CN111080339A

CN111080339A - Method and device for generating category preference data based on scene

Info

Publication number: CN111080339A
Application number: CN201911128690.8A
Authority: CN
Inventors: 桑梓森
Original assignee: Koukouxiangchuan Beijing Network Technology Co ltd
Current assignee: Koukouxiangchuan Beijing Network Technology Co ltd
Priority date: 2019-11-18
Filing date: 2019-11-18
Publication date: 2020-04-28
Anticipated expiration: 2039-11-18
Also published as: CN111080339B

Abstract

The invention discloses a category preference data generation method and device based on scenes, wherein the method comprises the following steps: collecting user behavior historical data related to search behaviors, wherein the user behavior historical data comprises user behavior scene historical data and corresponding user behavior target entity historical data; carrying out scene recognition processing according to the historical data of the user behavior scene to obtain user search scene data; determining category data and behavior preference data corresponding to the user behavior target entity according to the historical data of the user behavior target entity; and generating category preference data under the user search scene according to the user search scene data, the category data and the behavior preference data, and performing offline storage. The method achieves the purpose of mining the category preference of the user under different scenes by the big data, and the category preference is generated by combining the search scenes of the user, so that the mined user preference is more fit with the actual preference habit of the user, the accuracy is higher, and the search recommendation effect can be improved when search recommendation is subsequently performed.

Description

Method and device for generating category preference data based on scene

Technical Field

The invention relates to the technical field of internet, in particular to a category preference data generation method and device based on scenes.

Background

In a traditional search mode, user preference content is usually mined according to a search log so as to perform search recommendation, but in the field of O2O, because a user consumes on line, the content of interest of the user in different scenes is different, and therefore the search recommendation mode without considering the search scene of the user has poor effect.

The prior art (CN 105975522 a) discloses a multi-domain content recommendation method, which calculates the preference degree of each user for different domains in each situation through a server, and selects a domain with the preference degree greater than a set preference degree threshold as a preference domain; then obtaining the preference degree of each user for all contents in all fields, and sequencing according to the preference degree from high to low to obtain a preference content list; then, the content belonging to the preference field is selected from the preference content list, and a recommended content list of each user in each scene is obtained. And finally, the server searches a recommended content list corresponding to the current scene in the recommended list according to the current scene and sends the recommended content list to the client.

However, the inventor finds out in the process of implementing the invention that: in the prior art, for each user, the preferences of the user in different scenarios are mined based on the browsing records of the user, and then the user is recommended to search by referring to the preferences of the user in different scenarios. The method is characterized in that the portrait description is carried out on the user according to the historical browsing record of the user, and then the search recommendation is carried out according to the portrait of the user, for a certain user, the data source which is depended on is the browsing record of the user, the data source of the method is single, the mined user preference content is lack of diversity and poor in accuracy, and therefore the search recommendation effect is poor.

Disclosure of Invention

In view of the above, the present invention has been made to provide a scene-based category preference data generating method and apparatus that overcomes or at least partially solves the above-mentioned problems.

According to an aspect of the present invention, there is provided a method for generating category preference data based on a scene, including:

collecting user behavior historical data related to search behaviors, wherein the user behavior historical data comprises user behavior scene historical data and corresponding user behavior target entity historical data;

carrying out scene recognition processing according to the historical data of the user behavior scene to obtain user search scene data;

determining category data and behavior preference data corresponding to the user behavior target entity according to the historical data of the user behavior target entity;

and generating category preference data under the user search scene according to the user search scene data, the category data and the behavior preference data, and performing offline storage.

Optionally, the user behavior scenario history data includes: time data, location data, and/or search terms;

performing scene recognition processing according to the user behavior scene historical data, and obtaining user search scene data further comprises:

and obtaining a time interval label, a position label and/or a search intention label under the search scene of the user according to the time data, the position data and/or the search words.

Optionally, the time period tag comprises: a plurality of time period grading tags, whether a workday time period tag and/or a veto time period tag;

the location tag includes: a plurality of location type tags, whether a frequent location tag and/or a distance tag between a search location and a current location;

the search intent tag includes: an entity intent tag, a category intent tag, an address intent tag, and/or a content intent tag.

Optionally, the user behavior target entity history data includes: user browsing behavior data and user consumption behavior data of target entities related to the search behavior;

the behavior preference data includes: the amount of data that the user browses behavioral data and the user consumes behavioral data.

Optionally, the generating category preference data in the user search scenario according to the user search scenario data, the category data and the behavior preference data further includes:

generating category preference data under user search scenes with different priorities according to the user search scene data, the category data and the behavior preference data;

the user search scenes are as follows according to the priority from high to low: the search system comprises a user search scene containing a search intention label, a time period label and a position label, a user search scene containing a time period label and a position label, and a user search scene containing a time period label.

Optionally, after obtaining the user search scene data, the method further includes:

and aggregating the user search scene data, and screening out the user search scene data with the data volume smaller than a preset data volume threshold value.

According to another aspect of the present invention, there is provided a scene-based search recommendation method, including:

receiving user behavior real-time data related to the search behavior, wherein the user behavior real-time data comprises user behavior scene real-time data;

inquiring user real-time searching scene data matched with the user behavior scene real-time data;

inquiring category preference data which are stored in an off-line mode and correspond to the scene data searched by the user in real time;

recommending search results according to category preference data;

and returning the recommended search result to the client so that the client can display the search result.

Optionally, the user behavior scene real-time data includes: time data, location data, and/or search terms;

querying the user real-time search context data that matches the user behavior context real-time data further comprises:

and obtaining a time interval label, a position label and/or a search intention label under the real-time search scene of the user according to the time data, the position data and/or the search words.

Optionally, the querying of category preference data stored offline and corresponding to the real-time search scene data of the user specifically includes:

inquiring category preference data under a user search scene with the highest priority corresponding to the user real-time search scene data in category preference data under user search scenes with different priorities stored offline;

Optionally, the recommending search results according to category preference data specifically includes:

and recommending the search result by using the category preference data as a ranking factor.

