. This view will be regarded as the page browsing preference view of the target user. Next, the application uses the page browsing preference view of the target user to determine the user's professional evaluation tags. These tags can judge the user's professionalism in the field based on the content that the user is interested in. In this case, if the target user has a high page browsing preference for drought issues, the application can evaluate it as having a certain degree of professionalism in drought issues. Finally, according to the user's professional evaluation tags, the application combines multiple paraphrasing popular science text material units to generate paraphrasing text. These text material units may include information on drought and flood disasters, how to deal with drought or floods, etc. By selecting and integrating these materials based on the user's professional evaluation tags, the application generates popular science articles on meteorological disasters to be pushed to meet the needs and interests of target users.

[0207] It can be seen that step 160 describes that in the drought and flood disaster event scenario, the meteorological disaster popular science articles to be pushed are generated according to the page browsing preferences of the target user, and the user's professional evaluation tags are combined to provide content that meets the user's needs and interests. In this way, by analyzing the page browsing preferences of the target users, their interests in different topics and content can be understood. This enables the system to generate personalized popular science articles for the target users and provide information related to their browsing preferences. This personalized content push can improve the user experience and increase the user's participation in the application; by using the target user's page browsing preferences to determine the user's professional evaluation tags, the user's professionalism in a specific field can be more accurately judged. This helps the system better provide users with popular science information that meets their professional level and needs, and improves the accuracy and credibility of the information; combining the user's professional evaluation tags to generate multiple retelling popular science text material units, more targeted and effective popular science articles can be generated according to the user's needs and professional level. This helps to improve the quality and readability of the article, making it easier for users to understand and digest the meteorological disaster popular science knowledge provided.

[0208] In some independent embodiments, the step 160 of determining the target user's page browsing preference through the target user's page browsing behavior data may include steps 161 to 164.

[0209] Step 161: Obtain page browsing behavior data, perform behavior description recognition on the page browsing behavior data according to a first behavior description recognition strategy, and obtain a first behavior description knowledge vector, wherein the first behavior description recognition strategy is a knowledge mining strategy based on time attention.

[0210] Step 162: Based on the first behavior description knowledge vector of the page browsing behavior data, determine at least one reference page browsing behavior data that matches the page browsing behavior data, and perform behavior description identification on the page browsing behavior data and each reference page browsing behavior data according to a second behavior description identification strategy to obtain a second behavior description knowledge vector, wherein the second behavior description identification strategy is a knowledge mining strategy based on spatial attention.

[0211] Step 163: Based on the identified second behavior description knowledge vector, the page browsing behavior data is matched with each reference page browsing behavior data in turn for behavior operation items, and the target page browsing behavior data in the at least one reference page browsing behavior data is determined as the sample page browsing behavior data, and the number of behavior operation items that match the target page browsing behavior data and the page browsing behavior data is greater than the set number.

[0212] When a drought or flood disaster event scenario is involved, the following example may be used to illustrate the detailed process of step 161 to step 164 .

[0213] In step 164, the page browsing preference viewpoint corresponding to the sample page browsing behavior data is determined using the sample page browsing preference viewpoint corresponding to the sample page browsing behavior data.

[0214] Obtain page browsing behavior data and apply a knowledge mining strategy based on time attention to identify behavior descriptions. Assume that the target user is a farmer who has viewed several pages related to drought and flood disasters, such as drought index, rainfall statistics, etc., using a meteorological website in the past week. The system will analyze the time information of these pages, such as browsing time, frequency, and duration, and generate a first behavior description knowledge vector, which contains information about the user's time preference for drought and flood disaster information.

[0215] In step 162, based on the first behavior description knowledge vector, the system determines at least one reference page browsing behavior data that matches the page browsing behavior data of the target user. For example, the system can select browsing behavior data of several other farmer users as references. Then, through the knowledge mining strategy based on spatial attention, the system will compare the browsing behavior data of the target user with the browsing behavior data of the reference users. The system may find that one of the reference users browsed pages similar to the target user, such as farmland water conservation technology, the impact of climate change on crops, etc., and generate a second behavior description knowledge vector, which contains the spatial correlation between the target user and the reference user.

[0216] In step 163, based on the second behavior description knowledge vector, the system matches the browsing behavior data of the target user with the browsing behavior data of the sample page. Assume that the browsing behavior data of the sample page of the reference user shows browsing operation items for irrigation methods and dryland management. The system will check the browsing behavior data of the target user to see if there are behaviors related to these operation items. If the number of behavior operation items of the target user on similar topics exceeds the set threshold, for example, he also browses pages on farmland irrigation technology and drought response measures, then the browsing behavior data of the sample page of the reference user will be determined as sample data, indicating that the user has a browsing preference for these contents.

