Information processing method and device, electronic equipment and storage medium

Abstract

The present disclosure provides an information processing method and device, electronic equipment and storage medium, relates to the field of computers, in particular to the field of big data processing of electronic maps. The specific implementation scheme is: determining N semantic variable groups based on a historical query request and N historical points of interest (POIs) associated with the historical query request; N is an integer greater than or equal to 2; determining a correlation result corresponding to the historical query request and the N historical POIs based on the N semantic variable groups and image information corresponding to the N historical POIs respectively; determining a candidate POI from the N historical POIs based on the correlation result corresponding to the historical query request and the N historical POIs respectively; and generating a candidate recall set of candidate key information based on the candidate POI, with the historical query request as the candidate key information. The technical scheme of the present disclosure embodiment can improve the accuracy of the final candidate recall set, and further improve the accuracy of the recall result.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of computers, in particular to the field of big data processing of electronic maps, and in particular to an information processing method and device, electronic equipment and a storage medium. BACKGROUND

[0002] In the process of finding a location in a map application, a query request is obtained, key information corresponding to the query request is determined, a candidate point of interest (POI) set is recalled based on the key information, and finally the recalled candidate POI set is displayed to the user as a recall result. However, in the above process, the accuracy of the final recalled candidate POI set cannot be guaranteed, and thus the accuracy of the recall result cannot be guaranteed. SUMMARY

[0003] The present disclosure provides an information processing method and device, electronic equipment and a storage medium.

[0004] According to a first aspect of the present disclosure, an information processing method is provided, comprising:

[0005] Based on the historical query request and its associated N historical POIs, determine N semantic variable groups; N is an integer greater than or equal to 2;

[0006] Based on the N semantic variable groups and the image information corresponding to the N historical POIs, determine the relevance results of the historical query request and the N historical POIs respectively;

[0007] Based on the relevance results of the historical query request and the N historical POIs respectively, determine a candidate POI from the N historical POIs;

[0008] Take the historical query request as a candidate key information, and generate a candidate recall set of the candidate key information based on the candidate POI.

[0009] According to a second aspect of the present disclosure, an information processing device is provided, comprising:

[0010] The semantic processing module is configured to determine N semantic variable groups based on the historical query request and its associated N historical points of interest (POIs); N is an integer greater than or equal to 2;

[0011] The relevance processing module is configured to determine the relevance results of the historical query request and the N historical POIs respectively based on the N semantic variable groups and the image information corresponding to the N historical POIs;

[0012] A selecting module is configured to determine a candidate POI from the N historical POIs based on the relevance results corresponding to the historical query request and the N historical POIs respectively.

[0013] An indexing module is configured to generate a candidate recall set of the candidate key information based on the candidate POI, with the historical query request as the candidate key information.

[0014] According to a third aspect of the present disclosure, an electronic device is provided, comprising:

[0015] at least one processor; and

[0016] a memory connected with the at least one processor in communication; wherein

[0017] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the foregoing method.

[0018] According to a fourth aspect of the present disclosure, a non-transitory computer readable storage medium storing computer instructions is provided, the computer instructions being used to cause the computer to perform the foregoing method.

[0019] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the foregoing method.

[0020] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present disclosure, nor to limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description.

[0021] The scheme provided by the embodiment is based on a historical query request and a semantic variable group obtained based on an associated historical POI. Based on the semantic variable group and image information of the historical POI, a relevance result between the historical query request and the historical POI is obtained. According to the relevance result between the historical query request and the historical POI, a candidate POI is determined. Then, the historical query request is taken as candidate key information, and a candidate recall set of the candidate key information is generated based on the candidate POI. Thus, through multi-dimensional comprehensive analysis by fusing image information, the accuracy of the relevance result between the historical query request and the historical POI can be improved, the accuracy of the candidate recall set obtained finally can be improved, and the accuracy of the subsequent recall result can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0022] The accompanying drawings serve to better understand the present scheme and do not constitute limitations on the present disclosure. Among them:

[0023] Figure 1is a flowchart of an information processing method according to an embodiment of the present disclosure;

[0024] Figure 2 is a flowchart of obtaining corresponding recall results based on a query request according to another embodiment of the present disclosure;

[0025] Figure 3 is another flowchart of obtaining corresponding recall results based on a query request according to another embodiment of the present disclosure;

[0026] Figure 4 is a flowchart of determining N semantic variable groups according to another embodiment of the present disclosure;

[0027] Figure 5 is a flowchart of obtaining a relevance result according to another embodiment of the present disclosure;

[0028] Figure 6 is another flowchart of obtaining a relevance result according to another embodiment of the present disclosure;

[0029] Figure 7 is an exemplary illustration of obtaining a relevance result according to another embodiment of the present disclosure;

[0030] Figure 8 is a flowchart of corresponding relevance results between historical query information and associated historical POIs according to another embodiment of the present disclosure;

[0031] Figure 9 is a flowchart of different scenarios before and after using the information processing method provided by the present disclosure according to the present disclosure;

[0032] Figure 10 is a flowchart of an information processing method according to another embodiment of the present disclosure;

[0033] Figure 11 is a flowchart of an information processing device according to an embodiment of the present disclosure;

[0034] Figure 12 is another flowchart of an information processing device according to an embodiment of the present disclosure;

[0035] Figure 13 is a block diagram of an electronic device for implementing a request processing method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0036] Exemplary embodiments of the present disclosure are described herein with reference to the accompanying drawings, which are cited as illustrative examples. Various details of the embodiments of the present disclosure are described herein in order to provide a thorough understanding thereof. It will be understood by those of ordinary skill in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Also, in the following description, descriptions of well-known functions and constructions are omitted for clarity and conciseness.

[0037] The first aspect of the present disclosure provides an information processing method, comprising: Figure 1 as shown, comprising:

[0038] S101: determining N semantic variable groups based on a historical query request and N historical POIs (Point of Interest) associated with the historical query request; N is an integer greater than or equal to 2;

[0039] S102: determining a relevance result of the historical query request and the N historical POIs based on the N semantic variable groups and image information corresponding to the N historical POIs, respectively;

[0040] S103: determining a candidate POI from the N historical POIs based on the relevance result of the historical query request and the N historical POIs, respectively;

[0041] S104: generating a candidate recall set of the candidate key information based on the candidate POI, taking the historical query request as the candidate key information.

[0042] The above information processing method can be applied to an electronic device, which can be a server in particular.