According to another aspect of the present invention, there is provided a scene-based category preference data generating apparatus including:

the acquisition module is suitable for acquiring user behavior historical data related to the search behavior, and the user behavior historical data comprises user behavior scene historical data and corresponding user behavior target entity historical data;

the scene recognition module is suitable for carrying out scene recognition processing according to the historical data of the user behavior scene to obtain user search scene data;

the data analysis module is suitable for determining category data and behavior preference data corresponding to the user behavior target entity according to the historical data of the user behavior target entity;

and the scene preference analysis module is suitable for generating category preference data under the user search scene according to the user search scene data, the category data and the behavior preference data and storing the category preference data in an off-line mode.

the scene recognition module is further adapted to: and obtaining a time interval label, a position label and/or a search intention label under the search scene of the user according to the time data, the position data and/or the search words.

Optionally, the time period tag comprises: a plurality of time period grading tags, whether a workday time period tag and/or a veto time period tag; the location tag includes: a plurality of location type tags, whether a frequent location tag and/or a distance tag between a search location and a current location; the search intent tag includes: an entity intent tag, a category intent tag, an address intent tag, and/or a content intent tag.

Optionally, the scene preference analysis module is further adapted to: generating category preference data under user search scenes with different priorities according to the user search scene data, the category data and the behavior preference data;

Optionally, the apparatus further comprises:

and the aggregation processing module is suitable for aggregating the user search scene data and screening out the user search scene data with the data volume smaller than the preset data volume threshold value.

According to another aspect of the present invention, there is provided a scene-based search recommendation apparatus including:

the receiving module is suitable for receiving user behavior real-time data related to the searching behavior, and the user behavior real-time data comprises user behavior scene real-time data;

the matching module is suitable for inquiring the user real-time searching scene data matched with the user behavior scene real-time data;

the query module is suitable for querying category preference data which are stored in an off-line mode and correspond to the scene data searched by the user in real time;

the search recommending module is suitable for recommending search results according to the category preference data;

and the return module is suitable for returning the recommended search result to the client so that the client can display the search result.

the matching module is further adapted to: and obtaining a time interval label, a position label and/or a search intention label under the real-time search scene of the user according to the time data, the position data and/or the search words.

Optionally, the query module is further adapted to:

Optionally, the search recommendation module is further adapted to:

According to yet another aspect of the present invention, there is provided a computing device comprising: the system comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete mutual communication through the communication bus;

the memory is used for storing at least one executable instruction, and the executable instruction enables the processor to execute the operation corresponding to the scene-based category preference data generation method and the scene-based search recommendation method.

According to still another aspect of the present invention, there is provided a computer storage medium having at least one executable instruction stored therein, the executable instruction causing a processor to perform operations corresponding to the scene-based category preference data generating method and the scene-based search recommendation method as described above.

According to the method and the device for generating the category preference data based on the scene, user behavior historical data related to search behaviors are collected, and the user behavior historical data comprise user behavior scene historical data and corresponding user behavior target entity historical data; carrying out scene recognition processing according to the historical data of the user behavior scene to obtain user search scene data; determining category data and behavior preference data corresponding to the user behavior target entity according to the historical data of the user behavior target entity; and generating category preference data under the user search scene according to the user search scene data, the category data and the behavior preference data, and performing offline storage. According to the method, the historical data of the user behaviors related to the search behaviors are analyzed and processed, the search scene data, the category data and the user preference data of the user are generated, the category preference data under the search scene is generated according to the search scene data and the user preference data, the purpose of mining the category preference of the user under different scenes through big data is achieved, the category preference is generated according to the search scene, the user preference data can be better fit with the actual preference habit of the user after being mined, the accuracy is higher, and the search recommendation effect can be improved when the search recommendation is carried out subsequently.

According to the scene-based search recommendation method and device, user behavior real-time data related to search behaviors are received, and the user behavior real-time data comprise user behavior scene real-time data; inquiring user real-time searching scene data matched with the user behavior scene real-time data; and inquiring category preference data which is stored in an off-line mode and corresponds to the scene data searched by the user in real time. According to the method, the search scenes of the user are identified by combining the information of three dimensions of time, space and search intention, search sorting is carried out according to category preference data corresponding to the search scenes of the user, search recommendation combining the search scenes is achieved, and the search recommendation effect can be improved.

The foregoing description is only an overview of the technical solutions of the present invention, and the embodiments of the present invention are described below in order to make the technical means of the present invention more clearly understood and to make the above and other objects, features, and advantages of the present invention more clearly understandable.

Drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to refer to like parts throughout the drawings. In the drawings:

FIG. 1 illustrates a flow diagram of a method for generating category preference data based on a scene according to an embodiment of the present invention;

FIG. 2 illustrates a flow diagram of a method for generating category preference data based on a scene according to another embodiment of the present invention;

FIG. 3 is a flowchart illustrating a method for scene-based search recommendation according to another embodiment of the present invention;

FIG. 4 is a flow diagram illustrating a scenario-based search recommendation process in another embodiment of the present invention;

FIG. 5 illustrates a functional block diagram of a scene-based category preference data generating apparatus according to still another embodiment of the present invention;

FIG. 6 is a functional block diagram illustrating a scene-based search recommendation apparatus according to another embodiment of the present invention;

FIG. 7 illustrates a schematic structural diagram of a computing device in accordance with embodiments of the present invention;

FIG. 8 illustrates a schematic structural diagram of a computing device according to an embodiment of the present invention.