[0217] In step 164, the system determines the page browsing preference viewpoints of the target user using the sample page browsing preference viewpoints corresponding to the sample page browsing behavior data. Based on the sample page browsing behavior data, the system can infer the target user's preference for specific topics or content, for example, he may be interested in effective drought mitigation methods. Based on these preference viewpoints, the system can add professional evaluation tags to the target user, such as "interested in drought response measures." In this way, the system can customize personalized meteorological disaster popular science articles to be pushed based on the user's professional evaluation tags, such as providing articles on drought management technology and preventive measures to meet the specific needs of the target user.

[0218] It can be seen that by applying steps 161 to 164, the system can customize personalized popular science articles on meteorological disasters to be pushed by analyzing the page browsing behavior data and preferred views of the target user. In this way, users will receive information that matches their professional level and interests, providing more accurate and targeted services; by customizing articles according to the user's professional evaluation tags, the system can provide content that meets the user's specific needs and interests. This will improve the user experience, enable users to obtain more practical and relevant information, and increase their satisfaction with the system; by customizing personalized popular science articles, the system can ensure that the target user receives information suitable for their professional level. This will improve the effectiveness and influence of popular science work, make it easier for users to understand and respond to meteorological disasters, and thus enhance the public's ability to respond to meteorological disasters; by analyzing and mining the user's page browsing behavior data, the system can understand the user's needs and interests, and then provide relevant articles in a targeted manner. This will help optimize resource utilization, so that popular science content can be more accurately conveyed to users who really need it, and improve the effective utilization of resources.

[0219] On the basis of the above, an AI personalized customized push processing system is provided, comprising a processor and a memory communicating with each other, wherein the processor is used to retrieve a computer program from the memory and implement the above method by running the computer program.

[0220] Based on the above, a computer-readable storage medium is provided, on which a computer program is stored, and the computer program implements the above method when running.

[0221] The above are only 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 substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A method for pushing meteorological disaster text based on artificial intelligence, characterized in that: Applied to an AI personalized customized push processing system, the method includes: Obtaining original popular science text materials and a text word frequency distribution relationship network of meteorological disaster events; wherein the text word frequency distribution relationship network is used to characterize the distribution popularity of relevant words and sentences in drought and flood disaster events, and the text word frequency distribution relationship network is obtained based on word frequency statistics in the original popular science text materials; Performing meteorological disaster event attention analysis on the original popular science text material to obtain a meteorological disaster event material set in the original popular science text material; Through the text word frequency distribution relationship network and the meteorological disaster event material set, the original popular science text material is subjected to a text block disassembly operation to obtain a plurality of original popular science text material units with different text granularities; Parsing out the text retelling professionalism score corresponding to each original popular science text material unit and the text retelling weight factor corresponding to the text retelling professionalism score, wherein the text retelling professionalism score represents the influence coefficient of the text fine-grainedness contained in the original popular science text material unit; For each original popular science text material unit, a text retelling weight factor corresponding to the text retelling professionalism score is used to perform a text retelling operation on the original popular science text material unit to obtain a plurality of retelling popular science text material units; When a text push processing request for a target user is received, based on the page browsing preference of the target user, a paraphrase text is generated for the plurality of paraphrase popular science text material units to obtain a meteorological disaster popular science article to be pushed; The method comprises: performing a text block disassembly operation on the original popular science text material through the text word frequency distribution relationship network and the meteorological disaster event material set to obtain a plurality of original popular science text material units with different text fine-grainedness, including: determining a meteorological disaster event word frequency distribution relationship subnet corresponding to the meteorological disaster event in the text word frequency distribution relationship network based on the meteorological disaster event material set; performing word frequency classification on the meteorological disaster event word frequency distribution relationship subnet by setting a word frequency threshold to obtain a word frequency classification result; and disassembling the original popular science text material into a plurality of original popular science text material units with different text fine-grainedness through the meteorological disaster event word frequency distribution relationship subnet according to the word frequency classification result; The method comprises: for each original popular science text material unit, performing a text retelling operation on the original popular science text material unit using a text retelling weight factor corresponding to the text retelling professionalism score to obtain a plurality of retelling popular science text material units, including: adjusting a set strategy execution variable in a set text material retelling strategy based on the text retelling weight factor to obtain a target text material retelling strategy; performing a text retelling operation on the original popular science text material unit according to the target text material retelling strategy to obtain a plurality of retelling popular science text material units; Among them, according to the target text material retelling strategy, the original popular science text material unit is subjected to text retelling operation to obtain multiple retelling popular science text material units, including: semantic element mining of the original popular science text material unit to obtain basic semantic element feature variables of the original popular science text material unit; according to the target text material retelling strategy, weighted reinforcement is performed through the text retelling weight factor corresponding to the target text material retelling strategy and the basic semantic element feature variables to obtain semantic element feature weighted reinforcement variables; the basic semantic element feature variables of the original popular science text material unit are replaced by the semantic element feature weighted reinforcement variables to obtain retelling popular science text material units.

