CN111782748B - Map retrieval method, information point POI semantic vector calculation method and device - Google Patents

Map retrieval method, information point POI semantic vector calculation method and device Download PDF

Info

Publication number
CN111782748B
CN111782748B CN202010597660.8A CN202010597660A CN111782748B CN 111782748 B CN111782748 B CN 111782748B CN 202010597660 A CN202010597660 A CN 202010597660A CN 111782748 B CN111782748 B CN 111782748B
Authority
CN
China
Prior art keywords
semantic
target poi
poi
target
vector
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010597660.8A
Other languages
Chinese (zh)
Other versions
CN111782748A (en
Inventor
臧文华
范淼
卓安
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Baidu Netcom Science and Technology Co Ltd
Original Assignee
Beijing Baidu Netcom Science and Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Baidu Netcom Science and Technology Co Ltd filed Critical Beijing Baidu Netcom Science and Technology Co Ltd
Priority to CN202010597660.8A priority Critical patent/CN111782748B/en
Publication of CN111782748A publication Critical patent/CN111782748A/en
Application granted granted Critical
Publication of CN111782748B publication Critical patent/CN111782748B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/29—Geographical information databases
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Abstract

The application discloses a map retrieval method, a calculation method and a calculation device of information point POI semantic vectors, relates to the technical field of map retrieval, and can be applied to the cloud field and the deep learning field. The map retrieval method comprises the following steps: calculating semantic vectors of map retrieval information; searching a target POI semantic vector matched with the semantic vector of the map retrieval information according to a pre-established information point POI semantic vector index library; the semantic vector of the target POI is obtained by calculation according to the semantic features of the graph of the target POI, and the semantic features of the graph of the target POI aggregate the semantic features between the target POI and N neighbor nodes. The generation of the POI semantic vector is optimized, the semantic vector expression of the POI is richer, the accuracy and the confidence level of the POI semantic vector can be improved, and therefore the retrieval prompt can be recalled accurately, the basic recall rate is effectively improved, the recall effect is improved, and the problems existing in the prior art are solved.

Description

Map retrieval method, information point POI semantic vector calculation method and device
Technical Field
The application relates to a data processing technology, in particular to the technical field of map retrieval, and specifically relates to a map retrieval method, a method and a device for calculating semantic vectors of information Points (POIs).
Background
In the map retrieval technology, a semantic recall mode can be adopted to realize recall of a retrieval prompt (sug for short). However, in the current semantic recall method, a POI vector index is generally established according to POI names, and for map retrieval information input by a user, only the similarity between the semantic vector of the map retrieval information and the semantic vector of the POI name is considered, and when there is a difference in the map retrieval information, the recall effect is poor.
Disclosure of Invention
The application provides a map retrieval method, and a method and a device for calculating semantic vectors of POIs (point of information).
According to a first aspect, the present application provides a map retrieval method, the method comprising:
calculating semantic vectors of map retrieval information;
searching a target POI semantic vector matched with the semantic vector of the map retrieval information according to a pre-established information point POI semantic vector index library; the semantic vector of the target POI is obtained by calculation according to the semantic features of the target POI, the semantic features of the target POI aggregate the semantic features between the target POI and N neighbor nodes, and N is an integer greater than 1.
According to a second aspect, the present application provides a method for calculating a semantic vector of a POI, the method comprising:
obtaining the graph semantic features of a target POI, wherein the graph semantic features of the target POI aggregate the semantic features between the target POI and N neighbor nodes, and N is an integer greater than 1;
and calculating the semantic vector of the target POI according to the graph semantic features of the target POI.
According to a third aspect, the present application provides a map retrieval apparatus comprising:
the first calculation module is used for calculating semantic vectors of map retrieval information;
the searching module is used for searching the semantic vector of the target POI matched with the semantic vector of the map retrieval information according to the pre-established information point POI semantic vector index library; the semantic vector of the target POI is obtained by calculation according to the semantic features of the target POI, the semantic features of the target POI aggregate the semantic features between the target POI and N neighbor nodes, and N is an integer greater than 1.
According to a fourth aspect, the present application provides a computing device for semantic vectors of POI, including:
the first acquisition module is used for acquiring the graph semantic features of the target POI, wherein the graph semantic features of the target POI aggregate the semantic features between the target POI and N neighbor nodes, and N is an integer greater than 1;
And the calculating module is used for calculating the semantic vector of the target POI according to the graph semantic features of the target POI.
According to a fifth aspect, the present application provides an electronic device comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform any one of the methods of the first aspect or to enable the at least one processor to perform any one of the methods of the second aspect.
According to a sixth aspect, the present application provides a non-transitory computer readable storage medium storing computer instructions for causing the computer to perform any of the methods of the first aspect, or for causing the computer to perform any of the methods of the second aspect.
According to the technology of the application, the semantic vector of the POI is obtained through calculation according to the semantic features of the graph of the POI, and the semantic features of the graph of the POI aggregate the semantic features between the POI and N neighbor nodes, so that the generation of the semantic vector of the POI is optimized, the semantic vector expression of the POI is richer, the accuracy and the confidence of the semantic vector of the POI can be improved, and a more accurate and efficient POI semantic vector index library can be constructed. Therefore, even if the map retrieval information input by the user is different, the retrieval prompt can be recalled accurately, the basic recall rate is effectively improved, the recall effect is improved, and the problems in the prior art are solved.
It should be understood that the description of this section is not intended to identify key or critical features of the embodiments of the application or to delineate the scope of the application. Other features of the present application will become apparent from the description that follows.
Drawings
The drawings are for better understanding of the present solution and do not constitute a limitation of the present application. Wherein:
fig. 1 is a flow diagram of a map retrieval method according to a first embodiment of the present application;
FIG. 2 is a POI-query graph according to a first embodiment of the present application;
fig. 3 is a POI-POI map according to a first embodiment of the present application;
fig. 4 is a schematic diagram of obtaining semantic vectors of a target POI according to a first embodiment of the present application;
FIG. 5 is a schematic diagram of semantic vectors for deriving map retrieval information according to a first embodiment of the present application;
FIG. 6 is a schematic diagram of a map retrieval process according to a first embodiment of the present application;
FIG. 7 is a flow chart of a method of computing POI semantic vectors according to a second embodiment of the present application;
fig. 8 is a schematic structural view of a map retrieval device according to a third embodiment of the present application;
FIG. 9 is a schematic structural diagram of a computing device for POI semantic vectors according to a third embodiment of the present application;
Fig. 10 is a block diagram of an electronic device used to implement an embodiment of the present application.
Detailed Description
Exemplary embodiments of the present application are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present application to facilitate understanding, and should be considered as merely exemplary. Accordingly, one of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present application. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
First embodiment
As shown in fig. 1, the present application provides a map retrieval method, including the following steps:
step 101: semantic vectors of map retrieval information are calculated.
The map search information may be understood as map search information input by a user, and an english expression of the map search information may be "query".
In the step, the semantic vector of the map retrieval information can be obtained through calculation through a pre-trained semantic model, specifically, after the map retrieval information input by a user is received, the map retrieval information is input into the pre-trained semantic model for calculation, and the semantic vector of the map retrieval information is obtained. The semantic vector of map retrieval information may be, for example, but is not limited to, an embedded representation.
In the present application, the semantic vector of the map retrieval information may be calculated using either the existing semantic model at present or the new semantic model provided by the present application, which will be described later.
Step 102: searching a target POI semantic vector matched with the semantic vector of the map retrieval information according to a pre-established POI semantic vector index library; the semantic vector of the target POI is obtained by calculation according to the semantic features of the graph of the target POI, and the semantic features of the graph of the target POI aggregate the semantic features between the target POI and N neighbor nodes.
Wherein N is an integer greater than 1.
