CN113326450A - Interest point recall method and device, electronic equipment and storage medium - Google Patents
Interest point recall method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN113326450A CN113326450A CN202110600169.0A CN202110600169A CN113326450A CN 113326450 A CN113326450 A CN 113326450A CN 202110600169 A CN202110600169 A CN 202110600169A CN 113326450 A CN113326450 A CN 113326450A
- Authority
- CN
- China
- Prior art keywords
- user
- interest
- poi
- point
- vector representation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 57
- 239000013598 vector Substances 0.000 claims abstract description 153
- 238000012545 processing Methods 0.000 claims abstract description 41
- 238000004590 computer program Methods 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims description 75
- 230000007774 longterm Effects 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 239000000126 substance Substances 0.000 claims description 2
- 238000013473 artificial intelligence Methods 0.000 abstract description 3
- 238000005516 engineering process Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 13
- 230000006399 behavior Effects 0.000 description 10
- 238000004891 communication Methods 0.000 description 8
- 235000013305 food Nutrition 0.000 description 5
- 230000003993 interaction Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 230000000977 initiatory effect Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 1
- 241001122767 Theaceae Species 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000013475 authorization Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 235000021185 dessert Nutrition 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000008267 milk Substances 0.000 description 1
- 210000004080 milk Anatomy 0.000 description 1
- 235000013336 milk Nutrition 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000001953 sensory effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
- G06F16/90—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
- 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
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The disclosure provides a point of interest recall method, a point of interest recall device, electronic equipment, a storage medium and a computer program product, and relates to the field of artificial intelligence, in particular to an intelligent transportation technology. The specific implementation scheme is as follows: carrying out feature cross processing on user features of at least two dimensions of a user to obtain user vector representation of the user; determining POI vector representation of a target interest point matched with user vector representation of a user from POI vector representation of candidate interest points determined in advance; the POI vector representation of the candidate interest points is obtained by performing cross processing on POI characteristics of at least two dimensions of the candidate interest points; and recalling the target interest points according to the POI vector representation of the target interest points. The embodiment of the disclosure realizes fast and accurate recall of the interest points matched with the user characteristics.
Description
Technical Field
The present disclosure relates to the field of artificial intelligence, and in particular, to the field of intelligent transportation, and more particularly, to a method, an apparatus, an electronic device, a storage medium, and a computer program product for point of interest recall.
Background
POIs are abbreviations for points of interest that may serve as identifiers for banks, sights, companies, hospitals, government agencies, restaurants, malls, etc. in a map.
Currently, in daily life, more and more users obtain relevant information of places by inquiring about POIs using an electronic map, for example, looking up a museum in a certain area.
Disclosure of Invention
The present disclosure provides a point of interest recall method, apparatus, electronic device, storage medium, and computer program product.
According to an aspect of the present disclosure, there is provided a point of interest recall method including:
carrying out feature cross processing on user features of at least two dimensions of a user to obtain user vector representation of the user;
determining POI vector representation of a target interest point matched with user vector representation of a user from POI vector representation of candidate interest points determined in advance; the POI vector representation of the candidate interest points is obtained by performing cross processing on POI characteristics of at least two dimensions of the candidate interest points;
and recalling the target interest points according to the POI vector representation of the target interest points.
According to another aspect of the present disclosure, there is provided a point of interest recall apparatus including:
the first calculation module is used for performing feature cross processing on the user features of at least two dimensions of the user to obtain user vector representation of the user;
the matching module is used for determining POI vector representation of a target interest point matched with the user vector representation of the user from the POI vector representation of the candidate interest points determined in advance; the POI vector representation of the candidate interest points is obtained by performing cross processing on POI characteristics of at least two dimensions of the candidate interest points;
and the recall module is used for recalling the target interest points according to the POI vector representation of the target interest points. .
According to another aspect of the present disclosure, there is provided an electronic device including:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the point of interest recall method of any embodiment of the present disclosure.
According to another aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform a point of interest recall method of any embodiment of the present disclosure.
According to another aspect of the present disclosure, there is provided a computer program product comprising a computer program which, when executed by a processor, implements the point of interest recall method of any embodiment of the present disclosure.
According to the technology disclosed by the invention, the interest points matched with the user characteristics can be recalled quickly and accurately.
It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present disclosure, nor do they limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description.