Detailed Description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

In the existing search scenes, the first type of search scenes is that a user searches for an obvious target direction, if the user inputs a search word 'kendiry', the category direction of the search word 'kendiry' is obvious, the search target of the user is clear, namely 'kendiry' is searched, and then the system directly recommends a corresponding search result according to the search target of the user; the second type of search scene is that the user search target direction is not obvious enough, for example, the user inputs a search word "beef", under the search scenes of different time, space and intention, the category preferences of the user for the target are different, the user also finds beef, the user tends to a beef steamed stuffed bun with the light fast food category in the morning, the user tends to a beef chafing dish with dinner in the noon, and the like, the category direction of the search word "beef" is not obvious, and the search scenes of the user about time, space and intention need to be determined, so that the category preferences of the user under the search scenes can be further determined; when searching for addresses, the second category of search scenarios has greater diversity, for example, when searching for addresses with longer distances, it is more likely to find restaurants with higher consumption and higher quality, the category prefers to be relatively high, and when the search addresses are closer, the consumption is less obvious.

Based on the method, the device and the system, the scene of the time, the space and the search intention when the user searches is identified, particularly the search intention is identified aiming at the search words with unobvious category directions, category preferences under different search scenes are generated, and the obtained scene category preferences are applied to search sequencing.

Fig. 1 shows a flowchart of a method for generating category preference data based on a scene according to an embodiment of the present invention, where the method of the present embodiment is implemented based on big data, and as shown in fig. 1, the method includes:

step S101, collecting user behavior historical data related to search behaviors, wherein the user behavior historical data comprises user behavior scene historical data and corresponding user behavior target entity historical data.

In this embodiment, history data of user behaviors related to search behaviors, specifically, the search behaviors refer to historical search behaviors of the user. The historical data of the user behavior scene related to the search behavior may include time data, location data, and/or data related to scenes such as search words and the like corresponding to the historical search behavior of the user, that is, time, location, and/or search words used by the user when searching. The corresponding user behavior target entity historical data specifically refers to various behavior data, such as behavior data generated by a user clicking and browsing an entity behavior, behavior data generated by a user purchasing an entity behavior, and the like, for a target entity, generated after the user searches, where the entity refers to: things that exist objectively and are distinguishable from each other include, in particular, commodities, stores, tickets, and the like.

And S102, carrying out scene identification processing according to the historical data of the user behavior scene to obtain user search scene data.

Through the steps, the search scene where the user searches is identified, for example, historical data of the user behavior scene is labeled based on a scene understanding method, and the search scene data of the user is obtained.

Step S103, determining category data and behavior preference data corresponding to the user behavior target entity according to the historical data of the user behavior target entity.

And counting the historical data of the target entity of the user behavior to obtain category data and behavior preference data corresponding to the target entity for which the user historical behavior aims. For example, the categories to which the target entity for which the user's purchasing behavior belongs and the number of times each category is purchased are counted.

And step S104, generating category preference data under the user search scene according to the user search scene data, the category data and the behavior preference data, and performing offline storage.

According to the content, the corresponding relation exists between the user search scene data and the category data and the behavior preference data, the user search scene is determined according to the corresponding relation between the data and the user search scene data, the category data is subjected to preference grading processing according to the category preference data under the user search scene, so that the category preference data under the user search scene is obtained, and then the generated category preference data under the user search scene is stored in an off-line mode.

According to the method for generating the category preference data based on the scene, provided by the embodiment, user behavior historical data related to search behaviors are collected, wherein the user behavior historical data comprises user behavior scene historical data and corresponding user behavior target entity historical data; carrying out scene recognition processing according to the historical data of the user behavior scene to obtain user search scene data; determining category data and behavior preference data corresponding to the user behavior target entity according to the historical data of the user behavior target entity; and generating category preference data under the user search scene according to the user search scene data, the category data and the behavior preference data, and performing offline storage. According to the method, the historical data of the user behaviors related to the search behaviors are analyzed and processed, the search scene data, the category data and the user preference data of the user are generated, the category preference data under the search scene is generated according to the search scene data and the user preference data, the purpose of mining the category preference of the user under different scenes through big data is achieved, the category preference is generated according to the search scene, the user preference data can be better fit with the actual preference habit of the user after being mined, the accuracy is higher, and the search recommendation effect can be improved when the search recommendation is carried out subsequently.

Fig. 2 is a flowchart illustrating a method for generating category preference data based on a scene according to another embodiment of the present invention, as shown in fig. 2, the method includes:

step S201, collecting user behavior history data related to search behavior, wherein the user behavior history data comprises user behavior scene history data and corresponding user behavior target entity history data, and the user behavior scene history data comprises: time data, location data, and/or search terms.

In this embodiment, history data of user behaviors related to search behaviors, specifically, the search behaviors refer to historical search behaviors of the user. The historical data of the user behavior scene related to the search behavior comprises time data, position data and/or search words corresponding to the historical search behavior of the user, namely the time, the position and/or the used search words when the user searches. The user behavior target entity historical data corresponding to the corresponding user behavior scene historical data specifically refers to various behavior data which are generated after the user searches and are aimed at the target entity.

For example, if a user searches in an office building with the search word "eat nearby" seven times in the morning, and finally purchases the coupon entity "kendeki breakfast coupon" in the search result after browsing the store entities "breakfast porridge shop" and "happy breakfast" in the search result, the collected user behavior scene history data related to the current search behavior includes: time data- "seven points in the morning", position data- "office building" and search term "nearby eaten", the corresponding user behavior target entity history data contains: browsing behavior data generated by the store entities "breakfast porridge store" and "breakfast at fun", and purchasing behavior data generated by the coupon entity "kendyy breakfast coupon" behavior.

Optionally, the historical data of the user behavior scene related to the location data and the historical data of the user behavior target entity corresponding to the historical data are collected, where the location data is related to indicate that the location corresponding to the location data is located in the same area, for example, the same city, the same administrative area, the same country, and the like. In practical application, in an O2O scenario, a city is used as a basic unit of people's life, and category distribution can be produced in a city dimension, and then category preference is produced based on the city category distribution.

Step S202, obtaining a time interval label, a position label and/or a search intention label under a user search scene according to the time data, the position data and/or the search words.

In this embodiment, the user search scene data is obtained by tagging the historical data of the user behavior scene related to the historical search behavior, that is, tagging the time, the position, and the search word when the user searches.

Aiming at the understanding of the time interval of the time data, namely identifying the time interval to which the time data corresponding to the search behavior belongs, and determining a time interval label of the search behavior, wherein the time interval label comprises: a multiple period rating tag, a workday period tag, and/or a veto period tag. Wherein the plurality of period grading labels comprise: the time interval labels correspond to a time interval, and when the time interval labels are printed, the time interval label corresponding to the time interval is printed after the time data belong to which time interval is determined. And judging by combining two pieces of time information according to the decision period label, namely generating time information of click browse behavior data aiming at a target entity and generating time information of consumption behavior data aiming at the target entity, such as time for a user to click a browse shop and time for consumption in the shop, determining whether the time is in the decision period according to the time interval between the time information and the time, if the time interval exceeds a preset threshold, determining the period label to be the decision period label, and if the time interval does not exceed the preset threshold, determining the period label to be the non-decision period label.

And performing spatial understanding on the position data, namely identifying a place to which the position data corresponding to the search behavior belongs, and determining the position label of the search behavior. When a user searches, based on a POI engine, the name, type and other information of the position where the user is located are obtained according to the geographical position information of the user, WiFi information connected with a terminal and the like, so that position data corresponding to a search behavior is obtained, and further a position label is determined according to the position data. The location tag includes: a plurality of location type tags, a frequent or no premises tag, and/or a distance tag of a search location from a current location. Wherein the plurality of location type tags comprises: office area tags, residential district tags, mall tags, office building tags, school zone tags, transportation hub tags, and the like. For the permanent station label, the permanent station data can be determined according to the position data contained in the historical behavior data of the user, and the position data corresponding to the searching behavior is compared with the permanent station data, so that whether the position data is the permanent station label can be determined. And aiming at the distance tags of the search place and the current place, setting a short-distance tag, a medium-distance tag and a long-distance tag in advance according to different distance intervals, wherein the short-distance tag corresponds to a distance within 500 meters, the medium-distance tag corresponds to 500 meters to 1000 meters, the long-distance tag corresponds to more than 1000 meters, and the like.

And (4) aiming at the search words, understanding the search intention, namely identifying the search intention according to the search words corresponding to the search behavior, and obtaining the search intention label of the search behavior. The search intent tag includes: an entity intent tag, a category intent tag, an address intent tag, and/or a content intent tag. When a user searches, the user usually inputs search words, and the search words are identified to determine whether the search intention of the user is to search for shops, cuisine, categories, addresses or contents, so as to determine a search intention label. For example, if the search word is 'eating nearby', the search intention of the user is determined to be a search shop, the corresponding search intention tag is an entity intention tag, and if the search word is 'mini vocal bar', the user may want to search for the nearby 'mini vocal bar', the search intention of the user is determined to be an address for searching for the 'mini vocal bar', and the search intention address intention tag of the user is determined. Therefore, based on the understanding of the search intention, the category search intention label can be attached to the search words with fuzzy categories input by the user, and the search intention of the user can be accurately identified.

Preferably, after the user search scene data is obtained, aggregation processing is performed on the user search scene data, and user search scene data with a data volume smaller than a preset data volume threshold value is screened out. The data volume can represent the occurrence frequency of a user search scene, and many users search in a certain search scene, so that the data volume of the user search scene of the search scene is larger. For example, for search scene data "night time period tag + very-located tag + long-distance tag", generally, such search scene is less frequent, and the reference value of user behavior data in such search scene is not great, so that search scene data with a small amount of data is screened out, and the influence of a small amount of special search scene data is eliminated in this way, thereby improving the accuracy of the preference of the scene category.

Step S203, determining category data and behavior preference data corresponding to the user behavior target entity according to the historical data of the user behavior target entity.

The historical data of the target entity of the user behavior in this embodiment specifically refers to various behavior data generated after the user searches for the target entity, that is, the browsing behavior data of the user related to the historical search behavior.

According to the user browsing behavior data related to the historical search behavior, determining category data to which a target entity to which the user browsing behavior aims belongs, and according to the user consumption behavior data related to the historical search behavior, determining a category to which a target entity to which the user consumption behavior aims belongs, so that the category data corresponding to the target entity of the user behavior is obtained. The behavior preference data corresponding to the user behavior target entity comprises: the amount of data that the user browses behavioral data and the user consumes behavioral data. Correspondingly, the data volume of the user browsing behavior data can also represent the number of times of the user browsing behavior, the data volume of the user consumption behavior data can also represent the number of times of the user consumption behavior, and the larger the data volume, the more the number of times is indicated.

Briefly, the step is to count categories to which target entities browsed and consumed by the user after searching, and browsing and consuming times corresponding to each category.

Wherein, the categories can be divided into a first category, a second category, and a third category, the first category comprising: gourmet, recreational, etc., secondary categories include: chinese meal, western meal, cosmetic massage, etc., the third category comprising: jiangzhe vegetable, Benzhuan vegetable, Farpai, spa and the like. The specific content and classification mode of the classification purpose can be set according to actual needs, which is not limited by the invention.

Step S204, according to the time interval label, the position label and/or the search intention label, the category data and the behavior preference data in the user search scene, the category preference data in the user search scene with different priorities are generated and are stored in an off-line mode.

Determining a user search scene according to the user search scene data, determining category data and behavior preference data in the user search scene, and performing grading processing on each category contained in the category data according to the behavior preference data to obtain preference scores of each category in the user search scene, wherein the higher the preference score of the category is, the stronger the preference degree of the user for the category is. Optionally, since the user browsing behavior and the user purchasing behavior represent user preferences of different degrees, the browsing behavior represents a user may be interested in, and the purchasing behavior represents a real need of the user, when rating the categories under the user scene according to the behavior preference data, different weight values are respectively set for the user browsing behavior data and the user purchasing behavior data, and the preference score of each category is calculated according to the data amount and the weight value of the user browsing behavior data, and the data amount and the weight value of the user purchasing behavior data. Optionally, the weighted value of the user browsing behavior data is lower than the weighted value of the user purchasing behavior data.