POIs are abbreviations for "Point of Information," and Chinese can be translated into "information points" (or "points of interest"). In the field of geographic information systems and map retrieval technologies, a POI may be a house, a shop, a mailbox, a bus stop, etc. The traditional geographic information acquisition method requires that a map surveying staff adopts a precise surveying instrument to acquire the longitude and latitude of one POI, and then marks the POI. Each POI can comprise information of names, categories, coordinates and classifications, the POI can remind the user of detailed information of branches of road conditions and surrounding buildings, and can conveniently search various places required by the user in navigation, and route planning can be carried out by selecting the most convenient and unobstructed roads.
The POI semantic vector index library refers to an index library which is created in advance and is used for storing a plurality of POIs and semantic vectors corresponding to the POIs, and the POIs stored in the POI semantic vector index library can be full-quantity POIs. Any POI semantic vector stored in the POI semantic vector index library can be obtained through calculation through a pre-trained semantic model, specifically, taking a target POI as an example, the semantic features of the graph of the target POI can be obtained first, and the semantic features of the graph of the target POI are input into the pre-trained semantic model for calculation, so that the semantic vector of the target POI is obtained. After the semantic vector of the target POI is obtained, the target POI and the semantic vector corresponding to the target POI can be stored in the POI semantic vector index library. The semantic vector of the target POI may be, for example, but not limited to, an assembled representation.
In the present application, the semantic vector of the POI may be calculated using either the existing semantic model at present or the new semantic model provided by the present application, which will be described later.
It should be noted that, in order to improve the map retrieval efficiency, the POI semantic vector index library may be created in advance in an offline manner, that is, the map semantic features of the POI and the semantic vectors of the POI may be calculated in advance in an offline manner. Of course, it is not excluded to implement the above-described procedure in an online manner. The process of calculating the semantic vector of the map retrieval information in step 101 and searching the target POI semantic vector matched with the semantic vector of the map retrieval information according to the POI semantic vector index library in step 102 can be realized in an online mode.
The following provides a method for obtaining the semantic features of the graph of the target POI, namely, the semantic features of the graph of the target POI can be obtained through the following steps:
acquiring N neighbor nodes of the target POI;
semantic association between the target POI and each neighbor node is obtained respectively to obtain N semantic features;
and aggregating the N semantic features to obtain the graph semantic features of the target POI.
In the method, after N neighbor nodes of a target POI are acquired, composition can be performed by utilizing GCN (Graph Convolutional Network, graph convolution network), a graph model of the target POI and the N neighbor nodes is constructed, semantic features between the target POI and each neighbor node are acquired respectively, and the semantic features are aggregated by utilizing propagation characteristics of the graph model, so that graph semantic features of the target POI are obtained.
The neighbor nodes of the target POI can be determined according to the user history click data in the preset time, and the source of the user history click data is rich and various, so that the neighbor nodes of the target POI are enriched, and semantic vector expression of the POI is richer.
In summary, since the semantic vector of the POI is obtained by calculation according to the semantic features of the graph of the POI, and the semantic features of the graph of the POI aggregate the semantic features between the POI and a plurality of neighbor nodes, the generation of the semantic vector of the POI is optimized, the semantic vector expression of the POI is richer, the accuracy and the confidence of the semantic vector of the POI can be improved, and a more accurate and efficient POI semantic vector index library can be constructed.
In actual application, in the step, the target POI semantic vector matched with the semantic vector of the map retrieval information can be searched according to the POI semantic vector index library, so that recall of the retrieval prompt can be realized, and the POI corresponding to the map retrieval information is recommended to the user.
In the application, at least one POI most relevant to map retrieval information can be recalled from the full quantity POIs by using a K nearest neighbor algorithm, a hnsw search algorithm and the like.
Because the semantic vector expression of the POI is richer, the accuracy and the confidence of the semantic vector of the POI are higher, even if the map retrieval information input by the user has differences, the retrieval prompt can be recalled more accurately, the basic recall rate is effectively improved, the recall effect is improved, and the problems existing in the prior art are solved. From the perspective of the user, the user input step length can be reduced, the user input efficiency is improved, the user requirements can be better met, and the user experience is improved.
Optionally, the N neighboring nodes are N pieces of map retrieval information, and the N pieces of map retrieval information are responded to the selection operation of the target POI by the user; or,
the N neighbor nodes are N POIs sequentially selected by a user in preset time, and the target POI and the N POIs are of the same type.
The embodiment provides a graph model of two target POIs and N neighbor nodes, or provides a composition scheme of two POIs and neighbor nodes.
Firstly, a POI-query graph can be formed by counting map search information (i.e. query) input by a user clicking a target POI in a preset time, as shown in fig. 2. Secondly, a POI-POI diagram can be formed by counting the fact that a user clicks other POIs which are the same as the target POI in sequence within a preset time, and the POI-POI diagram is shown in fig. 3.
Considering that the application scene of the method is based on semantic association between POIs and map retrieval information, the POI-query graph can be preferentially selected and constructed, so that semantic information of historical map retrieval information of target POIs can be aggregated, and richer cross features are provided for association of the POIs and map retrieval information in online map retrieval.
In this embodiment, the N POIs that are responded by the user and sequentially selected by the user in the preset time may be all understood as user history click data, and the neighboring node of the target POI is determined according to the user history click data.
Optionally, the semantic vector of the target POI is obtained by calculation according to the graph semantic feature of the target POI and the basic information of the target POI; the basic information of the target POI comprises at least one of the name of the target POI, the address of the target POI, the tag of the target POI and the geographic position of the target POI.
In this embodiment, the semantic vector of the target POI further considers basic information of the target POI, such as a name (POI name), address (address), tag (tag), and geographic location (e.g., city), in addition to the semantic feature of the map of the target POI.
The basic information of the target POI plays an important role in representing the semantic vector of the target POI. For example, the labels of POIs are divided into hundreds of categories such as local communities, hospitals, banks, buildings and the like, when a user searches for a specific POI, a tag tendency is implied, if the relation between query and POI name and tag can be established, the POI of a type of tag can be defined, the search range is reduced, and other tag impurities are reduced. Assuming search "living delegation", the tag of the POI is a living delegation class, not a traffic class. For example, the geographical location information of the POI is input into the same query at different geographical locations (such as different cities), the user needs are greatly different, and if the user searches for "chinese university of geology" in beijing, the user's main needs are likely to be "chinese university of geology" (beijing), and if the user searches for "chinese university of geology" in wuhan, the user's main needs are likely to be "chinese university of geology" (wuhan).
Therefore, the implementation mode can realize multisource fusion by fusing vector representations of multiple domains of the POI, and semantic vector expression of the POI can be more accurate and richer.
Optionally, the semantic vector of the target POI is calculated by the following steps:
calculating a first vector representation of the target POI according to the graph semantic features of the target POI;
calculating a second vector representation of the target POI according to the basic information of the target POI;
and calculating the semantic vector of the target POI according to the first vector representation and the second vector representation.
In this embodiment, when various basic information of the target POI is comprehensively considered, corresponding vector representations, such as tag vector representation, city vector representation, POI name vector representation, address vector representation, and the like, may be calculated respectively according to different basic information.
In this embodiment, the first vector representation of the target POI may be calculated by GCN and the POIname vector representation may be calculated by the semantic pre-training model ERNIE-TINY.
As shown in fig. 4, fig. 4 shows a schematic diagram of fusing vector representations of three basic information (POI name, tag and city) and the semantic features of the map together in four domains to obtain a semantic vector of the target POI.
The method and the device can map the tag of the POI clicked by the query into a tag vector representation by using the user click log (namely user historical click data), and splice the tag vector representation to the vector representation of the POI name by using the click relation between the query and the POI.
Optionally, the calculating the semantic vector of the map retrieval information includes:
calculating a first vector representation of the map retrieval information according to the map retrieval information;
after receiving the map retrieval information, calculating a second vector representation of the map retrieval information according to the current geographic location information;
and calculating a semantic vector of the map retrieval information according to the first vector representation and the second vector representation.
In this embodiment, the first vector representation of map retrieval information may be calculated by CNN (Convolutional Neural Networks, convolutional neural network).
In this embodiment, the semantic vector of the map retrieval information further considers the current geographic location (e.g., city) information in addition to the map retrieval information itself. Because the current geographic position information can well reflect the main requirement of the user, the current geographic position plays an important role in representing the semantic vector of the map retrieval information. The current geographical location information may be understood as scene information (side information) of map retrieval information, and the map retrieval information itself and the current geographical location information may be regarded as a plurality of domains of the map retrieval information.
As shown in fig. 5, fig. 5 shows a schematic diagram of a semantic vector that fuses vector representations of two domains, namely query and city, to obtain map retrieval information.
According to the embodiment, the vector representations of the domains of the map retrieval information are fused, so that multi-source fusion can be realized, and the semantic vector expression of the POI is more accurate and richer.
Optionally, the semantic vector of the target POI and the semantic vector of the map retrieval information are obtained through calculation of a pre-trained semantic model;
the training function of the semantic model is as follows:
wherein Q represents the feature matrix of the map retrieval information sample, P represents the feature matrix of the POI sample,represents the Cronecker product, R pos Representing positive sample feature matrix, R neg Representing a negative sample feature matrix, m representing the spacing between the positive and negative sample feature matrices, and l representing the loss function.
The penalty function used by the semantic model may be referred to as matrix-wise loss, which is different from the existing penalty function, in that the training data used by the penalty function is to construct positive and negative samples in each batch of training data, in which the positive samples are the kronecker product of the vector of the query and the vector of the POI clicked by the query, and the negative samples are the inner products of the vector of the query and the vector of the POI not clicked by the query. The loss function does not need label, and m is used for increasing the distance between positive and negative samples during calculation, so that the positive and negative samples are differentiated. The loss function used by the semantic model has better robustness, so that the robustness of the semantic model can be improved.
For example, in the training process, it is assumed that a batch of training data includes two pieces of data, which are respectively:
query: hundred degrees (denoted as Q1), the POI clicked by the user is: hundred degree mansion (denoted as P1);
query: west two flag (noted Q2), the POI clicked by the user is: a western two-flag subway station (denoted as P2).
Then in the batch of training data, the inner product of the Q1 and P1 vectors corresponds to the positive sample feature matrix and the inner product of the Q1 and P2 vectors corresponds to the negative sample feature matrix. The loss function is applied to enable the similarity of the positive sample query and the POI to be high, the similarity of the negative sample query and the POI to be low, and the difference between the positive sample query and the POI is not greater than m, so that the positive sample and the negative sample are distinguished.
After the semantic model is obtained by using the training function, the calculation of the POI semantic vector and the query semantic vector can be realized through the semantic model.
As shown in fig. 6, fig. 6 illustrates a map retrieval process that uses semantic models to implement the computation of POI semantic vectors and query semantic vectors, and semantic recall.
It should be noted that, the various optional embodiments of the map searching method in the present application may be implemented in combination with each other, or may be implemented separately, which is not limited to this application.
The above embodiments of the present application have at least the following advantages or benefits:
in the method, the semantic vector of the POI is obtained through calculation according to the semantic features of the graph of the POI, and the semantic features of the graph of the POI aggregate the semantic features between the POI and N neighbor nodes, so that the method optimizes the generation of the semantic vector of the POI, the semantic vector expression of the POI is richer, the accuracy and the confidence of the semantic vector of the POI can be improved, and a more accurate and efficient POI semantic vector index library can be constructed. Therefore, even if the map retrieval information input by the user is different, the retrieval prompt can be recalled accurately, the basic recall rate is effectively improved, the recall effect is improved, and the problems in the prior art are solved. The application can be applied to the cloud field, such as cloud service or cloud platform, and can also be applied to the deep learning field.
Second embodiment
As shown in fig. 7, the present application provides a method for calculating a POI semantic vector, which includes the following steps:
step 201: obtaining the graph semantic features of a target POI, wherein the graph semantic features of the target POI aggregate the semantic features between the target POI and N neighbor nodes, and N is an integer greater than 1;
Step 202: and calculating the semantic vector of the target POI according to the graph semantic features of the target POI.
Optionally, the obtaining the semantic features of the map of the target POI includes:
acquiring N neighbor nodes of the target POI;
semantic association between the target POI and each neighbor node is obtained respectively to obtain N semantic features;
and aggregating the N semantic features to obtain the graph semantic features of the target POI.
Optionally, the N neighboring nodes are N pieces of map retrieval information, and the N pieces of map retrieval information are responded to the selection operation of the target POI by the user; or,
the N neighbor nodes are N POIs sequentially selected by a user in preset time, and the target POI and the N POIs are of the same type.
Optionally, the method further comprises:
acquiring basic information of the target POI, wherein the basic information of the target POI comprises at least one of a name of the target POI, an address of the target POI, a tag to which the target POI belongs and a geographic position of the target POI;
the calculating the semantic vector of the target POI according to the graph semantic features of the target POI comprises the following steps:
and calculating the semantic vector of the target POI according to the graph semantic features of the target POI and the basic information of the target POI.
Optionally, the calculating the semantic vector of the target POI according to the semantic feature of the map of the target POI and the basic information of the target POI includes:
calculating a first vector representation of the target POI according to the graph semantic features of the target POI;
calculating a second vector representation of the target POI according to the basic information of the target POI;
and calculating the semantic vector of the target POI according to the first vector representation and the second vector representation.
Optionally, the semantic vector of the target POI is obtained through calculation of a pre-trained semantic model;
the training function of the semantic model is as follows:
wherein Q represents the feature matrix of the map retrieval information sample, P represents the feature matrix of the POI sample,represents the Cronecker product, R pos Representing positive sample feature matrix, R neg Representing a negative sample feature matrix, m representing the spacing between the positive and negative sample feature matrices, and l representing the loss function.
In this application, the related implementation manners of the second embodiment may refer to any implementation manner related to the calculation of the POI semantic vector in the first embodiment, and any implementation manner related to the calculation of the POI semantic vector in the first embodiment is applicable to the second embodiment, and can achieve the same beneficial effects, so that repetition is avoided and no redundant description is given here.
Third embodiment
As shown in fig. 8, the present application provides a map retrieval apparatus 300, including:
a first calculation module 301, configured to calculate a semantic vector of map retrieval information;
the searching module 302 is configured to search a semantic vector of a target POI that matches the semantic vector of the map retrieval information according to a pre-created semantic vector index library of the information point POI; the semantic vector of the target POI is obtained by calculation according to the semantic features of the target POI, the semantic features of the target POI aggregate the semantic features between the target POI and N neighbor nodes, and N is an integer greater than 1.
Optionally, the map retrieval device 300 further includes:
the first acquisition module is used for acquiring N neighbor nodes of the target POI;
the second acquisition module is used for respectively acquiring semantic association between the target POI and each neighbor node so as to obtain N semantic features;
and the aggregation module is used for aggregating the N semantic features to obtain the graph semantic features of the target POI.
Optionally, the N neighboring nodes are N pieces of map retrieval information, and the N pieces of map retrieval information are responded to the selection operation of the target POI by the user; or,
The N neighbor nodes are N POIs sequentially selected by a user in preset time, and the target POI and the N POIs are of the same type.
Optionally, the semantic vector of the target POI is obtained by calculation according to the graph semantic feature of the target POI and the basic information of the target POI; the basic information of the target POI comprises at least one of the name of the target POI, the address of the target POI, the tag of the target POI and the geographic position of the target POI.
Optionally, the map retrieval device 300 further includes a second calculation module for:
calculating a first vector representation of the target POI according to the graph semantic features of the target POI;
calculating a second vector representation of the target POI according to the basic information of the target POI;
and calculating the semantic vector of the target POI according to the first vector representation and the second vector representation.
Optionally, the first computing module 301 is specifically configured to:
calculating a first vector representation of the map retrieval information according to the map retrieval information;
after receiving the map retrieval information, calculating a second vector representation of the map retrieval information according to the current geographic location information;
And calculating a semantic vector of the map retrieval information according to the first vector representation and the second vector representation.
Optionally, the semantic vector of the target POI is obtained through calculation of a pre-trained semantic model;
the training function of the semantic model is as follows:
wherein Q represents the feature matrix of the map retrieval information sample, P represents the feature matrix of the POI sample,represents the Cronecker product, R pos Representing positive sample feature matrix, R neg Representing a negative sample feature matrix, m representing the spacing between the positive and negative sample feature matrices, and l representing the loss function.
The map retrieval device 300 provided in the present application can implement each process in the above map retrieval method embodiment, and can achieve the same beneficial effects, so that repetition is avoided, and no further description is provided here.
Fourth embodiment
As shown in fig. 9, the present application provides a computing device 400 for POI semantic vector, including:
a first obtaining module 401, configured to obtain a graph semantic feature of a target POI, where the graph semantic feature of the target POI aggregates semantic features between the target POI and N neighboring nodes, and N is an integer greater than 1;
the calculating module 402 is configured to calculate a semantic vector of a target POI according to a graph semantic feature of the target POI.
Optionally, the first obtaining module 401 includes:
the first acquisition submodule is used for acquiring N neighbor nodes of the target POI;
the second acquisition sub-module is used for respectively acquiring semantic association between the target POI and each neighbor node so as to obtain N semantic features;
and the aggregation sub-module is used for aggregating the N semantic features to obtain the graph semantic features of the target POI.
Optionally, the N neighboring nodes are N pieces of map retrieval information, and the N pieces of map retrieval information are responded to the selection operation of the target POI by the user; or,
the N neighbor nodes are N POIs sequentially selected by a user in preset time, and the target POI and the N POIs are of the same type.
Optionally, the computing device 400 for POI semantic vector further includes:
the second acquisition module is used for acquiring basic information of the target POI, wherein the basic information of the target POI comprises at least one of a name of the target POI, a tag to which the target POI belongs and a geographic position of the target POI;
