A multi-platform integrated intelligent topic inspiration generation method and topic inspiration engine
By using a multi-platform integrated intelligent topic inspiration generation method, which utilizes channel interfaces and weighting algorithms to calculate keyword popularity, the problem of finding topic inspiration across multiple platforms for new media professionals has been solved. This method enables rapid and accurate topic generation, reducing workload and improving the accuracy of topic selection.
Patent Information
- Application Number
- CN202411248796.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-06
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-09-06
AI Technical Summary
In the current technology, new media professionals find it difficult to quickly and accurately find the hottest topic inspirations from a large number of news and information platforms when creating new media content, and they lack systematic tools to integrate information from multiple platforms, resulting in a large workload and inaccurate topic selection.
By using a multi-platform integrated intelligent topic inspiration generation method, information rankings are crawled through channel interfaces, and keyword popularity is calculated by combining a keyword library and weighting algorithm to generate a topic list, including first topic weight, second topic weight and third topic weight, and finally constructing a topic list with final weight.
It enables the rapid and accurate acquisition of the latest and hottest keywords across all platforms as inspiration for topics, reducing the workload for new media professionals in searching for topic inspiration across multiple platforms and improving the accuracy and efficiency of topic selection.
Smart Images

Figure CN118780280B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of new media applications, and in particular to a multi-platform integrated intelligent topic inspiration generation method and a topic inspiration engine. Background Art
[0002] At present, with the continuous development of new media technology, more and more traditional media forms are being replaced by short videos and online media platforms. Therefore, when various bloggers and new media people are creating and producing corresponding short videos, articles, public accounts and other new media content, how to quickly and accurately find the hottest topics is related to the playback volume, reprint volume and other data of the new media content they create and produce.
[0003] There are currently a large number of news and information platforms and short video platforms on the market, and a large number of platforms are producing huge amounts of information content every day. These information include hot information, daily information, industry information, financial information, etc. Before producing new media content, new media people need to combine current hot information or their corresponding industry information to select corresponding topics. How to get the best topic inspiration from such a large amount of information is an important part of the work of new media people.
[0004] Currently, each media platform has a hot list, but there is no systematic tool that can integrate the hot information from several media platforms. This results in new media practitioners only being able to select topics on one or a few platforms based on the hot lists integrated by their platforms. On the one hand, due to the different fields and key considerations corresponding to each platform, the hottest information on some platforms may not be the hottest information on all current platforms. On the other hand, new media practitioners are required to manually search for topic inspiration on multiple platforms, which requires a lot of work. Summary of the invention
[0005] In order to improve the effect of multi-platform overall optimization in the topic selection process, the present application provides a multi-platform integrated intelligent topic inspiration generation method and a topic inspiration engine.
[0006] In the first aspect, the present application provides a multi-platform integrated intelligent topic inspiration generation method, which adopts the following technical solutions:
[0007] A method for generating topic inspiration includes the following steps:
[0008] Crawl the information lists on the corresponding platforms through several channel interfaces;
[0009] Adding a number of the above-mentioned information lists to the reserved topic pool, and each of the above-mentioned information lists is regarded as an independent topic list;
[0010] Calculate the first topic weight corresponding to each information based on the position of each information in the topic list;
[0011] Selecting keywords or keyword groups consisting of several keywords corresponding to the titles of the information based on the keyword library, and calculating corresponding second topic selection weights based on the content popularity of the keywords or keyword groups;
[0012] Calculate the third topic weight corresponding to the keyword or the keyword group based on the number of times the keyword or the keyword group appears in the text of the corresponding information and the number of times the keyword or the keyword group appears in the text of other information in the topic list;
[0013] Calculate the final weight corresponding to each of the keywords or the keyword groups based on the first topic selection weight, the second topic selection weight, and the third topic selection weight;
[0014] A preset number of the keywords or keyword groups are selected from the topic pool in order from high to low along the final weight to construct a topic list, wherein the topic list includes the value of the final weight corresponding to each keyword or keyword group and the corresponding link to the information.
[0015] In some of the embodiments, after crawling the information list on the corresponding platform through several channel interfaces, the following steps are also included:
[0016] Determine whether some of the information on each of the information lists is negative information through a preset negative filtering model, and filter the negative information;
[0017] It is determined through a preset specific word library whether there are specific filtering words in the information on each of the information lists, and the information containing the specific filtering words is filtered.
[0018] In some of the embodiments, calculating the first topic weight corresponding to each piece of information based on the position of each piece of information in the topic list includes the following steps:
[0019] Generate a preliminary weight based on the position of the information in the topic list, wherein the higher the position, the greater the preliminary weight;
[0020] generating a maximum weight based on the length of the topic list;
[0021] The platform generation effect addition coefficient corresponding to the topic list is based on the platform generation effect addition coefficient;
[0022] The first topic selection weight is calculated based on the ratio of the preliminary weight to the maximum weight multiplied by the effect addition coefficient.
[0023] In some of the embodiments, calculating the corresponding second topic weight based on the content popularity of each keyword or keyword group includes the following steps:
[0024] Selecting a daily heat range and a real-time heat range from the keyword library, wherein both the daily heat range and the real-time heat range contain a number of keywords;
[0025] Acquire the heat retention value of each of the keywords in the keyword library, and add the keywords whose heat retention value is greater than a first preset value into the daily heat range;
[0026] Obtaining a sudden change value of heat of each keyword in the keyword library, and taking the keyword whose sudden change value increase is greater than a second preset value as a candidate keyword;
[0027] Perform a global online search based on the candidate keyword, and obtain the event time corresponding to the candidate keyword. If the event time matches the current time, add the candidate keyword to the real-time popularity range;
[0028] The corresponding second topic selection weight is generated based on whether the keyword or the keyword group is within the daily popularity range or the real-time popularity range.