In the embodiment, in order to prevent the problem of sparseness of category preference data, the priority of the user search scene is set in a hierarchical category preference construction mode, so that the category preference data under the user search scenes with different priorities are generated. Optionally, the user search scenes are, in order of priority from high to low: the search system comprises a user search scene containing a search intention label, a time period label and a position label, a user search scene containing a time period label and a position label, and a user search scene containing a time period label.

For example, a user search scene containing a search intention tag, a time slot tag and a position tag is determined as a primary user search scene, a user search scene containing a time slot tag and a position tag is determined as a secondary user search scene, and a user search scene containing a time slot tag is determined as a tertiary user search scene.

The method of this embodiment will be described below by way of example. Firstly, acquiring user behavior historical data of a user A related to search behaviors, searching by using 'eating nearby' at 7 am on a weekday, browsing 'Qingfeng bunyao', and determining user search scene data as follows according to the steps: "breakfast time period tag + office building tag + entity intent tag". According to the search scene data of the user A, the determined user search scenes with different priorities comprise: first-level user search scenario: "breakfast time slot tag + office building tag + entity intention tag", secondary user search scenario: "breakfast time slot tag + office building tag", third-level user search scenario: "weekday period label". The user behavior target entity historical data corresponding to the user search scene data comprises browsing data generated when the user A browses the Qingfeng steamed bun and bunk.

For another example, if the user a searches at twelve am on weekdays using "eat nearby" and purchases a "kendyi coupon", then it is determined according to the above steps that the user search scenario data is: "lunch period label + office building label + entity intention label". The determined user search scenarios with different priorities comprise: first-level user search scenario: "lunch period label + office building label + entity intention label", secondary user search scene: "lunch period label + office building label", tertiary user search scene: "lunch period label". That is, the user behavior target entity history data corresponding to the user search scenario data includes browsing data generated when the user a purchases the "kendy coupon". Therefore, the user has different searching behaviors in different searching scenes, the historical searching scenes of the user can be identified according to the historical data of the user behaviors of the historical searching behaviors, and the entity data of the user behavior target in the historical searching scenes of the user is obtained.

Then, according to the mapping relation between the user searching scene data and the user behavior target entity historical data, determining category data and behavior preference data under each user searching scene, namely, according to browsing behaviors and purchasing behaviors which are actually generated by the user after the user searches under each searching scene, determining the category preference of the user under the searching scene.

For example, in a search scenario including "breakfast time slot tag + office building tag + entity intention tag", statistics of user behavior target entity data are obtained as follows: the purchasing behavior data of 300 users purchasing the 'Qingfeng steamed stuffed bun coupon', the browsing behavior data of 100 users browsing the 'Qingfeng steamed stuffed bun bunk', the purchasing behavior data of 90 users purchasing the 'kender base package', and the browsing behavior data of 10 users browsing the 'kender base shop'. The category corresponding to the "qingfeng steamed stuffed bun coupon" and the "qingfeng steamed stuffed bun bunk" is Chinese food, the category corresponding to the "kendir package meal" and the "kendir shop" is fast food, and then the category data under the search scene includes: chinese food and fast food. Obtaining user preference data corresponding to the category data in the search scene according to the analysis method comprises: the data volume of the browsing behavior of the user corresponding to the meal in the category is 100, the data volume proportion is 100/(100+300+10+90) ═ 20%, the data volume of the purchasing behavior of the user is 300, and the data volume proportion is 300/(100+300+10+90) ═ 60%; the data volume of the browsing behavior of the user corresponding to the category fast food is 10, the data volume proportion is 10/(100+300+10+90) ═ 2%, the data volume of the purchasing behavior of the user is 90, and the data volume proportion is 90/(100+300+10+90) ═ 18%. And setting the weight value of the browsing behavior data of the user to be 0.5 and the weight value of the purchasing behavior data of the user to be 0.8. Then, the finally calculated preference of the category Chinese food under the first-level user search scene is divided into: 20% 0.5+ 60% 0.8-0.58, the preferred scores for the category snack are: 2% by 0.5+ 18% by 0.8 ═ 0.154. Thus, category preference data of the first-level user search scene is obtained. The analysis modes of the category preference data of other user search scenes are consistent, and are not described herein any more. Of course, this is merely an example of the present invention and the aspects of the present invention are not limited thereto.

Further, after the category preference scores of the categories in the user search scene are obtained, screening can be performed according to a preset threshold value, so that the category preference scores of the categories in the category preference data are higher than the preset threshold value.

In summary, the method of the embodiment is implemented in a machine learning manner based on big data, learning is performed according to historical search data of a user to obtain user category preference data, different search scenes are identified by analyzing scene data including scene information of three dimensions of time, geographic position and search terms when the user searches, behavior data of a target entity is analyzed according to behavior data of the user when the user searches, preferences of the user for various categories under different search scenes are mined, the purpose of mining user scene preferences in a big data manner is achieved, the mined user scene preferences can better fit with actual preferences of the user, and accuracy is higher.

Fig. 3 is a schematic flowchart illustrating a method for recommending search based on a scene according to another embodiment of the present invention, where the method of this embodiment is implemented based on a method for generating category preference data based on a scene in the above embodiment, that is, the method according to the first two embodiments can generate category preference data in a user search scene, and the method of this embodiment applies the category preference data in the user search scene to perform search recommendation. As shown in fig. 3, the method includes:

step S301, receiving user behavior real-time data related to the search behavior, wherein the user behavior real-time data comprises user behavior scene real-time data.

The application scenario of the embodiment is a real-time search scenario. In order to distinguish the historical search behaviors, the search behavior of the real-time search scenario in this embodiment is referred to as a real-time search behavior, and when a user performs a search, the real-time data of the user behavior related to the real-time search behavior is received, and includes the real-time data of the user behavior scenario, such as time data, location data and/or search terms.

Step S302, inquiring the real-time searching scene data of the user matched with the real-time data of the user behavior scene.

The purpose of this step is to identify the user real-time search scenario from the user behavior scenario real-time data. Specifically, a time interval label, a position label and/or a search intention label in a user search scene are obtained according to the time data, the position data and/or the search words. Referring to the description of the above embodiment, the time period to which the time data corresponding to the real-time search behavior belongs is identified, the time period tag of the real-time search behavior is determined, the location to which the position data corresponding to the real-time search behavior belongs is identified, the position tag of the real-time search behavior is determined, and/or the search intention is identified according to the search word corresponding to the real-time search behavior, so as to obtain the search intention tag of the real-time search behavior. For specific meanings of the time period tag, the position tag and the search intention tag, reference is made to the description in the above embodiments, and details are not repeated here. Specifically, for the decision period label, when a real-time search scene is identified, whether the search scene is in the decision period is determined according to the distance between the search place and the current place, if the distance exceeds a preset threshold, the search scene is determined to be in the decision period, otherwise, the search scene is not in the decision period, and therefore whether the decision period label is determined.

Step S303, inquiring category preference data which are stored in an off-line mode and correspond to the scene data searched by the user in real time.

The step determines category preference data in a real-time search scene. For example, a user searches breakfast in an office building seven morning, and receiving real-time data of user behavior scenes related to real-time searching behaviors comprises the following steps: the time data- "seven points in the morning", the position data- "office building" and the search term "nearby eating", and the scene recognition processing in step S302, it is obtained that the user real-time search scene data includes "breakfast time slot tag + office building tag + entity intention tag". Then, in the category preference data stored offline, the category preference data in the user search scenario including "breakfast time period tag + office building tag + entity intention tag" is searched.

Further, when the category preference data with different priorities are stored offline, the category preference data in the user search scene with the highest priority corresponding to the user real-time search scene data in the category preference data with different priorities stored offline is queried. The user search scenes are as follows according to the priority from high to low: the search system comprises a user search scene containing a search intention label, a time period label and a position label, a user search scene containing a time period label and a position label, and a user search scene containing a time period label.

That is, the category preference data of the user search scene with the highest priority is searched first, if the category preference data corresponding to the real-time search scene is not found, the category preference data of the user search scene with the next priority is searched, and if the category preference data of the user search scene with the next priority is not found, the category preference data of the user search scene with the next priority is searched. The reason that the category preference data of the user search scene with the highest priority is matched with the real scene where the user is located is to search and sort as far as possible, the search recommendation effect can be improved in the mode, and the category preference data of the user search scene with the lower priority are to prevent the category preference data from being vacant due to data sparseness.

Following the above example, if the user searches for scene data in real time as "breakfast time slot tag + office building tag + entity intention tag", the user first searches for category preference data corresponding to "breakfast time slot tag + office building tag + entity intention tag", if the search is not successful, the user searches for category preference data corresponding to "breakfast time slot tag + office building tag", and if the search is not successful, the user continues to search for category preference data of "breakfast time slot tag".

And step S304, recommending search results according to the category preference data.

According to the above, the category preference data in the search scene is substantially the category preference scores of the categories in the user search scene, and then the search result recommendation is performed according to the category preference scores of the categories. That is, the category preference data is used as a ranking factor for search result recommendation.

For example, the ranking score of the category is increased for categories whose preference score is higher, and the ranking score of the category is appropriately decreased for categories whose preference score is lower. An alternative embodiment is: and presetting initial sequencing scores of all categories, and calculating the sequencing scores of all categories according to the category preference scores of all categories and the initial sequencing scores of all categories under a real-time search scene. And finally, sorting the search results according to the sorting scores of all categories. In practical applications, the category preference data may be used as a unique sorting factor, or may be used as one of the sorting factors, which is not limited in the present invention.

Step S305, the recommended search result is returned to the client side, so that the client side can display the search result.

And returning the recommended search result to the client, and displaying the recommended search result by the client.

According to the scene-based search recommendation method provided by the embodiment, the real-time search scene of the user is identified by combining the information of three dimensions of time, space and search intention, search sequencing is performed according to category preference data corresponding to the real-time search scene of the user, search recommendation combining the search scene is realized, and the search recommendation effect can be improved.

The method according to the embodiment of the present invention is described below with reference to a specific scenario, and fig. 4 is a schematic flow chart illustrating a search recommendation process based on a scenario according to another embodiment of the present invention, where as shown in fig. 4, the specific flow includes:

the method comprises the following steps that when a user initiates a search, real-time data of a user behavior scene related to a user search behavior are obtained, and the method comprises the following steps: location data, time data, and search terms. In this example, the location data includes: the method comprises the steps that longitude and latitude information reported by a client and WiFi information connected with user equipment during user searching, time data refer to searching time during user searching, and searching words refer to query input by the user during searching.

And secondly, understanding of the user search scene is conducted based on the real-time behavior scene data of the user.

In one aspect, geographic location understanding. The method specifically comprises the following steps: the method comprises the steps of identifying longitude and latitude information based on a POI engine, identifying WiFi information connected with user equipment by utilizing a WiFi identification model, and obtaining information such as the name and the type of the position where a user is located, wherein the POI engine refers to an engine related to geographic position information, such as an engine of a map website or an engine of a map APP, and the WiFi identification model is an identification model constructed based on a WiFi positioning technology. The identified location information is then tagged to obtain a geographic location understanding result, which is shown in fig. 4 as "city: beijing City; the area is as follows: shopping mall a "is geographical location information of the user," type of area in: mall is a geographical location tag.

On the other hand, temporal understanding. The method specifically comprises the following steps: it is determined to which time slot the user search time belongs based on the five-degree time-binning rule, and it is determined whether the user search time is a weekday, thereby determining a time stamp, i.e., "time slot: the afternoon tea time; whether it is a workday: NO ".

In yet another aspect, intent recognition. The method specifically comprises the following steps: the query search of the user is subjected to understanding identification by using the query understanding model and the store-to-store identification model, whether the user is in a decision stage or a store-to-consume stage is determined, and an intention understanding result is obtained, wherein the intention understanding result shown in fig. 4 represents that: the user's search term is hot pot, in a decision period, and the search intent is a category intent.

And thirdly, inquiring a scene understanding engine according to the real-time search scene of the user, and searching category preference matched with the real-time search scene of the user, wherein the scene understanding engine is used for generating category preference data and storing the category preference data. The category preference for finding a search scene matching the location (beijing city, shopping mall a, business circle/market) + time (afternoon tea time, non-workday) + search intention (hot pot, decision period, category intention) is: sichuan chafing dish and chafing dish string.

And fourthly, recommending search results, namely floating the shops with the preference categories, namely arranging the shops with the categories of Sichuan hotpot and hotpot string at the front position in the search results.

And fifthly, returning the recommended search result to the client, and displaying the search result by the client. By this, the whole process of search recommendation based on the user search scene is completed.

Fig. 5 is a functional block diagram illustrating a scene-based category preference data generating apparatus according to still another embodiment of the present invention, as shown in fig. 5, the apparatus including:

the acquisition module 51 is adapted to acquire user behavior history data related to the search behavior, where the user behavior history data includes user behavior scene history data and corresponding user behavior target entity history data;

the scene recognition module 52 is adapted to perform scene recognition processing according to the historical data of the user behavior scene to obtain user search scene data;

the data analysis module 53 is adapted to determine category data and behavior preference data corresponding to the user behavior target entity according to the historical data of the user behavior target entity;

and the scene preference analysis module 54 is adapted to generate category preference data in the user search scene according to the user search scene data, the category data and the behavior preference data, and store the category preference data in an offline manner.

the scene recognition module 52 is further adapted to: and obtaining a time interval label, a position label and/or a search intention label under the search scene of the user according to the time data, the position data and/or the search words.

the behavior preference data includes: the data volume of the user browsing behavior data and the user consumption behavior data.

Optionally, the scene preference analysis module 54 is further adapted to: generating category preference data under user search scenes with different priorities according to the user search scene data, the category data and the behavior preference data;

Optionally, the apparatus further comprises:

Fig. 6 is a functional block diagram illustrating a scene-based search recommendation apparatus according to another embodiment of the present invention, as shown in fig. 6, the apparatus including:

a receiving module 61 adapted to receive user behavior real-time data related to the search behavior, the user behavior real-time data including user behavior scene real-time data;

a matching module 62 adapted to query the user real-time search scene data matched with the user behavior scene real-time data;

the query module 63 is adapted to query category preference data which are stored offline and correspond to the scene data searched by the user in real time;

a search recommendation module 64 adapted to perform search result recommendation according to category preference data;

a returning module 65 adapted to return the recommended search results to the client for the client to present the search results.

the matching module 62 is further adapted to: and obtaining a time interval label, a position label and/or a search intention label under the real-time search scene of the user according to the time data, the position data and/or the search words.

Optionally, the query module 63 is further adapted to:

Optionally, the search recommendation module 64 is further adapted to:

The embodiment of the application provides a non-volatile computer storage medium, wherein at least one executable instruction is stored in the computer storage medium, and the computer executable instruction can execute the method for generating the category preference data based on the scene in any method embodiment.

The embodiment of the application provides a non-volatile computer storage medium, wherein at least one executable instruction is stored in the computer storage medium, and the computer executable instruction can execute the scene-based search recommendation method in any method embodiment.

Fig. 7 is a schematic structural diagram of a computing device according to an embodiment of the present invention, and the specific embodiment of the present invention does not limit the specific implementation of the computing device.

As shown in fig. 7, the computing device may include: a processor (processor)702, a Communications Interface 704, a memory 706, and a communication bus 708.

Wherein:

the processor 702, communication interface 704, and memory 706 communicate with each other via a communication bus 708.

A communication interface 704 for communicating with network elements of other devices, such as clients or other servers.

The processor 702 is configured to execute the program 710, and may specifically execute relevant steps in the above-described method for generating category preference data based on a scene.

In particular, the program 710 may include program code that includes computer operating instructions.

The processor 702 may be a central processing unit CPU, or an Application Specific Integrated Circuit (ASIC), or one or more Integrated circuits configured to implement an embodiment of the invention. The computing device includes one or more processors, which may be the same type of processor, such as one or more CPUs; or may be different types of processors such as one or more CPUs and one or more ASICs.

The memory 706 stores a program 710. The memory 706 may comprise high-speed RAM memory, and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.

The program 710 may specifically be used to cause the processor 702 to perform the following operations: collecting user behavior historical data related to search behaviors, wherein the user behavior historical data comprises user behavior scene historical data and corresponding user behavior target entity historical data;

In an alternative approach, the user behavior scenario history data comprises: time data, location data, and/or search terms;

the program 710 may be further specifically configured to cause the processor 702 to perform the following operations: and obtaining a time interval label, a position label and/or a search intention label under the search scene of the user according to the time data, the position data and/or the search words.

In an alternative form, the time period tag comprises: a plurality of time period grading tags, whether a workday time period tag and/or a veto time period tag; the location tag includes: a plurality of location type tags, whether a frequent location tag and/or a distance tag between a search location and a current location; the search intent tag includes: an entity intent tag, a category intent tag, an address intent tag, and/or a content intent tag.

In an alternative mode, the user behavior target entity history data comprises: user browsing behavior data and user consumption behavior data of target entities related to the search behavior; the behavior preference data includes: the amount of data that the user browses behavioral data and the user consumes behavioral data.

In an optional manner, the program 710 may be further specifically configured to cause the processor 702 to perform the following operations: generating category preference data under user search scenes with different priorities according to the user search scene data, the category data and the behavior preference data; the user search scenes are as follows according to the priority from high to low: the search system comprises a user search scene containing a search intention label, a time period label and a position label, a user search scene containing a time period label and a position label, and a user search scene containing a time period label.

The program 710 may be further specifically configured to cause the processor 702 to perform the following operations: and after the user search scene data is obtained, carrying out aggregation processing on the user search scene data, and screening out the user search scene data with the data volume smaller than a preset data volume threshold value.

Fig. 8 is a schematic structural diagram of a computing device according to an embodiment of the present invention, and the specific embodiment of the present invention does not limit the specific implementation of the computing device.

As shown in fig. 8, the computing device may include: a processor (processor)802, a Communications Interface 804, a memory 806, and a communication bus 808.

Wherein:

the processor 802, communication interface 804, and memory 806 communicate with one another via a communication bus 808.

A communication interface 804 for communicating with network elements of other devices, such as clients or other servers.

The processor 802 is configured to execute the program 810, and may specifically execute relevant steps in the above-described scenario-based search recommendation method embodiment.

In particular, the program 810 may include program code comprising computer operating instructions.

The processor 802 may be a central processing unit CPU, or an Application Specific Integrated Circuit (ASIC), or one or more Integrated circuits configured to implement an embodiment of the invention. The computing device includes one or more processors, which may be the same type of processor, such as one or more CPUs; or may be different types of processors such as one or more CPUs and one or more ASICs.

The memory 806 stores a program 810. The memory 806 may comprise high-speed RAM memory, and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.

The program 810 may be specifically configured to cause the processor 802 to perform the following operations:

recommending search results according to category preference data;

In an alternative approach, the user behavior scene real-time data comprises: time data, location data, and/or search terms; the program 810 may be further specifically configured to cause the processor 802 to perform the following operations: and obtaining a time interval label, a position label and/or a search intention label under the real-time search scene of the user according to the time data, the position data and/or the search words.

In an alternative manner, the program 810 may be further specifically configured to cause the processor 802 to perform the following operations: inquiring category preference data under a user search scene with the highest priority corresponding to the user real-time search scene data in category preference data under user search scenes with different priorities stored offline;

In an alternative manner, the program 810 may be further specifically configured to cause the processor 802 to perform the following operations: and recommending the search result by using the category preference data as a ranking factor.

The algorithms and displays presented herein are not inherently related to any particular computer, virtual machine, or other apparatus. Various general purpose systems may also be used with the teachings herein. The required structure for constructing such a system will be apparent from the description above. Moreover, the present invention is not directed to any particular programming language. It is appreciated that a variety of programming languages may be used to implement the teachings of the present invention as described herein, and any descriptions of specific languages are provided above to disclose the best mode of the invention.

In the description provided herein, numerous specific details are set forth. It is understood, however, that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.

Similarly, it should be appreciated that in the foregoing description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. However, the disclosed method should not be interpreted as reflecting an intention that: that the invention as claimed requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention.

Those skilled in the art will appreciate that the modules in the device in an embodiment may be adaptively changed and disposed in one or more devices different from the embodiment. The modules or units or components of the embodiments may be combined into one module or unit or component, and furthermore they may be divided into a plurality of sub-modules or sub-units or sub-components. All of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or elements of any method or apparatus so disclosed, may be combined in any combination, except combinations where at least some of such features and/or processes or elements are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.

Furthermore, those skilled in the art will appreciate that while some embodiments described herein include some features included in other embodiments, rather than other features, combinations of features of different embodiments are meant to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments may be used in any combination.

The various component embodiments of the invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that a microprocessor or Digital Signal Processor (DSP) may be used in practice to implement some or all of the functionality of some or all of the components in a computing device according to embodiments of the present invention. The present invention may also be embodied as apparatus or device programs (e.g., computer programs and computer program products) for performing a portion or all of the methods described herein. Such programs implementing the present invention may be stored on computer-readable media or may be in the form of one or more signals. Such a signal may be downloaded from an internet website or provided on a carrier signal or in any other form.

It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The usage of the words first, second and third, etcetera do not indicate any ordering. These words may be interpreted as names.

Claims

1. A category preference data generation method based on scenes comprises the following steps:

2. The method of claim 1, wherein the user behavior scenario history data comprises: time data, location data, and/or search terms;

the scene recognition processing according to the user behavior scene historical data to obtain the user search scene data further comprises:

3. A scene-based search recommendation method includes:

receiving user behavior real-time data related to search behaviors, wherein the user behavior real-time data comprises user behavior scene real-time data;

recommending search results according to the category preference data;

4. A scene-based category preference data generating apparatus comprising:

the acquisition module is suitable for acquiring user behavior historical data related to search behaviors, and the user behavior historical data comprises user behavior scene historical data and corresponding user behavior target entity historical data;

5. The apparatus of claim 4, wherein the user behavior scenario history data includes: time data, location data, and/or search terms;

6. A scene-based search recommendation apparatus comprising:

the receiving module is suitable for receiving user behavior real-time data related to searching behaviors, and the user behavior real-time data comprises user behavior scene real-time data;

7. A computing device, comprising: the system comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete mutual communication through the communication bus;

the memory is used for storing at least one executable instruction, and the executable instruction causes the processor to execute the operation corresponding to the scene-based category preference data generation method according to any one of claims 1-2.

8. A computer storage medium having stored therein at least one executable instruction for causing a processor to perform operations corresponding to the method for generating scene-based category preference data according to any one of claims 1-2.

9. A computing device, comprising: the system comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete mutual communication through the communication bus;

the memory is used for storing at least one executable instruction, and the executable instruction causes the processor to execute the operation corresponding to the scene-based search recommendation method according to claim 3.

10. A computer storage medium having stored therein at least one executable instruction for causing a processor to perform operations corresponding to the method of claim 3.