WO2023124005A1 - 地图兴趣点查询方法、装置、设备、存储介质及程序产品 - Google Patents

地图兴趣点查询方法、装置、设备、存储介质及程序产品 Download PDF

Info

Publication number
WO2023124005A1
WO2023124005A1 PCT/CN2022/104877 CN2022104877W WO2023124005A1 WO 2023124005 A1 WO2023124005 A1 WO 2023124005A1 CN 2022104877 W CN2022104877 W CN 2022104877W WO 2023124005 A1 WO2023124005 A1 WO 2023124005A1
Authority
WO
WIPO (PCT)
Prior art keywords
interest
query
map
point
multivariate
Prior art date
Application number
PCT/CN2022/104877
Other languages
English (en)
French (fr)
Inventor
黄际洲
孙雅铭
卓安
王海峰
Original Assignee
北京百度网讯科技有限公司
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 北京百度网讯科技有限公司 filed Critical 北京百度网讯科技有限公司
Publication of WO2023124005A1 publication Critical patent/WO2023124005A1/zh

Links

Images

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
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • 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

Definitions

  • the present disclosure relates to the field of data processing technology, specifically to the field of artificial intelligence technology such as deep learning, natural language processing, knowledge graph, and smart map, and in particular to a method, device, electronic device, computer-readable storage medium, and computer for querying points of interest on a map. program product.
  • artificial intelligence technology such as deep learning, natural language processing, knowledge graph, and smart map
  • Navigation electronic map products not only provide travel services for hundreds of millions of users every day, but also serve as a digital base for new infrastructure, playing an increasingly important and indispensable role in various industries.
  • data is the infrastructure for survival
  • accuracy is the lifeline of maps
  • timeliness is the necessary ability for maps to depict the real world.
  • Embodiments of the present disclosure provide a map interest point query method, device, electronic equipment, computer-readable storage medium, and computer program product.
  • the embodiment of the present disclosure proposes a map POI query method, including: receiving an input POI query request; extracting target query words contained in the POI query request; using a predetermined query word and interest The semantic correspondence between points is determined to determine the target interest point corresponding to the target query word; wherein, the semantic correspondence is determined by the query word and interest point extracted from the preset multivariate map knowledge model, and the multivariate map knowledge model records the interest point and interest point Multivariate correspondence between multivariate knowledge, multivariate knowledge includes: geographic location information, knowledge map information and input query words.
  • an embodiment of the present disclosure provides a map POI query device, including: a POI query request receiving unit configured to receive an input POI query request; a target query word extraction unit configured to obtain a POI query The target query words contained in the query request are extracted; the target interest point determination unit is configured to determine the target interest points corresponding to the target query words by using the predetermined semantic correspondence between the query words and the interest points; wherein, the semantic The corresponding relationship is determined by query words and points of interest extracted from the preset multivariate map knowledge model.
  • the multivariate map knowledge model records the multivariate correspondence between points of interest and multivariate knowledge.
  • the multivariate knowledge includes: geographic location information, knowledge map information and input query words.
  • an embodiment of the present disclosure provides an electronic device, the electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions that can be executed by the at least one processor , the instruction is executed by at least one processor, so that the at least one processor can implement the map interest point query method described in any implementation manner in the first aspect when executed.
  • an embodiment of the present disclosure provides a non-transitory computer-readable storage medium storing computer instructions, the computer instructions are used to enable a computer to implement the point of interest on the map as described in any implementation manner in the first aspect. Query method.
  • an embodiment of the present disclosure provides a computer program product including a computer program.
  • the computer program When the computer program is executed by a processor, the method for querying a point of interest on a map as described in any implementation manner in the first aspect can be implemented.
  • this disclosure pre-constructs a multivariate map knowledge model based on interest points and corresponding multivariate knowledge. Since multivariate knowledge covers geographical location information, knowledge map information and input query words, therefore The trained multivariate map knowledge model can more comprehensively and accurately determine the query words that match the points of interest.
  • the material for semantic correlation modeling is the points of interest and matching query words extracted from the trained multivariate map knowledge model, and then the semantic correspondence that can accurately reflect the user's search habits at the semantic level is obtained. Finally, with the help of this The semantic correspondence can accurately determine the target point of interest, which improves the accuracy of the query result of the point of interest query service.
  • FIG. 1 is an exemplary system architecture to which the present disclosure may be applied;
  • FIG. 2 is a flow chart of a method for querying a map point of interest provided by an embodiment of the present disclosure
  • FIG. 3 is a flow chart of a method for constructing a multivariate map knowledge model provided by an embodiment of the present disclosure
  • Fig. 4 is a schematic diagram of node association constructed based on the scheme provided in Fig. 3;
  • FIG. 5 is a flowchart of a method for determining a node category provided by an embodiment of the present disclosure
  • FIG. 6 is a flow chart of a node labeling method provided by an embodiment of the present disclosure.
  • FIG. 7 is a structural block diagram of a map interest point query device provided by an embodiment of the present disclosure.
  • FIG. 8 is a schematic structural diagram of an electronic device suitable for performing a method for querying a point of interest in a map provided by an embodiment of the present disclosure.
  • Fig. 1 shows an exemplary system architecture 100 to which embodiments of the map POI query method, device, electronic device, and computer-readable storage medium of the present disclosure can be applied.
  • a system architecture 100 may include terminal devices 101 , 102 , 103 , a network 104 and a server 105 .
  • the network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105.
  • Network 104 may include various connection types, such as wires, wireless communication links, or fiber optic cables, among others.
  • Terminal devices 101 , 102 , 103 Users can use terminal devices 101 , 102 , 103 to interact with server 105 via network 104 to receive or send messages and the like.
  • Various applications for information communication between the terminal devices 101, 102, 103 and the server 105 such as map navigation applications, model training applications, and instant messaging applications, can be installed.
  • the terminal devices 101, 102, 103 and the server 105 may be hardware or software.
  • the terminal devices 101, 102, 103 are hardware, they can be various electronic devices with display screens, including but not limited to smart phones, tablet computers, laptop computers and desktop computers, etc.; when the terminal devices 101, 102 When , 103 is software, it can be installed in the electronic devices listed above, and it can be implemented as multiple software or software modules, or can be implemented as a single software or software module, which is not specifically limited here.
  • the server 105 is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server; when the server is software, it can be implemented as multiple software or software modules, or as a single software or software The module is not specifically limited here.
  • the server 105 can provide various services through various built-in applications. Taking a map navigation application that can provide map point-of-interest query services as an example, the server 105 can achieve the following effects when running the map navigation application: First, the receiving terminal device 101, 102, 103 through the incoming POI query request through the network 104; then, extract the included target query words from the POI query request; next, use the semantic correspondence between the predetermined query words and POIs , determine the target interest point corresponding to the target query word; finally, return the determined target interest point to the terminal devices 101 , 102 , 103 through the network 104 .
  • the semantic correspondence is determined by the query words and points of interest extracted from the preset multivariate map knowledge model.
  • the multivariate map knowledge model records the multivariate correspondence between the interest points and the multivariate knowledge.
  • the multivariate knowledge includes: geographic location information, knowledge map information and entered query terms.
  • the multivariate map knowledge model may be pre-trained by the server 105 through a built-in model training application.
  • the POI query request can also be pre-stored locally in the server 105 in various ways. Therefore, when the server 105 detects that these data are already stored locally (such as starting to process pending query tasks), it can choose to directly obtain these data locally. In this case, the exemplary system architecture 100 may not It includes terminal devices 101 , 102 , 103 and network 104 .
  • the map POI query methods provided in subsequent embodiments of the present disclosure are generally executed by the server 105 with relatively strong computing power and more computing resources.
  • the map POI query device is generally also set in the server 105 .
  • the terminal devices 101, 102, and 103 can also complete the above-mentioned tasks through the map navigation applications installed on them.
  • Various calculations performed by the server 105 further output the same results as the server 105 .
  • exemplary system architecture 100 may also exclude server 105 and network 104 .
  • terminal devices, networks and servers in Fig. 1 are only illustrative. According to the implementation needs, there can be any number of terminal devices, networks and servers.
  • FIG. 2 is a flow chart of a map interest point query method provided by an embodiment of the present disclosure, wherein the process 200 includes the following steps:
  • Step 201 receiving an input POI query request
  • This step is intended to receive the POI query from the user terminal (such as the terminal equipment 101, 102, 103 shown in FIG. 1 ) through the network 104 by the executive body (such as the server 105 shown in FIG. 1 ) of the map point of interest query method. ask.
  • the user terminal such as the terminal equipment 101, 102, 103 shown in FIG. 1
  • the executive body such as the server 105 shown in FIG. 1
  • the interest point query request is used to characterize a user's query demand for a certain point of interest, which at least includes a target query word used as a query basis, and the target query word can be directly expressed as plain text information or unencrypted Voice signals can also be expressed as ciphertext strings or encrypted voice signals to meet security requirements.
  • the target query word in addition to the basic information that characterizes the target query word, it can also include the terminal information of the user terminal, the current location when the user initiates the request, and possible query restrictions on points of interest, so as to combine the information that may affect the query results.
  • Various information is used to improve the accuracy of the obtained query results to better meet the needs of users.
  • Step 202 Extract the target query words contained in the POI query request
  • this step aims to extract the target query words contained in the POI query request by the above-mentioned execution subject.
  • the target query term may be expressed in different forms, it is also necessary to use a correct and matching extraction method during extraction. For example, when the target query term is included in the POI query request in a ciphertext or encrypted manner, It is also necessary to pre-determine the correct decryption key or decryption rule, so as to obtain the correct target query word; when the target query word is expressed as a voice signal, it is also necessary to convert the target query word into a text form that is convenient for subsequent processing through speech recognition technology ; When the POI query request does not directly contain the target query word, but contains relevant information that can guide the acquisition of the target query word from other places, it is also necessary to be able to correctly obtain the target query word from other places according to the relevant information .
  • Step 203 Using the predetermined semantic correspondence between the query word and the point of interest, determine the target point of interest corresponding to the target query word.
  • this step aims to use the determined semantic correspondence relationship to determine the target interest point that has a correspondence relationship with the target query word and that the correspondence relationship is consistent with the semantic correspondence relationship.
  • the semantic correspondence is determined by the query words and interest points extracted from the preset multivariate map knowledge model, and the multivariate map knowledge model records the multivariate correspondence between interest points and multivariate knowledge.
  • the multivariate knowledge includes: geographic location information, Knowledge graph information and input query words. Since the multivariate map knowledge model centers on points of interest and incorporates knowledge of various dimensions, not only can these multivariate knowledge be used to better understand points of interest, but multivariate knowledge in other dimensions can also help determine the relationship between a point of interest and a certain dimension. The relevance of knowledge.
  • this disclosure chooses to extract the matching interest points and query words from the model fused with multi-knowledge, and based on this to determine their semantic correspondence at the semantic level, while inheriting the benefits of multi-knowledge, it also The corresponding relationship can be simplified as much as possible.
  • the method of predetermining the correspondence between query words and interest points at the semantic level can be: extract real interest points and corresponding historical query words from the multivariate map knowledge model, and then use semantic correlation modeling technology, to determine the correspondence between the real point of interest and the corresponding historical query words at the semantic level, and obtain the semantic correspondence.
  • semantic correlation modeling technology can be used to obtain the semantic correspondence.
  • other technologies that can achieve the same or similar effects can also be used to obtain the semantic correspondence.
  • the reason for determining the semantic correspondence between the matching query words and the POI is to explain as much as possible the user's search for a POI when the user tries to search for a POI under the condition of integrating multiple knowledge.
  • Customary query words that is, in actual situations, the query words are often different from the official text of the POI, but the two have the same semantics, that is to say, it is often difficult for users to accurately remember the official name of the POI when searching, but It is often possible to "make up or name" a query term by itself based on the semantics expressed.
  • the map point of interest query method constructs a multivariate map knowledge model based on interest points and corresponding multivariate knowledge in advance. Since multivariate knowledge covers geographic location information, Knowledge map information and input query words, so the trained multivariate map knowledge model can more comprehensively and accurately determine the query words that match the points of interest.
  • the material for semantic correlation modeling is the points of interest and matching query words extracted from the trained multivariate map knowledge model, and then the semantic correspondence that can accurately reflect the user's search habits at the semantic level is obtained. Finally, with the help of this The semantic correspondence can accurately determine the target point of interest, which improves the accuracy of the query result of the point of interest query service.
  • this embodiment shows a flow chart of a method for constructing a multivariate map knowledge model through FIG. 3, wherein the process 300 includes the following steps:
  • Step 301 Obtain each point of interest within the target map area
  • Step 302 Obtain geographic location information corresponding to the point of interest, knowledge map information and input query words;
  • a specific way to obtain geographical location information, knowledge map information and input query words including:
  • the geographic location code can be embodied as a string, for example, it can be obtained based on Geohash (which is an address coding method that can encode two-dimensional spatial latitude and longitude data into a string) or based on Google-S2;
  • the knowledge map can be It is a map dedicated to recording various knowledge related to the entity of the map point of interest. For example, taking the point of interest representing a certain restaurant as an example, the map can record: customer evaluation, rating of a certain dish, per capita Consumption, recommended dishes, floors, queuing time, business hours, etc.; the query words entered by the user are used to establish the correspondence between the query words entered by the user and the official names of the actually selected points of interest.
  • Step 303 Use the point of interest as the master node, and use the geographical location information, knowledge map information and input query words as different slave nodes of the master node, and make the connection between the master node and each slave node;
  • Step 304 Determine the relationship between different points of interest according to the user behavior information, and connect the corresponding main nodes according to the relationship to obtain a node graph;
  • the line feature of the connection is determined based on the confidence between the two connected nodes, for example, the different confidence can be represented by the color of the line, or the length, thickness, etc. of the line.
  • Step 305 Based on the preset training task target, the node graph is pre-trained through the graph neural network to obtain a multivariate map knowledge model that meets the training task target requirements.
  • the method of defining various knowledge as nodes and then establishing connections between nodes can construct a node graph for learning through a graph neural network, so that under the guidance of the preset training task target , the required multivariate correspondence can be learned from the connection relations recorded in the nodes, and then a multivariate map knowledge model that meets the requirements of the training task objectives can be obtained.
  • FIG. 4 is a schematic diagram of node association constructed based on the scheme provided in FIG. 3 .
  • the goal of the training task may be expressed as: performing multi-classification on each master node, and the classification result accuracy of the multi-classification meets the preset requirements.
  • the classification referred to by multi-classification can be the classification of categories, types, and attributes, and the accuracy of classification results is used to indicate accuracy.
  • the graph neural network is used as the model architecture
  • the point of interest (usually expressed as a name in text form) is used as the main node
  • the geographical location information, knowledge graph information, and input query words corresponding to the point of interest are respectively used as the main node.
  • Different slave nodes of the node and establish a node graph showing the connection between the master node and the slave node, and between the master node and the master node, and use the node graph as a training sample to train according to the graph neural network, and then get as much as possible
  • a multivariate map knowledge model that embodies the multivariate knowledge association of points of interest.
  • a secondary pre-training technology may also be used.
  • the specific implementation process can be:
  • the constructed node graph is pre-trained through the graph neural network to be trained.
  • the reason why the above implementation process is described as a second pre-training is because the process of starting training of the model (such as BERT, ERNIE, GPT-3, etc. models commonly used in the field of natural language processing) from random parameters is called a pre-training, Since this embodiment inherits the first pre-training result of the model, there is no need to start pre-training from random parameters again, that is, the training of the graph neural network after inheriting the trained parameters is called the second pre-training. Since only the parameters related to the text content are inherited, it does not affect the network structure related to the non-text content in the graph neural network, thereby improving the training effect and shortening the training time while avoiding negative effects.
  • the model such as BERT, ERNIE, GPT-3, etc. models commonly used in the field of natural language processing
  • the multivariate map knowledge model provided by the embodiment shown in Figure 3 can also be used to determine the node category to help determine In the case of node categories, the node information is enriched, thereby further improving the correlation between nodes.
  • FIG. 5 is a flowchart of a method for determining a node category provided by an embodiment of the present disclosure, wherein the process 500 includes the following steps:
  • Step 501 For the first query word that has been included in the multivariate map knowledge model, obtain the node vector representation of the first query word, and determine the probability of belonging to different categories according to the node vector representation, and use the category with the highest corresponding probability as the first query word the actual class of
  • the node vector representation is jointly determined based on the information of other nodes that have established connections with itself or have adjacent relationships, so it can be combined with various information to correctly represent the nodes it belongs to.
  • Step 502 For the second query word not included in the multivariate map knowledge model, determine the target first query word similar to the second query word, and determine the category corresponding to the target first query word with the highest similarity as the second query word the actual category of .
  • This embodiment provides a method of determining similar target first query words according to similarity and inheriting the category of the target first query words. Method to realize.
  • the association between multi-dimensional knowledge and interest points can be used to solve other problems, such as the consistency between slave nodes connected to different master nodes. , to perform label completion or label adjustment according to the consistent results, so as to realize automatic labeling of unlabeled interest points.
  • FIG. 6 is a flowchart of a node labeling method provided by an embodiment of the present disclosure, wherein the process 600 includes the following steps:
  • Step 601 Based on the consistency between the slave nodes connected to different master nodes, perform a marked content completion operation for the master nodes and/or slave nodes that are not marked or the marked amount is less than the preset number;
  • Step 602 Based on the consistency among the slave nodes connected to different master nodes, determine the abnormal master node and/or abnormal slave node that has label errors, and initiate an error label inquiry or Press to perform corrective action.
  • the annotation content completion operation refers to supplementing the annotation information of nodes with annotations to nodes with no annotations or a small amount of annotations with consistency, so as to complete the annotation information; the wrong annotation query aims to Inquiry is used to determine whether there is an error in the consistency judgment.
  • this disclosure also provides an implementation plan for obtaining a multivariate map knowledge model through specific training in combination with the actual situation:
  • this embodiment chooses to fuse the following key information: 1) the basic information of POI (name, alias, address, category); 2 ) Geographic location information of POI; 3) Knowledge map information in POI field; 4) User behavior data related to POI.
  • the basic information of POI, geographic location information, knowledge map information and user behavior information can be expressed in the form of a graph, and in the graph, it is specifically represented as four different nodes, namely POI, query (query word), and geographic location Information, and tags associated with POIs in the knowledge graph.
  • Each node has its own features, mainly text features.
  • the nodes are connected by edges, and the graph mainly includes the following types of edges: 1) the path relationship between POIs and POIs, indicating that a user has traveled from one POI to another POI; 2) the click relationship between query and POI; 3) POI The relationship with the map label; 4) The relationship between POI and geographic location.
  • the node graph of this composition can be seen in Figure 4.
  • this embodiment proposes a pre-training model based on a graph neural network.
  • the pre-training model When the pre-training model is applied to downstream tasks, it can not only provide a pre-training model similar to the general domain for downstream tasks to model text, but also learn a vector representation containing rich information for each type of node, such as POI vector Indicates that the basic attribute information of POI, geographic location, POI-POI association, and POI-query association can be fully integrated. These vectors can be used for tasks such as semantic recall for map retrieval, query analysis, POI retrieval, and personalized recommendation of POIs.
  • the specific technical scheme is as follows:
  • the vector representation of each node is determined by its own feature information and its neighbor nodes in the graph.
  • the characteristics of the node itself are mainly text information.
  • the characteristics of POI nodes mainly include name, alias, address, and category;
  • the characteristics of query nodes are mainly query words;
  • the characteristics of graph label nodes include label type, content, and related description text, geographical
  • the location information may be expressed as a character string, for example, obtained based on Geohash or obtained based on Google-S2.
  • the pre-training model in the general field has a very good effect on text modeling, so you can use the existing pre-training model such as BERT, ERNIE, GPT-3 (all are commonly used in natural language Processing deep learning model) and so on to model the text features of the node, the simplest one can take the vector corresponding to the output layer CLS (English full name is Classification, which can be understood as used for downstream classification tasks) as the initialization vector representation of the node.
  • CLS English full name is Classification, which can be understood as used for downstream classification tasks
  • the update of the node vector is based on the vector representation of the node itself and the vector representation of the neighbor nodes in the last iteration, which can be expressed as the following formula:
  • N(u) represents the neighbor nodes of u in the graph
  • h u represents the vector representation of node u
  • the aggregate function in the formula is responsible for the representation of the neighbor nodes of the converging node u in the graph.
  • the aggregate it is necessary to consider the specific characteristics of the heterogeneous graph in the map field. Since the graph contains different types of nodes and relationships, when aggregating the representations of neighbor nodes, in order to better retain the information of different types of nodes, each node information can be aggregated separately, and then the different types of nodes information is summarized.
  • the representations of the neighbor nodes can be summed and averaged, or an attention mechanism can be introduced to weight the current node according to the importance of different nodes.
  • the update function updates the vector representation of each node according to the vector representation of the node itself and the vector representations of the gathered neighbor nodes in the last iteration.
  • the design of this function can have a variety of schemes, such as adding the node representation and the neighbor node representation through a linear layer, and then going through a layer of nonlinear transformation to obtain an updated vector representation.
  • the number of iterations k indicates that at most one node's k-hop neighbor information is used.
  • the representation of each node in the final graph can be independently represented by the node after the kth iteration can also be with the initial combined (such as splicing), or the arrive The representations of the nodes are combined as the final node representation.
  • the advantage of this method is that it can retain more important feature information of the node itself, such as text description information.
  • the pre-training task designed in this embodiment is node prediction, and the goal is to predict which category a POI node belongs to (the map field divides POIs into categories, and each POI is marked with a category).
  • node prediction it can be regarded as a classification task.
  • the node to be predicted is represented through a simple network structure, such as a linear layer, and then passed through a softmax (normalized) layer to predict the category it belongs to.
  • the loss function of model training can adopt the cross entropy loss function.
  • the present disclosure provides an embodiment of a device for querying points of interest on a map.
  • This device embodiment corresponds to the method embodiment shown in FIG. 2 , and the device Specifically, it can be applied to various electronic devices.
  • the apparatus 700 for querying a point of interest on a map in this embodiment may include: a point of interest query request receiving unit 701 , a target query word extraction unit 702 , and a target point of interest determination unit 703 .
  • the point of interest query request receiving unit 701 is configured to receive the input point of interest query request;
  • the target query word extraction unit 702 is configured to extract the target query word contained in the point of interest query request;
  • the target point of interest determination unit 703, configured to determine a target point of interest corresponding to the target query word by using a predetermined semantic correspondence between the query word and the point of interest; wherein the semantic correspondence is extracted from the query word and the preset multivariate map knowledge model
  • the points of interest are determined, and the multivariate map knowledge model records the multivariate correspondence between the points of interest and the multivariate knowledge.
  • the multivariate knowledge includes: geographic location information, knowledge map information, and input query words.
  • the specific processing of the point of interest query request receiving unit 701, the target query word extraction unit 702, the target point of interest determination unit 703 and the technical effects brought by it can be referred to FIG. 2 corresponds to the relevant descriptions of steps 201-203 in the embodiment, and details are not repeated here.
  • the device 700 for querying map POIs may also include:
  • An information extraction unit configured to extract real points of interest and corresponding historical query words from the multivariate map knowledge model
  • the semantic correlation modeling unit is configured to use the semantic correlation modeling technology to determine the semantic correspondence between the real point of interest and the corresponding historical query words, and obtain the semantic correspondence.
  • the device 700 for querying map POIs may also include:
  • a point of interest text information acquisition unit configured to acquire each point of interest within the target map area
  • a multivariate knowledge acquisition unit configured to acquire geographical location information corresponding to the point of interest, knowledge map information and input query words
  • the master-slave node determination and connection unit is configured to use the point of interest as the master node, and use the geographical location information, knowledge map information, and input query words as different slave nodes of the master node, and act as a link between the master node and each slave node. connection between
  • connection unit between main nodes is configured to determine the relationship between different points of interest according to the user behavior information, and make the corresponding connection between the main nodes according to the relationship to obtain a node graph; wherein, the line characteristics of the connection Determined based on the confidence between the two connected nodes;
  • the pre-training unit is configured to pre-train the node graph through the graph neural network based on the preset training task target to obtain a multivariate map knowledge model that meets the training task target requirements.
  • the multivariate knowledge acquisition unit can be further configured to:
  • the query words input by the user before the point of interest is selected are obtained.
  • the training task target includes: performing multi-classification on each master node, and the classification result accuracy of the multi-classification meets a preset requirement.
  • the device 700 for querying map POIs may also include:
  • the category determination unit of the included query word is configured to obtain the node vector representation of the first query word for the first query word included in the multivariate map knowledge model, and determine the probability of belonging to different categories according to the node vector representation, and the corresponding probability The largest category is used as the actual category of the first query term; wherein, the node vector representation is jointly determined based on the information of other nodes that are connected or have adjacent relationships with themselves;
  • the non-collected query word category determination unit is configured to determine the target first query word similar to the second query word for the second query word that is not included in the multivariate map knowledge model, and will correspond to the target first query word with the highest similarity
  • the category of is determined as the actual category of the second query word.
  • the device 700 for querying map POIs may also include:
  • the node annotation content completion unit is configured to perform annotation content completion for the master nodes and/or slave nodes that are not marked or the amount of annotation is less than the preset number based on the consistency between the slave nodes connected to different master nodes operate;
  • the abnormal label content processing unit is configured to determine the abnormal master node and/or the abnormal slave node with labeling errors based on the consistency between the slave nodes connected to different master nodes, and the abnormal master node and/or the abnormal slave node Initiate an error flag query or press to take corrective action.
  • the pre-training unit may be further configured as:
  • the parameters after training are used as the initial parameters of the network structure related to the text content in the graph neural network to obtain the graph neural network to be trained;
  • the constructed node graph is pre-trained through the graph neural network to be trained.
  • This embodiment exists as an apparatus embodiment corresponding to the method embodiment described above.
  • the map point of interest query device constructs a multivariate map knowledge model based on interest points and corresponding multivariate knowledge in advance. Since multivariate knowledge covers geographic location information, Knowledge map information and input query words, so the trained multivariate map knowledge model can more comprehensively and accurately determine the query words that match the points of interest.
  • the material for semantic correlation modeling is the points of interest and matching query words extracted from the trained multivariate map knowledge model, and then the semantic correspondence that can accurately reflect the user's search habits at the semantic level is obtained. Finally, with the help of this The semantic correspondence can accurately determine the target point of interest, which improves the accuracy of the query result of the point of interest query service.
  • the present disclosure also provides an electronic device, the electronic device includes: at least one processor; and a memory connected in communication with the at least one processor; wherein, the memory stores information executable by the at least one processor. Instructions, the instructions are executed by at least one processor, so that the at least one processor can implement the map interest point query method described in any of the above embodiments when executed.
  • the present disclosure also provides a readable storage medium, the readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to implement the map interest points described in any of the above embodiments. Query method.
  • the present disclosure further provides a computer program product, when the computer program is executed by a processor, the method for querying a point of interest on a map described in any of the above embodiments can be implemented.
  • FIG. 8 shows a schematic block diagram of an example electronic device 800 that may be used to implement embodiments of the present disclosure.
  • Electronic device is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers.
  • Electronic devices may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smart phones, wearable devices, and other similar computing devices.
  • the components shown herein, their connections and relationships, and their functions, are by way of example only, and are not intended to limit implementations of the disclosure described and/or claimed herein.
  • the device 800 includes a computing unit 801 that can execute according to a computer program stored in a read-only memory (ROM) 802 or loaded from a storage unit 808 into a random access memory (RAM) 803. Various appropriate actions and treatments. In the RAM 803, various programs and data necessary for the operation of the device 800 can also be stored.
  • the computing unit 801, ROM 802, and RAM 803 are connected to each other through a bus 804.
  • An input/output (I/O) interface 805 is also connected to the bus 804 .
  • the I/O interface 805 includes: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a magnetic disk, an optical disk, etc. ; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, and the like.
  • the communication unit 809 allows the device 800 to exchange information/data with other devices over a computer network such as the Internet and/or various telecommunication networks.
  • the computing unit 801 may be various general-purpose and/or special-purpose processing components having processing and computing capabilities. Some examples of computing units 801 include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, digital signal processing processor (DSP), and any suitable processor, controller, microcontroller, etc.
  • the calculation unit 801 executes various methods and processes described above, such as a method for querying points of interest on a map.
  • the map POI query method may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as the storage unit 808 .
  • part or all of the computer program may be loaded and/or installed on the device 800 via the ROM 802 and/or the communication unit 809.
  • the computer program When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the map point of interest query method described above can be executed.
  • the computing unit 801 may be configured in any other appropriate way (for example, by means of firmware) to execute the map POI query method.
  • Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips Implemented in a system of systems (SOC), load programmable logic device (CPLD), computer hardware, firmware, software, and/or combinations thereof.
  • FPGAs field programmable gate arrays
  • ASICs application specific integrated circuits
  • ASSPs application specific standard products
  • SOC system of systems
  • CPLD load programmable logic device
  • computer hardware firmware, software, and/or combinations thereof.
  • programmable processor can be special-purpose or general-purpose programmable processor, can receive data and instruction from storage system, at least one input device, and at least one output device, and transmit data and instruction to this storage system, this at least one input device, and this at least one output device an output device.
  • Program codes for implementing the methods of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special purpose computer, or other programmable data processing devices, so that the program codes, when executed by the processor or controller, make the functions/functions specified in the flow diagrams and/or block diagrams Action is implemented.
  • the program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.
  • a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device.
  • a machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium.
  • a machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing.
  • machine-readable storage media would include one or more wire-based electrical connections, portable computer discs, hard drives, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, compact disk read only memory (CD-ROM), optical storage, magnetic storage, or any suitable combination of the foregoing.
  • RAM random access memory
  • ROM read only memory
  • EPROM or flash memory erasable programmable read only memory
  • CD-ROM compact disk read only memory
  • magnetic storage or any suitable combination of the foregoing.
  • the systems and techniques described herein 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 the user. ); and a keyboard and pointing device (eg, a mouse or a trackball) through which a user can provide input to the computer.
  • a display device e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor
  • a keyboard and pointing device eg, a mouse or a trackball
  • Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and can be in any form (including Acoustic input, speech input or, tactile input) to receive input from the user.
  • the systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., as a a user computer having a graphical user interface or web browser through which a user can interact with embodiments of the systems and techniques described herein), or including such backend components, middleware components, Or any combination of front-end components in a computing system.
  • the components of the system can be interconnected by any form or medium of digital data communication, eg, a communication network. Examples of communication networks include: Local Area Network (LAN), Wide Area Network (WAN) and the Internet.
  • a computer system may include clients and servers.
  • Clients and servers are generally remote from each other and typically interact through a communication network.
  • the relationship of client and server arises by computer programs running on the respective computers and having a client-server relationship to each other.
  • the server can be a cloud server, also known as cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the management difficulties in traditional physical host and virtual private server (VPS, Virtual Private Server) services Large and weak business expansion.
  • cloud server also known as cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the management difficulties in traditional physical host and virtual private server (VPS, Virtual Private Server) services Large and weak business expansion.
  • VPN Virtual Private Server
  • the embodiments of the present disclosure pre-construct a multivariate map knowledge model based on interest points and corresponding multivariate knowledge, because multivariate knowledge covers geographical location information, knowledge map information and input query words , so the trained multivariate map knowledge model can more comprehensively and accurately determine the query words that match the points of interest.
  • the material for semantic correlation modeling is the points of interest and matching query words extracted from the trained multivariate map knowledge model, and then the semantic correspondence that can accurately reflect the user's search habits at the semantic level is obtained. Finally, with the help of this The semantic correspondence can accurately determine the target point of interest, which improves the accuracy of the query result of the point of interest query service.
  • steps may be reordered, added or deleted using the various forms of flow shown above.
  • each step described in the present disclosure may be executed in parallel, sequentially, or in a different order, as long as the desired result of the technical solution disclosed in the present disclosure can be achieved, no limitation is imposed herein.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Remote Sensing (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

本公开提供了一种地图兴趣点查询方法、装置、电子设备、计算机可读存储介质及计算机程序产品，涉及深度学习、自然语言处理、知识图谱、智能地图等人工智能技术领域。该方法包括：接收输入的兴趣点查询请求；从兴趣点查询请求中提取出包含的目标查询词；利用预先确定的查询词与兴趣点之间的语义对应关系，确定与目标查询词对应的目标兴趣点；其中，语义对应关系由提取自预设多元地图知识模型的查询词和兴趣点确定，多元地图知识模型记录有兴趣点与多元知识之间的多元对应关系，多元知识包括：地理位置信息、知识图谱信息和输入的查询词。应用该方法可以提升兴趣点查询服务的查询结果准确性。

Description

地图兴趣点查询方法、装置、设备、存储介质及程序产品
本专利申请要求于2021年12月28日提交的、申请号为202111626640.X、发明名称为“地图兴趣点查询方法、装置、设备、存储介质及程序产品”的中国专利申请的优先权，该申请的全文以引用的方式并入本申请中。
技术领域
本公开涉及数据处理技术领域，具体涉及深度学习、自然语言处理、知识图谱、智能地图等人工智能技术领域，尤其涉及一种地图兴趣点查询方法、装置、电子设备、计算机可读存储介质及计算机程序产品。
背景技术
导航电子地图产品，不仅每天为亿万用户提供出行服务同时也作为新基建数字底座，在各个行业发挥着越来越重要，不可或缺的作用。对于地图产品来说，数据是赖以生存的基础设施，准确是地图的生命线，时效则是地图刻画真实世界的必备能力。
而除了大量的真实数据，为地图领域涉及的多种知识进行科学的处理，才能够使以地图数据为基础构建的产品更好的满足用户的需求。
发明内容
本公开实施例提出了一种地图兴趣点查询方法、装置、电子设备、计算机可读存储介质及计算机程序产品。
第一方面，本公开实施例提出了一种地图兴趣点查询方法，包括：接收输入的兴趣点查询请求；从兴趣点查询请求中提取出包含的目标查询词；利用预先确定的查询词与兴趣点之间的语义对应关系，确定与目标查询词对应的目标兴趣点；其中，语义对应关系由提取自预设多元地图知识模型的查询词和兴趣点确定，多元地图知识模型记录有兴趣点与多元知识之间的多元对应关系，多元知识包括：地理位置信息、知识图谱信息和输入的查询词。
第二方面，本公开实施例提出了一种地图兴趣点查询装置，包括：兴趣点查询请求接收单元，被配置成接收输入的兴趣点查询请求；目标查询词提取单元，被配置成从兴趣点查询请求中提取出包含的目标查询词；目标兴趣点确定单元，被配置成利用预先确定的查询词与兴趣点之间的语义对应关系，确定与目标查询词对应的目标兴趣点；其中，语义对应关系由提取自预设多元地图知识模型的查询词和兴趣点确定，多元地图知识模型记录有兴趣点与多元知识之间的多元对应关系，多元知识包括：地理位置信息、知识图谱信息和输入的查询词。
第三方面，本公开实施例提供了一种电子设备，该电子设备包括：至少一个处理器；以及与至少一个处理器通信连接的存储器；其中，存储器存储有可被至少一个处理器执行的指令，该指令被至少一个处理器执行，以使至少一个处理器执行时能够实现如第一方面中任一实现方式描述的地图兴趣点查询方法。
第四方面，本公开实施例提供了一种存储有计算机指令的非瞬时计算机可读存储介质，该计算机指令用于使计算机执行时能够实现如第一方面中任一实现方式描述的地图兴趣点查询方法。
第五方面，本公开实施例提供了一种包括计算机程序的计算机程序产品，该计算机程序在被处理器执行时能够实现如第一方面中任一实现方式描述的地图兴趣点查询方法。
为了向用户提供更好的地图兴趣点查询服务，本公开预先基于兴趣点和对应的多元知识构建了多元地图知识模型，由于多元知识覆盖了地理位置信息、知识图谱信息和输入的查询词，因此经过训练的多元地图知识模型可以更加全面、准确的确定与兴趣点匹配的查询词。而进行语义相关性建模的素材正是从训练好的多元地图知识模型中提取出的兴趣点和匹配的查询词，进而得到能够从语义层面准确体现用户搜索习惯的语义对应关系，最终借助该语义对应关系得以准确的确定出目标兴趣点，提升了兴趣点查询服务的查询结果准确性。
应当理解，本部分所描述的内容并非旨在标识本公开的实施例的关键或重要特征，也不用于限制本公开的范围。本公开的其它特征将通过以下的说明书而变得容易理解。
附图说明
通过阅读参照以下附图所作的对非限制性实施例所作的详细描述，本公开的其它特征、目的和优点将会变得更明显：
图1是本公开可以应用于其中的示例性系统架构；
图2为本公开实施例提供的一种地图兴趣点查询方法的流程图；
图3为本公开实施例提供的一种构建多元地图知识模型的方法的流程图；
图4为基于图3所提供方案构建出的节点关联示意图；
图5为本公开实施例提供的一种节点类别确定方法的流程图；
图6为本公开实施例提供的一种节点标注方法的流程图；
图7为本公开实施例提供的一种地图兴趣点查询装置的结构框图；
图8为本公开实施例提供的一种适用于执行地图兴趣点查询方法的电子设备的结构示意图。
具体实施方式
以下结合附图对本公开的示范性实施例做出说明，其中包括本公开实施例的各种细节以助于理解，应当将它们认为仅仅是示范性的。因此，本领域普通技术人员应当认识到，可以对这里描述的实施例做出各种改变和修改，而不会背离本公开的范围和精神。同样，为了清楚和简明，以下的描述中省略了对公知功能和结构的描述。需要说明的是，在不冲突的情况下，本公开中的实施例及实施例中的特征可以相互组合。
本公开的技术方案中，所涉及的用户个人信息的收集、存储、使用、加工、传输、提供和公开等处理，均符合相关法律法规的规定，且不违背公序良俗。
图1示出了可以应用本公开的地图兴趣点查询方法、装置、电子设备及计算机可读存储介质的实施例的示例性系统架构100。
如图1所示，系统架构100可以包括终端设备101、102、103，网络104和服务器105。网络104用以在终端设备101、102、103和服务 器105之间提供通信链路的介质。网络104可以包括各种连接类型，例如有线、无线通信链路或者光纤电缆等等。
用户可以使用终端设备101、102、103通过网络104与服务器105交互，以接收或发送消息等。终端设备101、102、103和服务器105上可以安装有各种用于实现两者之间进行信息通讯的应用，例如地图导航类应用、模型训练类应用、即时通讯类应用等。
终端设备101、102、103和服务器105可以是硬件，也可以是软件。当终端设备101、102、103为硬件时，可以是具有显示屏的各种电子设备，包括但不限于智能手机、平板电脑、膝上型便携计算机和台式计算机等等；当终端设备101、102、103为软件时，可以安装在上述所列举的电子设备中，其可以实现成多个软件或软件模块，也可以实现成单个软件或软件模块，在此不做具体限定。当服务器105为硬件时，可以实现成多个服务器组成的分布式服务器集群，也可以实现成单个服务器；服务器为软件时，可以实现成多个软件或软件模块，也可以实现成单个软件或软件模块，在此不做具体限定。
服务器105通过内置的各种应用可以提供各种服务，以可以提供地图兴趣点查询服务的地图导航类应用为例，服务器105在运行该地图导航类应用时可实现如下效果：首先，接收终端设备101、102、103通过网络104传入的兴趣点查询请求；然后，从兴趣点查询请求中提取出包含的目标查询词；接下来，利用预先确定的查询词与兴趣点之间的语义对应关系，确定与目标查询词对应的目标兴趣点；最后，将确定出的目标兴趣点通过网络104返回至终端设备101、102、103。
其中，语义对应关系由提取自预设多元地图知识模型的查询词和兴趣点确定，多元地图知识模型记录有兴趣点与多元知识之间的多元对应关系，多元知识包括：地理位置信息、知识图谱信息和输入的查询词。具体的，多元地图知识模型可以由服务器105预先通过内置的模型训练类应用训练得到。
需要指出的是，兴趣点查询请求除可以从终端设备101、102、103通过网络104获取到之外，也可以通过各种方式预先存储在服务器105本地。因此，当服务器105检测到本地已经存储有这些数据时(例如开 始处理之前留存的待处理查询任务)，可选择直接从本地获取这些数据，在此种情况下，示例性系统架构100也可以不包括终端设备101、102、103和网络104。
本公开后续各实施例所提供的地图兴趣点查询方法一般由拥有较强运算能力、较多运算资源的服务器105来执行，相应地，地图兴趣点查询装置一般也设置于服务器105中。但同时也需要指出的是，在终端设备101、102、103也具有满足要求的运算能力和运算资源时，终端设备101、102、103也可以通过其上安装的地图导航类应用完成上述本交由服务器105做的各项运算，进而输出与服务器105同样的结果。尤其是在同时存在多种具有不同运算能力的终端设备的情况下，但地图导航类应用判断所在的终端设备拥有较强的运算能力和剩余较多的运算资源时，可以让终端设备来执行上述运算，从而适当减轻服务器105的运算压力，相应的，地图兴趣点查询装置也可以设置于终端设备101、102、103中。在此种情况下，示例性系统架构100也可以不包括服务器105和网络104。
应该理解，图1中的终端设备、网络和服务器的数目仅仅是示意性的。根据实现需要，可以具有任意数目的终端设备、网络和服务器。
请参考图2，图2为本公开实施例提供的一种地图兴趣点查询方法的流程图，其中流程200包括以下步骤：
步骤201：接收输入的兴趣点查询请求；
本步骤旨在由地图兴趣点查询方法的执行主体(例如图1所示的服务器105)接收用户终端(例如图1所示的终端设备101、102、103)通过网络104传入的兴趣点查询请求。
其中，该兴趣点查询请求用于表征某用户对某个兴趣点的查询需求，其中至少包含有用于作为查询基础的目标查询词，该目标查询词可以直接表示为明文的文本信息或未加密的语音信号，也可以表现为密文的字符串或加密的语音信号，以满足安全性需求。
当然，除了包含有表征目标查询词的基本信息外，还可以包含用户终端的终端信息、用户发起该请求时的当前位置、以及可能包含的兴趣点查询限 缩条件，以结合可能影响查询结果的各种信息来提升所得到的查询结果的准确性，以更好的满足用户的需求。
步骤202：从兴趣点查询请求中提取出包含的目标查询词；
在步骤201的基础上，本步骤旨在由上述执行主体从兴趣点查询请求中提取出包含的目标查询词。
具体的，考虑到目标查询词可能表示为不同的形式，在提取时也需要采用正确、匹配的提取方式，例如在目标查询词是以密文或加密的方式包含于兴趣点查询请求中时，还需要预先确定正确的解密密钥或解密规则，从而得到正确的目标查询词；当目标查询词表现为语音信号时，还需要通过语音识别技术来转换得到便于后续处理的文本形式的目标查询词；当兴趣点查询请求中未直接包含目标查询词，而是包含了能够指导从他处获取到目标查询词的相关信息时，还需要能够按照该相关信息从他处正确的获取到目标查询词。
步骤203：利用预先确定的查询词与兴趣点之间的语义对应关系，确定与目标查询词对应的目标兴趣点。
在步骤202的基础上，本步骤旨在由上述执行主体利用确定的语义对应关系，确定与目标查询词存在对应关系、且该对应关系与该语义对应关系一致的目标兴趣点。
其中，语义对应关系由提取自预设多元地图知识模型的查询词和兴趣点确定，而该多元地图知识模型记录有兴趣点与多元知识之间的多元对应关系，多元知识包括：地理位置信息、知识图谱信息和输入的查询词。由于该多元地图知识模型以兴趣点为中心还融合有多种不同维度的知识，不仅能够借助这些多元知识更好的理解兴趣点，其它维度的多元知识也能够帮助确定兴趣点与某个维度的知识的关联性。因此本公开选择从融合有多元知识的模型中提取存在匹配关系的兴趣点和查询词，并基于此来确定其在语义层面上的语义对应关系，继承了多元知识所带来好处的同时，也能够尽可能的精简对应关系。
具体的，预先确定描述查询词与兴趣点在语义层面上的对应关系的方式，可以为：从多元地图知识模型中提取出真实兴趣点和对应的历史查询词，然后，利用语义相关性建模技术，确定真实兴趣点和对应的历史查询词在语 义层面的对应关系，得到语义对应关系。当然，也可以利用其它能够起到相同或类似效果的技术来得到该语义对应关系。
应当理解的是，之所以要确定存在匹配关系的查询词和兴趣点之间的语义对应关系，是为了在融合多元知识的情况下，尽可能解释用户在尝试搜索某个兴趣点时体现其搜索习惯的查询词，即实际情况下，查询词往往与兴趣点的官方文本不同，但两者具有相同的语义，也就说是用户在搜索时往往难以准确的记住兴趣点的官方名称，但往往能够基于所表达的语义来自行“编造或命名”一个查询词。
为了向用户提供更好的地图兴趣点查询服务，本公开实施例提供的地图兴趣点查询方法，预先基于兴趣点和对应的多元知识构建了多元地图知识模型，由于多元知识覆盖了地理位置信息、知识图谱信息和输入的查询词，因此经过训练的多元地图知识模型可以更加全面、准确的确定与兴趣点匹配的查询词。而进行语义相关性建模的素材正是从训练好的多元地图知识模型中提取出的兴趣点和匹配的查询词，进而得到能够从语义层面准确体现用户搜索习惯的语义对应关系，最终借助该语义对应关系得以准确的确定出目标兴趣点，提升了兴趣点查询服务的查询结果准确性。
为了更好的理解流程200中所使用的多元地图知识模型，本实施例通过图3示出了一种构建多元地图知识模型的方法的流程图，其中流程300包括以下步骤：
步骤301：获取处于目标地图区域内的各兴趣点；
步骤302：获取与兴趣点对应的地理位置信息、知识图谱信息和输入的查询词；
一种具体的获取地理位置信息、知识图谱信息和输入的查询词的方式，包括：
获取与兴趣点对应的地理位置编码；在预设的知识图谱中，获取与由兴趣点作为待查询实体对应的兴趣点实体信息；在记录用户操作信息的操作日志(该操作日志的读取已事先获取相应用户的授权)中，获取兴趣点被选择时刻之前用户输入的查询词。
其中，地理位置编码可以具体表现为字符串，例如可以是基于Geohash(是一种地址编码方法，能够把二维的空间经纬度数据编码成一个字符串)得到或者基于Google-S2得到；知识图谱可以是专用于记载与地图兴趣点的实体相关的各种知识的图谱，例如以代表某个餐饮店铺的兴趣点为例，该图谱中可以记载有：客人的评价、对某道菜的评分、人均消费、推荐菜、所在楼层、排队时长、营业时间等等；用户输入的查询词，则用于建立用户输入的查询词与所实际选择的兴趣点的官方名称之间的对应关系。
步骤303：将兴趣点作为主节点，并将地理位置信息、知识图谱信息和输入的查询词分别作为主节点的不同从节点，且做主节点与各从节点之间的连线；
步骤304：根据用户行为信息，确定不同兴趣点之间的关联关系，并根据关联关系做相应的主节点间的连线，得到节点图；
其中，连线的线特征基于所连接的两个节点之间的置信度确定，例如可以通过线的颜色不同来表征不同的置信度，或者通过线的长短、粗细等。
步骤305：基于预设的训练任务目标，将节点图通过图神经网络进行预训练，得到满足训练任务目标要求的多元地图知识模型。
即本实施例所提供的将各种知识定义为节点、然后建立节点之间连线的方式，能够构建出用于通过图神经网络进行学习的节点图，使得在预设的训练任务目标指导下，可以从节点中记载的连线关系中学习到所需的多元对应关系，进而得到满足训练任务目标要求的多元地图知识模型。图4为基于图3所提供方案构建出的节点关联示意图。
具体的，该训练任务目标可以表现为：对每个主节点进行多分类、且多分类的分类结果精度满足预设要求。多分类所指的分类可以是类别、类型、属性的分类，分类结果精度则用于表示准确性。
本实施例采用图神经网络作为模型架构，将兴趣点(通常表现为文本形式的名称)作为主节点，将与该兴趣点对应的地理位置信息、知识图谱信息、输入的查询词分别作为该主节点的不同从节点，并建立表现为主节点与从节点之间、主节点与主节点之间的连线的节点图，将节点图作为训练样本按照图神经网络进行训练，进而得到尽可能的体现兴趣点多元知识关联的多 元地图知识模型。
应当理解的是，除本实施例所采用的图神经网络外，也可以采用能够起到类似效果的其它模型作为架构，本实施例仅以图神经网络作为一个优选例子来描述整个过程，其它应用场景可自行根据场景下所有可能存在的限制条件或约束，选择其它模型，此处不做具体限定。
在流程300所示实施例的基础上，为了尽可能的提升模型训练效果、缩短模型训练耗时，还可以采用二次预训练技术。具体实现过程可以为：
从已训练好的用于自然语言处理的模型中，获取与文本内容相关的网络结构的训练后参数；
将训练后参数作为图神经网络中与文本内容相关的网络结构的初始参数，得到待训练图神经网络；
将构建出的节点图通过待训练图神经网络进行预训练。
之所以将上述实现过程描述为二次预训练，是因为将该模型(例如常用于自然语言处理领域的BERT、ERNIE、GPT-3等模型)从随机参数开始训练的过程称为一次预训练，而由于本实施例继承了该模型的一次预训练结果，无需再重新从随机参数开始预训练，即图神经网络在继承训练后参数后开始的训练被称为二次预训练。由于继承的仅是与文本内容相关的参数，并不影响图神经网络中与非文本内容相关的网络结构，从而在避免带来负面影响的情况下，提升训练效果、缩短训练耗时。
考虑到真实世界中的各项实物不断发展、更新，兴趣点不断增加、替换、更换，基于图3所示实施例所提供的多元地图知识模型，还可以用于确定节点类别，以在帮助确定节点类别的情况下，丰富节点信息，从而进一步的提升各节点之间的关联性。
请参考图5，图5为本公开实施例提供的一种节点类别确定方法的流程图，其中流程500包括以下步骤：
步骤501：针对多元地图知识模型已收录的第一查询词，获取第一查询词的节点向量表示，并根据节点向量表示确定所属不同类别的概率，且将对应概率最大的类别作为第一查询词的实际类别；
其中，节点向量表示基于与自身建立有连接或存在相邻关系的其它 节点的信息共同确定得到，因此可以结合各方面信息来正确的表示所属的节点。
步骤502：针对多元地图知识模型未收录的第二查询词，确定与第二查询词相似的目标第一查询词，并将对应相似度最高的目标第一查询词的类别确定为第二查询词的实际类别。
区别于已收录的第一查询词，第二查询词大多数新产生的查询词，本实施例提供了按照相似性来确定相似的目标第一查询词、并继承目标第一查询词的类别的实现方式。
由于多元地图知识模型融合了多维度的知识，而多维度的知识所表现出的与兴趣点的关联，又可以用于解决其它问题，例如基于连接于不同主节点的各从节点间的一致性，来根据一致性的结果进行标注补全或标注调整，以实现无标注兴趣点的自动标注。
请参考图6，图6为本公开实施例提供的一种节点标注方法的流程图，其中流程600包括以下步骤：
步骤601：基于连接于不同主节点的各从节点间的一致性，对未标注或标注量少于预设数量的主节点和/或从节点，执行标注内容补全操作；
步骤602：基于连接于不同主节点的各从节点间的一致性，确定存在标注错误的异常主节点和/或异常从节点，并对异常主节点和/或异常从节点发起错误标注问询或按执行纠正操作。
其中，标注内容补全操作是指将具有标注的节点的标注信息补充至具有一致性的不具有标注或具有少量标注的节点上，以此实现标注信息的补全；错误标注问询旨在通过以问询的方式来确定一致性判断是否存在错误。
为加深理解，本公开还结合实际情况，给出了一种具体训练得到多元地图知识模型的实现方案：
为了在通用领域预训练模型的基础上，训练出包含多元知识的地图领域的预训练模型，本实施例选择融合以下关键信息：1)POI的基础信息(名称、别名、地址、类别)；2)POI的地理位置信息；3)POI领域的知识图谱信息；4)POI相关的用户行为数据。
其中，POI的基础信息、地理位置信息、知识图谱信息以及用户行为信息可以表示成图的形式，并在图中具体表现为不同的4中节点，分别是POI、query(查询词)、地理位置信息、以及知识图谱中与POI有关联的标签。
每个节点具有自己的特征，主要是文本特征。节点之间通过边相连接，图中主要包含以下几种边：1)POI与POI的路径关系，表示有用户曾从一个POI到另一个POI；2)query和POI的点击关系；3)POI和图谱标签的关系；4)POI和地理位置的关系。该构成的节点图可参见图4。
为了更好的建模图中节点之间的关系，本实施例提出了一种基于图神经网络的预训练模型。预训练模型在应用到下游任务时，既可以提供一个类似通用领域的预训练模型供下游任务对文本进行建模，也可以为每个类型的节点学习一个包含丰富信息的向量表示，例如POI向量表示可以充分融合POI的基本属性信息，地理位置，POI与POI的关联，以及POI和query的关联。这些向量可用于地图检索的语义召回、query分析、POI检索和POI的个性化推荐等任务。具体的技术方案如下：
在图神经网络中，每个节点的向量表示由其本身的特征信息以及其在图中的邻居节点共同决定。节点本身的特征主要是文本信息，例如POI节点的特征主要有名称、别名、地址、类别，query节点的特征主要是查询词，图谱标签节点的特征包括标签类型、内容、以及相关描述文本，地理位置信息可以表现为字符串，例如基于Geohash得到或者基于Google-S2得到。
由于节点的主要特征为文本，而通用领域的预训练模型对于文本的建模有着非常好的效果，所以可以采用已有的预训练模型例如BERT、ERNIE、GPT-3(均是常用于自然语言处理的深度学习模型)等对节点的文本特征进行建模，最简单的可以取输出层CLS(英文全称为Classification，可理解为用于下游的分类任务)对应的向量作为节点的初始化向量表示。
节点向量的更新基于上一次迭代该节点自身的向量表示以及邻居节点的向量表示，可以表示为如下公式：
Figure PCTCN2022104877-appb-000001
其中N(u)表示u在图中的邻居节点，h u表示节点u的向量表示，
Figure PCTCN2022104877-appb-000002
表示节点u的初始化向量表示，
Figure PCTCN2022104877-appb-000003
表示节点u经过k次迭代之后的向量表示。
公式中的aggregate函数负责汇聚节点u在图中的邻居节点的表示，在 设计aggregate时，需要考虑地图领域异构图的具体特点。由于图中包含了不同类型的节点以及关系，因此在对邻居节点的表示进行聚合时，为了更好的保留不同类型节点的信息，可以分别对每种节点信息进行聚合，然后再将不同类型节点的信息进行汇总。在对每种类型的邻居节点进行信息聚合时，可以将邻居节点的表示相加，取平均，或者引入attention(注意力)机制根据不同节点对当前节点的重要程度进行加权。由于图中的边天然具有置信度，所以也可以根据边的权重对邻居节点的向量表示进行加权求和。update函数根据上一次迭代时节点本身的向量表示以及汇聚的邻居节点的向量表示来对每个节点的向量表示进行更新。该函数的设计可以有多种方案，例如将节点表示和邻居节点的表示分别经过线性层之后进行相加，再经过一层非线性变换得到更新后的向量表示。
迭代次数k表示最多用到一个节点的k跳邻居信息。最终图中每个节点的表示可以单独取第k次迭代后的节点表示
Figure PCTCN2022104877-appb-000004
也可以将
Figure PCTCN2022104877-appb-000005
与初始的
Figure PCTCN2022104877-appb-000006
进行结合(例如拼接)，或者将
Figure PCTCN2022104877-appb-000007
到
Figure PCTCN2022104877-appb-000008
的表示进行结合来作为最终的节点表示，这种方式的优点是可以更多的保留节点自身重要的特征信息，如文本描述信息。
本实施例设计的预训练任务为节点预测，目标为预测POI节点属于哪个类别(地图领域对POI进行了类别划分，每个POI都被打上了一个类别)。对于节点预测，可以将其看作一个分类任务，将待预测的节点表示经过一个简单的网络结构，例如线性层，再经过一个softmax(归一化)层，预测所属的类别。模型训练的损失函数可以采用交叉熵损失函数。
进一步参考图7，作为对上述各图所示方法的实现，本公开提供了一种地图兴趣点查询装置的一个实施例，该装置实施例与图2所示的方法实施例相对应，该装置具体可以应用于各种电子设备中。
如图7所示，本实施例的地图兴趣点查询装置700可以包括：兴趣点查询请求接收单元701、目标查询词提取单元702、目标兴趣点确定单元703。其中，兴趣点查询请求接收单元701，被配置成接收输入的兴趣点查询请求；目标查询词提取单元702，被配置成从兴趣点查询请求中提取出包含的目标查询词；目标兴趣点确定单元703，被配置成利用预先 确定的查询词与兴趣点之间的语义对应关系，确定与目标查询词对应的目标兴趣点；其中，语义对应关系由提取自预设多元地图知识模型的查询词和兴趣点确定，多元地图知识模型记录有兴趣点与多元知识之间的多元对应关系，多元知识包括：地理位置信息、知识图谱信息和输入的查询词。
在本实施例中，地图兴趣点查询装置700中：兴趣点查询请求接收单元701、目标查询词提取单元702、目标兴趣点确定单元703的具体处理及其所带来的技术效果可分别参考图2对应实施例中的步骤201-203的相关说明，在此不再赘述。
在本实施例的一些可选的实现方式中，地图兴趣点查询装置700还可以包括：
信息提取单元，被配置成从多元地图知识模型中提取出真实兴趣点和对应的历史查询词；
语义相关性建模单元，被配置成利用语义相关性建模技术，确定真实兴趣点和对应的历史查询词在语义层面的对应关系，得到语义对应关系。
在本实施例的一些可选的实现方式中，地图兴趣点查询装置700还可以包括：
兴趣点文本信息获取单元，被配置成获取处于目标地图区域内的各兴趣点；
多元知识获取单元，被配置成获取与兴趣点对应的地理位置信息、知识图谱信息和输入的查询词；
主从节点确定及连线单元，被配置成将兴趣点作为主节点、并将地理位置信息、知识图谱信息、输入的查询词分别作为主节点的不同从节点，且做主节点与各从节点之间的连线；
主节点间连线单元，被配置成根据用户行为信息，确定不同兴趣点之间的关联关系，并根据关联关系做相应的主节点间的连线，得到节点图；其中，连线的线特征基于所连接的两个节点之间的置信度确定；
预训练单元，被配置成基于预设的训练任务目标，将节点图通过图神经网络进行预训练，得到满足训练任务目标要求的多元地图知识模型。
在本实施例的一些可选的实现方式中，多元知识获取单元可以被进一 步配置成：
获取与兴趣点对应的地理位置编码；
在预设的知识图谱中，获取与由兴趣点作为待查询实体对应的兴趣点实体信息；
在操作日志中，获取兴趣点被选择时刻之前用户输入的查询词。
在本实施例的一些可选的实现方式中，训练任务目标包括：对每个主节点进行多分类、且多分类的分类结果精度满足预设要求。
在本实施例的一些可选的实现方式中，地图兴趣点查询装置700还可以包括：
已收录查询词类别确定单元，被配置成针对多元地图知识模型已收录的第一查询词，获取第一查询词的节点向量表示，并根据节点向量表示确定所属不同类别的概率，且将对应概率最大的类别作为第一查询词的实际类别；其中，节点向量表示基于与自身建立有连接或存在相邻关系的其它节点的信息共同确定得到；
未收录查询词类别确定单元，被配置成针对多元地图知识模型未收录的第二查询词，确定与第二查询词相似的目标第一查询词，并将对应相似度最高的目标第一查询词的类别确定为第二查询词的实际类别。
在本实施例的一些可选的实现方式中，地图兴趣点查询装置700还可以包括：
节点标注内容补全单元，被配置成基于连接于不同主节点的各从节点间的一致性，对未标注或标注量少于预设数量的主节点和/或从节点，执行标注内容补全操作；
异常标注内容处理单元，被配置成基于连接于不同主节点的各从节点间的一致性，确定存在标注错误的异常主节点和/或异常从节点，并对异常主节点和/或异常从节点发起错误标注问询或按执行纠正操作。
在本实施例的一些可选的实现方式中，预训练单元可以被进一步配置成：
从已训练好的用于自然语言处理的模型中，获取与文本内容相关的网络结构的训练后参数；
将训练后参数作为图神经网络中与文本内容相关的网络结构的初始参 数，得到待训练图神经网络；
将构建出的节点图通过待训练图神经网络进行预训练。
本实施例作为对应于上述方法实施例的装置实施例存在。
为了向用户提供更好的地图兴趣点查询服务，本公开实施例提供的地图兴趣点查询装置，预先基于兴趣点和对应的多元知识构建了多元地图知识模型，由于多元知识覆盖了地理位置信息、知识图谱信息和输入的查询词，因此经过训练的多元地图知识模型可以更加全面、准确的确定与兴趣点匹配的查询词。而进行语义相关性建模的素材正是从训练好的多元地图知识模型中提取出的兴趣点和匹配的查询词，进而得到能够从语义层面准确体现用户搜索习惯的语义对应关系，最终借助该语义对应关系得以准确的确定出目标兴趣点，提升了兴趣点查询服务的查询结果准确性。
根据本公开的实施例，本公开还提供了一种电子设备，该电子设备包括：至少一个处理器；以及与至少一个处理器通信连接的存储器；其中，存储器存储有可被至少一个处理器执行的指令，该指令被至少一个处理器执行，以使至少一个处理器执行时能够实现上述任意实施例所描述的地图兴趣点查询方法。
根据本公开的实施例，本公开还提供了一种可读存储介质，该可读存储介质存储有计算机指令，该计算机指令用于使计算机执行时能够实现上述任意实施例所描述的地图兴趣点查询方法。
根据本公开的实施例，本公开还提供了一种计算机程序产品，该计算机程序在被处理器执行时能够实现上述任意实施例所描述的地图兴趣点查询方法。
图8示出了可以用来实施本公开的实施例的示例电子设备800的示意性框图。电子设备旨在表示各种形式的数字计算机，诸如，膝上型计算机、台式计算机、工作台、个人数字助理、服务器、刀片式服务器、大型计算机、和其它适合的计算机。电子设备还可以表示各种形式的移动装置，诸如，个人数字处理、蜂窝电话、智能电话、可穿戴设备和其 它类似的计算装置。本文所示的部件、它们的连接和关系、以及它们的功能仅仅作为示例，并且不意在限制本文中描述的和/或者要求的本公开的实现。
如图8所示，设备800包括计算单元801，其可以根据存储在只读存储器(ROM)802中的计算机程序或者从存储单元808加载到随机访问存储器(RAM)803中的计算机程序，来执行各种适当的动作和处理。在RAM 803中，还可存储设备800操作所需的各种程序和数据。计算单元801、ROM 802以及RAM 803通过总线804彼此相连。输入/输出(I/O)接口805也连接至总线804。
设备800中的多个部件连接至I/O接口805，包括：输入单元806，例如键盘、鼠标等；输出单元807，例如各种类型的显示器、扬声器等；存储单元808，例如磁盘、光盘等；以及通信单元809，例如网卡、调制解调器、无线通信收发机等。通信单元809允许设备800通过诸如因特网的计算机网络和/或各种电信网络与其他设备交换信息/数据。
计算单元801可以是各种具有处理和计算能力的通用和/或专用处理组件。计算单元801的一些示例包括但不限于中央处理单元(CPU)、图形处理单元(GPU)、各种专用的人工智能(AI)计算芯片、各种运行机器学习模型算法的计算单元、数字信号处理器(DSP)、以及任何适当的处理器、控制器、微控制器等。计算单元801执行上文所描述的各个方法和处理，例如地图兴趣点查询方法。例如，在一些实施例中，地图兴趣点查询方法可被实现为计算机软件程序，其被有形地包含于机器可读介质，例如存储单元808。在一些实施例中，计算机程序的部分或者全部可以经由ROM 802和/或通信单元809而被载入和/或安装到设备800上。当计算机程序加载到RAM 803并由计算单元801执行时，可以执行上文描述的地图兴趣点查询方法的一个或多个步骤。备选地，在其他实施例中，计算单元801可以通过其他任何适当的方式(例如，借助于固件)而被配置为执行地图兴趣点查询方法。
本文中以上描述的系统和技术的各种实施方式可以在数字电子电路系统、集成电路系统、场可编程门阵列(FPGA)、专用集成电路(ASIC)、专用标准产品(ASSP)、芯片上系统的系统(SOC)、负载可编程逻辑 设备(CPLD)、计算机硬件、固件、软件、和/或它们的组合中实现。这些各种实施方式可以包括：实施在一个或者多个计算机程序中，该一个或者多个计算机程序可在包括至少一个可编程处理器的可编程系统上执行和/或解释，该可编程处理器可以是专用或者通用可编程处理器，可以从存储系统、至少一个输入装置、和至少一个输出装置接收数据和指令，并且将数据和指令传输至该存储系统、该至少一个输入装置、和该至少一个输出装置。
用于实施本公开的方法的程序代码可以采用一个或多个编程语言的任何组合来编写。这些程序代码可以提供给通用计算机、专用计算机或其他可编程数据处理装置的处理器或控制器，使得程序代码当由处理器或控制器执行时使流程图和/或框图中所规定的功能/操作被实施。程序代码可以完全在机器上执行、部分地在机器上执行，作为独立软件包部分地在机器上执行且部分地在远程机器上执行或完全在远程机器或服务器上执行。
在本公开的上下文中，机器可读介质可以是有形的介质，其可以包含或存储以供指令执行系统、装置或设备使用或与指令执行系统、装置或设备结合地使用的程序。机器可读介质可以是机器可读信号介质或机器可读储存介质。机器可读介质可以包括但不限于电子的、磁性的、光学的、电磁的、红外的、或半导体系统、装置或设备，或者上述内容的任何合适组合。机器可读存储介质的更具体示例会包括基于一个或多个线的电气连接、便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦除可编程只读存储器(EPROM或快闪存储器)、光纤、便捷式紧凑盘只读存储器(CD-ROM)、光学储存设备、磁储存设备、或上述内容的任何合适组合。
为了提供与用户的交互，可以在计算机上实施此处描述的系统和技术，该计算机具有：用于向用户显示信息的显示装置(例如，CRT(阴极射线管)或者LCD(液晶显示器)监视器)；以及键盘和指向装置(例如，鼠标或者轨迹球)，用户可以通过该键盘和该指向装置来将输入提供给计算机。其它种类的装置还可以用于提供与用户的交互；例如，提供给用户的反馈可以是任何形式的传感反馈(例如，视觉反馈、听觉反 馈、或者触觉反馈)；并且可以用任何形式(包括声输入、语音输入或者、触觉输入)来接收来自用户的输入。
可以将此处描述的系统和技术实施在包括后台部件的计算系统(例如，作为数据服务器)、或者包括中间件部件的计算系统(例如，应用服务器)、或者包括前端部件的计算系统(例如，具有图形用户界面或者网络浏览器的用户计算机，用户可以通过该图形用户界面或者该网络浏览器来与此处描述的系统和技术的实施方式交互)、或者包括这种后台部件、中间件部件、或者前端部件的任何组合的计算系统中。可以通过任何形式或者介质的数字数据通信(例如，通信网络)来将系统的部件相互连接。通信网络的示例包括：局域网(LAN)、广域网(WAN)和互联网。
计算机系统可以包括客户端和服务器。客户端和服务器一般远离彼此并且通常通过通信网络进行交互。通过在相应的计算机上运行并且彼此具有客户端-服务器关系的计算机程序来产生客户端和服务器的关系。服务器可以是云服务器，又称为云计算服务器或云主机，是云计算服务体系中的一项主机产品，以解决传统物理主机与虚拟专用服务器(VPS，Virtual Private Server)服务中存在的管理难度大，业务扩展性弱的缺陷。
为了向用户提供更好的地图兴趣点查询服务，本公开实施例预先基于兴趣点和对应的多元知识构建了多元地图知识模型，由于多元知识覆盖了地理位置信息、知识图谱信息和输入的查询词，因此经过训练的多元地图知识模型可以更加全面、准确的确定与兴趣点匹配的查询词。而进行语义相关性建模的素材正是从训练好的多元地图知识模型中提取出的兴趣点和匹配的查询词，进而得到能够从语义层面准确体现用户搜索习惯的语义对应关系，最终借助该语义对应关系得以准确的确定出目标兴趣点，提升了兴趣点查询服务的查询结果准确性。
应该理解，可以使用上面所示的各种形式的流程，重新排序、增加或删除步骤。例如，本发公开中记载的各步骤可以并行地执行也可以顺序地执行也可以不同的次序执行，只要能够实现本公开公开的技术方案所期望的结果，本文在此不进行限制。
上述具体实施方式，并不构成对本公开保护范围的限制。本领域技术人员应该明白的是，根据设计要求和其他因素，可以进行各种修改、组合、子组合和替代。任何在本公开的精神和原则之内所作的修改、等同替换和改进等，均应包含在本公开保护范围之内。

Claims (19)

  1. 一种地图兴趣点查询方法，包括：
    接收输入的兴趣点查询请求；
    从所述兴趣点查询请求中提取出包含的目标查询词；
    利用预先确定的查询词与兴趣点之间的语义对应关系，确定与所述目标查询词对应的目标兴趣点；其中，所述语义对应关系由提取自预设多元地图知识模型的查询词和兴趣点确定，所述多元地图知识模型记录有兴趣点与多元知识之间的多元对应关系，所述多元知识包括：地理位置信息、知识图谱信息和输入的查询词。
  2. 根据权利要求1所述的方法，还包括：
    从所述多元地图知识模型中提取出真实兴趣点和对应的历史查询词；
    利用语义相关性建模技术，确定真实兴趣点和对应的历史查询词在语义层面的对应关系，得到所述语义对应关系。
  3. 根据权利要求1所述的方法，还包括：
    获取处于目标地图区域内的各兴趣点；
    获取与所述兴趣点对应的地理位置信息、知识图谱信息和输入的查询词；
    将所述兴趣点作为主节点，并将所述地理位置信息、所述知识图谱信息和所述输入的查询词分别作为所述主节点的不同从节点，且做所述主节点与各所述从节点之间的连线；
    根据用户行为信息，确定不同兴趣点之间的关联关系，并根据所述关联关系做相应的主节点间的连线，得到节点图；其中，所述连线的线特征基于所连接的两个节点之间的置信度确定；
    基于预设的训练任务目标，将所述节点图通过图神经网络进行预训练，得到满足所述训练任务目标要求的多元地图知识模型。
  4. 根据权利要求3所述的方法，其中，所述获取与所述兴趣点对应的地理位置信息、知识图谱信息和输入的查询词，包括：
    获取与所述兴趣点对应的地理位置编码；
    在预设的知识图谱中，获取与由所述兴趣点作为待查询实体对应的兴趣点实体信息；
    在操作日志中，获取所述兴趣点被选择时刻之前用户输入的查询词。
  5. 根据权利要求3所述的方法，其中，所述训练任务目标包括：对每个所述主节点进行多分类、且多分类的分类结果精度满足预设要求。
  6. 根据权利要求3所述的方法，还包括：
    针对所述多元地图知识模型已收录的第一查询词，获取所述第一查询词的节点向量表示，并根据所述节点向量表示确定所属不同类别的概率，且将对应概率最大的类别作为所述第一查询词的实际类别；其中，所述节点向量表示基于与自身建立有连接或存在相邻关系的其它节点的信息共同确定得到；
    针对所述多元地图知识模型未收录的第二查询词，确定与所述第二查询词相似的目标第一查询词，并将对应相似度最高的目标第一查询词的类别确定为所述第二查询词的实际类别。
  7. 根据权利要求3所述的方法，还包括：
    基于连接于不同主节点的各从节点间的一致性，对未标注或标注量少于预设数量的主节点和/或从节点，执行标注内容补全操作；
    基于连接于不同主节点的各从节点间的一致性，确定存在标注错误的异常主节点和/或异常从节点，并对所述异常主节点和/或所述异常从节点发起错误标注问询或按执行纠正操作。
  8. 根据权利要求3-7任一项所述的方法，其中，所述将构建出的节点图通过图神经网络进行预训练，包括：
    从已训练好的用于自然语言处理的模型中，获取与文本内容相关的网络结构的训练后参数；
    将所述训练后参数作为所述图神经网络中与文本内容相关的网络结构 的初始参数，得到待训练图神经网络；
    将构建出的节点图通过所述待训练图神经网络进行预训练。
  9. 一种地图兴趣点查询装置，包括：
    兴趣点查询请求接收单元，被配置成接收输入的兴趣点查询请求；
    目标查询词提取单元，被配置成从所述兴趣点查询请求中提取出包含的目标查询词；
    目标兴趣点确定单元，被配置成利用预先确定的查询词与兴趣点之间的语义对应关系，确定与所述目标查询词对应的目标兴趣点；其中，所述语义对应关系由提取自预设多元地图知识模型的查询词和兴趣点确定，所述多元地图知识模型记录有兴趣点与多元知识之间的多元对应关系，所述多元知识包括：地理位置信息、知识图谱信息和输入的查询词。
  10. 根据权利要求9所述的装置，还包括：
    信息提取单元，被配置成从所述多元地图知识模型中提取出真实兴趣点和对应的历史查询词；
    语义相关性建模单元，被配置成利用语义相关性建模技术，确定真实兴趣点和对应的历史查询词在语义层面的对应关系，得到所述语义对应关系。
  11. 根据权利要求9所述的装置，还包括：
    兴趣点文本信息获取单元，被配置成获取处于目标地图区域内的各兴趣点；
    多元知识获取单元，被配置成获取与所述兴趣点对应的地理位置信息、知识图谱信息和输入的查询词；
    主从节点确定及连线单元，被配置成将所述兴趣点作为主节点，并将所述地理位置信息、所述知识图谱信息和所述输入的查询词分别作为所述主节点的不同从节点，且做所述主节点与各所述从节点之间的连线；
    主节点间连线单元，被配置成根据用户行为信息，确定不同兴趣点之间的关联关系，并根据所述关联关系做相应的主节点间的连线，得到节点图；其中，所述连线的线特征基于所连接的两个节点之间的置信度确定；
    预训练单元，被配置成基于预设的训练任务目标，将所述节点图通过图神经网络进行预训练，得到满足所述训练任务目标要求的多元地图知识模型。
  12. 根据权利要求11所述的装置，其中，所述多元知识获取单元被进一步配置成：
    获取与所述兴趣点对应的地理位置编码；
    在预设的知识图谱中，获取与由所述兴趣点作为待查询实体对应的兴趣点实体信息；
    在操作日志中，获取所述兴趣点被选择时刻之前用户输入的查询词。
  13. 根据权利要求11所述的装置，其中，所述训练任务目标包括：对每个所述主节点进行多分类、且多分类的分类结果精度满足预设要求。
  14. 根据权利要求11所述的装置，还包括：
    已收录查询词类别确定单元，被配置成针对所述多元地图知识模型已收录的第一查询词，获取所述第一查询词的节点向量表示，并根据所述节点向量表示确定所属不同类别的概率，且将对应概率最大的类别作为所述第一查询词的实际类别；其中，所述节点向量表示基于与自身建立有连接或存在相邻关系的其它节点的信息共同确定得到；
    未收录查询词类别确定单元，被配置成针对所述多元地图知识模型未收录的第二查询词，确定与所述第二查询词相似的目标第一查询词，并将对应相似度最高的目标第一查询词的类别确定为所述第二查询词的实际类别。
  15. 根据权利要求11所述的装置，还包括：
    节点标注内容补全单元，被配置成基于连接于不同主节点的各从节点间的一致性，对未标注或标注量少于预设数量的主节点和/或从节点，执行标注内容补全操作；
    异常标注内容处理单元，被配置成基于连接于不同主节点的各从节点间的一致性，确定存在标注错误的异常主节点和/或异常从节点，并对所述 异常主节点和/或所述异常从节点发起错误标注问询或按执行纠正操作。
  16. 根据权利要求11-15任一项所述的装置，其中，所述预训练单元被进一步配置成：
    从已训练好的用于自然语言处理的模型中，获取与文本内容相关的网络结构的训练后参数；
    将所述训练后参数作为所述图神经网络中与文本内容相关的网络结构的初始参数，得到待训练图神经网络；
    将构建出的节点图通过所述待训练图神经网络进行预训练。
  17. 一种电子设备，包括：
    至少一个处理器；以及
    与所述至少一个处理器通信连接的存储器；其中，
    所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行权利要求1-8中任一项所述的地图兴趣点查询方法。
  18. 一种存储有计算机指令的非瞬时计算机可读存储介质，所述计算机指令用于使所述计算机执行权利要求1-8中任一项所述的地图兴趣点查询方法。
  19. 一种计算机程序产品，包括计算机程序，所述计算机程序被处理器执行时实现根据权利要求1-8中任一项所述地图兴趣点查询方法的步骤。
PCT/CN2022/104877 2021-12-28 2022-07-11 地图兴趣点查询方法、装置、设备、存储介质及程序产品 WO2023124005A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111626640.XA CN114329244A (zh) 2021-12-28 2021-12-28 地图兴趣点查询方法、装置、设备、存储介质及程序产品
CN202111626640.X 2021-12-28

Publications (1)

Publication Number Publication Date
WO2023124005A1 true WO2023124005A1 (zh) 2023-07-06

Family

ID=81015973

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/104877 WO2023124005A1 (zh) 2021-12-28 2022-07-11 地图兴趣点查询方法、装置、设备、存储介质及程序产品

Country Status (2)

Country Link
CN (1) CN114329244A (zh)
WO (1) WO2023124005A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117370672A (zh) * 2023-12-06 2024-01-09 烟台大学 基于混合结构图的用户兴趣点推荐方法、系统和设备

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114329244A (zh) * 2021-12-28 2022-04-12 北京百度网讯科技有限公司 地图兴趣点查询方法、装置、设备、存储介质及程序产品
CN114969556B (zh) * 2022-07-29 2022-10-25 北京百度网讯科技有限公司 基于用户个性化的兴趣点呈现和交互方法、相关装置
CN115098789B (zh) * 2022-08-05 2024-04-16 湖南工商大学 基于神经网络的多维兴趣融合推荐方法、装置及相关设备

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101685021A (zh) * 2008-09-24 2010-03-31 高德软件有限公司 一种兴趣点信息获取方法及装置
US20180349755A1 (en) * 2017-06-02 2018-12-06 Microsoft Technology Licensing, Llc Modeling an action completion conversation using a knowledge graph
CN110413886A (zh) * 2019-07-17 2019-11-05 海信集团有限公司 一种兴趣点展示方法及装置
CN112507040A (zh) * 2020-12-21 2021-03-16 北京百度网讯科技有限公司 多元关系生成模型的训练方法、装置、电子设备及介质
CN114329244A (zh) * 2021-12-28 2022-04-12 北京百度网讯科技有限公司 地图兴趣点查询方法、装置、设备、存储介质及程序产品

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101685021A (zh) * 2008-09-24 2010-03-31 高德软件有限公司 一种兴趣点信息获取方法及装置
US20180349755A1 (en) * 2017-06-02 2018-12-06 Microsoft Technology Licensing, Llc Modeling an action completion conversation using a knowledge graph
CN110413886A (zh) * 2019-07-17 2019-11-05 海信集团有限公司 一种兴趣点展示方法及装置
CN112507040A (zh) * 2020-12-21 2021-03-16 北京百度网讯科技有限公司 多元关系生成模型的训练方法、装置、电子设备及介质
CN114329244A (zh) * 2021-12-28 2022-04-12 北京百度网讯科技有限公司 地图兴趣点查询方法、装置、设备、存储介质及程序产品

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117370672A (zh) * 2023-12-06 2024-01-09 烟台大学 基于混合结构图的用户兴趣点推荐方法、系统和设备
CN117370672B (zh) * 2023-12-06 2024-02-23 烟台大学 基于混合结构图的用户兴趣点推荐方法、系统和设备

Also Published As

Publication number Publication date
CN114329244A (zh) 2022-04-12

Similar Documents

Publication Publication Date Title
WO2023124005A1 (zh) 地图兴趣点查询方法、装置、设备、存储介质及程序产品
WO2021082397A1 (zh) 获取poi状态信息的方法、装置、设备和计算机存储介质
WO2021232724A1 (zh) 提取地理位置点空间关系的方法、训练提取模型的方法及装置
US20220100963A1 (en) Event extraction from documents with co-reference
US11977574B2 (en) Method and apparatus for acquiring POI state information
US20220100772A1 (en) Context-sensitive linking of entities to private databases
CN114861889B (zh) 深度学习模型的训练方法、目标对象检测方法和装置
CN108228567B (zh) 用于提取组织机构的简称的方法和装置
WO2023168909A1 (zh) 地理预训练模型的预训练方法及模型微调方法
CN112100332A (zh) 词嵌入表示学习方法及装置、文本召回方法及装置
WO2023024975A1 (zh) 文本处理方法、装置和电子设备
WO2022105121A1 (zh) 一种应用于bert模型的蒸馏方法、装置、设备及存储介质
US11170765B2 (en) Contextual multi-channel speech to text
US20240062253A1 (en) Advertisement title rewriting method, apparatus and device, and storage medium
EP4222635A1 (en) Lifecycle management for customized natural language processing
WO2024082827A1 (zh) 文本相似性度量方法、装置、设备、存储介质和程序产品
US20220100967A1 (en) Lifecycle management for customized natural language processing
JP2023517518A (ja) ヌル値又は同等の値を有するリレーショナル・テーブルのためのベクトル埋込モデル
WO2023130960A1 (zh) 服务资源确定方法、装置及服务资源确定系统
WO2023087667A1 (zh) 用于智能推荐的排序模型训练方法、智能推荐方法及装置
CN114328956B (zh) 文本信息的确定方法、装置、电子设备及存储介质
CN113807102B (zh) 建立语义表示模型的方法、装置、设备和计算机存储介质
CN113435523B (zh) 预测内容点击率的方法、装置、电子设备以及存储介质
WO2022271369A1 (en) Training of an object linking model
CN114281990A (zh) 文档分类方法及装置、电子设备和介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22913294

Country of ref document: EP

Kind code of ref document: A1