Drawings
The drawings are included to provide a better understanding of the present solution and are not to be construed as limiting the present disclosure. Wherein:
FIG. 1 is a schematic diagram of a point of interest recall method according to an embodiment of the present disclosure;
FIG. 2 is a schematic diagram of another point of interest recall method according to an embodiment of the present disclosure;
FIG. 3a is a schematic diagram of a training process of a user and a point of interest model according to an embodiment of the present disclosure;
FIG. 3b is a schematic diagram of a user and a point of interest model, according to an embodiment of the present disclosure;
FIG. 4 is a schematic diagram of yet another point of interest recall method according to an embodiment of the present disclosure;
FIG. 5 is a logic diagram of a point of interest recall method according to an embodiment of the present disclosure;
FIG. 6 is a schematic diagram of a point of interest recall apparatus in accordance with an embodiment of the present disclosure;
FIG. 7 is a block diagram of an electronic device used to implement a method of point of interest recall of embodiments of the present disclosure.
Detailed Description
Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, in which various details of the embodiments of the disclosure are included to assist understanding, and which are to be considered as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
The inventor finds out through research that when a user searches for a point of interest (POI) meeting general requirements of the user in a map application, such as searching for gourmet food, scenic spots or initiating surrounding exploration, a spatial range concerned by the user changes along with behavior types of the user, for example, when the user searches for the scenic spots, the POI within dozens of or even dozens of kilometers of the surrounding can be accepted, and when the user searches for the gourmet food, the user often only focuses on results within 1-2km of the surrounding; when a user manually zooms a region to a different level, the spatial extent of interest is also different. In addition, the user needs are various and are related to the current scene, the portrait of the user, the interest of the user, the real-time needs, and the like. For example, for a user scenario, users in a displaced row may be more relevant to a surrounding hotel than local users; for user portraits, young people prefer bars rather than middle-aged and elderly people, and female users prefer milk tea and dessert shops rather than male users; aiming at the real-time requirements of users, the users who just search scenic spots hope to see more travel recommendations with high probability; for long-term interest of users, users who have a preference for daily food are willing to see more daily food gourmet.
When a user searches for a point of interest POI meeting the general requirement of the user on a map, for example, inputting a general search word or initiating a surrounding exploration and the like, a point of interest list matched with the behavior, scene, portrait and interest of the user is recalled to help the user to quickly find the surrounding point of interest POI. See the following specific examples for specific point of interest recall methods.
Fig. 1 is a schematic flow chart of a point of interest recall method according to an embodiment of the present disclosure, which is applicable to a situation where a user accurately recalls a point of interest that meets a user's demand when performing a general demand search in a map application. The method may be performed by a point of interest recall apparatus implemented in software and/or hardware and integrated on an electronic device.
Specifically, referring to fig. 1, the point of interest recall method is as follows:
s101, performing feature cross processing on user features of at least two dimensions of a user to obtain user vector representation of the user.
In the implementation of the present disclosure, the user refers to a user who triggers a universal demand search in a map application, and the user characteristics include at least one of a scene where the user is located (time, holiday, local place, and the like), a spatial grid where the user is located, portrait characteristics (age, gender, life stage, and the like), long-term interest characteristics, and short-term interest characteristics.
The scene where the user is located can be determined according to time and position information when the user triggers the extensive demand retrieval; the portrait characteristics of the user are obtained by querying a portrait dictionary using the user id as a key. The spatial grid in which the user is located may also be determined according to where the user is located.
The long-term interest characteristics of the user can count all clicking behaviors of the user within two months through a retrieval log of the map, and the characteristic vectors of the clicked interest points are added and averaged to be used as the long-term interest characteristics of the user. The short-term interest features of the user can be obtained by counting the latest interest points (for example, the latest 5 interest points) interacted by the user within three days, and connecting the vector features of the interest points in series to serve as the short-term interest features of the user. In the implementation of the application, the long-term interest features are used for modeling the stable preference of a user, for example, when the user frequently goes to a park, the score of the park is improved when the model recommends surrounding scenic spots. The real-time interest features are used for modeling current preferences of a user, for example, the user just clicks a submarine scoop, and the model can subsequently recommend other similar hot pot restaurants.
It should be noted here that the acquired user characteristics are authorized by the user.
In the embodiment of the present disclosure, the inventor finds that, when recalling an interest point, user features of different dimensions are mutually affected, for example, scenes and interests of users are mutually affected, and if the interest point is recalled according to user features of only one dimension, the accuracy of the recalled interest point is low. In order to ensure that the recalled interest points fully meet the user requirements, after the user features are obtained, feature cross processing is performed on the user features of at least two dimensions of the user, for example, feature vectors of different user features are subjected to multiplication processing, and then user vector representation of the user is obtained. Therefore, when the interest points are recalled based on the user vector representation, the user vector representation considers the relationship among different user characteristics, and the interest points matched with the scene, the long-term interest, the short-term interest and the portrait of the user can be accurately recalled subsequently.
S102, POI vector representation of the target interest point matched with the user vector representation of the user is determined from the POI vector representation of the candidate interest points determined in advance.
The candidate interest points can be the total interest points in the map application, the POI vector representation of the candidate interest points is obtained by performing cross processing on the POI features of at least two dimensions of the candidate interest points, and the POI features include interest point features including interest point attribute tags, interest point IDs, a spatial grid to which the interest points belong, interest point heat and the like. In determining the POI vector representation of the target point of interest that matches the user vector representation of the user, optionally, the similarity may be determined by calculating a similarity between the user vector representation and the POI vector representation of the candidate point of interest, for example, calculating a cosine similarity, or may be determined by other means, which is not limited herein. It should be noted that, the POI vector representation of the candidate interest point is determined before the user search, and the purpose is to ensure the efficiency of the interest point recall.
S103, recalling the target interest points according to POI vector representation of the target interest points.
After the POI vector representation of the target point of interest is determined through the step S102, the corresponding target point of interest may be recalled according to the association relationship between the POI vector representation and the point of interest.
In the embodiment of the disclosure, the user vector representation is obtained on the basis of cross processing of a plurality of user features, the relationship among different user features is considered, and then when the interest points are recalled on the basis of the user vector representation, the interest points matched with scenes, long-term and short-term interests and figures where the users are located can be recalled accurately; and the efficiency of recalling the target interest point can be improved through the matching calculation between the user vector representation and the POI vector representation of the candidate interest point.
Fig. 2 is a schematic flowchart of another point of interest recall method according to an embodiment of the present disclosure, which is optimized based on the above embodiments, and determines a user vector representation and a POI vector representation of a candidate point of interest by using a trained user-point of interest model, where the user-point of interest model is exemplarily a two-tower model including a user sub-model with displayed feature intersection capability and a POI sub-model with displayed feature intersection capability, and the user sub-model and the POI sub-model implement the feature intersection capability through feature intersection sub-networks in respective models, and the user sub-model uses user features as input, and the POI sub-model uses POI features as input.
Referring to fig. 2, the point of interest recall method is embodied as follows:
s201, inputting at least two dimensional user characteristics of a user into a user sub-model of the user and the interest point model.
The user refers to a user who triggers the extensive demand retrieval in the map at present. After a user triggers general demand retrieval, user characteristics of the user are obtained, at least two dimensionality user characteristics of the user are spliced, and the spliced user characteristics are input into a user sub-model of the user and an interest point model.
S202, carrying out cross processing on at least two dimensional user characteristics through a user characteristic cross sub-network in the user sub-model to obtain user vector representation of the user.
Optionally, the user feature cross sub-network in the user sub-model performs cross processing on the user features of at least two dimensions, for example, performs vector multiplication operation on the user features of at least two dimensions to obtain user vector representation of the user. It should be noted that, the trained user and interest point model is used to determine the user vector representation, which can improve the efficiency of calculating the user vector representation.
S203, POI vector representation of the target interest point matched with the user vector representation of the user is determined from the POI vector representation of the candidate interest points determined in advance.
The POI vector representation of the candidate interest points is obtained by performing cross processing on POI characteristics of at least two dimensions of the candidate interest points; the interest point characteristics comprise interest point attribute labels, interest point IDs, a spatial grid to which the interest points belong and interest point heat. In an alternative embodiment, predetermining a POI vector representation of a candidate point of interest comprises: inputting at least two dimensions POI characteristics of the candidate interest points into POI submodels of the user and interest point model; and carrying out cross processing on the POI characteristics of at least two dimensions through a POI characteristic cross sub-network in the POI sub-model to obtain POI vector representation of the candidate interest points. It should be noted that, the trained user and the point of interest model are used to determine the POI vector representation of the candidate point of interest, which can improve the efficiency of calculating the POI vector representation.
And S204, recalling the target interest points according to POI vector representation of the target interest points.
In the embodiment of the disclosure, the trained user and interest point models are used for determining the user vector representation and the POI vector representation of the candidate interest points, so that the efficiency of calculating the user vector representation and the POI vector representation can be improved, and the efficiency of recalling the target POI is further ensured.
Fig. 3a is a schematic flow chart of a training process of a user and a point of interest model according to an embodiment of the present disclosure, and referring to fig. 3a, the training process is specifically as follows:
s301, determining an initial training sample according to historical data of the point of interest clicked by the user.
The historical data of the click interest points of the user can be optionally obtained from a map retrieval log, and the initial training samples comprise initial positive training samples and initial negative training samples. Optionally, the initial positive training sample refers to an interest point clicked by a user when initiating a general demand search in the map search log; the initial negative training sample refers to a preset number of interest points extracted from the interest points which are not clicked when the user initiates the extensive demand retrieval according to the map retrieval log. The ratio of the number of initial positive and negative examples is illustratively 1: 4.
s302, constructing a target training sample with the user characteristics and the interest point characteristics according to the user characteristics, the interest point characteristics and the initial training sample of the user.
In the embodiment of the disclosure, the user characteristics of the user comprise space-time scene class characteristics, portrait characteristics and long-term and short-term interest characteristics. The characteristics of the space-time scene (such as time, holidays, a spatial grid where the user is located and the like) are calculated according to the time and the positioning information recorded in the log of the interest points recalled by the user. The portrait characteristics of the user are obtained by querying a portrait dictionary using the user ID as a key value. And counting all clicking behaviors of the historical user within two months through a retrieval log of the map, and adding and averaging feature vectors of clicked interest points to serve as the long-term interest features of the user. The short-term interest features of the user are the latest interest points (such as the latest 5 interest points) interacted by the user within three days, and the vector features of the interest points are connected in series to serve as the short-term interest features of the user. It should be noted that, the user characteristics of the user are obtained under the authorization of the user.
The interest point features refer to features of interest points serving as initial training samples, and mainly include interest point attribute labels, interest point IDs, spatial grids to which the interest points belong, interest point heat, and the like.
In the embodiment of the disclosure, the target training samples comprise a target positive training sample and a target negative training sample, and the target positive training sample is formed by splicing an initial positive training sample, interest point characteristics corresponding to the initial positive training sample and user characteristics; the target negative training sample is formed by splicing an initial negative training sample, interest point characteristics corresponding to the initial negative training sample and user characteristics.
S303, training the user and the interest point model by using the target training sample.
After a target training sample is obtained, the target sample is directly trained and input into a user and interest point model, and the user and interest point model comprises a user sub-model and a POI sub-model, so that the user characteristics in the target training sample are input into a network corresponding to the user sub-model during actual training to obtain user vector representation; and inputting the interest point characteristics in the target training sample into a network corresponding to the POI sub-model to obtain POI vector representation, and further calculating the similarity and loss between the two vector representations so as to adjust the weight parameters of the user and the interest point model according to the loss until the user and the interest point model are converged.
Illustratively, referring to fig. 3b, a schematic diagram of a user and a point of interest model is shown, which can be divided into an input layer, a presentation layer and a matching layer. In the training stage, user features in a target training sample are processed and spliced on an input layer, and meanwhile, interest Point (POI) features in the target training sample are processed and spliced; performing feature cross processing on the spliced user features on the presentation layer to obtain user vector presentation, and performing feature cross processing on the spliced interest point features on the presentation layer to obtain POI vector presentation; and calculating the similarity and loss of the user vector representation and the POI vector representation at a matching layer, and further adjusting the weight parameters of the model according to the loss until the model converges.
Further, since the exploration of the periphery is a requirement with strong space constraint, it is required that there is a definite correlation between the spatial grid features and the interest points, for example, there is a strong correlation between the spatial grids of the palace courtyard and the palace. To model this correlation, the feature vectors of the spatial grid and the points of interest are pre-trained from the search logs of the map.
In an alternative embodiment, a sample of the spatial grid and the point of interest ID is constructed (e.g., based on a map retrieval log), and a spatial grid feature vector and a point of interest ID feature vector are generated. Optionally, the retrieval interaction behaviors of all users recorded in the retrieval log of the map are segmented according to sessions, and the behaviors of the historical users in each session are continuously related. And taking the interest points interacted by the user in each session as a behavior sequence of the user, and adding the spatial grid ID corresponding to each interest point to the sequence to form a new sequence. For example, extracting the points of interest of the user after search interaction as POI a, POI B, and POI C, and adding the grid of each point of interest to the sequence to form a new sequence: POIA, GRID A, POI B, GRID B, POI C, GRID C. And further generating a word vector for each interest point and grid by using a word vector tool (such as word2vec) and the behavior sequence added with the spatial grid information. The word vector tool can generate vector expression with definite semantics according to a large number of user behavior sequences. The semantics are mainly reflected in words with high adjacent frequency in the sequence, and the similarity of the words is high. For example, the grid where the Imperial palace is located has higher similarity with the surrounding Imperial palace courtyard and Tiananmen square.
Further, after the space grid feature vector and the interest point ID feature vector are obtained, the target training sample is used, the space grid feature vector and the interest point ID feature vector are combined, the user and the interest point model are trained, namely when the user and the interest point model are trained, if the space grid feature vector and the interest point ID feature vector are used, the pre-trained space grid feature vector and the pre-trained interest point ID feature vector can be directly called, and therefore training efficiency of the user and interest point double-tower model is improved.
In the embodiment of the disclosure, the user and interest point models are trained by using the samples with the user characteristics and the interest point characteristics, so that the user vector representation of the user and the POI vector representation of the candidate interest points can be calculated by directly using the trained models, and the interest point recall efficiency is ensured.
Fig. 4 is a schematic flowchart of another point of interest recall method according to an embodiment of the present disclosure, which is optimized based on the foregoing embodiment, and referring to fig. 4, the point of interest recall method is specifically as follows:
s401, performing feature cross processing on the user features of at least two dimensions of the user to obtain user vector representation of the user.
Wherein the user characteristics include at least one of a scene where the user is located, a spatial grid where the user is located, portrait characteristics, long-term interest characteristics, and short-term interest characteristics.
S402, determining POI vector representation of the target interest point matched with the user vector representation of the user from the POI vector representation of the candidate interest points determined in advance based on the approximate nearest neighbor searching mode.
The POI vector representation of the candidate interest point is obtained by performing cross processing on the POI features of at least two dimensions of the candidate interest point, and the process of determining the candidate interest point in advance may refer to the above embodiments, which are not described herein again. When the target POI is recalled based on the approximate nearest neighbor searching mode, optionally, the POI vector representation of the predetermined candidate interest point is used as an input value of an approximate nearest neighbor searching tool to construct an approximate nearest neighbor dictionary, and the process can be completed before the user initiates the general demand retrieval; when the user retrieves the recalled points of interest, the user vector representation of the user is input into the approximate nearest neighbor lookup tool so that the approximate nearest neighbor lookup tool determines the POI vector representation of the target point of interest matched with the approximate nearest neighbor lookup tool according to the approximate nearest neighbor dictionary. It should be noted that the core idea of approximate nearest neighbor search is: searching for data items that may be neighbors is no longer limited to returning the most likely item, improving retrieval efficiency at the expense of accuracy within an acceptable range.
And S403, recalling the target interest points according to POI vector representation of the target interest points.
In the embodiment of the disclosure, the POI vector representation of the target interest point matched with the user vector representation of the user is determined based on the approximate nearest neighbor searching mode, so that the searching efficiency can be improved, and the speed of recalling the interest point is further ensured.
Fig. 5 is a logic diagram of a point of interest recall method according to an embodiment of the present disclosure, which is optimized based on the above-described embodiment, and referring to fig. 5, the logic of the point of interest recall method includes three stages, namely a sample and feature processing stage, an offline training stage, and an online service stage.
A sample and feature processing stage, which is mainly used for constructing a positive and negative training sample according to a log of general demand-type behaviors of a map user (such as words of searching gourmet food, scenic spots and the like by the user or a peripheral exploration request is initiated); splicing positive and negative training samples and characteristics (user characteristics and POI characteristics); and constructing a pre-training sample of the spatial grid and the interest point ID according to the full-scale retrieval log of the map, and pre-training the feature vectors of the spatial grid and the interest point ID. For the specific process, reference may be made to the above embodiments, which are not described herein again
The offline training stage is mainly to train the user and the point of interest model (an exemplary two-tower model), and to construct an approximate nearest neighbor dictionary (or ANN dictionary) required for online Approximate Nearest Neighbor (ANN) recall. For the process of training the user and the interest point model, refer to the above embodiments, and are not described herein again. Optionally, the trained user and interest point model is used to calculate the POI vector representation of the candidate interest point, and the POI vector representation of the candidate interest point is used as an input value of the approximate nearest neighbor searching tool to construct the approximate nearest neighbor dictionary.
And in the online service stage, based on the trained user and interest point models, calculating user vector representation according to the user characteristics of the users, and recalling N interest points closest to the user vector representation from the ANN dictionary as a final recall result.
In the embodiment of the disclosure, the interest points meeting the requirements of the users are recalled through the three-stage processing.
Fig. 6 is a schematic structural diagram of an interest point recall device according to an embodiment of the present disclosure, which is applicable to a situation where a user accurately recalls an interest point meeting a user requirement when performing a general requirement search in a map application. As shown in fig. 6, the apparatus specifically includes:
a first calculation module 601, configured to perform feature cross processing on user features of at least two dimensions of a user to obtain a user vector representation of the user;
a matching module 602, configured to determine, from the predetermined POI vector representations of the candidate points of interest, a POI vector representation of a target point of interest that matches the user vector representation of the user; the POI vector representation of the candidate interest points is obtained by performing cross processing on POI characteristics of at least two dimensions of the candidate interest points;
the recalling module 603 is configured to recall the target interest point according to the POI vector representation of the target interest point.
On the basis of the foregoing embodiment, optionally, the first calculating module includes:
the input unit is used for inputting at least two dimensional user characteristics of a user into user submodels of the user and the interest point model; the user characteristics comprise at least one of a scene where the user is located, a spatial grid where the user is located, portrait characteristics, long-term interest characteristics and short-term interest characteristics;
and the calculating unit is used for performing cross processing on the user characteristics of at least two dimensions through the user characteristic cross sub-network in the user sub-model to obtain user vector representation of the user.