[0043] The determination of the N semantic variable groups based on the historical query request and the N historical POIs associated with the historical query request can be: obtaining a first group of variables based on relevant information of the historical query request, and obtaining N second groups of variables based on relevant information of the N historical POIs associated with the historical query request; generating the N semantic variable groups based on the first group of variables and the N second groups of variables.

[0044] The relevant information of the historical query request can include the historical query request, region information corresponding to the historical query request, and relevant information of a query geographic location corresponding to the historical query request. The relevant information of each historical POI in the N historical POIs associated with the historical query request can include the name of the historical POI, region information where the historical POI is located, and geographic location relevant information of the historical POI.

[0045] The image information corresponding to the N historical POIs respectively can be specifically the satellite image information corresponding to each historical POI in the N historical POIs. The acquisition manner of the satellite image information corresponding to each historical POI is not limited in the embodiment.

[0046] The historical query request can be specifically included as the candidate key information, or the specific content included in the historical query request and the area information corresponding to the historical query request can be included as the candidate key information.

[0047] The candidate POI can be sorted based on the correlation result corresponding to the candidate POI to obtain a sorted candidate POI, and the sorted candidate POI can be included as the candidate recall set of the candidate key information.

[0048] It can be seen that by using the above scheme, the semantic variable group is obtained based on the historical query request and the historical POI associated therewith, the correlation result between the historical query request and the historical POI is obtained based on the semantic variable group and the image information of the historical POI, the candidate POI is determined according to the correlation result between the historical query request and the historical POI, the historical query request is included as the candidate key information, and the candidate recall set of the candidate key information is generated based on the candidate POI. Therefore, by fusing the image information for multi-dimensional comprehensive analysis, the accuracy of the correlation result between the historical query request and the historical POI can be improved, the accuracy of the candidate recall set obtained finally can be improved, and the accuracy of the subsequent recall result can be improved.

[0049] On the basis of the foregoing embodiment, the acquisition manner of the foregoing historical query request and the N historical POIs associated therewith is further described.

[0050] The foregoing historical query request and the N historical POIs associated therewith can be generated based on historical logs. The acquisition manner of the historical query request and any one of the historical POIs associated therewith can include: acquiring interaction information of a historical operation behavior, including the query request in the interaction information of the historical operation behavior as the historical query request; and including the POI selected by a user in the interaction information of the historical operation behavior as one historical POI associated with the historical query request. The historical operation behavior can be a query operation behavior of the user, and the POI selected by the user can be a POI selected by the user from the recall result corresponding to the query request, the recall result including one or more POIs.

[0051] In the process of generating the interaction information of the historical operation behavior, the following can be included: obtaining a corresponding recall result based on the received query request, sending the recall result to a user device, and receiving a POI selected by a user fed back by the user device.

[0052] One way of obtaining a corresponding recall result based on a query request is shown in combination with Figure 2 The following can be included:

[0053] S201: receiving a first query request.

[0054] For example, as shown in Figure 2 The first query request is Figure 2 “Word A” as shown in

[0055] S202: obtaining a first recall result. Specifically, one or more POIs with the first query request as a key value can be extracted from an inverted index, and the one or more POIs are taken as the first recall result.

[0056] For example, referring to Figure 2 The first recall result contains five POIs with “Word A” as a key value, and the five POIs include “Place 1, Place 2, Place 3, Place 4, and Place 5”. The five POIs “Place 1, Place 2, Place 3, Place 4, and Place 5” can be places in different cities, such as Place 1 to Place 3 being POIs in City 1 and Place 4 to Place 5 being POIs in City 2.

[0057] S203: sorting the first recall result to obtain a sorted first recall result. The sorted first recall result contains one or more sorted POIs, such as Figure 2 “Place 2” as shown in

[0058] Another way of obtaining a corresponding recall result based on a query request is shown in combination with Figure 3 The following can be included:

[0059] S301: receiving a second query request.

[0060] The second query request can include specific content contained in the second query request and area information corresponding to the second query request.

[0061] The specific content contained in the second query request can be multiple characters. At least one character in the multiple characters can be a process character generated by a five-stroke input method, for example, as shown inFigure 3 As shown, the specific content contained in the second query request is "Senmu", wherein "mu" is a process character generated by the five-stroke input method of "lin".

[0062] The area information corresponding to the second query request can be information of an area where a terminal device used by the user is located, such as Figure 3 As shown, the name of the city, City A, can be included. Further, the area information corresponding to the second query request can also include the location coordinates of the location of the terminal device used by the user, such as Figure 3 As shown, the location coordinates are "longitude a, latitude a".

[0063] S302: Obtain a second recall result.

[0064] Specifically, one or more POIs with the second query request as the key value can be obtained, and the one or more POIs are taken as the second recall result. For example, referring to Figure 3 The second recall result contains three POIs with "Senmu-City A" as the key value, and the three POIs are "Location 1, Location 2, and Location 3".

[0065] S303: Sort the second recall result to obtain a sorted second recall result.

[0066] The sorted second recall result can contain one or more sorted POIs, such as Figure 3 As shown in the above, "Location 3" is the highest in the sorting, and "Location 1" and "Location 2" are in the following order. It should be understood that the sorted second recall result can be sent to the terminal device on the user side for display.

[0067] By using the above method, the historical query request and the N associated historical POIs can be obtained. On the basis of obtaining the historical query request and the N associated historical POIs, the processing of S101-S104 can be performed according to actual needs to obtain a more accurate candidate recall set with the historical query request as the candidate key information.

[0068] In an embodiment, the N semantic variable groups are determined based on the historical query request and the N associated historical POIs, such as Figure 4 As shown, the N semantic variable groups can include:

[0069] S401: Obtain first geographic coding information based on the query geographic location of the historical query request, and obtain the i-th second geographic coding information based on the geographic location of the i-th historical POI associated with the historical query request; i is an integer greater than or equal to 1 and less than or equal to N.

[0070] S402: generate a first group of variables based on the first geographic coding information, and generate an i-th second group of variables based on the i-th second geographic coding information;

[0071] S403: obtain an i-th semantic variable group in the N semantic variable groups based on the first group of variables and the i-th second group of variables.

[0072] The first geographic coding information obtained based on the query geographic location of the historical query request can specifically include: encoding the query geographic location of the historical query request based on a first encoding mode to obtain the first geographic coding information. The first encoding mode can be set according to actual conditions. For example, the first encoding mode can be a GeoHash encoding mode.

[0073] The query geographic location of the historical query request can specifically refer to the position coordinates of the location where the user device is located when the historical query request is sent. The position coordinates can specifically include longitude and latitude. The user device can refer to a terminal device used by a user, such as a smartphone, a tablet computer, a desktop computer, a notebook computer, or any one of the foregoing. The geographic location where the user device is located when the historical query request is sent can be obtained by the user device based on a sensor set by the user device. The sensor can be a Global Positioning System (GPS) sensor or other types of sensors as long as the geographic location where the user device is located can be obtained.

[0074] The i-th second geographic coding information obtained based on the geographic location of the i-th historical POI associated with the historical query request can specifically include: encoding the geographic location of the i-th historical POI associated with the historical query request based on a second encoding mode to obtain the i-th second geographic coding information. The geographic location of the i-th historical POI can refer to the position coordinates of the i-th historical POI, such as the longitude and latitude of the i-th historical POI. The second encoding mode can be the same as or different from the first encoding mode, and the second encoding mode can also be set according to actual conditions. In an example, the second encoding mode is the same as the first encoding mode, which can also be a GeoHash encoding mode.

[0075] The i-th historical POI in the N historical POIs can be any one of the N historical POIs. Since the specific processing manner for the N historical POIs is the same, the processing manner of the N historical POIs will not be described one by one.

[0076] The generating the first set of variables based on the first geographic coding information and the i-th second set of variables based on the i-th second geographic coding information can include: generating the first set of variables based on the related information of the historical query request and the first geographic coding information; and generating the i-th second set of variables based on the related information of the i-th historical POI and the i-th second set of variables.

[0077] The related information of the historical query request can include at least one of the following: specific content contained in the historical query request, and area information corresponding to the historical query request. The area information corresponding to the historical query request can specifically refer to the area information where the user equipment is located when sending the historical query request. Here, the area information can specifically include the name of the area; the name of the area can refer to the name of the city, such as A1 city, or the name of the administrative region in the city, such as B1 district of A1 city, and the like.

[0078] Correspondingly, the generating the first set of variables based on the related information of the historical query request and the first geographic coding information can include: taking the specific content contained in the historical query request, the area information corresponding to the historical query request, and the first geographic coding information as the first set of variables.

[0079] The i-th historical POI can refer to a certain position on the map, such as a house, a shop, a mailbox, a bus stop, and the like. The related information of the i-th historical POI can include the name of the i-th historical POI, the area information where the i-th historical POI is located, the geographic position of the i-th historical POI, and the like. The name of the i-th historical POI can refer to the name of the corresponding geographic position at the position where the i-th historical POI is located, such as C1 mansion, C2 mall, and the like. The area information where the i-th historical POI is located can be the name of the area range where the i-th historical POI is located, such as the name of the city, or the name of the administrative region in the city, or the name of the street in the city, and the like. It should be understood that the i-th historical POI is any one of the N historical POIs, that is, any one of the N historical POIs can include the above content, but will not be described one by one.

[0080] Correspondingly, the generating the i-th second group of variables based on the i-th historical POI-related information and the i-th second group of variables can include: taking the name of the i-th historical POI, the area information where the i-th historical POI is located, and the i-th second geographic coding information as the i-th second group of variables.

[0081] The obtaining the i-th semantic variable group in the N semantic variable groups based on the first group of variables and the i-th second group of variables can include: taking the first group of variables and the i-th second group of variables as the i-th semantic variable group in the N semantic variable groups.

[0082] It can be seen that by using the above scheme, the first geographic coding information is obtained by encoding the geographic location of the historical query request, the i-th second geographic coding information is obtained by encoding the geographic location of the i-th historical POI, and the i-th semantic variable group is obtained based on the first geographic coding information and the i-th second geographic coding information. In this way, the geographic location is more accurately obtained by using geographic location coding information, so that the semantic variable group obtained subsequently is also more accurate, and the accuracy of the correlation result obtained by analyzing the semantic variable group and the image information corresponding to the N historical POIs is ensured.

[0083] The determining the correlation result of the historical query request and the N historical POIs based on the N semantic variable groups and the image information corresponding to the N historical POIs can have two processing modes, which will be described below.

[0084] In one embodiment, the determining the correlation result of the historical query request and the N historical POIs based on the N semantic variable groups and the image information corresponding to the N historical POIs includes: Figure 5

[0085] S501: generating i-th first input information based on the i-th semantic variable group in the N semantic variable groups and the image information corresponding to the i-th historical POI in the N historical POIs;

[0086] S502: inputting the i-th first input information into the first model to obtain an i-th first output result of the first model;

[0087] S503: determining the correlation result of the historical query request and the i-th historical POI in the N historical POIs based on the i-th first output result.

[0088] ​The first input information can be generated based on the i-th semantic variable group in the N semantic variable groups and image information corresponding to the i-th historical POI in the N historical POIs. The image information corresponding to the i-th historical POI can be satellite image information (or satellite image data) corresponding to the i-th historical POI. The satellite image information (or satellite image data) can be obtained from a satellite communication server or by using other methods.

[0089] Alternatively, the first input information can be generated based on the i-th semantic variable group in the N semantic variable groups and image information corresponding to the i-th historical POI in the N historical POIs. The image information corresponding to the i-th historical POI can be satellite image information (or satellite image data) corresponding to the i-th historical POI. The satellite image information (or satellite image data) can be obtained from a satellite communication server or by using other methods.

[0090] The image information corresponding to the i-th historical POI can be satellite image information (or satellite image data) corresponding to the i-th historical POI. The satellite image information (or satellite image data) can be obtained from a satellite communication server or by using other methods.

[0091] The first model can be a pre-trained model, and the first model is used to obtain a probability value. In an example, the first model can be a model trained in the electronic device performing the information processing method provided in the embodiment. In another example, the first model can be a model trained in another electronic device. In this example, the first model can be pre-acquired by the electronic device performing the information processing method provided in the embodiment from the other electronic device. The other electronic device can refer to another server used to train the first model.

[0092] The determining the relevance result of the historical query request and the i-th historical POI in the N historical POIs based on the i-th first output result can include: taking a probability value contained in the i-th first output result as the relevance result of the historical query request and the i-th historical POI in the N historical POIs.

[0093] It can be seen that by using the above scheme, the i-th semantic variable group and the image information corresponding to the i-th historical POI can be directly used as input information, and the input information can be input into the first model to obtain the relevance result of the historical query request and the i-th historical POI in the N historical POIs. In this way, only a small amount of preprocessing is required to obtain the input information of the first model, and the relevance result can be directly obtained through the first model, which can improve the processing efficiency while ensuring the accuracy of the final relevance result.

[0094] In another embodiment, the determining the relevance result of the historical query request and the N historical POIs based on the N semantic variable groups and the image information corresponding to the N historical POIs can include: Figure 6

[0095] S601: determining i-th semantic vector information based on an i-th semantic variable group in the N semantic variable groups, and determining an i-th image feature vector based on image information corresponding to an i-th historical POI in the N historical POIs;

[0096] S602: obtaining an i-th multi-modal vector based on the i-th semantic vector information and the i-th image feature vector;

[0097] S603: generating the relevance result of the historical query request and the i-th historical POI in the N historical POIs based on the i-th multi-modal vector.

[0098] The determining the i-th semantic vector information based on the i-th semantic variable group in the N semantic variable groups can refer to directly taking the i-th semantic variable group in the N semantic variable groups as the i-th semantic vector information.

[0099] Alternatively, the determining the i-th semantic vector information based on the i-th semantic variable group in the N semantic variable groups can include: inputting the i-th semantic variable group in the N semantic variable groups into a second model to obtain the i-th semantic vector information output by the second model.

[0100] ​More specifically, the i-th semantic variable group in the N semantic variable groups is input into a second model to obtain the i-th semantic vector information output by the second model, which can be:

[0101] The i-th semantic variable group in the N semantic variable groups is taken as i-th second input information, the i-th second input information is input into a second model to obtain i-th second output result output by the second model, and the i-th semantic vector information is determined based on the i-th second output result.

[0102] The i-th semantic variable group in the N semantic variable groups can be taken as i-th second input information, which means that all variables in the i-th semantic variable group in the N semantic variable groups are directly taken as i-th second input information.

[0103] The i-th second output result can include one or more vector information, and accordingly, the i-th semantic vector information is determined based on the i-th second output result, which can include taking the one or more vector information included in the i-th second output result as the i-th semantic vector information.

[0104] The second model can be ERNIE (Enhanced Representation through Knowledge Integration, semantic representation model enhanced by knowledge). The input of the second model is any one of the N semantic variable groups, and the output of the second model is the semantic vector information corresponding to the any one semantic variable group. The second model can be a pre-trained model. In an example, the second model can be a model trained in the electronic device performing the information processing method provided in the embodiment. In another example, the second model can be a model trained in another electronic device. In this example, the second model can be pre-acquired by the electronic device performing the information processing method provided in the embodiment from the other electronic device, and the other electronic device can be another server used to train the second model.

[0105] The i-th image feature vector can be determined based on the image information corresponding to the i-th historical POI in the N historical POIs by inputting the image information corresponding to the i-th historical POI in the N historical POIs into a third model to obtain the i-th image feature vector output by the third model.

[0106] The image information corresponding to the i-th historical POI in the N historical POIs can be directly taken as the i-th third input information, and the i-th third input information is input into the third model.

[0107] The i-th image feature vector output by the third model can be any one of the following two processing modes:

[0108] In a first processing mode, an image semantic vector included in the i-th third output result output by the third model is directly taken as the i-th image feature vector.

[0109] In a second processing mode, an image semantic vector included in the i-th third output result output by the third model is input into a full connection layer to obtain a processed image semantic vector output by the full connection layer; and the processed image semantic vector is taken as the i-th image feature vector. The full connection layer can be used to perform weighted summation calculation on the image semantic vector included in the i-th third output result to obtain the processed image semantic vector. In the second processing mode, the image semantic vector of the i-th third output result output by the third model is further processed by the full connection layer, so that the processed image semantic vector is more accurate.

[0110] The third model can be a convolutional neural network model (CNN) model, a deep neural network model (DNN), a region proposal network (RPN), or the like, which will not be listed exhaustively. The input of the third model is the image information corresponding to any one of the N historical POIs, and the output of the third model is an image feature vector corresponding to the any one of the historical POIs, which can also be referred to as an image semantic vector. The third model can be a pre-trained model. In an example, the third model can be a model trained in the electronic device performing the information processing method provided in the embodiment. In another example, the third model can be a model trained in another electronic device; in this example, the third model can be pre-acquired by the electronic device performing the information processing method provided in the embodiment from the other electronic device; wherein the other electronic device can refer to another server used to train the third model.

[0111] In the above embodiments, the i-th semantic vector information is obtained by processing the i-th semantic variable group in the N semantic variable groups using the second model, and the i-th image feature vector is obtained by processing image information corresponding to the i-th historical POI in the N historical POIs using the third model, so that the semantic vector information corresponding to the historical query information and the image feature vector of the image information corresponding to the historical POI are more accurate and the processing is more efficient. Furthermore, based on the more efficient and accurate processing, the processing of the relevance result of the historical query request and the historical POI is also more efficient and accurate.

[0112] The i-th group of multi-modal vectors is obtained based on the i-th semantic vector information and the i-th image feature vector, which can include fusing the i-th semantic vector information and the i-th image feature vector to obtain the i-th group of multi-modal vectors. The fusion of the i-th semantic vector information and the i-th image feature vector can be realized by a fusion layer.

[0113] Specifically, the fusion of the i-th semantic vector information and the i-th image feature vector to obtain the i-th group of multi-modal vectors can include one of the following: splicing the i-th semantic vector information and the i-th image feature vector to obtain the i-th group of multi-modal vectors; or adding the i-th semantic vector information and the i-th image feature vector to obtain the i-th group of multi-modal vectors.

[0114] Here, the processing mode of splicing the i-th semantic vector information and the i-th image feature vector to obtain the i-th group of multi-modal vectors is particularly suitable for scenarios where the i-th semantic vector information and the i-th image feature vector contain different numbers of vectors. For example, the i-th semantic vector information contains C1 text semantic vectors, the i-th image feature vector contains C2 image semantic vectors, and C1 and C2 are different. By splicing the i-th semantic vector information and the i-th image feature vector, the i-th group of multi-modal vectors ultimately obtained can contain C1+C2 vectors.

[0115] The processing manner of adding the i-th semantic vector information and the i-th image feature vector to obtain the i-th multi-modal vector is especially suitable for a scenario where the i-th semantic vector information and the i-th image feature vector contain the same number of vectors, for example, the i-th semantic vector information contains C3 text semantic vectors, and the i-th image feature vector also contains C3 image semantic vectors. By adding the i-th semantic vector information and the i-th image feature vector, the i-th multi-modal vector obtained finally also contains C3 vectors.

[0116] It should be noted that the aforementioned adding the i-th semantic vector information and the i-th image feature vector can be adding each text semantic vector in the i-th semantic vector information and each corresponding image semantic vector in the i-th image feature vector respectively. For example, the i-th semantic vector information contains two text semantic vectors, which are referred to as text semantic vector 1 and text semantic vector 2 respectively, and the i-th image feature vector contains two image semantic vectors, which are referred to as image semantic vector 1 and image semantic vector 2 respectively. Adding the i-th semantic vector information and the i-th image feature vector means adding the text semantic vector 1 and the image semantic vector 1 to obtain a multi-modal vector 1 in the i-th multi-modal vector, and adding the text semantic vector 2 and the image semantic vector 2 to obtain a multi-modal vector 2 in the i-th multi-modal vector.

[0117] The generating, based on the i-th multi-modal vector, of the relevance result of the historical query request and the i-th historical POI in the N historical POIs can be: obtaining a target value based on the i-th multi-modal vector; and taking the target value as the relevance result of the historical query request and the i-th historical POI in the N historical POIs.

[0118] The obtaining, based on the i-th multi-modal vector, of the target value can mean: in the case where the i-th multi-modal vector contains one multi-modal vector, directly taking the one multi-modal vector as the target value; or in the case where the i-th multi-modal vector contains multiple multi-modal vectors, performing weighted addition calculation on the multiple multi-modal vectors to obtain the target value. In the processing of the weighted addition calculation on the multiple multi-modal vectors, the weighting values of different multi-modal vectors can be preset, which is not limited herein.

[0119] In combination with Figure 7 The processing of determining the relevance result of the historical query request and the N historical POIs based on the N semantic variable groups and the image information corresponding to the N historical POIs is described exemplarily as follows:

[0120] The specific content 711 contained in the historical query request, the area information 712 corresponding to the historical query request (such as a city name), and the first geographic coding information 713 are taken as the first group of variables 710; the name 721 of the i-th historical POI, the area information 722 (such as a city name) where the i-th historical POI is located, and the i-th second geographic coding information 723 are taken as the i-th second group of variables 720.

[0121] The first group of variables 710 and the i-th second group of variables 720 are taken as the i-th semantic variable group in the N semantic variable groups;

[0122] The i-th semantic variable group is input into the second model 730 to obtain the i-th semantic vector information 741 output by the second model 730;

[0123] The image information corresponding to the i-th historical POI is input into the third model to obtain an i-th third output result output by the third model, the image semantic vector 750 contained in the i-th third output result output by the third model is input into the full connection layer 760 to obtain a processed image semantic vector output by the full connection layer 760, and the processed image semantic vector is taken as the i-th image feature vector 742;

[0124] Based on the i-th semantic vector information 741 and the i-th image feature vector 742, an i-th multi-modal vector 770 is obtained, and based on the i-th multi-modal vector 770, the relevance result of the historical query request and the i-th historical POI is generated.

[0125] It can be seen that by using the above scheme, the i-th semantic variable group based on the historical query information and the i-th historical POI is processed to obtain the i-th semantic vector information, and the i-th image feature vector is obtained based on the image information corresponding to the i-th historical POI; then the i-th multi-modal vector obtained from the i-th semantic vector information and the i-th image feature vector is used to determine the relevance result of the historical query request and the i-th historical POI in the N historical POIs. In this way, by extracting the semantic vector and extracting the image feature, the information can be more accurate, and the relevance result obtained finally can also be more accurate.

[0126] In an implementation manner, the determining the candidate POI from the N historical POIs based on the relevance result of the historical query request and the N historical POIs respectively can include:

[0127] acquiring a relevance result corresponding to the jth historical POI from the historical query request, where j is an integer greater than or equal to 1 and less than or equal to N;

[0128] In a case where the relevance result corresponding to the jth historical POI from the historical query request is greater than a preset threshold value, the jth historical POI is taken as one of the candidate POIs.

[0129] The jth historical POI refers to any one of the N historical POIs. The jth historical POI can be the same as the ith historical POI, or the jth historical POI can be different from the ith historical POI. It should be understood that, regardless of whether the jth historical POI is the same as or different from the ith historical POI, the determination manner of the relevance result corresponding to the jth historical POI from the historical query request is the same as the determination manner of the relevance result corresponding to the ith historical POI from the historical query request, and the present embodiment does not repeat the same.

[0130] The preset threshold value can be set according to actual conditions.

[0131] In addition, the method can further include: in a case where the relevance result corresponding to the jth historical POI from the historical query request is less than or equal to the preset threshold value, the jth historical POI is not taken as the candidate POI.

[0132] The number of candidate POIs can be one or more, and the number of candidate POIs is represented as M, where M can be less than or equal to N. In an example, M can be equal to N, that is, the relevance results corresponding to the N historical POIs from the historical query request are all greater than the preset threshold value, and the N historical POIs can all be taken as the candidate POIs. In another example, M can be zero, that is, the relevance results corresponding to the N historical POIs from the historical query request are all less than or equal to the preset threshold value, and the N historical POIs are not taken as the candidate POIs. In another example, M is a positive integer less than N, that is, among the relevance results corresponding to the N historical POIs from the historical query request, a part of the relevance results are greater than the preset threshold value, and the part of the historical POIs (i.e., M historical POIs) are taken as the candidate POIs.

[0133] In combination Figure 8 Exemplary description is performed, in which two historical query information are shown, which are "word 1-city A" and "word 2-city A". The historical POI corresponding to "word 1-city A" and the relevance result are as follows:Figure 8 As shown, respectively, are location A1, 0.997, location A2, 0.885, and location A3, 0.854. The corresponding historical POI and correlation results of "Word 2-City A" are as shown in the following table: Figure 8 As shown, respectively, are location A4, 0.988, and location A5, 0.823. If the preset threshold value is set to 0.86, "Word 1-City A" is a candidate key information, and the corresponding candidate POI is location A1 and location A2. "Word 1-City A" is a candidate key information, and the corresponding candidate POI is location A4.

[0134] In the scheme provided by the foregoing embodiments, by using the image information corresponding to the N historical POIs respectively and the N semantic variable groups of the N historical POIs to determine the correlation results, the determination manner of the correlation between the historical query request and the N historical POIs can be upgraded from the original single-mode understanding to multi-mode understanding, so that more accurate correlation results can be obtained.

[0135] As can be seen, by using the above scheme, the N historical POIs can be filtered based on the foregoing correlation results, so as to ensure the accuracy of the final candidate recall set, and make the target recall set obtained in response to the target query request more accurate.

[0136] For the foregoing various embodiments, it is also necessary to point out that the number of the foregoing historical query requests can be one or more, and the same applies to the candidate key information, which can also be one or more, without limiting the number. The foregoing various embodiments are directed to the implementation manner that each historical query request in one or more historical query requests can be executed, and finally each historical query request can be taken as the candidate key information to obtain the candidate recall set corresponding to each candidate key information.

[0137] In an embodiment, it can also include receiving a target query request sent by a target device, searching for target key information matching the target query request from the candidate key information, and obtaining the candidate recall set corresponding to the target key information; generating a target recall result based on the candidate recall set, and sending the target recall result to the target device.

[0138] The target device can be any terminal device (or user device) used by a user, which can be any one of a smart phone, a tablet computer, a notebook computer, a desktop computer, and the like. For example, when a user needs to find a location, the user can open a map application on a target device used by the user, and input query information in a search box of the map application; the target device takes the query information in the search box as the target query request.

[0139] It has been described that the number of the candidate key information can be one or more. The searching for the target key information from the candidate key information can refer to matching the target query request with one or more of the candidate key information respectively, and taking one of the candidate key information matched with the target query request as the target key information.

[0140] Alternatively, the searching for the target key information from the candidate key information can refer to matching the specific content contained in the target query request and the area information corresponding to the target query request with one or more of the candidate key information respectively, and taking one of the candidate key information matched with the target query request as the target key information. This is because any one of the candidate key information can include the specific content contained in the candidate key information and the area information corresponding thereto, and accordingly, the aforementioned processing can be based on the matching of the specific content contained in the target query request and the area information corresponding to the target query request with the specific content contained in the candidate key information and the area information corresponding thereto. In this way, the target key information matched finally can be more accurate.

[0141] The generating of the target recall result based on the candidate recall set and the sending of the target recall result to the target device can include taking the candidate POI contained in the candidate recall set as a target POI, and obtaining the target recall result based on the target POI.

[0142] The obtaining of the target recall result based on the target POI can include two exemplary processing modes, respectively.

[0143] In one exemplary processing mode, the obtaining of the target recall result based on the target POI can include sorting the target POI based on the relevance result corresponding to the target POI to obtain a sorted target POI, and taking the sorted target POI as the target recall result.

[0144] In another example processing manner, the method can further include: obtaining, based on the target query request, an other recall set from an other data set; the other data set can be constructed in another manner, can include other candidate key information and an other candidate recall set associated therewith, and the other candidate recall set can include other POIs and corresponding relevance results thereof, and the embodiments are not limited thereto. Correspondingly, the obtaining, based on the target POI, of the target recall result can include: obtaining the relevance results corresponding to the other POIs in the other recall set and the relevance result corresponding to the target POI; sorting the other POIs and the target POI based on the relevance results corresponding to the other POIs and the relevance result corresponding to the target POI to obtain a sorted POI set; and taking the sorted POI set as the target recall result.

[0145] In combination Figure 9 With the information processing method provided in the embodiments, there can be a large number of incorrect locations in the candidate key information and the candidate recall set associated therewith constructed in processing without using the information processing method, and thus there can be a large number of incorrect locations in the corresponding recall result obtained in response to a target query request. For example Figure 9 As shown on the left side, when the target query request A is received, the recall result matched with the target query request A includes "Location B1, region b1 of City 01", "Location B2, region b2 of City 03", and "Location B3, region b3 of City 02", and it can be seen that multiple locations in multiple cities are included, and when a user performs a search for a location, the user usually does not need to obtain different POIs in multiple cities, that is, the above recall result includes a large amount of impurities, and the accuracy of the recall result is low. After the information processing method is used, the image feature vector is obtained based on the image information of the historical POI, the semantic vector information is obtained by processing the historical POI and the historical query information, and the more accurate relevance result between the historical query information and the historical POI is obtained by processing based on the image feature vector and the semantic vector information, and thus a more accurate recall result can be obtained. For example Figure 9 As shown on the right side, when the target query request A is received, the target recall result matched with the target query request A includes "Location B1, region b1 of City 01", "Location B4, region b1 of City 01", and "Location B5, region b2 of City 01", and it can be seen that the locations in the same region of a city are included, and compared with Figure 9 the recall result on the left side, the recall result is more accurate.

[0146] It can be seen that, by adopting the above scheme, the target key information matched with the received target query request can be determined based on the candidate key information, and the corresponding target recall result can be generated according to the candidate recall set corresponding to the target key information, and then the target recall result is sent to the target device. Since the accuracy of the candidate recall set of the candidate key information is improved, the accuracy of the final target recall result fed back for the target query request is improved.

[0147] Finally, taking the electronic device executing the foregoing information processing method as an example, the combination of Figure 10 The foregoing information processing method is exemplarily described:

[0148] S1011: The server acquires a historical query request and N historical points of interest POIs associated therewith;

[0149] Here,

[0150] S1012: The server determines a relevance result corresponding to the historical query request and the N historical POIs respectively.

[0151] Specifically, N semantic variable groups can be determined based on the historical query request and the N historical points of interest POIs associated therewith, and the relevance result corresponding to the historical query request and the N historical POIs respectively can be determined based on the N semantic variable groups and image information corresponding to the N historical POIs respectively.

[0152] S1013: The server determines a candidate POI from the N historical POIs based on the relevance result corresponding to the historical query request and the N historical POIs respectively.

[0153] S1014: The server generates a candidate recall set of the candidate key information based on the candidate POI, taking the historical query request as the candidate key information.

[0154] S1021: The server receives a target query request sent by a target device;

[0155] S1022: The server obtains a candidate recall set;

[0156] Specifically, the target key information matched with the target query request can be found from the candidate key information obtained from S1014, and the candidate recall set corresponding to the target key information is acquired.

[0157] In addition, other recall sets can be acquired from other data sets based on the target query request.

[0158] S1023: The server sends the sorted POI set as the target recall result.

[0159] Specifically, the method can comprise: sorting the other POIs and the target POI based on the relevance results corresponding to the other POIs and the relevance result corresponding to the target POI to obtain a sorted POI set; and sending the sorted POI set as the target recall result.

[0160] After obtaining the foregoing recall result, the server can send the target recall result to the target device.

[0161] S1024: The target device acquires and presents the target recall result.

[0162] The embodiments of the present disclosure further provide an information processing apparatus, as shown in the accompanying drawings, comprising: Figure 11

[0163] The semantic processing module 1101 is configured to determine N semantic variable groups based on a historical query request and N historical points of interest (POIs) associated with the historical query request; N is an integer greater than or equal to 2.

[0164] The relevance processing module 1102 is configured to determine relevance results of the historical query request and the N historical POIs based on the N semantic variable groups and image information corresponding to the N historical POIs.

[0165] The selection module 1103 is configured to determine a candidate POI from the N historical POIs based on the relevance results of the historical query request and the N historical POIs.

[0166] The index module 1104 is configured to generate a candidate recall set of candidate key information based on the candidate POI, with the historical query request as the candidate key information.

[0167] The semantic processing module 1101 is configured to obtain first geocoding information based on a query geographic location of the historical query request, obtain an i th second geocoding information based on a geographic location of an i th historical POI in the N historical POIs associated with the historical query request; i is an integer greater than or equal to 1 and less than or equal to N; generate a first group of variables based on the first geocoding information, generate an i th second group of variables based on the i th second geocoding information; and obtain an i th semantic variable group in the N semantic variable groups based on the first group of variables and the i th second group of variables.

[0168] ​The correlation processing module 1102 is configured to generate i th first input information based on the i th semantic variable group in the N semantic variable groups and image information corresponding to the i th historical POI in the N historical POIs; input the i th first input information into a first model to obtain an i th first output result of the first model; and determine the correlation result of the historical query request and the i th historical POI in the N historical POIs based on the i th first output result.

[0169] The correlation processing module 1102 is configured to determine i th semantic vector information based on an i th semantic variable group in the N semantic variable groups, determine an i th image feature vector based on image information corresponding to the i th historical POI in the N historical POIs, obtain an i th multi-modal vector based on the i th semantic vector information and the i th image feature vector, and generate the correlation result of the historical query request and the i th historical POI in the N historical POIs based on the i th multi-modal vector.

[0170] The correlation processing module 1102 is configured to input the i th semantic variable group in the N semantic variable groups into a second model to obtain the i th semantic vector information output by the second model.

[0171] The correlation processing module 1102 is configured to input the image information corresponding to the i th historical POI in the N historical POIs into a third model to obtain the i th image feature vector output by the third model.

[0172] The selection module 1103 is configured to obtain a correlation result corresponding to the j th historical POI in the historical query request and the N historical POIs from the correlation results corresponding to the historical query request and the N historical POIs respectively; j is an integer greater than or equal to 1 and less than or equal to N; and in a case where the correlation result corresponding to the historical query request and the j th historical POI is greater than a preset threshold value, the j th historical POI is taken as one of the candidate POIs.

[0173] Further, as shown in Figure 12 The device further includes:

[0174] The receiving module 1201 is configured to receive a target query request sent by a target device.

[0175] The query module 1202 is configured to find target key information matching the target query request from the candidate key information, and obtain a candidate recall set corresponding to the target key information.

[0176] The sending module 1203 is configured to generate a target recall result based on the candidate recall set, and send the target recall result to the target device.

[0177] In the technical solutions of the present disclosure, the acquisition, storage and application of user personal information are in line with relevant laws and regulations and do not violate public order and good customs.

[0178] According to embodiments of the present disclosure, the present disclosure further provides an electronic device, a readable storage medium and a computer program product.

[0179] Figure 13 A schematic block diagram of an example electronic device 1300 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, tablets, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the present disclosure described and / or claimed in this document.

[0180] As shown in Figure 13 The electronic device 1300 includes a computing unit 1301 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 1302 or a computer program loaded into a random access memory (RAM) 1303 from a storage unit 1308. Various programs and data required for the operation of the electronic device 1300 can also be stored in the RAM 1303. The computing unit 1301, the ROM 1302, and the RAM 1303 are connected to each other through a bus 1304. An input / output (I / O) interface 1305 is also connected to the bus 1304.

[0181] Various components in the electronic device 1300 are connected to the I / O interface 1305, including an input unit 1306, such as a keyboard, a mouse, etc., an output unit 1307, such as various types of displays, a speaker, etc., a storage unit 1308, such as a magnetic disk, an optical disk, etc., and a communication unit 1309, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1309 allows the electronic device 1300 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0182] The computing unit 1301 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 1301 performs various methods and processes described above. For example, in some embodiments, the various methods described above can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 1308. In some embodiments, portions or all of the computer program can be loaded and / or installed onto the electronic device 1300 via the ROM 1302 and / or the communication unit 1309. When the computer program is loaded onto the RAM 1303 and executed by the computing unit 1301, one or more steps of the various methods described above can be performed. Alternatively, in other embodiments, the computing unit 1301 can be configured to perform the various methods described above by other any suitable means, such as by means of firmware.

[0183] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a programmable logic device (PLD), a computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0184] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces the functions / operations specified in the flowcharts and / or the block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package, partially on a machine and partially on a remote machine or entirely on a remote machine or server.

[0185] In the context of this disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0186] 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 the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0187] The systems and techniques described here can be implemented in a computing system that includes a back end 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 back end, 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 a local area network (LAN), a wide area network (WAN), and the Internet.

[0188] The computer system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0189] It should be understood that the various forms of flow shown above can be used to reorder, add, or delete steps. For example, the steps described in the present disclosure can be performed in parallel, in series, or in a different order, as long as the desired results of the technology disclosed in the present disclosure can be achieved, which is not limited herein.

[0190] The above detailed description does not constitute a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. An information processing method comprising: determining N semantic variable groups based on a historical query request and N historical points of interest (POIs) associated therewith; N being an integer greater than or equal to 2; determining relevance results of the historical query request and the N historical POIs based on the N semantic variable groups and image information corresponding to the N historical POIs, respectively; determining a candidate POI from the N historical POIs based on the relevance results of the historical query request and the N historical POIs, respectively; generating a candidate recall set of candidate key information based on the candidate POI, with the historical query request as the candidate key information; wherein the determining of the relevance results of the historical query request and the N historical POIs based on the N semantic variable groups and the image information corresponding to the N historical POIs, respectively, comprises: adding an i-th semantic vector information and an i-th image feature vector to obtain a plurality of multi-modal vectors in an i-th multi-modal vector group, wherein the i-th semantic vector information is determined based on an i-th semantic variable group in the N semantic variable groups, the i-th image feature vector is determined based on image information corresponding to an i-th historical POI in the N historical POIs, i is an integer greater than or equal to 1 and less than or equal to N; and performing weighted addition calculation on the plurality of multi-modal vectors in the i-th multi-modal vector group to obtain a target value; and taking the target value as the relevance result of the historical query request and the i-th historical POI in the N historical POIs.

2. The method of claim 1, wherein, The determining of the N semantic variable groups based on the historical query request and the N historical POIs associated therewith comprises: obtaining first geocoding information based on a query geographic location of the historical query request, and obtaining an i-th second geocoding information based on a geographic location of an i-th historical POI in the N historical POIs associated with the historical query request; generating a first group of variables based on the first geocoding information, and generating an i-th second group of variables based on the i-th second geocoding information; obtaining an i-th semantic variable group in the N semantic variable groups based on the first group of variables and the i-th second group of variables.

3. The method of claim 2, wherein, The determining of the relevance results of the historical query request and the N historical POIs based on the N semantic variable groups and the image information corresponding to the N historical POIs, respectively, comprises: generating an i-th first input information based on the i-th semantic variable group in the N semantic variable groups and the image information corresponding to the i-th historical POI in the N historical POIs; inputting the i-th first input information into a first model to obtain an i-th first output result of the first model; determining the relevance result of the historical query request and the i-th historical POI in the N historical POIs based on the i-th first output result.

4. The method of claim 2, further comprising: determine an i-th semantic vector information based on an i-th semantic variable group in the N semantic variable groups, and determine an i-th image feature vector based on image information corresponding to the i-th historical POI in the N historical POIs.

5. The method of claim 4, wherein, The determining of the i-th semantic vector information based on the i-th semantic variable group in the N semantic variable groups comprises inputting the i-th semantic variable group in the N semantic variable groups into a second model to obtain the i-th semantic vector information output by the second model. The determining of the i-th image feature vector based on the image information corresponding to the i-th historical POI in the N historical POIs comprises inputting the image information corresponding to the i-th historical POI in the N historical POIs into a third model to obtain the i-th image feature vector output by the third model.

6. The method according to any one of claims 1 to 5, wherein, The determining of the candidate POI from the N historical POIs based on the relevance results corresponding to the historical query request and the N historical POIs respectively comprises: obtaining a relevance result corresponding to a j-th historical POI in the historical query request and the N historical POIs; j is an integer greater than or equal to 1 and less than or equal to N; in a case where the relevance result corresponding to the j-th historical POI in the historical query request is greater than a preset threshold value, taking the j-th historical POI as one of the candidate POIs.

7. The method of any one of claims 1-6, further comprising: receiving a target query request sent by a target device; finding target key information matching the target query request from the candidate key information, and obtaining a candidate recall set corresponding to the target key information; generating a target recall result based on the candidate recall set, and sending the target recall result to the target device.

8. An information processing apparatus, comprising: a semantic processing module configured to determine N semantic variable groups based on a historical query request and N historical points of interest (POIs) associated with the historical query request; N is an integer greater than or equal to 2; a relevance processing module configured to determine relevance results corresponding to the historical query request and the N historical POIs respectively based on the N semantic variable groups and image information corresponding to the N historical POIs respectively; a selection module configured to determine candidate POIs from the N historical POIs based on the relevance results corresponding to the historical query request and the N historical POIs respectively; an index module configured to take the historical query request as candidate key information, and generate a candidate recall set of the candidate key information based on the candidate POIs. The correlation processing module is configured to add the i-th semantic vector information and the i-th image feature vector to obtain a plurality of multi-modal vectors in the i-th group of multi-modal vectors, wherein the i-th semantic vector information is determined based on an i-th semantic variable group in the N semantic variable groups, the i-th image feature vector is determined based on image information corresponding to the i-th historical POI in the N historical POIs, and i is an integer greater than or equal to 1 and less than or equal to N; the plurality of multi-modal vectors in the i-th group of multi-modal vectors are weighted and added to obtain a target value; and the target value is taken as the relevance result of the historical query request and the i-th historical POI in the N historical POIs.

9. The apparatus of claim 8, wherein, The semantic processing module is configured to obtain first geographic coding information based on a query geographic location of the historical query request, and obtain i-th second geographic coding information based on a geographic location of an i-th historical POI in the N historical POIs associated with the historical query request. The semantic processing module is configured to generate a first group of variables based on the first geographic coding information, and generate an i-th second group of variables based on the i-th second geographic coding information. The semantic processing module is configured to generate an i-th semantic variable group in the N semantic variable groups based on the first group of variables and the i-th second group of variables.

10. The apparatus of claim 9, wherein, The correlation processing module is configured to generate i-th first input information based on the i-th semantic variable group in the N semantic variable groups and image information corresponding to the i-th historical POI in the N historical POIs, input the i-th first input information into a first model to obtain an i-th first output result of the first model, and determine the relevance result of the historical query request and the i-th historical POI in the N historical POIs based on the i-th first output result. The correlation processing module is configured to determine i-th semantic vector information based on an i-th semantic variable group in the N semantic variable groups, and determine an i-th image feature vector based on image information corresponding to the i-th historical POI in the N historical POIs.

11. The apparatus of claim 9, wherein, The correlation processing module is configured to input the i-th semantic variable group in the N semantic variable groups into a second model to obtain the i-th semantic vector information output by the second model.

12. The apparatus of claim 11, wherein, The correlation processing module is configured to input image information corresponding to the i-th historical POI in the N historical POIs into a third model to obtain the i-th image feature vector output by the third model. The selection module is configured to obtain a relevance result corresponding to a j-th historical POI in the historical query request and the N historical POIs from the relevance results corresponding to the historical query request and the N historical POIs respectively, j is an integer greater than or equal to 1 and less than or equal to N.

13. The apparatus of any one of claims 8-12, wherein, In a case where the relevance result corresponding to the j-th historical POI in the historical query request is greater than a preset threshold value, the j-th historical POI is taken as one of the candidate POIs.

14. The apparatus according to any one of claims 8-13, further comprising: a receiving module configured to receive a target query request sent by a target device. ​ The query module is configured to search, from the candidate key information, target key information matching the target query request, and obtain a candidate recall set corresponding to the target key information. The sending module is configured to generate a target recall result based on the candidate recall set, and send the target recall result to the target device.

15. An electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

16. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to enable the computer to perform the method of any one of claims 1-7.

17. A computer program product comprising a computer program which, when executed by a processor, implements the method of any one of claims 1-7.

Images