2. The method according to claim 1, characterized in that The word frequency classification is performed on the meteorological disaster event word frequency distribution relationship subnet by setting a word frequency threshold to obtain a word frequency classification result, including: Performing semantic element mining on the meteorological disaster event word frequency distribution relationship subnet to obtain semantic element feature variables corresponding to the meteorological disaster event word frequency distribution relationship subnet; Determining a word frequency difference quantitative index of the meteorological disaster event word frequency distribution relationship subnet according to the semantic element characteristic variable; The word frequency difference quantification index and the set word frequency threshold are compared and analyzed to obtain the word frequency classification result.

3. The method according to claim 1, characterized in that According to the word frequency classification result, the original popular science text material is disassembled into a plurality of original popular science text material units with different text granularities through the meteorological disaster event word frequency distribution relationship subnet, including: According to the word frequency classification result, a feature coding operation is performed through the semantic element feature variables of the meteorological disaster event word frequency distribution relationship subnet and the semantic element feature variables of the original popular science text material to obtain a feature coding result; According to the feature coding result, the original popular science text material is disassembled into a plurality of original popular science text material units; A fine-grained parsing operation is performed on each original popular science text material unit to obtain multiple original popular science text material units with different text fine-grainedness.

4. The method according to claim 3, characterized in that The step of disassembling the original popular science text material into a plurality of original popular science text material units according to the feature coding result includes: Converting the feature encoding result to obtain a mutually exclusive variable set of local material features; Based on the mutually exclusive variable set of local material features, the meteorological disaster event material set is disassembled into a first popular science text material set and a second popular science text material set; Based on the first popular science text material set and the second popular science text material set, a third popular science text material set is extracted from the original popular science text material to obtain a plurality of original popular science text material units.

5. The method according to claim 1, characterized in that The determining, based on the meteorological disaster event material set, in the text word frequency distribution relationship network, a meteorological disaster event word frequency distribution relationship subnet corresponding to the meteorological disaster event comprises identifying material text position data corresponding to the meteorological disaster event material set; Determine the basic meteorological disaster event word frequency distribution relationship subnet corresponding to the meteorological disaster event in the text word frequency distribution relationship network through the material text position data; The distribution relationship of the basic meteorological disaster event word frequency distribution relationship subnet is updated to obtain the meteorological disaster event word frequency distribution relationship subnet.

6. The method according to claim 1, characterized in that The parsing of the text retelling professionalism score corresponding to each original popular science text material unit and the text retelling weight factor corresponding to the text retelling professionalism score, wherein the text retelling professionalism score represents the influence coefficient of the text fine-grainedness contained in the original popular science text material unit, includes: Compare and analyze the fine-grained text of each original popular science text material unit to obtain contrasting viewpoints; According to the comparative viewpoint, the text fine-grainedness of each original popular science text material unit is sequentially sorted to obtain a sequential sorting result; According to the results of the order arrangement, the text retelling professionalism score corresponding to each original popular science text material unit and the text retelling weight factor corresponding to the text retelling professionalism score are determined.

7. The method according to claim 1, characterized in that The original popular science text materials and the text word frequency distribution relationship network of meteorological disaster events are obtained, including: Obtain the basic original popular science text materials and the basic text word frequency distribution relationship network of meteorological disaster events; Determine a first semantic description vector corresponding to the basic original popular science text material and a second semantic description vector of the basic text word frequency distribution relationship network; According to the first semantic description vector and the second semantic description vector, a semantic matching operation is performed on the basic original popular science text material and the basic text word frequency distribution relationship network to obtain the original popular science text material and the text word frequency distribution relationship network; The step of performing a semantic matching operation on the basic original popular science text material and the basic text word frequency distribution relationship network according to the first semantic description vector and the second semantic description vector to obtain the original popular science text material and the text word frequency distribution relationship network includes: Determine a mapping list between the basic original popular science text material and the basic text word frequency distribution relationship network according to the first semantic description vector and the second semantic description vector; Performing semantic element mining on the basic original popular science text material and the basic text word frequency distribution relationship network to obtain semantic element feature variables corresponding to the basic original popular science text material and semantic element feature variables corresponding to the basic text word frequency distribution relationship network; Based on the mapping list, the semantic element feature variables corresponding to the basic original popular science text material and the semantic element feature variables corresponding to the basic text word frequency distribution relationship network are mapped to obtain the original popular science text material and the text word frequency distribution relationship network.

8. An AI personalized customized push processing system, characterized in that: The AI ​​personalized customized push processing system includes a processor and a memory that communicate with each other, and the processor is used to call a computer program from the memory and implement the method described in any one of claims 1 to 7 by running the computer program.

Images