the calculation module 402 is specifically configured to:
and calculating the semantic vector of the target POI according to the graph semantic features of the target POI and the basic information of the target POI.
Optionally, the computing module 402 is specifically configured to:
calculating a first vector representation of the target POI according to the graph semantic features of the target POI;
calculating a second vector representation of the target POI according to the basic information of the target POI;
and calculating the semantic vector of the target POI according to the first vector representation and the second vector representation.
Optionally, the semantic vector of the target POI is obtained through calculation of a pre-trained semantic model;
the training function of the semantic model is as follows:
wherein Q represents the feature matrix of the map retrieval information sample, P represents the feature matrix of the POI sample,represents the Cronecker product, R pos Representing positive sample feature matrix, R neg Representing a negative sample feature matrix, m representing the spacing between the positive and negative sample feature matrices, and l representing the loss function.
The calculation device 400 for POI semantic vector provided in the present application can implement each process in the above map retrieval method embodiment, and can achieve the same beneficial effects, so that repetition is avoided, and no redundant description is provided herein.
According to embodiments of the present application, an electronic device and a readable storage medium are also provided.
As shown in fig. 10, a block diagram of an electronic device according to an embodiment of the method of the present application is provided. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the application described and/or claimed herein.
As shown in fig. 10, the electronic device includes: one or more processors 701, memory 702, and interfaces for connecting the various components, including high-speed interfaces and low-speed interfaces. The various components are interconnected using different buses and may be mounted on a common motherboard or in other manners as desired. The processor may process instructions executing within the electronic device, including instructions stored in or on memory to display graphical information of the GUI on an external input/output device, such as a display device coupled to the interface. In other embodiments, multiple processors and/or multiple buses may be used, if desired, along with multiple memories and multiple memories. Also, multiple electronic devices may be connected, each providing a portion of the necessary operations (e.g., as a server array, a set of blade servers, or a multiprocessor system). One processor 701 is illustrated in fig. 10.
Memory 702 is a non-transitory computer-readable storage medium provided herein. The memory stores instructions executable by the at least one processor to cause the at least one processor to perform the map retrieval method or the POI semantic vector calculation method provided by the present application. The non-transitory computer-readable storage medium of the present application stores computer instructions for causing a computer to execute the map retrieval method or the calculation method of POI semantic vectors provided by the present application.
The memory 702 is used as a non-transitory computer readable storage medium for storing non-transitory software programs, non-transitory computer executable programs, and modules, such as program instructions/modules corresponding to the map retrieval method in the embodiments of the present application (e.g., the first module 301 and the search module 302 shown in fig. 8), and program instructions/modules corresponding to the POI semantic vector calculation method in the embodiments of the present application (e.g., the first acquisition module 401 and the calculation module 402 shown in fig. 9). The processor 701 executes various functional applications of the problem solving apparatus and data processing, that is, implements the map retrieval method or the calculation method of the POI semantic vector in the above-described method embodiment by executing the non-transitory software programs, instructions, and modules stored in the memory 702.
Memory 702 may include a storage program area that may store an operating system, at least one application program required for functionality, and a storage data area; the storage data area may store data created by use of the electronic device, and the like. In addition, the memory 702 may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 702 may optionally include memory located remotely from processor 701, which may be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The electronic device may further include: an input device 703 and an output device 704. The processor 701, the memory 702, the input device 703 and the output device 704 may be connected by a bus or otherwise, in fig. 10 by way of example.
The input device 703 may receive input numeric or character information and generate key signal inputs related to user settings and function control of the electronic device, such as a touch screen, keypad, mouse, trackpad, touchpad, pointer stick, one or more mouse buttons, trackball, joystick, and like input devices. The output device 704 may include a display apparatus, auxiliary lighting devices (e.g., LEDs), and haptic feedback devices (e.g., vibration motors), among others. The display device may include, but is not limited to, a Liquid Crystal Display (LCD), a Light Emitting Diode (LED) display, and a plasma display. In some implementations, the display device may be a touch screen.
Various implementations of the systems and techniques described here can be realized in digital electronic circuitry, integrated circuitry, application specific ASIC (application specific integrated circuit), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.
These computing programs (also referred to as programs, software applications, or code) include machine instructions for a programmable processor, and may be implemented in a high-level procedural and/or object-oriented programming language, and/or in assembly/machine language. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and/or device (e.g., magnetic discs, optical disks, memory, programmable Logic Devices (PLDs)) used to provide machine instructions and/or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal used to provide machine instructions and/or data to a programmable processor.
To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and pointing device (e.g., a mouse or trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic input, speech input, or tactile input.
The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), the internet, and blockchain networks.
The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.
According to the technical scheme of the embodiment of the application, the semantic vector of the POI is obtained through calculation according to the semantic features of the graph of the POI, and the semantic features of the graph of the POI aggregate the semantic features between the POI and N neighbor nodes, so that the generation of the semantic vector of the POI is optimized, the semantic vector expression of the POI is richer, the accuracy and the confidence of the semantic vector of the POI can be improved, and a more accurate and efficient POI semantic vector index library can be constructed. Therefore, even if the map retrieval information input by the user is different, the retrieval prompt can be recalled accurately, the basic recall rate is effectively improved, the recall effect is improved, and the problems in the prior art are solved.
It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps described in the present application may be performed in parallel, sequentially, or in a different order, provided that the desired results of the technical solutions disclosed in the present application can be achieved, and are not limited herein.
The above embodiments do not limit the scope of the application. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application are intended to be included within the scope of the present application.

Claims (15)

1. A map retrieval method, the method comprising:
calculating semantic vectors of map retrieval information;
searching a target POI semantic vector matched with the semantic vector of the map retrieval information according to a pre-established information point POI semantic vector index library; the semantic vector of the target POI is obtained by calculation according to the semantic features of the graph of the target POI, the semantic features of the graph of the target POI aggregate the semantic features between the target POI and N neighbor nodes, and N is an integer greater than 1;
The N neighbor nodes are N pieces of map retrieval information, and the N pieces of map retrieval information are responded by a user to the selection operation of the target POI; or,
the N neighbor nodes are N POIs sequentially selected by a user in preset time, and the target POI and the N POIs are of the same type.
2. The method according to claim 1, wherein the graph semantic features of the target POI are obtained by:
acquiring N neighbor nodes of the target POI;
semantic association between the target POI and each neighbor node is obtained respectively to obtain N semantic features;
and aggregating the N semantic features to obtain the graph semantic features of the target POI.
3. The method according to claim 1 or 2, wherein the semantic vector of the target POI is calculated according to the graph semantic feature of the target POI and the basic information of the target POI; the basic information of the target POI comprises at least one of the name of the target POI, the address of the target POI, the tag of the target POI and the geographic position of the target POI.
4. A method according to claim 3, wherein the semantic vector of the target POI is calculated by:
Calculating a first vector representation of the target POI according to the graph semantic features of the target POI;
calculating a second vector representation of the target POI according to the basic information of the target POI;
and calculating the semantic vector of the target POI according to the first vector representation and the second vector representation.
5. The method of claim 1 or 2, wherein the calculating semantic vectors of map retrieval information comprises:
calculating a first vector representation of the map retrieval information according to the map retrieval information;
after receiving the map retrieval information, calculating a second vector representation of the map retrieval information according to the current geographic location information;
and calculating a semantic vector of the map retrieval information according to the first vector representation and the second vector representation.
6. The method according to claim 1 or 2, wherein the semantic vector of the target POI and the semantic vector of the map retrieval information are calculated by a pre-trained semantic model;
the training function of the semantic model is as follows:
R neg ＝Q·P T
l＝max(0,m-R pos +R neg )
wherein Q represents the feature matrix of the map retrieval information sample, P represents the feature matrix of the POI sample, Represents the Cronecker product, R pos Representing positive sample feature matrix, R neg Representing a negative sample feature matrix, m representing the spacing between the positive and negative sample feature matrices, and l representing the loss function.
7. A method for calculating semantic vectors of information points POI, the method comprising:
obtaining the graph semantic features of a target POI, wherein the graph semantic features of the target POI aggregate the semantic features between the target POI and N neighbor nodes, and N is an integer greater than 1;
calculating semantic vectors of the target POIs according to the graph semantic features of the target POIs;
the N neighbor nodes are N pieces of map retrieval information, and the N pieces of map retrieval information are responded by a user to the selection operation of the target POI; or,
the N neighbor nodes are N POIs sequentially selected by a user in preset time, and the target POI and the N POIs are of the same type.
8. The method of claim 7, wherein the obtaining graph semantic features of the target POI comprises:
acquiring N neighbor nodes of the target POI;
semantic association between the target POI and each neighbor node is obtained respectively to obtain N semantic features;
And aggregating the N semantic features to obtain the graph semantic features of the target POI.
9. The method according to claim 7 or 8, characterized in that the method further comprises:
acquiring basic information of the target POI, wherein the basic information of the target POI comprises at least one of a name of the target POI, an address of the target POI, a tag to which the target POI belongs and a geographic position of the target POI;
the calculating the semantic vector of the target POI according to the graph semantic features of the target POI comprises the following steps:
and calculating the semantic vector of the target POI according to the graph semantic features of the target POI and the basic information of the target POI.
10. The method according to claim 9, wherein calculating the semantic vector of the target POI from the graph semantic features of the target POI and the base information of the target POI comprises:
calculating a first vector representation of the target POI according to the graph semantic features of the target POI;
calculating a second vector representation of the target POI according to the basic information of the target POI;
and calculating the semantic vector of the target POI according to the first vector representation and the second vector representation.
11. The method according to claim 7 or 8, wherein the semantic vector of the target POI is calculated by means of a pre-trained semantic model;
the training function of the semantic model is as follows:
R neg ＝Q·P T
l＝max(0,m-R pos +R neg )
wherein Q represents the feature matrix of the map retrieval information sample, P represents the feature matrix of the POI sample,represents the Cronecker product, R pos Representing positive sample feature matrix, R neg Representing a negative sample feature matrix, m representing the spacing between the positive and negative sample feature matrices, and l representing the loss function.
12. A map retrieval apparatus, comprising:
the first calculation module is used for calculating semantic vectors of map retrieval information;
the searching module is used for searching the semantic vector of the target POI matched with the semantic vector of the map retrieval information according to the pre-established information point POI semantic vector index library; the semantic vector of the target POI is obtained by calculation according to the semantic features of the graph of the target POI, the semantic features of the graph of the target POI aggregate the semantic features between the target POI and N neighbor nodes, and N is an integer greater than 1;
the N neighbor nodes are N pieces of map retrieval information, and the N pieces of map retrieval information are responded by a user to the selection operation of the target POI; or,
The N neighbor nodes are N POIs sequentially selected by a user in preset time, and the target POI and the N POIs are of the same type.
13. A computing device for semantic vectors of information points, comprising:
the first acquisition module is used for acquiring the graph semantic features of the target POI, wherein the graph semantic features of the target POI aggregate the semantic features between the target POI and N neighbor nodes, and N is an integer greater than 1;
the computing module is used for computing semantic vectors of the target POIs according to the graph semantic features of the target POIs;
the N neighbor nodes are N pieces of map retrieval information, and the N pieces of map retrieval information are responded by a user to the selection operation of the target POI; or,
the N neighbor nodes are N POIs sequentially selected by a user in preset time, and the target POI and the N POIs are of the same type.
14. An electronic device, comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 6 or to enable the at least one processor to perform the method of any one of claims 7 to 11.
15. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1 to 6 or for causing the computer to perform the method of any one of claims 7 to 11.
CN202010597660.8A 2020-06-28 2020-06-28 Map retrieval method, information point POI semantic vector calculation method and device Active CN111782748B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010597660.8A CN111782748B (en) 2020-06-28 2020-06-28 Map retrieval method, information point POI semantic vector calculation method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010597660.8A CN111782748B (en) 2020-06-28 2020-06-28 Map retrieval method, information point POI semantic vector calculation method and device

Publications (2)

Publication Number Publication Date
CN111782748A CN111782748A (en) 2020-10-16
CN111782748B true CN111782748B (en) 2024-01-12

Family

ID=72760658

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010597660.8A Active CN111782748B (en) 2020-06-28 2020-06-28 Map retrieval method, information point POI semantic vector calculation method and device

Country Status (1)

Country Link
CN (1) CN111782748B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112328890B (en) * 2020-11-23 2024-04-12 北京百度网讯科技有限公司 Method, device, equipment and storage medium for searching geographic position point

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108763293A (en) * 2018-04-17 2018-11-06 平安科技（深圳）有限公司 Point of interest querying method, device and computer equipment based on semantic understanding
CN108846126A (en) * 2018-06-29 2018-11-20 北京百度网讯科技有限公司 Generation, question and answer mode polymerization, device and the equipment of related question polymerization model
CN110083688A (en) * 2019-05-10 2019-08-02 北京百度网讯科技有限公司 Search result recalls method, apparatus, server and storage medium
CN110309268A (en) * 2019-07-12 2019-10-08 中电科大数据研究院有限公司 A kind of cross-language information retrieval method based on concept map

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140108383A1 (en) * 2012-10-16 2014-04-17 Xtreme Labs Inc. Method and System for Filtering Search Results for Maps Using Social Graph
US9904932B2 (en) * 2014-12-29 2018-02-27 Google Llc Analyzing semantic places and related data from a plurality of location data reports

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108763293A (en) * 2018-04-17 2018-11-06 平安科技（深圳）有限公司 Point of interest querying method, device and computer equipment based on semantic understanding
CN108846126A (en) * 2018-06-29 2018-11-20 北京百度网讯科技有限公司 Generation, question and answer mode polymerization, device and the equipment of related question polymerization model
CN110083688A (en) * 2019-05-10 2019-08-02 北京百度网讯科技有限公司 Search result recalls method, apparatus, server and storage medium
CN110309268A (en) * 2019-07-12 2019-10-08 中电科大数据研究院有限公司 A kind of cross-language information retrieval method based on concept map

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
一种语义增强的空间关键词搜索方法;韩军;范举;周立柱;;计算机研究与发展(09);全文 *
个性化跨语言信息检索中结果重排序研究;周栋;赵文玉;伍璇;刘建勋;;计算机工程与科学(10);全文 *

Also Published As

Publication number Publication date
CN111782748A (en) 2020-10-16

Similar Documents

Publication Publication Date Title
US11630560B2 (en) Map information display method and apparatus, electronic device, and computer storage medium
US20210239486A1 (en) Method and apparatus for predicting destination, electronic device and storage medium
WO2021232724A1 (en) Method for extracting geographic location point spatial relationship, method for training extraction model, and devices
CN111666292B (en) Similarity model establishment method and device for retrieving geographic position
CN111814077B (en) Information point query method, device, equipment and medium
KR102509814B1 (en) Method and apparatus for determining traffic checkpoint, electronic device, and medium
JP2022530690A (en) Query auto-completion methods, appliances, equipment, and computer storage media
CN112328890B (en) Method, device, equipment and storage medium for searching geographic position point
JP2022530688A (en) Sort model construction method, query auto-completion method and corresponding device
CN113160693B (en) Road intersection processing method, device, equipment and storage medium
CN111666461B (en) Method, apparatus, device and computer storage medium for retrieving geographic location
CN111782748B (en) Map retrieval method, information point POI semantic vector calculation method and device
CN111047107B (en) Road traffic time prediction method, device, electronic equipment and storage medium
CN110781657A (en) Management method, device and equipment for navigation broadcasting
CN112269925B (en) Method and device for obtaining geographic position point information
CN112861023A (en) Map information processing method, map information processing apparatus, map information processing device, storage medium, and program product
CN112818024A (en) Map data processing method, device, equipment and storage medium
CN111782750B (en) Method and device for determining map retrieval information trend region and electronic equipment
CN112381166B (en) Information point identification method and device and electronic equipment
CN112528156B (en) Method for establishing sorting model, method for inquiring automatic completion and corresponding device
CN112530161B (en) Road data processing method, device and equipment and computer storage medium
CN112419728B (en) Method, device, equipment and storage medium for determining road condition information
CN113221025B (en) Point-of-interest recall method, device, equipment and medium
KR20220122565A (en) Multipath generation method, apparatus, device and storage medium
CN114036414A (en) Method and device for processing interest points, electronic equipment, medium and program product

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant