A theme pushing method, device and electronic equipment

Abstract

The one or more embodiments of the specification provide a subject pushing method, device and electronic equipment, wherein the method comprises: acquiring at least one target seasonal point of interest (POI) corresponding to a current time, the target seasonal POI being a popular POI at the current time determined based on the traffic of the POI; respectively acquiring user comment texts corresponding to each target seasonal POI; extracting candidate subjects from the user comment texts corresponding to each target seasonal POI to obtain a subject recall set, the subject recall set including a plurality of candidate subjects; and selecting a subject from the plurality of candidate subjects included in the subject recall set for pushing.

Description

Technical Field

[0001] This specification relates to one or more embodiments in the field of search recommendation technology, and more particularly to a topic recommendation method, apparatus, and electronic device. Background Technology

[0002] Topic mining is a popular technique, and the topics mentioned here can be informational keywords. By pushing the mined topics to users within the application interface, users can be attracted and guided to click and learn more about topic-related content, thereby driving user traffic to the application.

[0003] For example, if the topic "museum" is recommended to a user, those interested in museums will click to browse a series of related museums. This demonstrates that topic-based recommendations are crucial for driving traffic to a business. However, in some related technologies, the recommended topics are not sufficiently attractive to users, and their effectiveness in driving traffic needs improvement. Summary of the Invention

[0004] In view of the above, one or more embodiments of this specification provide a topic push method, apparatus and electronic device.

[0005] To achieve the above objectives, one or more embodiments of this specification provide the following technical solutions:

[0006] According to a first aspect of the embodiments of this specification, a topic push method is provided, the method comprising:

[0007] Obtain at least one target seasonal point of interest (POI) corresponding to the current time, wherein the target seasonal POI is determined as a popular POI at the current time based on the traffic of the POI;

[0008] Retrieve the user comment text corresponding to each target seasonal POI;

[0009] Extract candidate topics from the user comment text corresponding to each target seasonal POI to obtain a topic recall set, which includes multiple candidate topics;

[0010] Select a topic from the multiple candidate topics included in the topic recall set and push it.

[0011] According to a second aspect of the embodiments of this specification, a topic push device is provided, the device comprising:

[0012] The POI acquisition module is used to acquire at least one target seasonal point of interest (POI) corresponding to the current time. The target seasonal POI is determined as a popular POI for the current time based on the traffic of the POI.

[0013] The text acquisition module is used to acquire the user comment text corresponding to each target seasonal POI;

[0014] The topic recall module is used to extract candidate topics from the user comment text corresponding to each target seasonal POI to obtain a topic recall set, which includes multiple candidate topics;

[0015] The topic selection module is used to select a topic from multiple candidate topics included in the topic recall set for push notification.

[0016] According to a third aspect of the embodiments of this specification, an electronic device is provided, comprising:

[0017] processor;

[0018] Memory used to store processor-executable instructions;

[0019] The processor executes the executable instructions to implement the method described in any embodiment of this specification.

[0020] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores computer instructions thereon, which, when executed by a processor, implement the methods described in any embodiment of this specification.

[0021] The topic push method, apparatus, and electronic device in the embodiments of this specification mine topics from user comment text of target seasonal POIs. Since target seasonal POIs are some popular POIs at the current time, such as Baiwang Park being particularly popular with many users visiting, or POIs having a high search volume, it indicates that many users are interested in the POI. Therefore, extracting topics from user comment text of such high-traffic target seasonal POIs makes it easier to obtain topic keywords that attract users, thereby improving the effectiveness of such push topic keywords in attracting traffic. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in one or more embodiments or related technologies of this disclosure, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments recorded in one or more embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of a topic push interface provided in an exemplary embodiment.

[0024] Figure 2This is a schematic diagram of a topic push interface provided in an exemplary embodiment.

[0025] Figure 3 This is a flowchart of a topic push method provided in an exemplary embodiment.

[0026] Figure 4 This is a flowchart of a method for obtaining a target seasonal point of interest (POI) as provided in an exemplary embodiment.

[0027] Figure 5 This is an exemplary embodiment that provides a time-series distribution map of POI traffic data within a preset time period.

[0028] Figure 6 It is based on Figure 5 A schematic diagram of the seasonal distribution data obtained by decomposing the flow data.

[0029] Figure 7 This is a flowchart illustrating the process of obtaining a topic recall set as provided in an exemplary embodiment.

[0030] Figure 8 This is a schematic diagram of the structure of a topic push device provided in an exemplary embodiment.

[0031] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an exemplary embodiment. Detailed Implementation

[0032] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with one or more embodiments of this specification. Rather, they are merely examples of apparatuses and methods consistent with some aspects of one or more embodiments of this specification as detailed in the appended claims.

[0033] It should be noted that the steps of the corresponding methods are not necessarily performed in the order shown and described in this specification in other embodiments. In some other embodiments, the methods may include more or fewer steps than described in this specification. Furthermore, a single step described in this specification may be broken down into multiple steps in other embodiments; and multiple steps described in this specification may be combined into a single step in other embodiments.

[0034] The topic push method provided in the embodiments of this specification can be applied to various scenarios. For example, topics can be pushed from the interface of an application client or from a web page. Taking the application client as an example, Figure 1and Figure 2 Two themed push interfaces were displayed.

[0035] like Figure 1 As shown, users can search for attractions within the mobile app. The app displays a range of information about the attractions, including their distance from the user and ratings. Furthermore, while showcasing attractions, the app can also display multiple themes on the interface. For example... Figure 1 The theme set 11 shown lists multiple topics, such as zoos, museums, ancient towns, and temples. If a user is interested in a particular topic, such as "museums," they can click on that topic to learn more about museums. Therefore, pushing these themes to users can guide them to focus on those topics and increase traffic.

[0036] Within the mobile app, you can also push themed content using background keywords. For example... Figure 2 As shown, taking a map app as an example, when a user opens the app, a search box 21 will be displayed, within which topics can be pushed. For example, if the topic "famous attractions" is pushed, and the user wants to learn more about attractions, they can directly click to search based on the pushed topic. This can also increase search volume.

[0037] Understandable, Figure 1 and Figure 2 These are just two examples of how to display push topics. This specification does not limit the display method of topics, and they can also be pushed in other locations and in other ways.

[0038] The topic recommendation method provided in the embodiments of this specification aims to discover topic keywords that can bring a stronger user experience and more user traffic. For example, October and November are the season for appreciating autumn leaves. If the topic "museum" is recommended, recommending "autumn leaf appreciation" will be more attractive to users than recommending "museum". Naturally, the topic "autumn leaf appreciation" will bring more user traffic than the topic "museum".

[0039] The topic push method of the embodiments of this specification will be described in detail below. Figure 3 This is a flowchart illustrating a topic push method as provided in an exemplary embodiment. This topic push method can be executed on a server. For example, by executing this method on the server side, after obtaining the topic to be pushed, the topic can be sent to the client for display.

[0040] like Figure 3 As shown, the method may include the following processing:

[0041] In step 300, at least one target seasonal POI corresponding to the current time is obtained. The target seasonal POI is determined to be a popular POI at the current time based on the traffic of the POI.

[0042] In this context, a POI (Point of Interest) can be a house, a shop, a mailbox, a bus stop, a tourist attraction, etc. The target seasonal POI referred to in this embodiment refers to a POI that is currently popular. This "popularity" can be determined based on the POI's search volume, reach, etc. If a POI has a high search volume or reach at the current time, it is considered a popular POI. Furthermore, the "current time" can refer to the time when the topic recommendation method of this embodiment is used to discover and recommend topics.

[0043] For example, it can be done by Figure 4 The method shown is used to obtain the target seasonal POI corresponding to the current time. Figure 4 This is a flowchart illustrating a method for obtaining a target seasonal point of interest (POI) as provided in an exemplary embodiment, which may include:

[0044] In step 3000, traffic data corresponding to each candidate POI within a preset time period is obtained.

[0045] In this step, the candidate POIs can be POIs used to subsequently filter out target seasonal POIs.

[0046] In one example, considering that the popularity of candidate POIs follows a similar cycle each year—for example, viewing autumn leaves is in October and November, and eating qingtuan (a type of glutinous rice dumpling) is around Qingming Festival—the preset time period mentioned in this step can be at least one year. This allows for a more accurate determination of whether there are periods of concentrated traffic for candidate POIs within a year, which can then be used in subsequent steps to determine whether a POI is seasonal.

[0047] Furthermore, it's understandable that the popularity of candidate POIs can fluctuate. For example, a candidate POI might not have been a popular attraction initially, but it could become very popular after extensive and effective promotion. Therefore, even though the popularity of candidate POIs follows a similar cycle each year, traffic data can be dynamically updated. For instance, traffic to candidate POIs can be statistically analyzed every week for a given period, ensuring the latest POI traffic statistics are obtained and allowing for more timely detection of changes in POI traffic. Moreover, due to the regular updates to POI traffic, the method described in this embodiment can be executed based on the regularly updated traffic to obtain the target seasonal POI. The preset time period corresponding to the traffic can be different or the same for different executions of the method. For example, the first execution of the method might obtain traffic for the preset time period from January 2021 to December 2021. When the method is executed a month later, the preset time period corresponding to the obtained traffic might be from January 2021 to January 2022, meaning the second preset time period is one month longer than the first.

[0048] Furthermore, in this step, traffic data corresponding to the candidate POI is obtained. This traffic data may include, but is not limited to, search volume, click volume, and navigation reach for the candidate POI. For example, traffic acquisition can be achieved by the application client collecting the aforementioned user behavior data such as searches and clicks for the candidate POI and binding this user behavior data to the candidate POI.

[0049] For example, if a user searches for a tourist attraction on the application client, one visitor's traffic data is recorded. If the user clicks on the attraction or navigates to it, one visitor's traffic data is also recorded. However, it's important to note that deduplication is necessary in traffic statistics. For instance, if the same user sequentially searches, clicks, and navigates to a tourist attraction, since it's the same user, only one visitor's traffic data is recorded to avoid duplicate recording. That is, the traffic data obtained in this embodiment indicates how many users have followed the candidate POI; searches, clicks, and navigation visits all count as following. In one example, the deduplication mentioned above can be achieved by identifying the user behind the user behavior data using the user's device identifier. For example, when collecting traffic data, the device ID of the device where the user's behavior occurred can be collected, and the same user's actions can be identified based on this device ID.

[0050] As shown above, this step can obtain the traffic data corresponding to each candidate POI. For example, POI-1 had f1 traffic visits within a preset time period, and POI-2 had f2 traffic visits within the same preset time period.

[0051] In step 3002, the traffic data corresponding to each candidate POI is processed to obtain the seasonal confidence level corresponding to the candidate POI. The seasonal confidence level is used to represent the probability of determining the existence of a traffic concentration period based on the traffic distribution within a preset time period.

[0052] Taking one candidate POI as an example, the traffic data of that candidate POI within a preset time period can be decomposed. Specifically, a combined model can be used. The combined model is a method in time series forecasting, usually divided into additive or multiplicative models. That is, by decomposing the main components of the traffic data within the preset time period, the initial traffic data can be obtained by adding or multiplying these decomposed components.

[0053] In this step, the traffic data for the preset time period can be decomposed using a combined model to obtain seasonal, trend, and noise distribution data. Furthermore, this combined model assumes that the trend distribution data changes linearly and the seasonal distribution data changes periodically. For example, in one practice, the combined model can be a linear function with a Fourier characteristic of 365.25 periods, and by optimizing a least-squares loss, the aforementioned seasonal, trend, and noise distribution data can be obtained.

[0054] In this embodiment, only the seasonal distribution data can be used, so the trend distribution data and noise distribution data will not be described in detail. Figure 5 This is an exemplary embodiment of a time-series distribution chart of POI traffic data within a preset time period, where the horizontal axis represents time and the vertical axis represents traffic (number of people). Figure 6 It is based on Figure 5 The seasonal distribution data obtained by decomposing traffic flow data shows that the horizontal axis represents time and the vertical axis represents traffic flow (person-times). Figure 5 and Figure 6 It can also be seen that the seasonal distribution data is used to represent the time-series distribution of traffic flow obtained by smoothly fitting the traffic flow data within a preset time period, i.e. Figure 6 The traffic data curve is smoother.

[0055] After obtaining the seasonal distribution data mentioned above, the seasonal confidence level of candidate POIs can be calculated based on the flow rates corresponding to each time period in the seasonal distribution data. See formula (1) below:

[0056]

[0057] In formula (1) above, F s That is, the seasonal confidence level of candidate POIs, where var represents the variance and R0 represents the variance. t S represents the residual.t express Figure 6 The seasonal distribution data of the flow rate is used to calculate the seasonal confidence level F of the candidate POI using formula (1). s .

[0058] If the seasonal confidence level F of the candidate POI is s The closer a value is to 1, the greater the volatility of the traffic data's temporal distribution within the preset time period. For example, there are periods of concentrated traffic, and the temporal impact is the strongest compared to other factors affecting traffic. In other words, the temporal fluctuations in traffic distribution are primarily due to time factors. For instance, taking... Figure 6 For example, F s The higher the value, the more it can be understood as follows: the total traffic of the candidate POI is 2,000 visits in the entire year from January 2022 to January 2023 (the value is for illustration only). However, the traffic during the period from March 2022 to May 2022 accounts for 1,200 visits, which means that the distribution in time is very volatile. It can also be understood as that there are periods of concentrated traffic in time, and this fluctuation in time is caused by time factors.

[0059] Therefore, it can be understood that the seasonal confidence level F s F is used to represent the probability of determining the existence of a period of concentrated traffic based on the traffic distribution within the preset time period. s The higher the value, the more likely there are periods of concentrated traffic. These periods of concentrated traffic refer to times when traffic is particularly high. For example, in the example above, the total traffic from January 2022 to January 2023 is 2000 visits (this figure is for illustrative purposes only). However, 1200 visits occur between March 2022 and May 2022. That is, the traffic distribution curve shows a peak between March 2022 and May 2022, and this time period can be called a period of concentrated traffic.

[0060] In step 3004, if the seasonality confidence level reaches the confidence threshold, then the candidate POI is determined to be a seasonal POI.

[0061] In this embodiment, a confidence threshold can be set. For example, the confidence threshold can be set to 0.8. If the seasonal confidence F of the candidate POI calculated in step 3002 is... s A confidence score of 0.8 or higher is sufficient to determine if a candidate POI is a seasonal POI. It's important to understand that 0.8 is just an example; it can be adjusted based on actual circumstances. For instance, in a small location with few candidate POIs, the seasonality confidence score F can be increased. sThe confidence threshold should be set slightly lower, for example, to 0.7.

[0062] It's important to clarify that the "seasonal POIs" mentioned in this embodiment do not mean that the popularity of a POI is related to a specific season, such as a POI being more popular in spring or summer. Rather, it means that a seasonal POI is one that experiences a significant increase in traffic within a certain period. For example, in the previously mentioned example, Baiwang Park sees a large number of visitors in October and November each year to admire the autumn leaves, so "Baiwang Park" is a seasonal POI, with its peak traffic period being "October and November." Similarly, Yuyuantan Park sees a large number of visitors in April each year to admire the cherry blossoms, so "Yuyuantan Park" is also a seasonal POI, with its peak traffic period being "April."

[0063] Using the methods described above, multiple seasonal POIs can be selected.

[0064] In step 3006, for any seasonal POI, if the current time is within the peak traffic period of the seasonal POI, then the seasonal POI is determined as the target seasonal POI corresponding to the current time.

[0065] For example, for the seasonal POI "Baiwang Park", its peak traffic period is "October and November". If the topic mining and push of this embodiment of the specification takes place in October, then "Baiwang Park" is a target seasonal POI.

[0066] For example, for the seasonal POI "Yuyuantan Park", its peak traffic period is "April". If the topic mining and push of this embodiment of the specification takes place in October, then "Yuyuantan Park" is not a target seasonal POI.

[0067] In step 302, the user comment text corresponding to each target seasonal POI is obtained.

[0068] After obtaining at least one target seasonal POI corresponding to the current time in step 300 above, the user comment text corresponding to each target seasonal POI can be obtained. The user comment text is obtained in order to extract the topic to be pushed from it later.

[0069] In this step, when obtaining user review text for each target seasonal POI, it is possible to obtain recent user review text for that target seasonal POI. For example, for the seasonal POI "Baiwang Park", its peak traffic period is "October and November". If the topic mining and push in this embodiment of the specification takes place in October, then "Baiwang Park" is a target seasonal POI, and user review text for Baiwang Park in October can be extracted.

[0070] Optionally, after obtaining at least one target seasonal Point of Interest (POI) corresponding to the current time, the at least one target seasonal POI can be displayed on a map, which can be called a "seasonal map." For example, the location, name, and other information of each target seasonal POI can be displayed on the seasonal map. By displaying it on the map, users can more intuitively observe where each seasonal POI is located.

[0071] In step 304, candidate topics are extracted from the user comment text corresponding to each target seasonal POI to obtain a topic recall set, which includes multiple candidate topics.

[0072] This step involves recalling a topic recall set, which may include multiple candidate topics.

[0073] Specifically, the topic recall set can be categorized as follows: Figure 7 The process shown is to obtain, Figure 7 This is an exemplary embodiment of a flowchart for obtaining a topic recall set, which may include:

[0074] In step 700, initial topics are extracted from user comment texts corresponding to each target seasonal POI based on the topic library.

[0075] In this embodiment, a topic library can be preset, which can include many keywords, such as "picnic," "flower viewing," and "spring outing." Based on these keywords in the topic library, text matching and extraction are performed on user comment texts. For example, words synonymous with the keywords in the topic library can be extracted from user comment texts.

[0076] For example, a user review might read, "The hairy crabs at this restaurant are so delicious!" By comparing this text with keywords in the topic library, we find that the topic "eating hairy crabs" is similar to "the hairy crabs are so delicious" in the review. Therefore, the extracted topic is "eating hairy crabs." In other words, when the content of a user review is not entirely consistent with, but is similar to, keywords in the topic library, the final extracted topic can use keywords from the topic library.

[0077] In this embodiment, the topics extracted from user comment text are called initial topics. For example, "eating hairy crabs" in the example above is an initial topic.

[0078] Furthermore, after extracting themes from user comment text, it's important to remove duplicates. For example, if the same theme appears twice or more in the same user comment text, only one extraction is needed; that is, only one identical theme word should be extracted from the same user comment text. Additionally, the extracted theme words can be processed, such as removing sensitive words and adjectives. For instance, if the theme is "tomb sweeping," it can be considered a sensitive word and discarded. If the theme word is "suitable for travel," the word "of" can be removed.

[0079] In step 702, for any initial topic, the traffic of each first seasonal POI corresponding to the initial topic is aggregated within a preset time period to obtain the traffic data corresponding to the initial topic within the preset time period. The initial topic is extracted from the user comment text of the first seasonal POI.

[0080] In this step, the first seasonal POI refers to the initial topic extracted from the user comment text of the first seasonal POI. For example, taking the initial topic "eating hairy crabs" as an example, this initial topic was extracted from the user comment text of target seasonal POI-A, and it was also extracted from the user comment text of target seasonal POI-B and target seasonal POI-C. Therefore, target seasonal POI-A, target seasonal POI-B, and target seasonal POI-C can all be referred to as the first seasonal POI.

[0081] In this step, for a given initial topic, the traffic of each first-season POI corresponding to that initial topic within a preset time period can be aggregated to obtain the traffic data for that initial topic within the preset time period. The number of first-season POIs corresponding to the initial topic can be at least one.

[0082] For example, in the aforementioned example, the initial topic "eating hairy crabs" is associated with three first-season POIs: POI-A, POI-B, and POI-C. The traffic from these three first-season POIs between January 2022 and January 2023 can be aggregated to obtain the traffic data corresponding to the initial topic "eating hairy crabs" from January 2022 to January 2023. This aggregation of traffic from the three first-season POIs between January 2022 and January 2023 can be achieved by simply adding the traffic from each of the three first-season POIs together.

[0083] In step 704, based on the traffic data corresponding to the initial topic within a preset time period, the seasonality confidence level of the topic corresponding to the initial topic is calculated. The seasonality confidence level of the topic is used to represent the probability that there is a period of concentrated traffic distribution within the preset time period.

[0084] Based on the traffic data, the seasonality confidence of the initial topic can be calculated according to the methods described in steps 3002 and 3004 of the previous example. For example, the traffic data corresponding to the initial topic "eating hairy crabs" from January 2022 to January 2023 can be decomposed to obtain the seasonal distribution data, and then the confidence can be calculated according to formula (1), which is called the topic seasonality confidence. This topic seasonality confidence can be used to determine the probability that there is a concentrated period of traffic distribution in the traffic distribution corresponding to the initial topic within a preset time period. For specific methods, please refer to [link to relevant documentation]. Figure 4 The process will not be detailed here.

[0085] In step 706, if the seasonality confidence of the topic reaches the first threshold and the current time is within the peak traffic period of the initial topic, then the initial topic is determined as a candidate topic and added to the topic recall set.

[0086] This embodiment can set a first threshold. If the seasonality confidence of a topic reaches the first threshold, and if the current time falls within the peak traffic period of the initial topic, then the initial topic is determined as a candidate topic. In other words, the candidate topic exhibits a peak traffic period within a preset time frame; for example, traffic is high in certain time intervals. Looking at the traffic curve, there is a traffic peak within a certain time interval, and the current time also falls within that time interval. Therefore, the candidate topics in the topic recall set obtained after the above steps are actually popular keywords at the current time, which can also be called seasonal topics. For example, "eating qingtuan" (a type of glutinous rice dumpling) might be a seasonal topic because this keyword usually has higher search volume around Qingming Festival and relatively lower volume at other times. However, "Forbidden City" is unlikely to be a seasonal topic because this keyword is popular all year round, and there is no obvious peak in the traffic distribution over time.

[0087] In step 306, a topic is selected from the multiple candidate topics included in the topic recall set for push notification.

[0088] In this step, for each candidate topic, the seasonal weight corresponding to the candidate topic can be calculated based on the popularity data of the candidate topic. The seasonal weight is used to represent the popularity characteristics of the candidate topic at the current time, and candidate topics whose seasonal weights meet the preset weight conditions are selected as push topics.

[0089] For example, the popularity data corresponding to the candidate topic may include at least one of the following:

[0090] 1) Expression intensity index

[0091] The expression intensity index can be calculated based on the user expression volume of candidate topics. The user expression volume refers to the number of user comment texts related to the candidate topics.

[0092] For example, as mentioned in the previous examples, the same keyword can be extracted only once from the same user comment text. For instance, if three identical keywords are extracted from the same user comment text, deduplication is required to leave only one keyword. Therefore, the same keyword can be extracted from multiple user comment texts. For example, if a certain topic is extracted from 10 user comment texts, then the corresponding user expression count is 10.

[0093] After obtaining the user expression volume of all candidate topics, the values ​​can be normalized to the mean. The larger the final value, the more user expression volume of the candidate topic and the higher the expression intensity index, which means that the candidate topic has been expressed by a large number of users in the comments.

[0094] For example, suppose there are two keywords T1 and T2. User mentions of T1 occur 10 times, and user mentions of T2 occur 20 times. Then, the "10 / 20" can be normalized to obtain a normalized value. This normalized value can be called the expression intensity index. A higher expression intensity index indicates that the keyword is strongly expressed by a large number of users.

[0095] 2) Seasonal growth rate

[0096] When calculating this metric, we can first aggregate the traffic of multiple first-season POIs corresponding to the candidate topic within a preset time period. For example, for topic T1, if topic T1 is extracted from user comment texts of POI-D, POI-E, and POI-F, then the traffic of POI-D, POI-E, and POI-F within the preset time period can be aggregated. After aggregation, we can obtain the traffic data corresponding to the candidate topic within the preset time period. Here, the first-season POIs are the POIs of the candidate topic extracted from their user comment texts. For example, in the above example, POI-D, POI-E, and POI-F are equivalent to multiple first-season POIs associated with topic T1.

[0097] Based on the traffic data obtained above, the seasonal growth rate corresponding to the candidate topic can be calculated. This seasonal growth rate can be calculated as follows: the traffic within the preset time period after aggregation can exhibit a similar pattern. Figure 5 or Figure 6 The flow distribution curve shown can be smoothed by first fitting it to a smoother curve. For example, a combined model can be used to decompose the seasonal distribution data, resulting in a smoother flow distribution curve. Then, based on this flow distribution curve, the curve position corresponding to the current time is found, and the slope of the flow distribution curve at that position is calculated. This slope can be called the seasonal growth rate, which can be used to characterize whether the current curve is in an upward or downward trend. If it is in an upward trend, then the candidate topic has the potential to become popular.

[0098] 3) Current traffic

[0099] The calculation of this indicator can still be based on the seasonal distribution data of the candidate topics over a preset time period obtained above, to obtain the traffic value corresponding to the current time. If the traffic value is relatively high, it means that the candidate topic has been popular recently.

[0100] When calculating the seasonal weight, it can be based on the aforementioned popularity data. The following example uses the expression intensity index, seasonal growth rate, and current flow data mentioned above. Please refer to formula (2):

[0101]

[0102] In formula (2) above, W is the seasonal weight of the candidate topic, f is the current traffic, r is the seasonal growth rate, and s is the expression intensity index. It can be seen that the seasonal weight is positively correlated with the expression intensity index, the seasonal growth rate, and the current traffic. It is understood that formula (2) is only an example, and other methods can be used for calculation.

[0103] Each candidate topic can have its corresponding seasonal weight calculated using the method described above. A higher seasonal weight indicates that the candidate topic is more popular at the current time.

[0104] When selecting topics for push notifications based on the aforementioned seasonal weights, candidate topics whose seasonal weights meet preset weight conditions can be selected as push topics. One exemplary approach is to sort the candidate topics by seasonal weight from highest to lowest and select the top M topics as push topics, where M is a positive integer, and "top M" is equivalent to the preset weight condition. Optionally, other methods can also be used to select push topics. For example, a weight threshold can be set; if the seasonal weight of a candidate topic is greater than or equal to this weight threshold, it can be used as the push topic. In this case, "greater than or equal to the weight threshold" is equivalent to the preset weight condition.

[0105] Furthermore, taking a map application as an example, each time a user opens the application, they receive a push notification with specific topics. These topics can be calculated using the topic push method implemented in this manual. However, it should be noted that the pushed topics are not necessarily calculated in real time. For example, in one example, the topic to be pushed can be calculated in real time each time the user opens the map application. In another example, the topic can be pre-calculated and used directly when the user opens the map application. This is mainly because popular topics typically maintain their popularity for a period of time, so the topic can be calculated once within its popularity period and pushed during that time.

[0106] For example, November is the season for viewing autumn leaves, and many users search for the topic "viewing autumn leaves" during this month. If, at the beginning of November, the method described in this embodiment of the specification reveals that "viewing autumn leaves" is among the topics to be pushed, and based on the traffic curves obtained from the aggregation of traffic from multiple first-season POIs associated with the "viewing autumn leaves" topic, it is known that the traffic aggregation of POIs is relatively large during November, then "viewing autumn leaves" can be used as the topic for push notifications throughout November. As long as a user uses the map application in November, "viewing autumn leaves" can be pushed to them, eliminating the need for real-time calculations each time, thus saving computational resources. However, to calculate the topics to be pushed more accurately, the time interval for calculation using the method described in this embodiment of the specification can be set to a shorter interval, such as updating weekly.

[0107] The topic recommendation method in the embodiments of this specification provides a method for automatically discovering and recommending topics. This method ensures that the discovered topics are more likely to attract users in the following ways:

[0108] 1) Extract themes from user reviews of seasonal POIs. Since seasonal POIs are often popular—for example, Baiwang Park is currently very popular with many visitors—or if a POI has high search volume, it indicates widespread user interest. Extracting themes from user reviews of these high-traffic POIs makes it easier to find keywords that will attract users.

[0109] 2) Furthermore, based on the above-mentioned extraction of topics from user comment text of seasonal POIs, these extracted topics can be further filtered to select more relevant candidate topics. For example, even when extracting topics from user comment text of seasonal POIs, some relatively common topics may still be selected. Therefore, further topic filtering can be performed, selecting topics with higher confidence in seasonality as candidate topics based on the traffic characteristics of the POIs associated with the topics. This increases the probability that the resulting candidate topics will be popular topics.

[0110] 3) Furthermore, based on the above-mentioned candidate topics, when recommending topics, it is not necessary to push all the selected candidate topics. These candidate topics can be sorted, and the candidate topics with the highest seasonal weight can be selected for recommendation. These candidate topics with higher seasonal weight are equivalent to the most popular topics, which can attract more user attention and thus increase traffic.

[0111] As can be seen, the topic push method in this embodiment of the specification, by extracting topics from user comment text of seasonal POIs in step 1), has enhanced the attractiveness of the topics and increased user traffic compared to the topics pushed in the prior art. Furthermore, by combining steps 2) and 3) above, the seasonality of the selected topics is gradually enhanced, resulting in push topics that are more attractive to users and have a better traffic-driving effect.

[0112] To implement the topic push method of any embodiment of this specification, this specification also provides a topic push device. Figure 8 This is a schematic diagram of the structure of a topic push device provided in an exemplary embodiment, such as... Figure 8 As shown, the device may include: a POI acquisition module 801, a text acquisition module 802, a topic recall module 803, and a topic selection module 804.

[0113] The POI acquisition module 801 is used to acquire at least one target seasonal point of interest (POI) corresponding to the current time. The target seasonal POI is determined as a popular POI for the current time based on the traffic of the POI.

[0114] The text acquisition module 802 is used to acquire the user comment text corresponding to each target seasonal POI.

[0115] The topic recall module 803 is used to extract candidate topics from the user comment text corresponding to each target seasonal POI to obtain a topic recall set, which includes multiple candidate topics.

[0116] The topic selection module 804 is used to select a topic from multiple candidate topics included in the topic recall set for push notification.

[0117] In one example, the POI acquisition module 801, when acquiring at least one target seasonal POI corresponding to the current time, includes: acquiring traffic data corresponding to each candidate POI within a preset time period; processing the traffic data corresponding to each candidate POI to obtain the seasonal confidence score corresponding to the candidate POI, wherein the seasonal confidence score is used to represent the probability of determining the existence of a traffic concentration period based on the traffic distribution within the preset time period; if the seasonal confidence score reaches a confidence threshold, then the candidate POI is determined to be a seasonal POI; for any seasonal POI, if the current time is within the traffic concentration period of the seasonal POI, then the seasonal POI is determined to be the target seasonal POI corresponding to the current time.

[0118] In one example, the POI acquisition module 801, when processing the traffic data corresponding to each candidate POI to obtain the seasonal confidence score of the candidate POI, includes: decomposing the traffic data of the candidate POI within a preset time period to obtain seasonal distribution data, wherein the seasonal distribution data is used to represent the traffic time series distribution obtained by smoothing and fitting the traffic data within the preset time period; and calculating the seasonal confidence score of the candidate POI based on the traffic corresponding to each time in the seasonal distribution data.

[0119] In one example, the topic recall module 803, when extracting candidate topics from user comment texts corresponding to each target seasonal POI to obtain a topic recall set, includes: extracting initial topics from user comment texts corresponding to each target seasonal POI based on a topic library; for any initial topic, aggregating the traffic of each first seasonal POI corresponding to the initial topic within a preset time period to obtain traffic data corresponding to the initial topic within the preset time period, wherein the initial topic is extracted from user comment texts of the first seasonal POI; calculating the topic seasonality confidence score corresponding to the initial topic based on the traffic data corresponding to the initial topic within the preset time period, wherein the topic seasonality confidence score is used to represent the probability that there is a traffic concentration period in the traffic distribution within the preset time period; if the topic seasonality confidence score reaches a first threshold, and the current time is within the traffic concentration period of the initial topic, then the initial topic is determined as a candidate topic; and added to the topic recall set.

[0120] In one example, the topic selection module 804, when selecting a topic for push from multiple candidate topics included in the topic recall set, includes: calculating the seasonal weight corresponding to the candidate topic based on the popularity data corresponding to the candidate topic, wherein the seasonal weight is used to represent the popularity characteristics of the candidate topic at the current time; and selecting the candidate topic whose seasonal weight satisfies the preset weight condition as the push topic.

[0121] In one example, the popularity data includes at least one of: expression intensity index, seasonal growth rate, and current traffic; the topic selection module 804, when calculating the seasonal weight corresponding to the candidate topic based on the popularity data corresponding to the candidate topic, includes: calculating the popularity data corresponding to the candidate topic, wherein the expression intensity index is calculated based on the user expression volume of the candidate topic, the user expression volume representing the number of user comment texts including the candidate topic; the seasonal growth rate and the current traffic corresponding to the current time are calculated based on the traffic data corresponding to the candidate topic, the traffic data being obtained by aggregating the traffic of multiple first seasonal POIs corresponding to the candidate topic within a preset time period; and determining the seasonal weight of the candidate topic based on the popularity data, the seasonal weight being positively correlated with the popularity data.

[0122] In one example, the POI acquisition module 801 is further configured to: after acquiring at least one target seasonal point of interest (POI) corresponding to the current time, display the at least one target seasonal POI on the map.

[0123] The specific implementation process of the functions and roles of each module in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0124] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of the solution in this specification according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0125] like Figure 9 As shown, Figure 9 This diagram illustrates a hardware structure of an electronic device containing a topic push device according to an embodiment of this specification. The device may include a processor 910, a memory 920, an input / output interface 930, a communication interface 940, and a bus 950. The processor 910, memory 920, input / output interface 930, and communication interface 940 are interconnected internally via the bus 950.

[0126] The processor 910 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification. The processor implements the above-described methods by running executable instructions.

[0127] The memory 920 for storing processor-executable instructions can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 920 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 920.

[0128] The input / output interface 930 is used to connect input / output modules to enable information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0129] The communication interface 940 is used to connect the communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0130] Bus 950 includes a pathway for transmitting information between various components of the device, such as processor 910, memory 920, input / output interface 930, and communication interface 940.

[0131] It should be noted that although the above-described device only shows the processor 910, memory 920, input / output interface 930, communication interface 940, and bus 950, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0132] This specification also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned topic push method.

[0133] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0134] This specification also provides a computer program that, when run, is used to implement the topic push method described above.