[0029] In some embodiments, if the information corresponds to the selected keyword, calculating the third topic weight corresponding to the keyword includes the following steps:
[0030] Obtain the number of times the keyword appears in the corresponding text of the information and define it as the original number;
[0031] Obtain the number of times the keyword appears in the text of each of the other information in the topic list, and select the maximum number and the minimum number;
[0032] The third topic weight is calculated based on the following formula:
[0033] The weight of the third question = (original number - minimum number) / (maximum number - minimum number).
[0034] In some embodiments, if the information corresponds to the selected keyword group, calculating the third topic weight corresponding to the keyword group includes the following steps:
[0035] Obtaining a number of the keywords in the keyword group;
[0036] Obtain the number of times each keyword appears in the corresponding text of the information and define it as the original number;
[0037] Respectively obtain the number of times each keyword appears in the text of each other information in the topic list, and respectively select the maximum number and the minimum number corresponding to each keyword;
[0038] The third topic weight corresponding to each keyword is calculated based on the following formula:
[0039] The weight of the third question = (original number - minimum number) / (maximum number - minimum number);
[0040] Determine whether the difference between the third topic selection weights of the plurality of keywords is greater than a preset value;
[0041] If it is greater than, selecting the third topic selection weight with the largest value as the third topic selection weight corresponding to the keyword group;
[0042] If not, the third topic weight corresponding to the keyword group is taken as the average value of several third topic weights.
[0043] In some of the embodiments, calculating the final weight corresponding to each of the keywords or the keyword groups based on the first topic selection weight, the second topic selection weight, and the third topic selection weight includes the following steps:
[0044] The final weight is calculated by the following formula:
[0045] ,
[0046] Among them, n represents the number of the topic list, k1, k2, k3 are all calculation coefficients, and the sum of k1, k2, k3 is 1.
[0047] In some of the embodiments, the final weight corresponding to each of the keywords or the keyword groups is calculated based on the first topic selection weight, the second topic selection weight, and the third topic selection weight, further comprising the following steps:
[0048] Determine whether there is a classified list for the information list on the corresponding platform;
[0049] If so, obtain the classification information corresponding to the classification list, and determine the number of each classification information in the topic pool;
[0050] If the keyword or the keyword group corresponds to the classification list, multiplying the calculated final weight by the corresponding classification coefficient;
[0051] Among them, the classification coefficient is greater than 1 and its size is inversely proportional to the amount of the classification information corresponding to its classification list in the topic pool.
[0052] In some of the embodiments, in the topic list, the method for generating the link of the information corresponding to each keyword or keyword group includes the following steps:
[0053] Acquire the information with the largest first topic weight among all the information containing the keyword or the keyword group as the target information;
[0054] Acquire the link corresponding to the target information based on the channel interface;
[0055] The link is sent and bound to the keyword or the keyword group.
[0056] In the second aspect, the present application provides a topic inspiration engine, which adopts the following technical solution:
[0057] A topic inspiration engine, comprising:
[0058] Channel interface, used to connect to each of the platforms;
[0059] A crawler module, used to crawl information lists on corresponding platforms connected to the channel interfaces;
[0060] A topic pool, used to store a plurality of the information lists, and convert each of the information lists into a corresponding plurality of independent topic lists;
[0061] A weight extraction module is used to calculate the first topic weight corresponding to each piece of information based on the position of each piece of information in the topic list; further select the keyword or keyword group consisting of several keywords corresponding to the title of each piece of information based on the keyword library, and calculate the corresponding second topic weight based on the content popularity of each keyword or keyword group; further calculate the third topic weight corresponding to the keyword or keyword group based on the number of times the keyword or keyword group appears in the text of the corresponding piece of information and the number of times it appears in the text of other pieces of information in the topic list;
[0062] A topic selection weight calculation module, used to calculate the final weight corresponding to each of the keywords or the keyword groups based on the first topic selection weight, the second topic selection weight and the third topic selection weight;
[0063] A topic generation module is used to select a preset number of the keywords or keyword groups in the topic pool in order from high to low along the final weight to construct a topic list, wherein the topic list includes the final weight value corresponding to each keyword or keyword group and the corresponding link to the information.
[0064] The technical solution provided by the embodiments of the present application has the following technical effects:
[0065] Integrate a large number of platforms, crawl the information lists on various platforms, and use weighted algorithms to make horizontal comparisons of keywords extracted from different hot information on multiple platforms. Judge the popularity of each keyword as a whole, so as to quickly and accurately obtain the latest and hottest keywords on the entire platform as the preferred topics. Users can directly get corresponding topic inspiration by calculating and screening out keywords, reducing workload. BRIEF DESCRIPTION OF THE DRAWINGS
[0066] Figure 1 It is a schematic diagram of the steps of a multi-platform integrated intelligent topic inspiration generation method provided in an embodiment of the present application.
[0067] Figure 2 This is a module connection diagram of a topic inspiration engine provided in an embodiment of the present application. DETAILED DESCRIPTION
[0068] To more clearly understand the purpose, technical solutions and advantages of the present application, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. However, it should be understood by those of ordinary skill in the art that the present application can be implemented without these details. In some cases, in order to avoid unnecessary descriptions that make the various aspects of the present application obscure, the well-known methods, processes, systems, components and / or circuits that have been described at a higher level will not be described in detail. For those of ordinary skill in the art, it is obvious that various changes can be made to the embodiments disclosed in the present application, and without departing from the principles and scope of the present application, the general principles defined in the present application can be applied to other embodiments and application scenarios. Therefore, the present application is not limited to the embodiments shown, but conforms to the broadest scope consistent with the scope claimed for protection of the present application.
[0069] It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention. In addition, the technical features involved in each embodiment of the present invention described below can be combined with each other as long as there is no conflict between them.
[0070] In the description of this application, "several" means one or more, "more" means more than two, "greater than", "less than", "exceed", etc. are understood to exclude the number itself, and "above", "below", "within", etc. are understood to include the number itself. If there is a description of "first" or "second", it is only used to distinguish the technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.
[0071] In the description of the present application, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples.
[0072] The embodiment of the present application discloses a multi-platform integrated intelligent topic selection inspiration generation method.
[0073] like Figure 1 As shown, a multi-platform integrated intelligent topic inspiration generation method includes the following steps:
[0074] S100 crawls the information lists on the corresponding platforms through several channel interfaces.
[0075] The channel interface is used to connect to the mainstream news platforms, self-media platforms, short video platforms, etc. on the Internet, and crawl the latest information lists on various mainstream popular channels through crawler algorithms. The information list is a hot list obtained by sorting on each platform based on the platform's own information heat calculation method. The crawled content includes the title of the information, the content of the information and the source of the information.
[0076] S200, adding a number of information lists to the reserved topic pool, and each information list is used as an independent topic list.
[0077] The topic pool is an integrated library that can store a predetermined number of information lists crawled from various platforms. At the same time, users can also view the information lists on various platforms in the topic pool.
[0078] When the information lists of each platform enter the topic pool, after subsequent information preprocessing, each information list becomes an independent topic list, and each piece of information on the topic list is arranged in the same order as in the information list.
[0079] S300, calculating the first topic weight corresponding to each information based on the position of each information in the topic list.
[0080] First, the first topic weight is obtained by ranking each piece of information on its topic list. Generally speaking, the higher the position of a piece of information in the information list, the higher the corresponding information popularity. Since the topic list is generated accordingly based on the information list, the higher the position of a piece of information on the topic list, the higher the corresponding information popularity.
[0081] The ranking of each piece of information is used as the first reference factor for judging the popularity of the topic.
[0082] S400, selecting keywords corresponding to the titles of each piece of information or a keyword library consisting of several keywords based on the keyword library, and calculating the corresponding second topic selection weight based on the content popularity of each keyword or the keyword library.
[0083] The keyword library is a word library generated by the keyword extraction model trained in advance by the large model and the neural network algorithm when the topic inspiration engine is created. It contains a large number of keywords that conform to common semantics and human language. It can use the keyword library commonly used in current technology, such as NLTK library, jieba library, etc. The object of keyword extraction is the title of the information, because the title of the information can often comprehensively summarize the key content of the information, and the number of words in the title is relatively small, so the keywords finally extracted are often more accurate and the results are relatively small.
[0084] Depending on the content of the information, the number of keywords selected in the title of each piece of information is not fixed. For example, in the information "Express delivery companies are accelerating their entry into villages, and the volume of parcels in some areas has increased tenfold", the selected keywords are "express delivery companies", and in the information "In the xx Sports Games, xxx athlete won the gold medal", the selected keywords are "xx Sports Games" and "xxx athlete".
[0085] The selection of keywords mainly includes the following aspects: major events, major conferences, major policies, major awards, major industrial enterprises, public figures, etc. Generally speaking, in order to reduce the workload of subsequent topic selection and improve the accuracy of topic selection, there are at most three keywords in the keyword group of a piece of information. When more than three keywords can be selected for a piece of information, the first three keywords in the word order of the information are selected by default.
[0086] After selecting the keywords or keyword groups corresponding to each piece of information, the content popularity of the keywords or keyword groups is used as the second reference factor to judge the popularity of the topic. The content popularity of a keyword or keyword group is characterized by the social activity of the event or object corresponding to the keyword during this period of time. For example, during holidays, the content popularity of keywords related to holiday travel is relatively high.
[0087] The second topic weight is mainly characterized by the level of activity of the social public opinion environment inferred by keywords of a piece of information.
[0088] S500, calculating a third topic weight corresponding to the keyword or keyword group based on the number of times the keyword or keyword group appears in the text of the corresponding information and the number of times the keyword or keyword group appears in the text of other information in the topic list.
[0089] The number of times a keyword or keyword group appears in each piece of information is used as the third reference factor to judge the popularity of the topic. If a keyword frequently appears in multiple pieces of information in an information list, then the event or person corresponding to the keyword is currently very popular, and the corresponding third topic weight will be greater.
[0090] S600: Calculate the final weight corresponding to each keyword or keyword group based on the first topic selection weight, the second topic selection weight, and the third topic selection weight.
[0091] The final weight corresponding to each keyword or keyword group is calculated by using the multiple sub-weights calculated above. The final weight is used to represent the overall weight of each keyword or keyword group in the information lists of all crawled platforms. The higher the final weight, the more popular it is on various platforms, and the more suitable it is as a topic content.
[0092] S700, select a preset number of keywords or keyword groups in the topic pool in order from high to low final weights to construct a topic list, wherein the topic list includes the final weight value corresponding to each keyword or keyword group and the link to the corresponding information.
[0093] After obtaining the final weight of each keyword or keyword group, several keywords or keyword groups with higher final weights can be selected to generate a topic list. The topic list displays several selected and more popular keywords for topic selection, and also displays the corresponding weight value of the keyword and the link to the information matched by the keyword.
[0094] Through the above method, a large number of platforms are connected and integrated, and the information lists on various platforms are crawled to make a horizontal comparison of the keywords extracted from different hot information on multiple platforms through a weight algorithm, so as to judge the popularity of each keyword as a whole, so as to quickly and accurately obtain the latest and hottest keywords on the entire platform as the preferred topics. Users can directly get the corresponding topic inspiration by calculating the selected keywords, thereby reducing the workload.
[0095] In some other embodiments, after crawling the information list on the corresponding platform through several channel interfaces, the following steps are also included:
[0096] S110, judging whether some information on each information list is negative information through a preset negative filtering model, and filtering the negative information.
[0097] Because some short video platforms and self-media platforms have an impact on the promotion and diversion of some sensitive negative content, in order to reduce the probability of generating negative keywords during the topic selection process and reduce misleading new media people, information pre-processing can be performed after crawling the information list.
[0098] First, compliance checks are carried out, and content such as illegal activities, wars, natural disasters, fraud, etc. are filtered out through big model technology.
[0099] The negative filtering model is obtained through the machine learning training algorithm in common technology. The model is trained through a large training set of negative and illegal words. The trained negative filtering model is used to classify and filter real-time text to automatically determine whether there is bad or illegal content in the text.
[0100] S111, judging whether there are specific filtering words in a number of information on each information list through a preset specific word library, and filtering the information containing the specific filtering words.
[0101] A large number of classification sets that conform to different industries, fields, and types are stored in the specific vocabulary. Each classification set contains keywords corresponding to the corresponding industry, field, and type. The specific vocabulary can be selected by the user himself. For example, if a new media user mainly produces information on the tram industry, he can check the classification sets corresponding to other unrelated industries in the specific vocabulary, such as stocks, economy, entertainment, etc., and through the selected filtering range, the information with corresponding specific filter words in some information can be deleted, so that the topic obtained by the user is more in line with the corresponding field, and the topics corresponding to the fields not involved will not occupy the position in the final topic list; at the same time, when calculating the topic weight in the subsequent topic, by deleting irrelevant information, the final calculation result of the topic weight required for certain specific fields and industries will also be more accurate.
[0102] In some other embodiments, calculating the first topic weight corresponding to each information based on the position of each information in the topic list includes the following steps:
[0103] S310, generating a preliminary weight based on the ranking of the information in the topic list, wherein the higher the ranking, the greater the preliminary weight.
[0104] Each topic list is an independent list, and the initial weight is calculated based on the position of each piece of information on the list. The information lists of each platform use certain algorithms, such as popularity priority, time priority, and unit time popularity ratio priority to sort the information. Often, the information with the highest popularity or the shortest release time will be at the top of the information list. The popularity of the information can be judged based on its position on the list.
[0105] S320, generating a maximum weight based on the length of the topic list.
[0106] The value of the maximum weight is related to the length of the topic list. If there are 10 pieces of information in the topic list, the maximum weight is 10. The larger the length, the greater the maximum weight.
[0107] S330, generating an effect bonus coefficient based on the platform corresponding to the topic list.
[0108] Different platforms correspond to different scales, such as popularity, number of users, activity, online volume, clicks, reposts, comments, etc., and different scales represent the size of the platform. The larger the size of the platform, the more popular the information on it is, and the greater its authority and audience, then the greater the effect bonus coefficient corresponding to the topic list.
[0109] Since the final weight calculated is used to represent the popularity of a keyword in an information, the effect addition coefficient is the addition amount to the weight of the first topic in the calculation process. The larger the platform and the more active the users are, the greater the addition to the popularity. In the embodiment of the present application, the effect addition coefficient is a value between 1 and 2. When the scale of the platform is extremely small, the effect addition coefficient is 1, indicating that there is no platform addition.
[0110] S340, calculating the first topic weight according to the ratio of the preliminary weight to the maximum weight multiplied by the effect addition coefficient.
[0111] The first topic weight is calculated by (initial weight / maximum weight)*effect addition coefficient. The first topic weight is represented by the information weight corresponding to a single piece of information.
[0112] In some other embodiments, calculating the corresponding second topic weight based on the content popularity of each keyword or keyword group includes the following steps:
[0113] S410, a daily heat range and a real-time heat range are selected in the keyword library, and the daily heat range and the real-time heat range both contain a number of keywords.
[0114] Two ranges are set in the reserved keyword library, that is, two empty sets, and any number of keywords that meet the requirements can be placed in each empty set. Among them, one empty set corresponds to the daily popularity range, in which the stored keywords are some keywords that have maintained a certain popularity for a long time, such as "semiconductor", "tram", etc., and the other empty set corresponds to the real-time popularity range, in which the stored keywords are keywords that suddenly increase in popularity within a specific period of time, such as "press conference", "concert", etc.
