CN112528145B - Information recommendation method, device, equipment and readable storage medium - Google Patents

Abstract

The application discloses an information recommendation method, an information recommendation device, information recommendation equipment and a readable storage medium, and relates to the technical field of electronic maps in artificial intelligence. The specific implementation scheme is as follows: after receiving a request instruction of a user for requesting to open the electronic map, the server responds to the request instruction to predict whether the user has a retrieval intention. When the user is predicted to have the search intention, generating a recommendation card for the user and outputting the recommendation card, wherein a plurality of search words are displayed on the recommendation card for the user to select, so that personalized recommendation is realized while the interaction efficiency is improved.

Description

Information recommendation method, device, equipment and readable storage medium

Technical Field

The present application relates to the field of electronic map technology in artificial intelligence, and in particular, to an information recommendation method, apparatus, device, and readable storage medium.

Background

The electronic map has navigation, positioning and other functions, and is an indispensable tool in life.

Users often find some type of point of interest (point of interest, POI) near the current anchor point or some POI using an electronic map. In the searching process, buttons such as 'find periphery' and the like on the electronic map are clicked, the electronic map popup window recommends a card, a series of search words are displayed on the recommends card, and each search word represents one type of POI. For example, if the search term is "food", the user clicks "food", and then the electronic map jumps to the food page, and the food page displays various kinds of food, such as nearby food, snack food, hot pot, etc., for the user to select. Meanwhile, some hot restaurants and the like are displayed on the food page for the user to select. Because of the limited screen, if the types of the delicious foods are various, some common types of the delicious foods are displayed on the delicious food pages, and some unusual types of the delicious foods are on the next-stage pages. For example, if the user clicks "more" on the food page, the next page is entered, and the next page lists more food categories for the user to select. In the process, each food category, such as chafing dish, is regarded as a search term.

The method of clicking the 'finding the periphery' can trigger the electronic map to directly initiate retrieval, such as retrieving food, retrieving scenic spots and the like. However, if the search term desired by the user is hidden deeply, the user needs to interact with the electronic map for multiple times to find the search term, which is time-consuming, labor-consuming and low in efficiency.

Disclosure of Invention

The application provides an information recommending method, device, equipment and readable storage medium, which actively pops up a recommending card conforming to the user behavior and the current scene when recognizing that a user has searching requirements, so that the user can click on a searching word on the recommending card, and a certain interest point can be quickly and conveniently found by the user.

In a first aspect, an embodiment of the present application provides an information recommendation method, including: after receiving a request instruction of a user for requesting to open the electronic map, the server responds to the request instruction to predict whether the user has a retrieval intention. And generating a recommended card for the user when the user is predicted to have the search intention, and outputting the recommended card, wherein a plurality of search terms are displayed on the recommended card for the user to select.

In a second aspect, an embodiment of the present application provides a model training method, including: the electronic equipment builds a training sample by utilizing historical data, and carries out model training to obtain a target model, wherein the target model is a model associated with a recommended card, a plurality of search terms are displayed on the recommended card, and different search terms in the plurality of search terms represent POIs of different categories.

In a third aspect, an embodiment of the present application provides an information recommendation apparatus based on an electronic map, including:

the receiving module is used for receiving a request instruction of a user, wherein the request instruction is used for requesting to open the electronic map;

a prediction module for predicting whether the user has a retrieval intention in response to the request instruction;

the processing module is used for generating a recommendation card for the user when the prediction module predicts that the user has the search intention, wherein a plurality of search terms are displayed on the recommendation card, and different search terms in the plurality of search terms represent points of interest POIs of different categories;

and the output module is used for outputting the recommended card.

In a fourth aspect, an embodiment of the present application provides a model training apparatus, including:

the construction module is used for constructing training samples according to historical data to obtain a training sample set, wherein the historical data is data generated when a historical user operates the electronic map;

the training module is used for training the initial model by using the training sample to obtain a target model, wherein the target model is a model associated with a recommended card, a plurality of search words are displayed on the recommended card, and different search words in the plurality of search words represent points of interest POIs of different categories.

In a fifth aspect, an embodiment of the present application provides an electronic device, including:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein, the liquid crystal display device comprises a liquid crystal display device,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the first aspect or any possible implementation of the first aspect.

In a sixth aspect, an embodiment of the present application provides an electronic device, including:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein, the liquid crystal display device comprises a liquid crystal display device,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the second aspect or any possible implementation of the second aspect.

In a seventh aspect, embodiments of the present application provide a computer program product comprising instructions which, when run on an electronic device, cause the electronic device computer to perform the method of the first aspect or various possible implementations of the first aspect.

In an eighth aspect, embodiments of the present application provide a computer program product comprising instructions which, when run on an electronic device, cause the electronic device computer to perform the method of the second aspect or various possible implementations of the second aspect described above.

In a ninth aspect, embodiments of the present application provide a non-transitory computer-readable storage medium storing computer instructions for causing the electronic device to perform the method of the first aspect or the various possible implementations of the first aspect.

In a tenth aspect, embodiments of the present application provide a non-transitory computer readable storage medium storing computer instructions for causing the electronic device to perform the method of the second aspect or the various possible implementations of the second aspect.

According to the technology provided by the application, when the user is identified to have search appeal, the recommended card conforming to the user behavior and the current scene is actively popped up for the user to click on the search word on the recommended card, so that the user can quickly and conveniently find a certain type of interest point, the input operation of the user on a search box is reduced, the user is prevented from being able to find the search word through clicking for multiple times, the interaction efficiency is improved, and meanwhile, personalized recommendation is realized.

It should be understood that the description in this section is not intended to identify key or critical features of the embodiments of the application or to delineate the scope of the application. Other features of the present application will become apparent from the description that follows.

Drawings

The drawings are included to provide a better understanding of the present application and are not to be construed as limiting the application. Wherein:

FIG. 1 is a schematic diagram of a process of a user searching a POI of a certain type near a current positioning point;

FIG. 2A is a large frame search schematic;

FIG. 2B is a diagram of a fixed list of terms landing pages;

fig. 3 is a schematic view of a scenario of an information recommendation method according to an embodiment of the present application;

FIG. 4 is a flowchart of an information recommendation method according to an embodiment of the present application;

FIG. 5A is a schematic diagram of an electronic map interface in an information recommendation method according to an embodiment of the present application;

FIG. 5B is another schematic diagram of an electronic map interface in the information recommendation method according to the embodiment of the present application;

FIG. 6 is a schematic diagram of a process for training and applying a first model in the information recommendation method according to the embodiment of the present application;

FIG. 7 is a schematic diagram of a process of training and applying a second model in the information recommendation method according to the embodiment of the present application;

FIG. 8 is a flow chart of a model training method provided by an embodiment of the present application;

Fig. 9 is a schematic structural diagram of an information recommendation device based on an electronic map according to an embodiment of the present application;

FIG. 10 is a schematic structural diagram of a model training device according to an embodiment of the present application;

fig. 11 is a block diagram of an electronic device for implementing an information recommendation method according to an embodiment of the present application.

Detailed Description

Exemplary embodiments of the present application will now be described with reference to the accompanying drawings, in which various details of the embodiments of the present application are included to facilitate understanding, and are to be considered merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the application. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

Currently, users often use electronic maps to find certain types of POIs. Fig. 1 is a schematic diagram of a process of searching a certain type of POI near a current positioning point by a user. Referring to fig. 1, after a user opens an electronic map, there is a request to search for surrounding amusement parks. In the searching process, the user clicks the periphery, then clicks more, and enters more search words to fall to the page. Then, the user slides the screen on the landing page to find the 'playground', and after clicking the 'playground', the electronic map locates a plurality of playgrounds.

If the user wants to find a certain type of POI near a certain position (not the current positioning point), the user needs to search the position on the electronic map and then quickly search by clicking the periphery. Or the user moves the electronic map, for example, the user is in the sea, the current positioning point is in the sea, the user wants to search for a coffee hall near a certain target position in the Beijing lake area, the electronic map displays Beijing, then the lake area is displayed, and then the target position is displayed through sliding operation. Finally, the user clicks the "periphery" to perform quick retrieval. And the electronic equipment locates and pushes the POI interested by the user according to the search of the user.

In the information recommendation process based on the electronic map, if the search word expected by the user is hidden deeply, the user needs to interact with the electronic map for multiple times to find the search word, which is time-consuming and labor-consuming. Resulting in the user being able to search within the search box. However, if the user cannot organize the valid search terms, the result obtained by the search frame search method is incorrect.

Fig. 2A is a large-frame search schematic. Fig. 2B is a schematic diagram of a fixed list search term landing page. Referring to fig. 2A, the large-frame search refers to a manner in which a user inputs a search term organized by himself in a search box to perform a search. If the user cannot organize the valid search terms, the valid structure may not be retrieved. Referring to fig. 2B, the search words in the landing page are fixed, and are a series of search words obtained by a developer according to the search behaviors of thousands of ordinary users. The user clicks the periphery and clicks more to enter the search term landing page, so that the interaction process is complicated. For example, when a user wants to find a hard cafe in a sunlight building, he needs to slide the electronic map to beijing through touch operation, find Yang Guangda building, click on "periphery", find "cafe" in a search term landing page, and click on "cafe", so that the POI of the cafe can be found. Moreover, since the search terms in the fixed list are fixed, the user cannot conduct scene and personalized recommendation according to the current scene and the use habit of the user.

The embodiment of the application aims to actively pop up the recommended card which accords with the user behavior and the current scene when the user is identified to have search appeal, so that the user can click on the search word on the recommended card, and the user can quickly and conveniently find a certain type of interest point.

Fig. 3 is a schematic view of a scenario of an information recommendation method according to an embodiment of the present application. Referring to fig. 3, the scenario includes: terminal device 301 and server 302, terminal device 301 and server 302 establish a network connection. Wherein the server 302 predicts whether the user has a retrieval intention. And after the server predicts that the user has the search intention, generating a recommendation card for the user and pushing the recommendation card, wherein the recommendation card is provided with a plurality of search terms, and each search term represents one type of POI. The recommended cards for different users are different. Therefore, the user can search without entering the search term landing page through multiple operations, and the efficiency is high. In addition, the search words on the recommendation card are related to the current scene of the user, the historical operation behavior of the user on the electronic map and the like, so that the search words on the recommendation card are more in line with the current appeal of the user, the efficiency is further improved, and personalized recommendation is realized.

In fig. 3, the terminal device 301 may be a desktop terminal or a mobile terminal, the desktop terminal may be a computer, etc., the mobile terminal may be a mobile phone, a tablet computer, a notebook computer, etc., and the server 302 may be an independent server or a server cluster formed by a plurality of servers, etc.

Fig. 4 is a flowchart of an information recommendation method provided in an embodiment of the present application, where the embodiment is illustrated from the perspective of a server, and includes:

401. and receiving a request instruction of a user.

The request instruction is used for requesting to open the electronic map.

Illustratively, in FIG. 3, terminal device 301 has an electronic map Application (APP) installed thereon. When the user clicks the APP to request opening of the electronic map, the terminal device 301 transmits a request instruction to the server 302.

402. In response to the request instruction, predicting whether the user has a search intention, and if the user has a detection intention, executing step 403; if the user does not detect intent, step 405 is performed.

After receiving the request instruction from the terminal device, the server judges whether the user has a search intention or not according to the current scene information of the scene where the user is currently located, the historical data of the electronic map used by the user before, and the like, namely, judges whether the user has the appeal of clicking search words such as food, snack, parking lots and the like. If the server predicts that the user has a stronger search intention, then step 403 is executed; if the server predicts that the user does not have a strong search intention, step 405 is performed.

403. A recommendation card is generated for the user.

And displaying a plurality of search terms on the recommendation card, wherein different search terms in the plurality of search terms represent points of interest (POIs) of different categories.

The server generates a recommendation card for the user according to the current scene information of the current scene of the user, the historical data of the electronic map used by the user before, and the like, wherein the search words on the recommendation card are search words with larger clicking probability of the user, and the recommendation cards of different users are different.

Fig. 5A is a schematic diagram of an electronic map interface in the information recommendation method according to the embodiment of the present application. Referring to fig. 5A, when a user travels to a foreign place, the historical data shows that the user frequently searches for a food in noon or searches for a POI of food in a large frame search mode, and after the user opens the electronic map, the electronic map actively pops up a recommended card, and the search words on the recommended card include special snack, food, fast food, hot pot, marinated and boiled fire and the like.

Fig. 5B is another schematic diagram of an electronic map interface in the information recommendation method according to the embodiment of the present application. Referring to fig. 5B, when a user travels to a foreign place, the historical data shows that the user frequently searches for POIs of hotel types and POIs of parking lots, gas stations, and the like, and after the user opens the electronic map, the electronic map pops up a recommended card, and the search terms on the recommended card include star hotels, quick hotels, service areas, gas stations, parking lots, and the like.

In addition, the electronic device may generate different recommended cards for different users, for example, for users who like snack, the search terms on the recommended cards may include snack, feature, etc., and for users who like western-style food, the search terms on the recommended cards may include western-style restaurants, cafes, etc.

The electronic device is different for the recommended cards generated by the same user under different scenes, for example, when the user travels to the outside, the search words on the recommended cards comprise hotels, parking lots and the like, and when the user is in the local place, the search words comprise food and the like.

404. And outputting the recommended card.

The server pushes the recommended card to the terminal equipment of the user, so that the recommended card is popped up on an electronic map interface of the terminal equipment, and the user can select the search term from the recommended card and click the recommended card. In this way, the user does not need to click "surrounding" or the like to perform the search.

405. And opening the electronic map, and not pushing the recommended card.

The exemplary server pushes only data and the like related to the electronic map interface, and does not push the recommended card, so that the electronic map is displayed on the terminal device but the recommended card is not displayed.

According to the information recommendation method provided by the embodiment of the application, after receiving a request instruction of a user for opening the electronic map, the server responds to the request instruction to predict whether the user has a retrieval intention. And generating a recommended card for the user when the user is predicted to have the search intention, and outputting the recommended card, wherein a plurality of search terms are displayed on the recommended card for the user to select. By adopting the scheme, when the user is identified to have search appeal, the recommended card conforming to the user behavior and the current scene is actively popped up for the user to click on the search word on the recommended card, so that the user can quickly and conveniently find out a certain type of interest point, the input operation of the user in a search box is reduced, the user is prevented from being able to find the search word through clicking for multiple times, the interaction efficiency is improved, and meanwhile, personalized recommendation is realized.

In the above embodiment, when the server responds to the request instruction and predicts whether the user has a search intention, first, the current scene information is determined, and the search probability of the user is determined according to the current scene information. And predicting that the user has the retrieval intention when the retrieval probability is larger than the preset retrieval probability. When the search probability is less than or equal to the preset search probability, the user is predicted to have no search intention.

Illustratively, the current context information includes a time, a place, a user gender, whether the user is in a foreign location, a historical operation behavior of the user, POI semantics, a behavior sequence of the user, etc. when the user requests to turn on the electronic map. The historical operation behavior of the user comprises the behavior that the user clicks the search word on the electronic map before, the behavior that the user inputs the search word in the search box of the electronic map and searches, the behavior that the user uses the electronic map to navigate, the behavior that the user uses the electronic map to locate, and the like. The user's historical operational behavior can reflect the user's interests, etc. The historical operation behavior of the user comprises the behavior that the user clicks the search word on the electronic map before, the behavior that the user inputs the search word in the search box of the electronic map and searches, the behavior that the user uses the electronic map to navigate, the behavior that the user uses the electronic map to locate, and the like. And the server acquires the current scene information. And then, the server determines the retrieval probability of the user according to the current scene information, and compares the retrieval probability with a preset retrieval probability. For example, if the preset search probability is 70%, the user is considered to have a search intention when the search probability is greater than 70%. When the search probability is less than or equal to 70%, the user is considered not to have a search intention, but to use the electronic map for navigation, positioning, and the like.

By adopting the scheme, the server acquires the current scene information and determines the retrieval probability of the user according to the current scene information, so that the purpose of determining whether the user has the retrieval intention is realized, the process is simple, and the accuracy is high. When the user is predicted to have the search intention, the recommended card is actively sprung, so that the path of the user for searching the expected search word is shortened, and the efficiency of using the electronic map by the user is improved.

When the server predicts that the user has the search intention and generates the recommended card for the user, the current scene information is determined first. And then, the server determines the click probability of each search term in the preset plurality of search terms clicked by the user according to the current scene information. And then, the server selects at least one search term with the click probability exceeding the preset click probability from the plurality of search terms, and generates the recommended card according to the at least one search term.

For example, a plurality of search terms are preset on the server, and each search term represents a POI of a certain class. The current scene information includes time, place, user gender, whether the user is in the foreign place, historical operation behavior of the user, POI semantics, etc. when the user requests to open the electronic map. The historical operation behavior of the user comprises the behavior that the user clicks the search word on the electronic map before, the behavior that the user inputs the search word in the search box of the electronic map and searches, the behavior that the user uses the electronic map to navigate, the behavior that the user uses the electronic map to locate, and the like. The user's historical operational behavior can reflect the user's interests, etc.

And the server acquires the current scene information. And then, the server determines the clicking probability of each search term in the plurality of search terms clicked by the user according to the current scene information. For example, if there are 100 preset terms, the server determines the click probability of each term in the 100 terms, and ranks the terms in order from high to low. Then, the server takes the search term of TOP N as the search term of the recommended card display, and generates the recommended card based on the search term of TOP N. Wherein N is, for example, an integer greater than or equal to 15, etc., embodiments of the present application are not limited.

By adopting the scheme, the server acquires the current scene information, determines the clicking probability of each search term in the plurality of search terms according to the current scene information, determines the search term displayed on the recommended card according to the clicking probability, and generates the recommended card which is more suitable for the use habit of the user, so that the interaction cost of the user and the electronic map is reduced.

In the above embodiment, the server needs to train a first model for determining the retrieval probability of the user before predicting the retrieval probability of the user. Before the server generates the recommended card for the user, a second model for calculating the click probability of each term in the plurality of terms needs to be trained. Next, a detailed description is given of how to train the two models.

First, a first model.

And the server constructs a plurality of first training samples according to historical data to obtain a first training sample set, wherein the historical data is data generated when a historical user operates the electronic map. And then, the server trains a first model by using the first training sample set, wherein the first model is used for determining the retrieval probability of the user.

Illustratively, when the server trains out the first model, historical data of the historical user using the electronic map is obtained, wherein the historical data comprises historical behaviors and scene information when the historical behaviors occur. The historical behavior is the operation behavior of the historical user on the electronic map, and comprises but is not limited to general search, positioning, navigation, clicking any search term after clicking the periphery, and the like. For example, the user inputs the behavior of the search term by a large-frame search mode, the behavior of clicking the search term by the history user clicking the "periphery" on the page where the search term falls, the behavior of inputting the POI location, the navigation behavior, and the like. The historical scene information includes: the time, place, the historical user's gender, the historical user's historical interest POI, the historical user's behavior sequence, etc. when the historical behavior occurs, the historical user's gender, local or foreign, the historical user's behavior sequence, etc. The scene information is also characteristic of the sample. The behavior sequence of the historical user is obtained by sequencing the behaviors of the historical user in a period of time according to the sequence. For example, the historical behavior of the historical user is to click on a food, and click on a hot pot, a food and a buffet in a week before clicking on the food, and the behavior sequence is to click on the hot pot, click on the food and click on the buffet.

After the historical data is obtained, the server marks the historical behaviors by utilizing the historical scene information corresponding to the historical behaviors, and the obtained first sample is a positive sample; and marking the historical behaviors by using other historical scene information, wherein the obtained first sample is a negative sample. For example, in the history data, when the history user a clicks on the food in the noon, the click on the food in the noon is a positive sample, and the click on the scenic spot in the noon is a negative sample.

In the model training process, if only one-sided factors are considered, if the user clicks to find the behaviors such as the periphery, the habit of the user cannot be reflected. Meanwhile, the trained model is restricted. For example, the search terms in the search term landing page are all fixed and only comprise scenic spots, delicacies and hotels, if only the peripheral behaviors found by clicking by a user are considered, the first model or the second model below cannot reflect the interest of the user in marinating and boiling. In the embodiment of the application, the server utilizes various operation behaviors of the historical user on the electronic map to construct a training sample set, and the training samples in the training sample set have comprehensive and single-entry influence, so that the trained first model and second model can adapt to more scenes.

After the server builds the first sample set, training a deep learning model, such as a convolutional neural network (Convolutional Neural Networks, CNN) model and a deep neural network (Deep Neural Networks, DNN), to obtain a first model, wherein the first model is used for predicting the retrieval probability of the current user.

After the server trains the first model, when the first model is used online, the current scene information of the current user is input into the first model, so that the retrieval probability of the user can be automatically and accurately obtained.

Fig. 6 is a schematic diagram of a process of training and applying a first model in the information recommendation method according to the embodiment of the present application. Referring to fig. 6, the history data includes a scene log and a behavior log. In the process of offline learning the model, the server determines a plurality of historical behaviors from a user log contained in the historical data, and different historical behaviors in the plurality of historical behaviors are used for indicating different operations of the user on the electronic map. And simultaneously, the server determines scene information of each historical behavior in the plurality of historical behaviors from scene logs contained in the historical data.

After the server acquires the historical behaviors and scene information of each historical behavior, log analysis fusion is carried out. In the fusion process, the server marks corresponding behavior samples by using scene information of each historical behavior in the plurality of historical behaviors to obtain a first training sample in the training sample set. The server then trains the deep learning model with a first sample in the first set of samples, resulting in a first model.

In the process of using the first model online, the server acquires the current scene information, and the current scene information is input into the first model, so that the retrieval probability of the current user can be obtained. The first model calculates the retrieval probability under the triggering of the user behavior, if the user clicks the electronic map to request to start the electronic map, the server performs online streaming analysis on the current scene information to obtain features, the analyzed features are input into the first model, and the retrieval probability is calculated based on the features. Assuming that 10 features are trained out of the first model, and that 7 features are included in the current scene information when the first model is used, the server sets the remaining 3 features to 0, and determines the retrieval probability based on the 7 features. The more features the current scene information contains, the more accurate the predicted probability of retrieval.

By adopting the scheme, the electronic equipment builds the first training sample set based on the historical data and performs model training to train the first model, and whether the user has the retrieval intention can be accurately predicted based on the first model.

Second, the second model.

When the user is predicted to have a retrieval intention, the server determines current scene information. And then, the server determines the click probability of each search term in the preset plurality of search terms clicked by the user according to the current scene information. Then, the server selects at least one search term with the click probability exceeding the preset click probability from the plurality of search terms, and generates the recommended card according to the at least one search term.

For example, when the server trains the second model, the server obtains the history data of the history user using the electronic map, and the history data can refer to the related description in the process of training the first model, which is not described herein. After the historical data is obtained, the server marks the search word by utilizing the historical scene information corresponding to the search word, and the obtained second sample is a positive sample; and labeling the search term by using other historical scene information, wherein the obtained second sample is a negative sample. For example, if historical user A clicks on a good at noon, then noon-good is a positive sample and noon-attractions are negative samples.

After the server builds the second sample set, training a learning ordering (Learning to Ranking, LTR) model to obtain a second model, where the second model is used to determine the probability that each search term is clicked in a certain scene.

After training the second model, the server inputs the current scene information of the current user into the second model when the second model is used online, so that the click probability of each search term can be obtained. And then, sorting the click probabilities, and generating recommended cards by using the search terms of the ranking top N.

By adopting the scheme, the server determines the click probability of each search term through the second model, so that the automation degree is high and the accuracy is high.

Fig. 7 is a schematic diagram of a process of training and applying a second model in the information recommendation method according to the embodiment of the present application. Referring to fig. 7, the history data includes a scene log and a behavior log. In the process of offline learning the model, the server determines a search term set from a user log contained in the historical data, wherein different search terms in the search term set are generated by different operations of the user on the electronic map. And simultaneously, the server determines the scene information of each search term in the plurality of search terms from the scene log contained in the historical data.

After obtaining the search words and the scene information of each search word, the server carries out log analysis fusion. In the fusion process, the server uses scene information of a plurality of search words to mark the corresponding search words to obtain a positive sample; and marking the search term by using the non-corresponding scene information to obtain a negative sample, wherein the positive sample and the negative sample form a second sample set.

In the process of using the second model online, the server acquires current scene information, and inputs the current scene information into the second model, so that the click probability of each search term can be obtained. And the second model calculates the clicking probability of each search term under the triggering of the user behavior, and if the user clicks the electronic map to request to start the electronic map, the server performs online streaming analysis on the current scene information to obtain features, inputs the analyzed features into the second model, and calculates the clicking probability of each search term based on the features.

In addition, after the second model is online, in the using process of the electronic map by the user, the server continues to generate a scene log and a behavior log according to the operation behavior of the user, and continues to learn the LTR model for a long time, so that the LTR model is updated.

By adopting the scheme, the electronic equipment builds the second training sample set based on the historical data and performs model training, a second model is trained, and a recommended card fitting the current scene of the user can be generated based on the second model, so that the interactive operation of the user and the electronic map is reduced.

In the above embodiment, after the server outputs the recommended card, the recommended card is pushed to the terminal device, so that the terminal device can pop up the recommended card on the electronic map. After clicking any search term on the recommended card, the user triggers the terminal equipment to send a search instruction. After receiving the search instruction, the server determines a plurality of POIs related to the search term according to the search term and outputs the POIs. And the terminal equipment displays the POIs on the electronic map and displays information such as scores of the POIs.

By adopting the scheme, after the server receives the recommendation instruction of the user, the POI expected by the user is determined according to the search term carried by the recommendation instruction and pushed to the terminal equipment, so that the interaction times of the user and the electronic map are reduced.

Fig. 8 is a flowchart of a model training method provided by an embodiment of the present application, where an execution body of the embodiment is an electronic device such as a server, and the embodiment includes:

801. and constructing training samples according to historical data to obtain a training sample set, wherein the historical data is data generated when a historical user operates the electronic map.

802. And training an initial model by using the training sample to obtain a target model, wherein the target model is a model associated with a recommended card, a plurality of search terms are displayed on the recommended card, and different search terms in the plurality of search terms represent points of interest (POIs) of different categories.

The target model is, for example, the first model described above, and the initial model is a deep learning model. Alternatively, the target model is the second model described above, and the initial model is the ranking model.

According to the model training method provided by the embodiment, the electronic equipment builds a training sample by utilizing historical data, performs model training to obtain a target model, predicts the retrieval probability of the user and the probability of each retrieval word clicked by the target model, and achieves the purpose of accurately identifying whether the user has retrieval complaints and the clicking probability of each retrieval word.

In a feasible design, when the initial model is a deep learning model and the target model is a first model, the electronic device constructs training samples according to historical data to obtain a training sample set, and constructs a plurality of first training samples according to the historical data to obtain a first training sample set. The electronic device then trains the deep learning model using the first set of training samples to obtain the first model, which is used to determine a retrieval probability for the user.

In a feasible design, when the electronic device constructs a plurality of first training samples according to the historical data to obtain a first training sample set, firstly, a plurality of historical behaviors are determined from a user log contained in the historical data, and different historical behaviors in the plurality of historical behaviors are used for indicating different operations of a historical user on the electronic map. Then, the electronic equipment determines scene information of each historical behavior in the plurality of historical behaviors from scene logs contained in the historical data, and marks corresponding behavior samples by using the scene information of each historical behavior in the plurality of historical behaviors to obtain a first training sample in the training sample set.

In the above embodiment, when the initial model is the sorting model and the target model is the second model, the electronic device constructs the training sample according to the history data to obtain the training sample set, and constructs the second training sample according to the history data to obtain the second training sample set. And then, the electronic equipment trains the sorting model by using the second training sample set to obtain a second model, wherein the second model is used for determining the click probability of each search term in the plurality of search terms.

In a feasible design, when the electronic device constructs a second training sample according to the historical data to obtain a second training sample set, firstly, a search word set is determined from a user log contained in the historical data, and different search words in the search word set are generated by different operations of a historical user on an electronic map. Then, the electronic equipment determines the scene information of each search word in the plurality of search words from the scene log contained in the historical data, and marks the corresponding search word by using the scene information of each search word in the plurality of search words, so as to obtain a second training sample in the second training sample set.

In the above embodiment, the operation of the electronic map by the history user includes at least one of the following operations: general search, positioning, navigation and clicking any search term after clicking the periphery.

The above description describes a specific implementation of the information recommendation method according to the embodiment of the present application, and the following is an embodiment of the apparatus of the present application, which may be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiments of the apparatus of the present application, please refer to the embodiments of the method of the present application.

Fig. 9 is a schematic structural diagram of an information recommendation device based on an electronic map according to an embodiment of the present application. The apparatus may be integrated in or implemented by an electronic device. As shown in fig. 9, in the present embodiment, the information recommendation apparatus 900 based on an electronic map may include:

a receiving module 91, configured to receive a request instruction of a user, where the request instruction is used to request to open an electronic map;

a prediction module 92 for predicting whether the user has a retrieval intention in response to the request instruction;

a processing module 93, configured to generate a recommendation card for the user when the prediction module 92 predicts that the user has a search intention, where a plurality of search terms are displayed on the recommendation card, and different search terms in the plurality of search terms represent points of interest POIs of different categories;

And the output module 94 is used for outputting the recommended card.

In a possible design, the prediction module 92 is specifically configured to determine current scene information in response to the request instruction, determine a search probability of the user according to the current scene information, and predict that the user has a search intention when the search probability is greater than a preset search probability.

In a feasible design, the prediction module 92 is configured to construct a plurality of first training samples according to historical data when determining the retrieval probability of the user according to the current scene information, so as to obtain a first training sample set, where the historical data is data generated when the historical user operates the electronic map; training a first model by using the first training sample set, wherein the first model is used for determining the retrieval probability of the user; and inputting the current scene information into the first model to obtain the retrieval probability of the user.

In a feasible design, the prediction module 92 constructs a plurality of first training samples according to historical data, and determines a plurality of historical behaviors from a user log included in the historical data when a first training sample set is obtained, wherein different historical behaviors in the plurality of historical behaviors are used for indicating different operations of a historical user on the electronic map; determining scene information of each historical behavior in the plurality of historical behaviors from scene logs contained in the historical data; and labeling corresponding behavior samples by using scene information of each historical behavior in the plurality of historical behaviors to obtain a first training sample in the training sample set.

In a possible design, the processing module 93 is configured to determine current scene information when the prediction module 92 predicts that the user has a search intention, and determine a click probability of each search term in a preset plurality of search terms by the user according to the current scene information; selecting at least one search term with the click probability exceeding the preset click probability from the plurality of search terms; and generating the recommended card according to the at least one search term.

In a feasible design, the processing module 93 constructs a second training sample according to historical data when determining a click probability of each search term in a plurality of preset search terms clicked by a user according to the current scene information, so as to obtain a second training sample set, wherein the historical data is data generated when a historical user operates the electronic map; training a second model by using the second training sample set, wherein the second model is used for determining the click probability of each search term in the plurality of search terms; and inputting the current scene information into the second model to obtain the click probability of each search term in the plurality of search terms.

In a possible design, the processing module 93 is configured to construct a second training sample according to the historical data, and determine a search term set from a user log included in the historical data when the second training sample set is obtained, where different search terms in the search term set are generated by different operations of the historical user on the electronic map; determining scene information of each search term in the plurality of search terms from a scene log contained in the historical data; and labeling the corresponding search words by using the scene information of each search word in the plurality of search words to obtain a second training sample in the second training sample set.

In a possible design, the operation of the electronic map by the history user includes at least one of: general search, positioning, navigation and clicking any search term after clicking the periphery.