[0135] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0136] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0137] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0138] The above description is merely a preferred embodiment of one or more embodiments of this specification and is not intended to limit the scope of one or more embodiments of this specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments of this specification should be included within the protection scope of one or more embodiments of this specification.

Claims

1. A topic-based push method, characterized in that, The method includes: Obtain at least one target seasonal POI corresponding to the current time, wherein the seasonal confidence of the target seasonal POI reaches a set confidence threshold, and the seasonal confidence is used to represent the probability of determining the existence of a traffic concentration period based on the traffic distribution within a preset time period; wherein the current time is located within the traffic concentration period of the target seasonal POI; Retrieve the user comment text corresponding to each target seasonal POI; Extract candidate topics from the user comment text corresponding to each target seasonal POI to obtain a topic recall set, which includes multiple candidate topics; Select a topic from the multiple candidate topics included in the topic recall set and push it.

2. The method according to claim 1, characterized in that, The step of obtaining at least one target seasonal POI corresponding to the current time includes: Obtain traffic data corresponding to each candidate POI within a preset time period; The traffic data corresponding to each candidate POI is processed to obtain the seasonal confidence level corresponding to the candidate POI; If the seasonality confidence level reaches the confidence threshold, then the candidate POI is determined to be a seasonal POI; For any of the seasonal POIs, if the current time is within the peak traffic period of the seasonal POI, then the seasonal POI is determined as the target seasonal POI corresponding to the current time.

3. The method according to claim 2, characterized in that, The process of processing the traffic data corresponding to each candidate POI to obtain the seasonal confidence level corresponding to the candidate POI includes: The traffic data of the candidate POI within a preset time period is decomposed to obtain seasonal distribution data, which is used to represent the traffic time series distribution obtained by smoothing and fitting the traffic data within the preset time period. The seasonal confidence level of the candidate POI is calculated based on the flow rate corresponding to each time period in the seasonal item distribution data.

4. The method according to claim 1, characterized in that, The step of extracting candidate topics from user comment texts corresponding to each target seasonal POI to obtain a topic recall set includes: Based on the topic library, initial topics are extracted from user comment texts corresponding to each target seasonal POI; For any initial topic, the traffic of each first seasonal POI corresponding to the initial topic is aggregated within a preset time period to obtain the traffic data corresponding to the initial topic within the preset time period. The initial topic is extracted from the user comment text of the first seasonal POI. Based on the traffic data corresponding to the initial topic within a preset time period, the seasonality confidence score of the topic corresponding to the initial topic is calculated. The seasonality confidence score is used to represent the probability that there is a period of concentrated traffic distribution within the preset time period. If the seasonality confidence of the topic reaches the first threshold, and the current time is within the peak traffic period of the initial topic, then the initial topic is determined as a candidate topic and added to the topic recall set.

5. The method according to claim 1, characterized in that, The step of selecting a topic for push notification from the multiple candidate topics included in the topic recall set includes: Based on the popularity data corresponding to the candidate topics, calculate the seasonal weight corresponding to the candidate topics. The seasonal weight is used to represent the popularity characteristics of the candidate topics at the current time. Select candidate topics that meet the preset weight conditions for seasonality as push topics.

6. The method according to claim 5, characterized in that, The heat data includes at least one of the following: expression intensity index, seasonal growth rate, and current flow. The step of calculating the seasonal weight of the candidate topics based on their popularity data includes: The popularity data corresponding to the candidate topics is calculated, wherein the expression intensity index is calculated based on the user expression volume of the candidate topics, and the user expression volume represents the number of user comment texts including the candidate topics; the seasonal item growth rate and the current traffic corresponding to the current time are calculated based on the traffic data corresponding to the candidate topics, and the traffic data is obtained by aggregating the traffic of multiple first seasonal POIs corresponding to the candidate topics within a preset time period; Based on the popularity data, the seasonal weight of the candidate topics is determined, and the seasonal weight is positively correlated with the popularity data.

7. The method according to claim 1, characterized in that, The method further includes: After obtaining at least one target seasonal point of interest (POI) corresponding to the current time, the at least one target seasonal POI is displayed on the map.

8. A topic push device, characterized in that, The device includes: The POI acquisition module is used to acquire at least one target seasonal POI corresponding to the current time. The seasonal confidence of the target seasonal POI reaches a set confidence threshold. The seasonal confidence is used to represent the probability of determining the existence of a traffic concentration period based on the traffic distribution within a preset time period. The current time is located within the traffic concentration period of the target seasonal POI. The text acquisition module is used to acquire the user comment text corresponding to each target seasonal POI; The topic recall module is used to extract candidate topics from the user comment text corresponding to each target seasonal POI to obtain a topic recall set, which includes multiple candidate topics; The topic selection module is used to select a topic from multiple candidate topics included in the topic recall set for push notification.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor implements the method as described in any one of claims 1-7 by executing the executable instructions.

10. A computer-readable storage medium storing computer instructions thereon, characterized in that, When executed by the processor, this instruction implements the steps of the method as described in any one of claims 1-7.

Images