[0115] S420, obtaining the heat retention value of each keyword in the keyword library, and adding the keywords whose heat retention value is greater than the first preset value into the daily heat range.
[0116] First, determine the heat retention value of each keyword, which is based on the number of times each keyword appears in each list in historical data, the time it is maintained, and other data. For example, if a keyword frequently appears in various information lists for 2-3 months, its overall heat is not high, but it is maintained for a long time. The heat retention value calculated for this type of keyword is larger. When the heat retention value is greater than the first preset value, the keyword is added to the daily heat range.
[0117] S430, obtaining the heat mutation value of each keyword in the keyword library, and taking the keyword whose heat mutation value increase is greater than the second preset value as a candidate keyword.
[0118] The heat mutation value is determined based on the number of times each keyword appears in each list in the historical data, the time it is maintained, and the number of times the keyword appears in the recent period of time and the time it is maintained. If a low-heat keyword that has appeared in the list very few times in the past suddenly appears frequently on the list in a period of time and maintains it for a period of time, it is considered that the heat mutation value is large, and when the heat mutation value is greater than the second preset value, the keyword is used as a candidate keyword and enters the next round of judgment.
[0119] S440, performing a global network search based on the candidate keyword and obtaining the event time corresponding to the candidate keyword. If the event time matches the current time, the candidate keyword is added to the real-time popularity range.
[0120] For example, if the candidate keyword is "concert", then the keyword is searched in the corresponding Internet search engine through the interface, and the event corresponding to the keyword and the event time corresponding to the event are obtained. For example, if the searched event is xxx holding a concert, then the event time is represented by the time period when the social opinion corresponding to the event is hotly discussed. For example, the time of the concert is 2024.5.20-2024.5.23, and its event time can be extended forward and backward by half a month, that is, 2024.5.5-2024.6.4, and the extended time is used as the public opinion increment corresponding to a highly discussed event, such as the pre-sale time before the concert and the hot discussion time after the concert.
[0121] If the current time is within an event, it is considered that the candidate keyword is within the occurrence of a certain hotly discussed event, and there is a possibility and rationality for a sudden change in heat, so the candidate keyword can be added to the real-time heat range at this time.
[0122] S450, generating a corresponding second topic selection weight based on whether the keyword or keyword group is within the daily popularity range or the real-time popularity range.
[0123] The second topic weights corresponding to keywords or keyword groups are different in different ranges. Generally speaking, if a keyword is within the daily popularity range, it means that the keyword is a reasonable and effective hot topic in a longer period of time. Although it does not completely match the optimal topic at the current time, it has a certain popularity at any time; and when a keyword is within the real-time popularity range, it means that the keyword is the most hotly discussed topic at the current time, and it can bring higher traffic and attention as a topic. Therefore, the second topic weight corresponding to the keyword within the real-time popularity range is greater than the second topic weight within the daily popularity range, and the specific size of the second topic weight is calculated based on the time and number of times the keyword has been on the information list in the historical data, and the calculation method can be modified according to the actual situation.
[0124] At the same time, if a keyword or keyword group is neither in the daily popularity range nor in the real-time popularity range, it means that the keyword is not a hot topic at present or in history, so the second topic weight generated at this time is smaller than the second topic weight generated in the other two cases.
[0125] In other embodiments, if the information corresponds to a keyword, calculating the third topic weight corresponding to the keyword includes the following steps:
[0126] S510, obtaining the number of times a keyword appears in the body of the corresponding information and defining it as the original number.
[0127] Count the number of times a keyword appears in the body of the news article.
[0128] S511, obtaining the number of times the keyword appears in the text of each other information in the topic list where the keyword is located, and selecting the maximum number and the minimum number.
[0129] Calculate the number of times the keyword appears in the text of each other information in the topic list, and select the maximum and minimum number of occurrences.
[0130] S512, calculating the third topic weight based on the following formula:
[0131] The weight of the third question = (original number - minimum number) / (maximum number - minimum number).
[0132] The weight represented by the number of occurrences of each keyword is calculated through normalization. For example, if the original number is 6, the minimum number of occurrences is 3, and the maximum number is 8, then the corresponding third topic weight is 3 / 5. The closer the original number is to the maximum number, the greater the corresponding third topic weight. Conversely, the closer the original number is to the minimum number, the smaller the corresponding third topic weight.
[0133] If a keyword exists in multiple information in its corresponding topic list, it is necessary to calculate the third topic weight corresponding to the keyword in each information separately, and add up the multiple calculated third topic weights. The result of the addition is used as the third topic weight value of the keyword in its corresponding topic list.
[0134] In other embodiments, if the information corresponds to a keyword group, calculating the third topic weight corresponding to the keyword group includes the following steps:
[0135] S520, obtaining a number of keywords in the keyword group.
[0136] If a keyword group is selected, the keyword group is decomposed into several keywords.
[0137] S521, respectively obtain the number of times a keyword appears in the body of the corresponding information and define it as the original number.
[0138] S522, respectively obtain the number of times each keyword appears in the text of each other information in the topic list where it is located, and respectively select the maximum number and the minimum number corresponding to each keyword.
[0139] S523, calculating the third topic weight corresponding to each keyword based on the following formula:
[0140] The weight of the third question = (original number - minimum number) / (maximum number - minimum number).
[0141] The third topic weight corresponding to each keyword is obtained through the above steps, which is the same as steps S510-S512. At the same time, when the number of information in which the keyword appears in the topic list is greater than 1, the final third topic weight is also obtained by adding.
[0142] S524, determining whether the difference between the third topic weights of a plurality of keywords is greater than a preset value.
[0143] Different from the third topic weight of the keyword corresponding to the information, if the information corresponds to a keyword group, it is also necessary to determine the difference between the third topic weights of each decomposed keyword. When the difference between multiple keywords is greater than or less than the preset value, different third topic weight selection methods correspond to them.
[0144] S525: If it is greater than, then select the third topic selection weight with the largest value as the third topic selection weight corresponding to the keyword group.
[0145] If the difference in the third topic weights between multiple keywords is greater than the preset value, it means that the independent heat of multiple keywords in a keyword group is not balanced. This may be caused by the different semantic effects of different words in the text. For example, the title of an information is: "In the x sports meeting, y athlete won the gold medal", and the keyword group is x sports meeting-y. In the text, x sports meeting represents the current overall event, and y is a person's name. According to the semantic logic and writing ideas, after the sports meeting is introduced at the beginning of the text, the subsequent content will be elaborated around the athlete y. Therefore, in this case, the number of times y appears is much higher than the number of times x sports appears. As a description of the overall event, the heat effect of x sports meeting will not be affected by the small number of appearances. Therefore, in order to ensure the accurate calculation of the weight for the number of appearances, the keyword with the largest number of appearances among multiple keywords, that is, the value with the largest third topic weight, can be selected as the third topic weight of the keyword group. At this time, keywords with very few appearances but high heat will not affect the overall weight value.
[0146] S526: If it is not greater than, then the average value of several third topic weights is used as the third topic weight corresponding to the keyword group.
[0147] If the difference between the third topic weights of multiple keywords in a keyword group is less than the preset value, it means that the number of times the multiple keywords appear is not much different, and at the same time, the weight influence of the multiple keywords on the calculation of the number factor is balanced. At this time, because there are multiple keywords in the keyword group, in order to ensure the numerical balance of the third topic weights between keyword groups composed of a single keyword or multiple keywords, it is necessary to calculate the average value of each calculated third topic weight to obtain the corresponding third topic weight.
[0148] In some other embodiments, calculating the final weight corresponding to each keyword or keyword group based on the first topic selection weight, the second topic selection weight, and the third topic selection weight includes the following steps:
[0149] S610, calculate the final weight using the following formula:
[0150] .
[0151] Among them, n represents the number of topic lists, k1, k2, and k3 are all calculation coefficients, and the sum of k1, k2, and k3 is 1.
[0152] k1, k2, and k3 are all corresponding numerical coefficients, which represent the proportion of each topic weight considered in the final weight. In the embodiment of the present application, k1 and k2 are both 0.3, and k3 is 0.4.
[0153] The final weight needs to add up the weights corresponding to each topic list, because the keywords will appear in each list. After the topic lists are compared horizontally, the values independently calculated for multiple topic lists need to be added together to obtain the overall popularity weight of the keyword or keyword group.
[0154] In some other embodiments, calculating the final weight corresponding to each keyword or keyword group based on the first topic selection weight, the second topic selection weight, and the third topic selection weight further includes the following steps:
[0155] S620, determining whether there is a classified list in the information list on the corresponding platform.
[0156] Classification lists are lists that are subdivided based on different specific fields and industries. For example, in addition to the overall list, some platforms’ information lists also have economic information lists, entertainment information lists, industrial information lists, etc.
[0157] S621, if it exists, obtain the classification information corresponding to the classification list, and determine the number of each classification information in the topic pool.
[0158] If it exists, you need to obtain the specific classification information corresponding to the classification list and calculate the number of lists of various categories in the topic pool.
[0159] S622: If the keyword or keyword group corresponds to a classification list, the calculated final weight is multiplied by the corresponding classification coefficient.
[0160] If a keyword or keyword group corresponds to a classified list, the final weight corresponding to the keyword or keyword group needs to be multiplied by a coefficient, which is the added value of lists with different detailed classifications. For example, if a keyword appears in an economics list, it is very likely to correspond to economics content. It is difficult for it to appear in other types of lists, but its absence in other lists does not mean that its current popularity is low. It just appears less frequently due to the special type of the list. Therefore, in order to avoid the accuracy of the weight value of keywords in specific fields and specific industries due to field limitations when calculating the final weight, it is necessary to multiply it by the classification coefficient to compensate for the topic limitations caused by the special type of the list.
[0161] Among them, the classification coefficient is greater than 1 and its size is inversely proportional to the number of classification information corresponding to its classification list in the topic pool. In other words, when the number of times a field corresponding to a classification list appears in the topic pool is less, the field represented by its classification list is more niche, the classification coefficient value that needs to be multiplied is larger, and the corresponding weight calculation support is greater. Conversely, when the number of times a field corresponding to a classification list appears in the topic pool is more, the field represented by its classification list is more popular, the classification coefficient value that needs to be multiplied is smaller, and the corresponding weight calculation support is smaller.
[0162] In other embodiments, in the topic list, a method for generating a link to information corresponding to each keyword or keyword group includes the following steps:
[0163] S710, obtaining information with the largest first topic weight among all information containing the keyword or keyword group as target information.
[0164] S720, obtaining a link corresponding to the target information based on the channel interface.
[0165] S730, sending the link bound to the keyword or keyword group.
[0166] When a keyword or a keyword group corresponds to multiple information, priority is given to selecting the information with the largest first-selection weight among the several information containing the keyword or keyword group. This information corresponds to the information that contains the keyword or keyword group and is ranked at the front, that is, the hottest information.
[0167] After selecting the corresponding information, the information is used as the target information, and the link of the target information is obtained through the channel interface. Later, when the user selects a keyword or keyword group and wants to view the corresponding information content, he can directly select the link corresponding to the keyword or keyword group to jump to the corresponding information.
[0168] In some other embodiments, after the user selects a keyword or keyword group corresponding to a corresponding topic, the following further includes:
[0169] Select the link corresponding to the keyword or keyword group and jump to the main text of the information. If the user wants to generate a topic based on the information, the main text of the information is passed in, and the negative opinions in the main text of the information are obtained and filtered through the negative filtering model. At the same time, the key sentences in the text are obtained through other big data models, including positive opinion sentences, summary sentences, etc., and the topic is generated based on the corresponding data information.
[0170] In other embodiments, after generating topic ideas, a topic script can be generated based on a preset trained big model. The big model learns and trains the writing style, popular Internet words, writing techniques, etc. of a specific article through a large number of script learning sets, and generates a spoken script of the corresponding style based on the incoming topic ideas brought into the big model.
[0171] like Figure 2 As shown, the present application also discloses a topic inspiration engine, including:
[0172] Channel interface, used to connect to various platforms.
[0173] The crawler module is used to crawl the information lists on the corresponding platforms connected to several channel interfaces;
[0174] The topic pool is used to place several information lists and convert each information list into several corresponding independent topic lists.
[0175] The weight extraction module is used to calculate the first topic weight corresponding to each piece of information based on the position of each piece of information in the topic list; it also selects the keyword corresponding to the title of each piece of information or a keyword group consisting of several keywords based on the keyword library, and calculates the corresponding second topic weight based on the content popularity of each keyword or keyword group; it also calculates the third topic weight corresponding to the keyword or keyword group based on the number of times the keyword or keyword group appears in the text of the corresponding information and the number of times it appears in the text of other information in the topic list.
[0176] The topic selection weight calculation module is used to calculate the final weight corresponding to each keyword or keyword group based on the first topic selection weight, the second topic selection weight and the third topic selection weight.
[0177] The topic generation module is used to select a preset number of keywords or keyword groups in the topic pool in order from high to low final weight to construct a topic list. The topic list contains the final weight value corresponding to each keyword or keyword group and the corresponding information link.
[0178] The implementation principle is:
[0179] Integrate a large number of platforms, crawl the information lists on various platforms, and use weighted algorithms to make horizontal comparisons of keywords extracted from different hot information on multiple platforms. Judge the popularity of each keyword as a whole, so as to quickly and accurately obtain the latest and hottest keywords on the entire platform as the preferred topics. Users can directly get corresponding topic inspiration by calculating and screening out keywords, reducing workload.
[0180] It should be understood that although the steps in the flowchart of the accompanying drawings are shown in sequence according to the instructions of the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise clearly stated in this document, the execution of these steps is not strictly limited in order and can be performed in other orders.
[0181] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.
Claims
1. A multi-platform integrated intelligent topic inspiration generation method, characterized in that: The following steps are involved: Crawl the information lists on the corresponding platforms through several channel interfaces; Adding a number of the above-mentioned information lists to the reserved topic pool, and each of the above-mentioned information lists is regarded as an independent topic list; Calculate the first topic weight corresponding to each information based on the position of each information in the topic list; Selecting keywords or keyword groups consisting of several keywords corresponding to the titles of the information based on the keyword library, and calculating corresponding second topic selection weights based on the content popularity of the keywords or keyword groups, wherein the content popularity is characterized by the degree of social activity of the events or objects corresponding to the keywords or keyword groups in the corresponding time period; Calculate the third topic weight corresponding to the keyword or the keyword group based on the number of times the keyword or the keyword group appears in the text of the corresponding information and the number of times the keyword or the keyword group appears in the text of other information in the topic list; Specifically, based on the keyword or the keyword group corresponding to the information, directly obtain the keyword or respectively obtain the number of times each keyword in the keyword group appears in the text of the corresponding information and define it as the original number, obtain the number of times the keyword or each keyword in the keyword group appears in the text of each other information in the topic list where it is located, and select the maximum number and the minimum number corresponding to each keyword; The third topic weight corresponding to the keyword is calculated based on the following formula: The weight of the third question = (original number - minimum number) / (maximum number - minimum number); Wherein, if the information corresponds to a keyword group, it is determined whether the difference between the third topic weights of several keywords is greater than a preset value. If it is greater, the third topic weight with the largest value is selected as the third topic weight corresponding to the keyword group. If it is not greater, the average value of several third topic weights is used as the third topic weight corresponding to the keyword group. Calculate the final weight corresponding to each of the keywords or the keyword groups based on the first topic selection weight, the second topic selection weight, and the third topic selection weight; A preset number of the keywords or keyword groups are selected from the topic pool in order from high to low along the final weight to construct a topic list, wherein the topic list includes the value of the final weight corresponding to each keyword or keyword group and the corresponding link to the information.
2. The multi-platform integrated intelligent topic inspiration generation method according to claim 1 is characterized in that: After crawling the information list on the corresponding platform through several channel interfaces, the following steps are also included: Determine whether some of the information on each of the information lists is negative information through a preset negative filtering model, and filter the negative information; It is determined through a preset specific word library whether there are specific filtering words in the information on each of the information lists, and the information containing the specific filtering words is filtered.
3. The multi-platform integrated intelligent topic inspiration generation method according to claim 1 is characterized in that: Calculating the first topic weight corresponding to each piece of information based on the position of each piece of information in the topic list includes the following steps: Generate a preliminary weight based on the position of the information in the topic list, wherein the higher the position, the greater the preliminary weight; generating a maximum weight based on the length of the topic list; The platform generation effect addition coefficient corresponding to the topic list is based on the platform generation effect addition coefficient; The first topic selection weight is calculated according to the ratio of the preliminary weight to the maximum weight multiplied by the effect addition coefficient.
4. The multi-platform integrated intelligent topic selection inspiration generation method according to claim 1 is characterized in that: Calculating the corresponding second topic weight based on the content popularity of each keyword or keyword group includes the following steps: Selecting a daily heat range and a real-time heat range from the keyword library, wherein both the daily heat range and the real-time heat range contain a number of keywords; Acquire the heat retention value of each of the keywords in the keyword library, and add the keywords whose heat retention value is greater than a first preset value into the daily heat range; Obtaining a sudden change value of heat of each keyword in the keyword library, and taking the keyword whose sudden change value increase is greater than a second preset value as a candidate keyword; Perform a global online search based on the candidate keyword, and obtain the event time corresponding to the candidate keyword. If the event time matches the current time, add the candidate keyword to the real-time popularity range; The corresponding second topic selection weight is generated based on whether the keyword or the keyword group is within the daily popularity range or the real-time popularity range.
5. The multi-platform integrated intelligent topic inspiration generation method according to claim 1 is characterized in that: Calculating the final weight corresponding to each of the keywords or the keyword groups based on the first topic selection weight, the second topic selection weight, and the third topic selection weight includes the following steps: The final weight is calculated by the following formula: , Among them, n represents the number of the topic list, k1, k2, k3 are all calculation coefficients, and the sum of k1, k2, k3 is 1.
6. The multi-platform integrated intelligent topic inspiration generation method according to claim 5 is characterized in that: The final weight corresponding to each of the keywords or the keyword groups is calculated based on the first topic selection weight, the second topic selection weight, and the third topic selection weight, and the following steps are also included: Determine whether there is a classified list for the information list on the corresponding platform; If so, obtain the classification information corresponding to the classification list, and determine the number of each classification information in the topic pool; If the keyword or the keyword group corresponds to the classification list, the calculated final weight is multiplied by the corresponding classification coefficient; wherein the classification coefficient is greater than 1 and its size is inversely proportional to the amount of the classification information corresponding to the classification list in the topic pool.
7. The multi-platform integrated intelligent topic selection inspiration generation method according to claim 1 is characterized in that: In the topic list, the method for generating the link of the information corresponding to each keyword or keyword group includes the following steps: Acquire the information with the largest first topic weight among all the information containing the keyword or the keyword group as the target information; Acquire the link corresponding to the target information based on the channel interface; The link is sent and bound to the keyword or the keyword group.
8. A topic inspiration engine, characterized in that: include: Channel interface, used to connect to various platforms; A crawler module, used to crawl information lists on corresponding platforms connected to the channel interfaces; A topic pool, used to store a plurality of the information lists, and convert each of the information lists into a corresponding plurality of independent topic lists; A weight extraction module is used to calculate the first topic weight corresponding to each piece of information based on the position of each piece of information in the topic list; further select the keyword or keyword group consisting of several keywords corresponding to the title of each piece of information based on the keyword library, and calculate the corresponding second topic weight based on the content popularity of each keyword or keyword group, wherein the content popularity is characterized by the social activity level of the event or object corresponding to the keyword or keyword group in the corresponding time period; further calculate the third topic weight corresponding to the keyword or keyword group based on the number of times the keyword or keyword group appears in the text of the corresponding piece of information and the number of times the keyword or keyword group appears in the text of other pieces of information in the topic list; It is also used to directly obtain the keyword or the keyword group corresponding to the information, respectively obtain the number of times each keyword in the keyword group appears in the text of the corresponding information and define it as the original number, obtain the number of times each keyword or the keyword in the keyword group appears in the text of each other information in the topic list, and select the maximum number and the minimum number corresponding to each keyword; The third topic weight corresponding to the keyword is calculated based on the following formula: The weight of the third question = (original number - minimum number) / (maximum number - minimum number); Wherein, if the information corresponds to a keyword group, it is determined whether the difference between the third topic weights of several keywords is greater than a preset value. If it is greater, the third topic weight with the largest value is selected as the third topic weight corresponding to the keyword group. If it is not greater, the average value of several third topic weights is used as the third topic weight corresponding to the keyword group. A topic selection weight calculation module, used to calculate the final weight corresponding to each of the keywords or the keyword groups based on the first topic selection weight, the second topic selection weight and the third topic selection weight; A topic generation module is used to select a preset number of the keywords or keyword groups in the topic pool in order from high to low along the final weight to construct a topic list, wherein the topic list includes the final weight value corresponding to each keyword or keyword group and the corresponding link to the information.