On the basis of the foregoing embodiment, optionally, the system further includes a second calculating module, where the second calculating module is configured to:
inputting at least two dimensions POI characteristics of the candidate interest points into POI submodels of the user and interest point model; the interest point characteristics comprise interest point attribute tags, interest point IDs, space grids to which the interest points belong and interest point heat degrees;
and carrying out cross processing on the POI characteristics of at least two dimensions through a POI characteristic cross sub-network in the POI sub-model to obtain POI vector representation of the candidate interest points.
On the basis of the above embodiment, optionally, the system further includes a model training module, where the model training module includes:
the first sample construction unit is used for determining an initial training sample according to historical data of a click interest point of a user;
the second sample construction unit is used for constructing a target training sample with the user characteristics and the interest point characteristics according to the user characteristics, the interest point characteristics and the initial training sample of the user;
and the training unit is used for training the user and the interest point model by utilizing the target training sample.
On the basis of the above embodiment, optionally, the method further includes:
the pre-training module is used for constructing samples of the spatial grids and the interest point IDs and generating spatial grid feature vectors and interest point ID feature vectors;
the training unit is further configured to:
and training the user and the interest point model by using the target training sample and combining the space grid feature vector and the interest point ID feature vector.
On the basis of the above embodiment, optionally, the matching module is specifically configured to:
and determining POI vector representation of the target interest point matched with the user vector representation of the user from the preset POI vector representation of the candidate interest point based on the approximate nearest neighbor searching mode.
The interest point recall device provided by the embodiment of the disclosure can execute the interest point recall method provided by any embodiment of the disclosure, and has corresponding functional modules and beneficial effects of the execution method. Reference may be made to the description of any method embodiment of the disclosure for a matter not explicitly described in this embodiment.
The present disclosure also provides an electronic device, a readable storage medium, and a computer program product according to embodiments of the present disclosure.
FIG. 7 illustrates a schematic block diagram of an example electronic device 700 that can be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the disclosure described and/or claimed herein.
As shown in fig. 7, the device 700 comprises a computing unit 701, which may perform various suitable actions and processes according to a computer program stored in a Read Only Memory (ROM)702 or a computer program loaded from a storage unit 708 into a Random Access Memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other by a bus 704. An input/output (I/O) interface 705 is also connected to bus 704.
Various components in the device 700 are connected to the I/O interface 705, including: an input unit 706 such as a keyboard, a mouse, or the like; an output unit 707 such as various types of displays, speakers, and the like; a storage unit 708 such as a magnetic disk, optical disk, or the like; and a communication unit 709 such as a network card, modem, wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information/data with other devices via a computer network, such as the internet, and/or various telecommunication networks.
Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), system on a chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.
Program code 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, special purpose computer, or other programmable data processing apparatus, such that the program codes, when executed by the processor or controller, cause the functions/operations specified in the flowchart and/or block diagram to be performed. 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.
In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The 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, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, 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.
To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.
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: local Area Networks (LANs), Wide Area Networks (WANs), blockchain networks, and the internet.
The computer system may 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, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so that the defects of high management difficulty and weak service expansibility in the traditional physical host and VPS service are overcome.
It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present disclosure may be executed in parallel, sequentially, or in different orders, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved, and the present disclosure is not limited herein.
In the technical scheme of the disclosure, the acquisition, storage, application and the like of the personal information of the related user all accord with the regulations of related laws and regulations, and do not violate the good customs of the public order.
The above detailed description should not be construed as limiting the scope of the disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present disclosure should be included in the scope of protection of the present disclosure.
Claims (15)
1. A point of interest recall method comprising:
carrying out feature cross processing on user features of at least two dimensions of a user to obtain user vector representation of the user;
determining POI vector representations of target interest points matched with the user vector representation of the user from the POI vector representations of the candidate interest points determined in advance; the POI vector representation of the candidate interest point is obtained by performing cross processing on POI characteristics of at least two dimensions of the candidate interest point;
and recalling the target interest points according to the POI vector representation of the target interest points.
2. The method of claim 1, wherein performing feature intersection processing on user features of at least two dimensions of a user to obtain a user vector representation of the user comprises:
inputting at least two dimensions of user characteristics of a user into user submodels of the user and the interest point model; wherein the user characteristics comprise at least one of a scene where the user is located, a spatial grid where the user is located, portrait characteristics, long-term interest characteristics and short-term interest characteristics;
and performing cross processing on the at least two dimensionality user characteristics through a user characteristic cross sub-network in the user sub-model to obtain user vector representation of the user.
3. The method of claim 1, wherein predetermining a POI vector representation of a candidate point of interest comprises:
inputting at least two dimensions POI characteristics of the candidate interest points into POI submodels of the user and interest point model; the interest point characteristics comprise interest point attribute tags, interest point IDs, a spatial grid to which the interest points belong and interest point heat degrees;
and performing cross processing on the POI characteristics of at least two dimensions through a POI characteristic cross sub-network in the POI sub-model to obtain POI vector representation of candidate interest points.
4. The method of claim 2 or 3, wherein the training process of the user and the interest point model comprises the following steps:
determining an initial training sample according to historical data of a point of interest clicked by a user;
constructing a target training sample with the user characteristics and the interest point characteristics according to the user characteristics, the interest point characteristics and the initial training sample of the user;
and training the user and the interest point model by using the target training sample.
5. The method of claim 4, prior to training a user and a point of interest model using the target training samples, the method further comprising:
constructing samples of a spatial grid and an interest point ID, and generating a spatial grid feature vector and an interest point ID feature vector;
training a user and an interest point model by using the target training sample, wherein the training comprises the following steps:
and training a user and an interest point model by using the target training sample and combining the space grid feature vector and the interest point ID feature vector.
6. The method of claim 1, wherein determining, from the predetermined POI vector representations of candidate points of interest, a POI vector representation of a target point of interest that matches the user vector representation of the user comprises:
and determining POI vector representation of the target interest point matched with the user vector representation of the user from the preset POI vector representations of the candidate interest points based on an approximate nearest neighbor searching mode.
7. A point of interest recall apparatus comprising:
the first calculation module is used for performing feature cross processing on the user features of at least two dimensions of the user to obtain user vector representation of the user;
a matching module for determining a POI vector representation of a target point of interest that matches a user vector representation of the user from POI vector representations of predetermined candidate points of interest; the POI vector representation of the candidate interest point is obtained by performing cross processing on POI characteristics of at least two dimensions of the candidate interest point;
and the recall module is used for recalling the target interest points according to the POI vector representation of the target interest points.
8. The apparatus of claim 7, wherein the first computing module comprises:
the input unit is used for inputting at least two dimensional user characteristics of a user into user submodels of the user and the interest point model; wherein the user characteristics comprise at least one of a scene where the user is located, a spatial grid where the user is located, portrait characteristics, long-term interest characteristics and short-term interest characteristics;
and the calculating unit is used for performing cross processing on the at least two dimensionality user characteristics through the user characteristic cross sub-network in the user sub-model to obtain user vector representation of the user.
9. The apparatus of claim 7, further comprising a second computing module to:
inputting at least two dimensions POI characteristics of the candidate interest points into POI submodels of the user and interest point model; the interest point characteristics comprise interest point attribute tags, interest point IDs, a spatial grid to which the interest points belong and interest point heat degrees;
and performing cross processing on the POI characteristics of at least two dimensions through a POI characteristic cross sub-network in the POI sub-model to obtain POI vector representation of candidate interest points.
10. The apparatus of claim 8 or 9, further comprising a model training module comprising:
the first sample construction unit is used for determining an initial training sample according to historical data of a click interest point of a user;
the second sample construction unit is used for constructing a target training sample with the user characteristics and the interest point characteristics according to the user characteristics, the interest point characteristics and the initial training sample of the user;
and the training unit is used for training the user and the interest point model by utilizing the target training sample.
11. The apparatus of claim 10, further comprising:
the pre-training module is used for constructing samples of the spatial grids and the interest point IDs and generating spatial grid feature vectors and interest point ID feature vectors;
the training unit is further configured to:
and training a user and an interest point model by using the target training sample and combining the space grid feature vector and the interest point ID feature vector.
12. The apparatus of claim 7, wherein the matching module is specifically configured to:
and determining POI vector representation of the target interest point matched with the user vector representation of the user from the preset POI vector representations of the candidate interest points based on an approximate nearest neighbor searching mode.
13. An electronic device, comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6.
14. A non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method of any one of claims 1-6.
15. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110600169.0A CN113326450B (en) | 2021-05-31 | 2021-05-31 | Point-of-interest recall method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110600169.0A CN113326450B (en) | 2021-05-31 | 2021-05-31 | Point-of-interest recall method and device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113326450A true CN113326450A (en) | 2021-08-31 |
| CN113326450B CN113326450B (en) | 2024-01-12 |
Family
ID=77422508
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110600169.0A Active CN113326450B (en) | 2021-05-31 | 2021-05-31 | Point-of-interest recall method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113326450B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114374881A (en) * | 2022-01-05 | 2022-04-19 | 北京百度网讯科技有限公司 | Method and device for distributing user flow, electronic equipment and storage medium |
| CN114625984A (en) * | 2022-03-31 | 2022-06-14 | 北京百度网讯科技有限公司 | Interest point verification method and device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108351876A (en) * | 2015-09-22 | 2018-07-31 | 纽昂斯通讯公司 | System and method for point of interest identification |
| CN111310074A (en) * | 2020-02-13 | 2020-06-19 | 北京百度网讯科技有限公司 | Interest point label optimization method and device, electronic equipment and computer readable medium |
| CN111444428A (en) * | 2020-03-27 | 2020-07-24 | 腾讯科技（深圳）有限公司 | Information recommendation method and device based on artificial intelligence, electronic equipment and storage medium |
| CN112559901A (en) * | 2020-12-11 | 2021-03-26 | 百度在线网络技术（北京）有限公司 | Resource recommendation method and device, electronic equipment, storage medium and computer program product |
| WO2021081823A1 (en) * | 2019-10-30 | 2021-05-06 | 深圳市欢太科技有限公司 | Information pushing method and apparatus, server, and storage medium |
-
2021
- 2021-05-31 CN CN202110600169.0A patent/CN113326450B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108351876A (en) * | 2015-09-22 | 2018-07-31 | 纽昂斯通讯公司 | System and method for point of interest identification |
| US20180349380A1 (en) * | 2015-09-22 | 2018-12-06 | Nuance Communications, Inc. | Systems and methods for point-of-interest recognition |
| WO2021081823A1 (en) * | 2019-10-30 | 2021-05-06 | 深圳市欢太科技有限公司 | Information pushing method and apparatus, server, and storage medium |
| CN111310074A (en) * | 2020-02-13 | 2020-06-19 | 北京百度网讯科技有限公司 | Interest point label optimization method and device, electronic equipment and computer readable medium |
| CN111444428A (en) * | 2020-03-27 | 2020-07-24 | 腾讯科技（深圳）有限公司 | Information recommendation method and device based on artificial intelligence, electronic equipment and storage medium |
| CN112559901A (en) * | 2020-12-11 | 2021-03-26 | 百度在线网络技术（北京）有限公司 | Resource recommendation method and device, electronic equipment, storage medium and computer program product |
Non-Patent Citations (1)
| Title |
|---|
| 丁松涛;曲仕茹;: "基于深度学习的交通目标感兴趣区域检测", 中国公路学报, no. 09 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114374881A (en) * | 2022-01-05 | 2022-04-19 | 北京百度网讯科技有限公司 | Method and device for distributing user flow, electronic equipment and storage medium |
| CN114374881B (en) * | 2022-01-05 | 2023-09-01 | 北京百度网讯科技有限公司 | Method and device for distributing user traffic, electronic equipment and storage medium |
| CN114625984A (en) * | 2022-03-31 | 2022-06-14 | 北京百度网讯科技有限公司 | Interest point verification method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113326450B (en) | 2024-01-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114549874B (en) | Training method of multi-target image-text matching model, image-text retrieval method and device | |
| CN112612957A (en) | Interest point recommendation method, interest point recommendation model training method and device | |
| CN112487173A (en) | Man-machine conversation method, device and storage medium | |
| CN113326450B (en) | Point-of-interest recall method and device, electronic equipment and storage medium | |
| CN113656698B (en) | Training method and device for interest feature extraction model and electronic equipment | |
| CN113407851B (en) | Method, device, equipment and medium for determining recommended information based on double-tower model | |
| CN113792154A (en) | Method and device for determining fault association relationship, electronic equipment and storage medium | |
| CN114329244A (en) | Map interest point query method, map interest point query device, map interest point query equipment, storage medium and program product | |
| CN114782719B (en) | Training method of feature extraction model, object retrieval method and device | |
| CN112860993A (en) | Method, device, equipment, storage medium and program product for classifying points of interest | |
| CN113010752B (en) | Recall content determining method, apparatus, device and storage medium | |
| CN112506864B (en) | File retrieval method, device, electronic equipment and readable storage medium | |
| CN113360683A (en) | Method for training cross-modal retrieval model and cross-modal retrieval method and device | |
| CN116228301A (en) | Method, device, equipment and medium for determining target user | |
| CN116049370A (en) | Information query method and training method and device of information generation model | |
| CN112784600B (en) | Information ordering method, device, electronic equipment and storage medium | |
| CN113360590B (en) | Method and device for updating interest point information, electronic equipment and storage medium | |
| CN114756774A (en) | Travel scheme recommendation method, travel scheme recommendation device, model training equipment and storage medium | |
| CN114422584B (en) | Method, device and storage medium for pushing resources | |
| CN113868532B (en) | Location recommendation method and device, electronic equipment and storage medium | |
| CN116167455B (en) | Model training and data deduplication method, device, equipment and storage medium | |
| CN113408661B (en) | Method, apparatus, device and medium for determining mismatching | |
| CN113377921B (en) | Method, device, electronic equipment and medium for matching information | |
| CN113987345A (en) | Data processing method, device, equipment, storage medium and program product | |
| CN117390276A (en) | Content recommendation method, device, equipment, computer readable storage medium and product |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |