CN112711712A - Landmark search result determination method and device, electronic equipment and readable storage medium - Google Patents

Abstract

The invention discloses a method and a device for determining landmark search results, electronic equipment and a readable storage medium. Wherein, the method comprises the following steps: acquiring a target entity matched with the search content and click distribution characteristics corresponding to the target entity; inputting the target entity and the click distribution characteristics corresponding to the target entity into a click search model which is trained in advance to obtain a first output; inputting the text features corresponding to the target entity into a text search model trained in advance to obtain a second output; and determining a search result corresponding to the search content according to the first output and the second output. The invention solves the technical problem that the landmark searching result is inaccurate because the landmark searching intention is complicated and the searching intention of the user can not be accurately determined in the related technology.

Description

Landmark search result determination method and device, electronic equipment and readable storage medium

Technical Field

The invention relates to the technical field of data search, in particular to a landmark search result determining method and device, electronic equipment and a readable storage medium.

Background

The existing map search and general search generally focus on directly providing information matched with search terms of a user aiming at a search address of the user, and the requirements of the user are not well met. However, in a local life service search scene, a user searches for a merchant/landmark point, which may be information of a search landmark, and may also want to search for merchants and services around the landmark point, for example, searching for XX university, and if the user wants to search for merchants and services around beijing university, existing search engines all return XX university, and do not recognize and return search requirements of the user for XX university periphery, nor dynamically determine a display style according to the strength of the search intention of the user address.

Meanwhile, in the complex field of local life, the landmark searching intention is complex, the landmark name may include commodities, the landmark name may include dishes, the landmark name may be matched with a certain merchant name, and the user may also search the commodities/dishes/the landmark name, so that whether the address searching intention is needed to be identified more accurately for the query of the user.

The existing technical scheme only considers the matching condition of the search query of the user and the landmark library, and does not consider the complexity of landmark search intention in the field of complex local life. In many cases, a search user in the local life field wants to search not only a target location for searching but also merchants or services around the target location, and therefore a specific front-end display style needs to be determined according to the strength of the user's demand for the merchants around the target location.

In view of the above problems, no effective solution has been proposed.

Disclosure of Invention

The embodiment of the invention provides a method and a device for determining landmark search results, electronic equipment and a readable storage medium, which are used for at least solving the technical problem that the landmark search results are inaccurate due to the fact that landmark search intentions in related technologies are complex and the search intentions of users cannot be accurately determined.

According to an aspect of an embodiment of the present invention, there is provided a landmark search result determination method, including: acquiring a target entity matched with search content and click distribution characteristics corresponding to the target entity; inputting the target entity and the click distribution characteristics corresponding to the target entity into a click search model trained in advance to obtain a first output; inputting the text features corresponding to the target entity into a text search model trained in advance to obtain a second output; and determining a search result corresponding to the search content according to the first output and the second output.

Further, the target entity includes a landmark entity, wherein obtaining the target entity matching the search content and the click distribution feature corresponding to the target entity includes: determining a plurality of candidate entities matching the search content; obtaining the matching result of the entity names of the candidate entities and the search content; sorting the candidate entities according to the matching result and the click distribution of the candidate entities; and determining at least two landmark entities with different styles in the candidate entities according to the search content and the click distribution characteristics.

Further, determining a landmark entity of a pair of the plurality of candidate entities according to the search content and the click distribution feature includes: and processing the candidate entities according to the search attributes corresponding to the search content, the knowledge graphs corresponding to the candidate entities and the click distribution characteristics to obtain the landmark entities with at least two different styles.

Further, the first output includes a first target entity, where inputting the target entity and the click distribution feature corresponding to the target entity into a click search model trained in advance includes: acquiring a click behavior record corresponding to the search content from a preset database; acquiring real-time click behavior in target equipment; inputting the landmark entities, the click behavior records and the real-time click behaviors of at least two different styles into the click search model to obtain the first target entity.

Further, the second output includes a second target entity, where inputting the text features corresponding to the target entity into a text search model that is trained in advance includes: inputting the at least two different styles of landmark entities into the text search model to obtain the second target entity.

Further, determining a search result corresponding to the search content according to the first output and the second output includes: under the condition that the first target entity is matched with the second target entity, determining a strong landmark entity and a weak landmark entity in the search result according to the first target entity and the second target entity; and under the condition that the first target entity is not matched with the second target entity, determining a strong landmark entity and a weak landmark entity in the search result from the landmark entities in the at least two different styles according to the real-time click behavior.

Further, after determining a search result corresponding to the search content according to the first output and the second output, the method further includes: presenting the strong landmark entity on the target device; receiving actual click behaviors acting on the strong landmark entities; determining a landmark search intention according to the actual clicking behavior, wherein the landmark search intention comprises a strong intention and a weak intention; if the landmark searching intention is the strong intention, showing POI around the strong landmark entity; if the landmark search intention is the weak intention, the weak landmark entity is shown.

Further, determining a search result corresponding to the search content according to the first output and the second output, further comprising: displaying the first target entity and the second target entity in the target device if the first target entity and the second target entity do not match; receiving a selection action acting on the first target entity or the second target entity; inputting the selection behavior, the first target entity, the second target entity and the click behavior record into a click search model to obtain a specified landmark entity; presenting the designated landmark entity in the user device.

According to another aspect of the embodiments of the present invention, there is also provided a landmark search result determination apparatus, including: the acquisition unit is used for acquiring a target entity matched with the search content and click distribution characteristics corresponding to the target entity; the first processing unit is used for inputting the target entity and the click distribution characteristics corresponding to the target entity into a click search model which is trained in advance so as to obtain a first output; the second processing unit is used for inputting the text features corresponding to the target entity into a text search model which is trained in advance so as to obtain a second output; and the determining unit is used for determining a search result corresponding to the search content according to the first output and the second output.

According to another aspect of the embodiments of the present invention, there is also provided an electronic device, including a processor, a memory, and a program or instructions stored on the memory and executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the landmark search result determination method as described above.

According to another aspect of the embodiments of the present invention, there is also provided a readable storage medium, on which a program or instructions are stored, which when executed by a processor, implement the steps of landmark search result determination as described above.

In the embodiment of the invention, the target entity matched with the search content and the click distribution characteristic corresponding to the target entity are obtained; inputting the target entity and the click distribution characteristics corresponding to the target entity into a click search model which is trained in advance to obtain a first output; inputting the text features corresponding to the target entity into a text search model trained in advance to obtain a second output; and determining a search result corresponding to the search content according to the first output and the second output. The search result is predicted according to the search content text matching content and the user click behavior distribution corresponding to the search content, the purpose of accurately identifying the search intention of the user is achieved, the technical effect of improving the accuracy of the search result is achieved, and the technical problem that the landmark search result is inaccurate due to the fact that the landmark search intention is complex in the related technology and the search intention of the user cannot be accurately determined is solved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:

FIG. 1 is a diagram illustrating an alternative landmark search result determination method according to an embodiment of the present invention;

FIG. 2a is a schematic diagram of an alternative search content according to an embodiment of the present invention;

FIG. 2b is a diagram of an alternative search result according to an embodiment of the present invention;

FIG. 3a is a schematic diagram of yet another alternative search content according to an embodiment of the present invention;

FIG. 3b is a schematic diagram of yet another alternative search result according to an embodiment of the present invention;

fig. 4 is a schematic structural diagram of an alternative landmark search result determining apparatus according to an embodiment of the present invention.

Detailed Description

In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

Example 1

According to an embodiment of the present invention, a landmark search result determining method is provided, as shown in fig. 1, the method may specifically include the following steps:

s102, acquiring a target entity matched with the search content and click distribution characteristics corresponding to the target entity;

the entities described in this embodiment include, but are not limited to, entities or objects that can be queried by a user on any network platform, such as a geographic location, a landmark, a merchant, a commodity, a dish, etc. The search content in this embodiment may be a query term and/or a selected query condition input by a user through a search page, or may be a query term and/or a query condition generated by a network platform according to a behavior or a record of the user. In this embodiment, the target entity is a landmark entity determined according to the search content and matching with the search content, for example, under some specified search of the user, any merchant may become a landmark candidate, such as "hotel near kend" as the search content, and actually the user wants to search for food near "kend" as the search content; for example, the search content "food in northern four rings of the home where the user is located" is actually a food that the user wishes to search for near the "store in northern four rings of the home where the user is located".

In this embodiment, the click characteristics of the target entities are obtained by distributing and acquiring user click behaviors respectively corresponding to a plurality of target entities, in which a user inputs search content through a search page, in a database of the network platform. Namely, after the user inputs the search content in the search page of the network platform, aiming at the click rate and other characteristics of a certain target entity.

In specific implementation, the search content for the current query can be acquired through a query entry of the network platform. Further, the network platform performs search query according to the acquired search content, and recalls a target entity matched with the search content and click distribution characteristics corresponding to the target entity. Specifically, each search content corresponds to a plurality of target entities, and each target entity has a corresponding click distribution characteristic (i.e., user click distribution). For example, in a certain opinion website platform, when a user inputs a search content query "restaurant around XX university," the platform will recall "XX university" and entities such as merchants, restaurants, dishes, and commodities near "XX university. In this scenario, the search content acquired by the platform is "restaurants around XX university," each recalled entity may be considered as a target entity, and the click rate or the selection rate of the user for each target entity in the search content may be used as a click distribution feature.

S104, inputting the target entity and the click distribution characteristics corresponding to the target entity into a click search model which is trained in advance to obtain a first output;

and then, for certain search content input to a query entry of the website platform, inputting the recalled target entity and the click distribution characteristics corresponding to the target entity into a click search model which is trained in advance, and obtaining a first output by the click search model according to the target entity matched with the search content and the click distribution characteristics corresponding to the target entity. The search content of the query input includes but is not limited to: a landmark text matching the search content, a shop text matching the search content, a goods text matching the search content, and a dishes text matching the search content, and so on.

In the embodiment, in a specific implementation, the click search model needs to be trained first.

In some embodiments of the present invention, a training sample set is constructed according to a search content of a network platform by a user, a target entity matched with the search content, and a click distribution feature corresponding to the target entity, where each training sample in the training sample set includes: searching the content, the target entity and the click distribution characteristics corresponding to the target entity.

First, user behavior data stored by a network platform is obtained. Generally, each query action of a user on a network platform generates a corresponding query record, and the query records stored on the network platform include, but are not limited to: the method comprises the steps of searching content, a target entity, click distribution characteristics corresponding to the target entity, whether the target entity is clicked or not and the like.

Then, training samples are constructed according to the user behavior data. In some embodiments of the present invention, a training sample may be obtained by processing the user behavior data. Each training sample comprises: the method comprises the following steps of searching content, a target entity, click distribution characteristics corresponding to the target entity, whether the target entity is clicked or not and the like. In some embodiments of the present invention, each training sample is represented as a triple, including < search content, a plurality of target entities associated with the search content, a click distribution characteristic corresponding to each target entity, and whether to click >.

Next, the click search model is trained based on the constructed training sample set. And taking the search content, the plurality of target entities and the corresponding click distribution characteristics as model input, and taking the clicked target entities as model targets to train a click search model.

In a preferred embodiment of this embodiment, the click search model comprises a DDPG model (Deep Deterministic Policy Gradient). In the DDPG model, the real-time click rate change of a user in a network platform and the click distribution of the user under a target entity are taken as evaluation criteria. If the click rate is increased and the click behavior of the user is close to the existing click distribution of the landmark style, the current landmark style is considered to be positive, and otherwise, the current landmark style is considered to be negative, and a penalty is given. DDPG introduces an empirical playback and dual-network method to improve the problem of difficult convergence of Actor-Critic.

S106, inputting the text features corresponding to the target entity into a text search model which is trained in advance to obtain a second output;

and then, for certain search content input to the query entrance of the website platform, inputting the recalled target entity to a text search model which is trained in advance, and obtaining a second output by the text search model according to the text content corresponding to the target entity matched with the search content. The search content of the query input includes but is not limited to: landmark text content matching the search content, shop text content matching the search content, merchandise text content matching the search content, dish text content matching the search content, and the like.

In the embodiment, when implementing, the text search model needs to be trained first.

In some embodiments of the present invention, a training sample set is constructed according to search content of a user on a network platform and text features corresponding to target entities matched with the search content, where each training sample in the training sample set includes: and searching the content and the text characteristics corresponding to the target entity.

First, user behavior data stored by a network platform is obtained. Generally, each query action of a user on a network platform generates a corresponding query record, and the query records stored on the network platform include, but are not limited to: search content, target entity, whether target entity is clicked, etc.

Then, training samples are constructed according to the user behavior data. In some embodiments of the present invention, a training sample may be obtained by processing the user behavior data. Each training sample comprises: searching content, text characteristics corresponding to the target entity, whether the target entity is clicked or not and the like. In some embodiments of the present invention, each training sample is represented as a binary group, and includes < search content, text features corresponding to a plurality of target entities associated with the search content, and whether to click > respectively.

Next, the text search model is trained based on the constructed training sample set. And taking the text characteristics respectively corresponding to the search content and the plurality of target entities as model input, and taking the clicked target entity as a model target to train a text search model.

In a preferred embodiment of this embodiment, the text search model comprises a Bert model (text classification). In the Bert model, the second output by the Bert model is used as a landmark text by adjusting the parameters of the Bert model, so that the purpose of shielding non-text non-landmark data is achieved.

And S108, determining a search result corresponding to the search content according to the first output and the second output.

Specifically, the corresponding search result is determined according to a first output of the click search model and a second output of the text search model. For example, according to a first landmark output by the click search model and a second landmark output by the text search model, a landmark search result corresponding to the search content is determined. And if the first landmark is the same as the second landmark or the text of the first landmark is matched with the second landmark, outputting a matched or consistent landmark searching result. And if the first landmark is not the same as the second landmark or the texts are not matched, determining a search result according to a preset rule or the clicking behavior of the user.

In one example, three target entities are processed through a Bert model to obtain that the second output is 2 landmark entities, the three target entities and click distribution characteristics corresponding to the three target entities are processed through a DDPG model to obtain that the first output is one target entity, the first output and the second output of the DDPG model are matched, and if a matched landmark entity exists, the landmark entity is output.

It should be noted that, in this embodiment, text matching is performed on search content to obtain a target entity, so as to implement preliminary screening on a landmark entity corresponding to the search content, and then, a landmark entity is further screened according to the target entity and a click distribution feature corresponding to the target entity, so as to implement determination of a corresponding landmark search result according to a search intention of a user.

According to the embodiment, the target entity matched with the search content and the click distribution characteristics corresponding to the target entity are obtained; inputting the target entity and the click distribution characteristics corresponding to the target entity into a click search model which is trained in advance to obtain a first output; inputting the text features corresponding to the target entity into a text search model trained in advance to obtain a second output; and determining a search result corresponding to the search content according to the first output and the second output. The search result is predicted according to the search content text matching content and the user click behavior distribution corresponding to the search content, the purpose of accurately identifying the search intention of the user is achieved, the technical effect of improving the accuracy of the search result is achieved, and the technical problem that the landmark search result is inaccurate due to the fact that the landmark search intention is complex in the related technology and the search intention of the user cannot be accurately determined is solved.

Optionally, in this embodiment, the target entity includes a landmark entity, where the target entity matched with the search content and the click distribution feature corresponding to the target entity are obtained, including but not limited to: determining a plurality of candidate entities matching the search content; obtaining the matching result of the entity names of a plurality of candidate entities and the search content; sorting the multiple candidate entities according to the matching result and the click distribution of the multiple candidate entities; and determining at least two landmark entities with different styles in the candidate entities according to the search content and the click distribution characteristics.

Specifically, text query or text matching is performed on the search content in a preset database to determine a plurality of candidate entities matched with the search content, wherein the landmark library of the preset database includes, but is not limited to, point landmarks, linear landmarks and planar landmarks. The point landmarks comprise landmark data of specific merchants, government agencies, traffic facilities, scenic spots, colleges, hospitals and the like; the surface landmarks comprise landmark data such as administrative districts, business circles, scenic spots and the like; the linear landmarks include landmark data such as roads and subway lines. Meanwhile, under some specified searches of the user, any merchant can become a landmark candidate point, namely a candidate entity. For example, the search content is "hotel near kentucky", which is actually a food that the user wishes to search for near "kentucky", wherein "kentucky", "hotel" are candidate entities.

The candidate entities are then ranked according to the text match similarity of the search content to the entity name of the candidate entity, and the user's click distribution under the search content. For example, the search content is "hotel near kendirk at XX university," and the candidate entity corresponding to the search content is "XX university," kendirk store, "or" hotel. Candidate entities "XX university", "kendiry store" and "hotel" are ranked according to the text match similarity of "XX university", "kendiry store", "hotel" and search content "hotels near kendiry of XX university", and the user's click behavior on "XX university", "kendiry store" and "hotel".

Optionally, in this embodiment, the landmark entity in a pair of the multiple candidate entities is determined according to the search content and the click distribution feature, including but not limited to: and processing the candidate entities according to the search attributes corresponding to the search content, the knowledge graph corresponding to the candidate entities and the click distribution characteristics to obtain at least two landmark entities with different styles.

Specifically, in the present embodiment, the search attribute corresponding to the search content includes, but is not limited to, a search intention corresponding to the search content, and a category distribution of the search content. Processing the multiple candidate entities corresponding to the search content according to the knowledge graph of the entity link corresponding to the multiple candidate entity locks, the search intention corresponding to the search content, the category distribution and the click distribution characteristic corresponding to the search content, and removing non-landmark entities in the multiple candidate entities to determine at least two landmark entities with different styles corresponding to the search content of the user.

It should be noted that different styles of landmark entities refer to different or unmatched landmark entities. In an actual application scenario, landmark entities matching with each other are landmark entities having different names for the same landmark. For the landmark entities which do not match or are not identical, the patterns of the two landmark entities are different.

Through the embodiment, the multiple candidate entities are processed according to the search attributes corresponding to the search content, the knowledge maps corresponding to the multiple candidate entities and the click distribution characteristics to obtain at least two landmark entities with different styles, the search content is subjected to text matching to obtain the target entity, and the landmark entities corresponding to the search content are primarily screened to obtain the landmark entities corresponding to the search content.

Optionally, in this embodiment, the first output includes a first target entity, where the target entity and the click distribution feature corresponding to the target entity are input into a click search model that is trained in advance, including but not limited to: acquiring a click behavior record corresponding to the search content from a preset database; acquiring real-time click behavior in target equipment; and inputting at least two landmark entities with different styles, click behavior records and real-time click behaviors into a click search model to obtain a first target entity.

In a specific application scenario, a query record corresponding to search content stored in a network platform is acquired, and a click behavior record corresponding to the search content is acquired from a preset database of the network platform. For example, the search result includes "XX elementary school" and "florist" in the query record corresponding to the search content "XX elementary school florist". And acquiring the click behavior records of the two types of the users, namely acquiring the selection rate or click rate of the users for the primary XX school and the florist.

In this embodiment, the target device is a terminal device that sends search content to the network platform or initiates a query request based on the search content, and the terminal device includes, but is not limited to, a mobile terminal, a PC, and other terminal devices that can communicate with the network platform and perform data transmission.

Then, the current real-time clicking behavior of the user in the target device is obtained, and at least two landmark entities with different styles matched with the search content are input into a pre-trained clicking search model according to the clicking behavior record stored in the cloud, the real-time clicking behavior in the local target device, and the first target entity is obtained.

Through the embodiment, whether the search content of the user is the search intention for searching the current merchant or surrounding the current merchant is identified according to the online real-time feedback, so that the phenomenon that poor merchants trigger surrounding search and the search intention of the user is misunderstood is avoided.

Optionally, in this embodiment, the second output includes a second target entity, where the text features corresponding to the target entity are input into a text search model that is trained in advance, and the text search model includes, but is not limited to: and inputting at least two different styles of landmark entities into the text search model to obtain a second target entity.

Specifically, search contents are matched in a preset database to obtain at least two landmark entities of different styles matched with the search contents, the two landmark entities of different styles are subjected to text classification through a text search model, and the landmark entities input into the text search model are classified to screen out non-text and non-landmark entities.

In one example, the search content "XX elementary school flower shop" includes 3 target entities "XX elementary school", "flower shop" and "flower", and the non-landmark entity "flower" is screened out by inputting "XX elementary school", "flower shop" and "flower" into the Bert model to get a second output "XX elementary school", "flower shop".

Optionally, in this embodiment, determining a search result corresponding to the search content according to the first output and the second output includes, but is not limited to: under the condition that the first target entity is matched with the second target entity, determining a strong landmark entity and a weak landmark entity in the search result according to the first target entity and the second target entity; and under the condition that the first target entity is not matched with the second target entity, determining a strong landmark entity and a weak landmark entity in the search result from the landmark entities in at least two different styles according to the real-time click behavior.

Specifically, the corresponding search result is determined according to a first target entity output by the click search model and a second target entity output by the text search model. For example, according to a first target entity output by the click search model and a second target entity output by the text search model, a landmark search result corresponding to the search content is determined. And if the first target entity is the same as the second target entity or the text of the first target entity is matched with the text of the second target entity, outputting a search result of the corresponding landmark entity. And if the first landmark is not the same as the second landmark or the text is not matched, determining the first target entity output by the click search model as a strong landmark entity, and determining the second target entity output by the text search model as a weak landmark entity. Preferentially displaying a strong landmark entity and then displaying a weak landmark entity in a target terminal, for example, in a landmark entity display page of a target entity, displaying the strong landmark entity at a first preset position of target equipment, displaying the weak landmark entity at a second preset position of the target equipment, wherein the display sequence of the first preset position is prior to the second preset position; or the display area of the first preset position is larger than that of the second preset position.

In one example, the three target entities are processed through a Bert model to obtain a second output which is a landmark entity a and a landmark entity B, the three target entities and click distribution characteristics corresponding to the three target entities are processed through a DDPG model to obtain a first output which is a landmark entity B, the first output and the second output are matched, and if a matched landmark entity B exists, the landmark entity B is displayed. In another example, if the first output is a landmark entity C, the landmark entity C is set as a strong landmark entity, and the landmark entities a and B are set as weak landmark entities.

Optionally, in this embodiment, after determining the search result corresponding to the search content according to the first output and the second output, the method further includes, but is not limited to: displaying a strong landmark entity on a target device; receiving actual click behaviors acting on the strong landmark entities; determining a landmark search intention according to the actual click behavior, wherein the landmark search intention comprises a strong intention and a weak intention; displaying POI around the strong landmark entity under the condition that the landmark searching intention is strong intention; and in the case that the landmark searching intention is a weak intention, displaying a weak landmark entity.

Specifically, in this embodiment, a strong landmark entity is displayed on the target device, an actual click behavior or a selection behavior of the user acting on the strong landmark entity is received, and if the user selects or clicks the strong landmark entity, it is considered that the search intention of the user on the landmark corresponding to the strong landmark entity is strong, and then POIs or POI navigation pages around the strong landmark entity are further displayed. If the user does not click the strong landmark entity all the time in the process of browsing the landmarks or the browsing time of the landmark detail page of the strong landmark entity is smaller than a preset time threshold after clicking the strong landmark entity, determining that the landmark searching intention of the current user is a weak intention, and displaying the weak landmark entity on the target device.

In one example, as shown in fig. 2a, a search is performed for a search content "XX" in a graphic user page 20 of a target device, after the search content "XX" is input, a landmark entity 201, a landmark entity 202 and a landmark entity 203 are displayed in the graphic user page 20, a selection operation of a user on the landmark entity 201 is received, and a search intention of the user is considered to be a strong intention. As shown in fig. 2b, a POI navigation page 210 near the landmark entity 201 is displayed in the graphical user page 20, and the POI navigation page 210 is configured to display POIs near the landmark entity 201 according to a selection operation of the user.

In another example, as shown in fig. 3a, a search is performed for the search content "XX seafood market" in the graphic user page 30 of the target device, a landmark entity 301, a landmark entity 302 and a landmark entity 303 are shown in the graphic user page 30, and if the user clicks the search control 32 without clicking any landmark entity, the search intention of the user is considered to be weak intention. As shown in fig. 3b, the weak landmark entity 311 and the weak landmark entity 312 corresponding to the landmark entity 301 are shown in the graphical user interface.

As a preferable technical solution, in this embodiment, determining a search result corresponding to the search content according to the first output and the second output further includes: under the condition that the first target entity is not matched with the second target entity, displaying the first target entity and the second target entity in the target equipment; receiving a selection action acting on a first target entity or a second target entity; inputting a click search model according to the selection behavior, the first target entity, the second target entity and the click behavior record to obtain a specified landmark entity; a designated landmark entity is presented in a user device.

Specifically, in this embodiment, under the condition that the first target entity and the second target entity are not matched, the first target entity and the second target entity are displayed in the target device, the first target entity and the second target entity are predicted again by clicking the search model according to the feedback of the click behavior of the user on the landmark entity, the search intention of the user is further determined according to the feedback of the selection behavior of the user, and thus the specified landmark entity which better meets the search intention of the user is obtained.

According to the embodiment, the target entity matched with the search content and the click distribution characteristics corresponding to the target entity are obtained; inputting the target entity and the click distribution characteristics corresponding to the target entity into a click search model which is trained in advance to obtain a first output; inputting the text features corresponding to the target entity into a text search model trained in advance to obtain a second output; and determining a search result corresponding to the search content according to the first output and the second output. The search result is predicted according to the search content text matching content and the user click behavior distribution corresponding to the search content, the purpose of accurately identifying the search intention of the user is achieved, the technical effect of improving the accuracy of the search result is achieved, and the technical problem that the landmark search result is inaccurate due to the fact that the landmark search intention is complex in the related technology and the search intention of the user cannot be accurately determined is solved.

It should be noted that, for simplicity of description, the above-mentioned method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present invention is not limited by the order of acts, as some steps may occur in other orders or concurrently in accordance with the invention. Further, those skilled in the art should also appreciate that the embodiments described in the specification are preferred embodiments and that the acts and modules referred to are not necessarily required by the invention.

Through the above description of the embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but the former is a better implementation mode in many cases. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (e.g., a mobile phone, a computer, a server, or a network device) to execute the method according to the embodiments of the present invention.

Example 2

According to an embodiment of the present invention, there is also provided a landmark search result determination device for implementing the above-described landmark search result determination device, as shown in fig. 4, the device including:

1) the acquiring unit 40 is configured to acquire a target entity matched with the search content and a click distribution feature corresponding to the target entity;

2) the first processing unit 42 is configured to input the target entity and the click distribution feature corresponding to the target entity into a click search model trained in advance to obtain a first output;

3) the second processing unit 44 is configured to input the text features corresponding to the target entity into a text search model that is trained in advance, so as to obtain a second output;

4) a determining unit 46, configured to determine a search result corresponding to the search content according to the first output and the second output.

Optionally, the specific example in this embodiment may refer to the example described in embodiment 1 above, and this embodiment is not described again here.

Example 3

According to an embodiment of the present invention, there is also provided an electronic device for implementing the above landmark search result determination method, the electronic device including a processor, a memory, and a program or an instruction stored on the memory and executable on the processor, the program or the instruction, when executed by the processor, implementing the steps of the landmark search result determination method according to embodiment 1.

The electronic device includes:

1) processor with a memory having a plurality of memory cells

2) Memory device

Optionally, in this embodiment, the memory is configured to store program code for performing the following steps:

s1, acquiring a target entity matched with the search content and click distribution characteristics corresponding to the target entity;

s2, inputting the target entity and the click distribution characteristics corresponding to the target entity into a click search model trained in advance to obtain a first output;

s3, inputting the text features corresponding to the target entity into a text search model trained in advance to obtain a second output;

and S4, determining a search result corresponding to the search content according to the first output and the second output.

Optionally, the specific example in this embodiment may refer to the example described in embodiment 1 above, and this embodiment is not described again here.

Example 4

Embodiments of the present invention also provide a readable storage medium on which a program or instructions are stored, which when executed by a processor implement the steps of the landmark search result determination method according to embodiment 1.

Optionally, in this embodiment, the readable storage medium is configured to store program code for performing the following steps:

s1, acquiring a target entity matched with the search content and click distribution characteristics corresponding to the target entity;

s2, inputting the target entity and the click distribution characteristics corresponding to the target entity into a click search model trained in advance to obtain a first output;

s3, inputting the text features corresponding to the target entity into a text search model trained in advance to obtain a second output;

and S4, determining a search result corresponding to the search content according to the first output and the second output.

Optionally, the storage medium is further configured to store program codes for executing the steps included in the method in embodiment 1, which is not described in detail in this embodiment.

Optionally, in this embodiment, the storage medium may include, but is not limited to: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.

Optionally, the specific example in this embodiment may refer to the example described in embodiment 1 above, and this embodiment is not described again here.

The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.

The integrated unit in the above embodiments, if implemented in the form of a software functional unit and sold or used as a separate product, may be stored in the above computer-readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing one or more computer devices (which may be personal computers, servers, network devices, etc.) to execute all or part of the steps of the method according to the embodiments of the present invention.

In the above embodiments of the present invention, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

In the several embodiments provided in the present application, it should be understood that the disclosed client may be implemented in other manners. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one type of division of logical functions, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

Claims (11)

1. A method for landmark search result determination, comprising:

acquiring a target entity matched with search content and click distribution characteristics corresponding to the target entity;

inputting the target entity and the click distribution characteristics corresponding to the target entity into a click search model trained in advance to obtain a first output;

inputting the text features corresponding to the target entity into a text search model trained in advance to obtain a second output;

and determining a search result corresponding to the search content according to the first output and the second output.

2. The method of claim 1, wherein the target entity comprises a landmark entity, wherein,

acquiring a target entity matched with the search content and click distribution characteristics corresponding to the target entity, wherein the click distribution characteristics comprise:

determining a plurality of candidate entities matching the search content;

obtaining the matching result of the entity names of the candidate entities and the search content;

sorting the candidate entities according to the matching result and the click distribution of the candidate entities;

and determining at least two landmark entities with different styles in the candidate entities according to the search content and the click distribution characteristics.

3. The method of claim 2, wherein determining a landmark entity from a pair of candidate entities based on the search content and the click distribution feature comprises:

and processing the candidate entities according to the search attributes corresponding to the search content, the knowledge graphs corresponding to the candidate entities and the click distribution characteristics to obtain the landmark entities with at least two different styles.

4. The method of claim 2, wherein the first output comprises a first target entity, wherein,

inputting the target entity and the click distribution characteristics corresponding to the target entity into a click search model which is trained in advance, wherein the click distribution characteristics comprise:

acquiring a click behavior record corresponding to the search content from a preset database;

acquiring real-time click behavior in target equipment;

inputting the landmark entities, the click behavior records and the real-time click behaviors of at least two different styles into the click search model to obtain the first target entity.

5. The method of claim 4, wherein the second output comprises a second target entity, wherein,

inputting the text features corresponding to the target entity into a text search model which is trained in advance, wherein the text search model comprises the following steps:

inputting the at least two different styles of landmark entities into the text search model to obtain the second target entity.

6. The method of claim 5, wherein determining search results corresponding to the search content based on the first output and the second output comprises:

under the condition that the first target entity is matched with the second target entity, determining a strong landmark entity and a weak landmark entity in the search result according to the first target entity and the second target entity;

and under the condition that the first target entity is not matched with the second target entity, determining a strong landmark entity and a weak landmark entity in the search result from the landmark entities in the at least two different styles according to the real-time click behavior.

7. The method of claim 6, further comprising, after determining a search result corresponding to the search content based on the first output and the second output:

presenting the strong landmark entity on the target device;

receiving actual click behaviors acting on the strong landmark entities;

determining a landmark search intention according to the actual clicking behavior, wherein the landmark search intention comprises a strong intention and a weak intention;

if the landmark searching intention is the strong intention, showing POI around the strong landmark entity;

if the landmark search intention is the weak intention, the weak landmark entity is shown.

8. The method of claim 6, wherein determining search results corresponding to the search content based on the first output and the second output further comprises:

displaying the first target entity and the second target entity in the target device if the first target entity and the second target entity do not match;

receiving a selection action acting on the first target entity or the second target entity;

inputting the selection behavior, the first target entity, the second target entity and the click behavior record into a click search model to obtain a specified landmark entity;

presenting the designated landmark entity in the target device.

9. A landmark search result determination apparatus, comprising:

the acquisition unit is used for acquiring a target entity matched with the search content and click distribution characteristics corresponding to the target entity;

the first processing unit is used for inputting the target entity and the click distribution characteristics corresponding to the target entity into a click search model which is trained in advance so as to obtain a first output;

the second processing unit is used for inputting the text features corresponding to the target entity into a text search model which is trained in advance so as to obtain a second output;

and the determining unit is used for determining a search result corresponding to the search content according to the first output and the second output.

10. An electronic device comprising a processor, a memory, and a program or instructions stored on the memory and executable on the processor, the program or instructions when executed by the processor implementing the steps of the landmark search result determination method according to any one of claims 1-8.

11. A readable storage medium, on which a program or instructions are stored, which when executed by a processor, carry out the steps of the landmark search result determination method according to any one of claims 1-8.

Images