In a possible design, the receiving module 91 is further configured to receive a search instruction of the user after the output module 94 outputs the recommended card, where the search instruction carries any one search term on the recommended card; the processing module 93 is further configured to determine, according to the search term, a plurality of POIs related to the search term; the output module 94 is further configured to output the multiple POIs.

Fig. 10 is a schematic structural diagram of a model training device according to an embodiment of the present application. The apparatus may be integrated in or implemented by an electronic device. As shown in fig. 10, in the present embodiment, the model training apparatus 1000 may include:

the construction module 1001 is configured to construct a training sample according to historical data, so as to obtain a training sample set, where the historical data is data generated when a historical user operates the electronic map;

the training module 1002 is configured to train the initial model by using the training sample to obtain a target model, where the target model is a model associated with a recommendation card, and a plurality of search terms are displayed on the recommendation card, and different search terms in the plurality of search terms represent points of interest POIs of different categories.

In a feasible design, when the initial model is a deep learning model and the target model is a first model, the building module 1001 is configured to build a plurality of first training samples according to historical data to obtain a first training sample set;

the training module 1002 is configured to train the deep learning model to obtain the first model by using the first training sample set, where the first model is used to determine a retrieval probability of the user.

In a possible design, the construction module 1001 is configured to determine, from a scenario log included in the historical data, scenario information of each historical behavior in the plurality of historical behaviors; and labeling corresponding behavior samples by using scene information of each historical behavior in the plurality of historical behaviors to obtain a first training sample in the training sample set.

In a feasible design, when the initial model is a ranking model and the target model is a second model, the construction module 1001 is configured to construct a second training sample according to the historical data to obtain a second training sample set, and train the ranking model to obtain the second model by using the second training sample set, where the second model is used to determine click probability of each term in the plurality of terms.

In a feasible design, the training module 1002 is configured to determine a set of search terms from a user log included in the history data, where different search terms in the set of search terms are generated by different operations of the history user on the electronic map, determine scene information of each search term in the plurality of search terms from a scene log included in the history data, and label a corresponding search term by using the scene information of each search term in the plurality of search terms, so as to obtain a second training sample in the second training sample set.

Fig. 11 is a block diagram of an electronic device for implementing an information recommendation method according to an embodiment of the present application. The electronic equipment can also execute the model training method provided by the embodiment of the application. The electronic device provided by the embodiment of the application is described in detail below by taking an electronic device execution information recommending method as an example. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 11, the electronic device 1100 includes a computing unit 1101 that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM) 1102 or a computer program loaded from a storage unit 1108 into a Random Access Memory (RAM) 1103. In the RAM 1103, various programs and data required for the operation of the device 1100 can also be stored. The computing unit 1101, ROM 1102, and RAM 1103 are connected to each other by a bus 1104. An input/output (I/O) interface 1105 is also connected to bus 1104.

Various components in device 1100 are connected to I/O interface 1105, including: an input unit 1106 such as a keyboard, a mouse, etc.; an output unit 1107 such as various types of displays, speakers, and the like; a storage unit 1108, such as a magnetic disk, optical disk, etc.; and a communication unit 1109 such as a network card, modem, wireless communication transceiver, or the like. The communication unit 1109 allows the device 1100 to exchange information/data with other devices through a computer network such as the internet and/or various telecommunication networks.

The computing unit 1101 may be a variety of general purpose and/or special purpose processing components having processing and computing capabilities. Some examples of computing unit 1101 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, etc. The calculation unit 1101 performs the respective methods and processes described above, such as an information recommendation method. For example, in some embodiments, the information recommendation method may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as storage unit 1108. In some embodiments, some or all of the computer programs may be loaded and/or installed onto device 1100 via ROM 1102 and/or communication unit 1109. When the computer program is loaded into the RAM 1103 and executed by the computing unit 1101, one or more steps of the information recommendation method described above may be performed. Alternatively, in other embodiments, the computing unit 1101 may be configured to perform the information recommendation method by any other suitable means (e.g. by means of firmware).

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for carrying out methods of the present disclosure may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus such that the program code, when executed by the processor or controller, causes the functions/operations specified in the flowchart and/or block diagram to be implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and pointing device (e.g., a mouse or trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic input, speech input, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), and the internet.

The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so that the defects of high management difficulty, weak service expansibility and the like in the traditional physical hosts and virtual special servers (Virtual Private Server, VPS) are overcome. The server may also be a server of a distributed system, or a server incorporating a blockchain

It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps described in the present application may be performed in parallel, sequentially, or in a different order, provided that the desired results of the disclosed embodiments are achieved, and are not limited herein.

The above embodiments do not limit the scope of the present application. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the scope of the present application.

Claims (16)

1. An information recommendation method is applied to a server and comprises the following steps:

receiving a request instruction of a user, wherein the request instruction is used for requesting to open an electronic map;

in response to the request instruction, determining current scene information,

determining the retrieval probability of the user according to the current scene information;

when the retrieval probability is larger than a preset retrieval probability, predicting that the user has a retrieval intention;

Determining current scene information when it is predicted that the user has a retrieval intention;

determining the click probability of each search term in a plurality of preset search terms clicked by a user according to the current scene information;

selecting at least one search term with the click probability exceeding the preset click probability from the plurality of search terms;

generating a recommended card according to the at least one search term, wherein a plurality of search terms are displayed on the recommended card, and different search terms in the plurality of search terms represent points of interest (POIs) of different categories;

and outputting the recommended card.

2. The method of claim 1, wherein the determining the search probability of the user from the current context information comprises:

constructing a plurality of first training samples according to historical data to obtain a first training sample set, wherein the historical data is data generated when a historical user operates the electronic map;

training a first model by using the first training sample set, wherein the first model is used for determining the retrieval probability of the user;

and inputting the current scene information into the first model to obtain the retrieval probability of the user.

3. The method of claim 2, wherein the constructing a plurality of first training samples from the historical data to obtain a first set of training samples comprises:

Determining a plurality of historical behaviors from a user log contained in the historical data, wherein different historical behaviors in the plurality of historical behaviors are used for indicating different operations of a historical user on the electronic map;

determining scene information of each historical behavior in the plurality of historical behaviors from scene logs contained in the historical data;

and labeling corresponding behavior samples by using scene information of each historical behavior in the plurality of historical behaviors to obtain a first training sample in the training sample set.

4. The method of claim 1, wherein the determining, according to the current scene information, a click probability of each of a plurality of preset terms being clicked by a user includes:

constructing a second training sample according to historical data, and obtaining a second training sample set, wherein the historical data is data generated when a historical user operates the electronic map;

training a second model by using the second training sample set, wherein the second model is used for determining the click probability of each search term in the plurality of search terms;

and inputting the current scene information into the second model to obtain the click probability of each search term in the plurality of search terms.

5. The method of claim 4, wherein constructing a second training sample from the historical data to obtain a second training sample set comprises:

determining a search term set from a user log contained in the historical data, wherein different search terms in the search term set are generated by different operations of a historical user on an electronic map;

determining scene information of each search term in the plurality of search terms from a scene log contained in the historical data;

and labeling the corresponding search words by using the scene information of each search word in the plurality of search words to obtain a second training sample in the second training sample set.

6. The method of claim 3 or 5, wherein the operation of the electronic map by the historic user comprises at least one of: general search, positioning, navigation and clicking any search term after clicking the periphery.

7. The method of any of claims 1-3, wherein the outputting the recommended card further comprises:

receiving a search instruction of the user, wherein the search instruction carries any search word on the recommended card;

determining a plurality of POIs related to the search term according to the search term;

Outputting the POIs.

8. An information recommendation device based on an electronic map is applied to a server and comprises:

the receiving module is used for receiving a request instruction of a user, wherein the request instruction is used for requesting to open the electronic map;

the prediction module is used for responding to the request instruction, determining current scene information, determining the retrieval probability of the user according to the current scene information, and predicting whether the user has a retrieval intention or not when the retrieval probability is larger than a preset retrieval probability;

the processing module is used for determining current scene information when the prediction module predicts that the user has the retrieval intention; determining the click probability of each search term in a plurality of preset search terms clicked by a user according to the current scene information; selecting at least one search term with the click probability exceeding the preset click probability from the plurality of search terms; generating a recommended card according to the at least one search term, wherein a plurality of search terms are displayed on the recommended card, and different search terms in the plurality of search terms represent points of interest (POIs) of different categories;

and the output module is used for outputting the recommended card.

9. The device of claim 8, wherein the prediction module is configured to construct a plurality of first training samples according to historical data when determining the retrieval probability of the user according to the current scene information, so as to obtain a first training sample set, wherein the historical data is data generated when the historical user operates the electronic map; training a first model by using the first training sample set, wherein the first model is used for determining the retrieval probability of the user; and inputting the current scene information into the first model to obtain the retrieval probability of the user.

10. The device of claim 9, wherein the prediction module constructs a plurality of first training samples according to historical data, and determines a plurality of historical behaviors from a user log contained in the historical data when a first training sample set is obtained, wherein different historical behaviors in the plurality of historical behaviors are used for indicating different operations of a historical user on the electronic map; determining scene information of each historical behavior in the plurality of historical behaviors from scene logs contained in the historical data; and labeling corresponding behavior samples by using scene information of each historical behavior in the plurality of historical behaviors to obtain a first training sample in the training sample set.

11. The device of claim 8, wherein the processing module constructs a second training sample according to historical data when determining a click probability of each search term in a plurality of preset search terms clicked by a user according to the current scene information, so as to obtain a second training sample set, wherein the historical data is data generated when a historical user operates the electronic map; training a second model by using the second training sample set, wherein the second model is used for determining the click probability of each search term in the plurality of search terms; and inputting the current scene information into the second model to obtain the click probability of each search term in the plurality of search terms.

12. The device of claim 11, wherein the processing module is configured to construct a second training sample according to the historical data, and determine a search term set from a user log included in the historical data when the second training sample set is obtained, where different search terms in the search term set are generated by different operations of the historical user on the electronic map; determining scene information of each search term in the plurality of search terms from a scene log contained in the historical data; and labeling the corresponding search words by using the scene information of each search word in the plurality of search words to obtain a second training sample in the second training sample set.

13. The apparatus of claim 10 or 12, wherein the operation of the electronic map by the historic user comprises at least one of: general search, positioning, navigation and clicking any search term after clicking the periphery.

14. The device according to any one of claims 8-10, wherein,

the receiving module is further used for receiving a search instruction of the user after the output module outputs the recommended card, wherein the search instruction carries any search word on the recommended card;

The processing module is further used for determining a plurality of POIs related to the search term according to the search term;

the output module is further used for outputting the POIs.

15. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein, the liquid crystal display device comprises a liquid crystal display device,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

16. A non-transitory computer readable storage medium storing computer instructions for causing an electronic device to perform the method of any one of claims 1-7.