CN111523962A - Searching method, searching device, electronic equipment and storage medium - Google Patents
Searching method, searching device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN111523962A CN111523962A CN202010209482.7A CN202010209482A CN111523962A CN 111523962 A CN111523962 A CN 111523962A CN 202010209482 A CN202010209482 A CN 202010209482A CN 111523962 A CN111523962 A CN 111523962A
- Authority
- CN
- China
- Prior art keywords
- scene
- vector
- user
- preset
- similarity
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 239000013598 vector Substances 0.000 claims abstract description 206
- 238000012163 sequencing technique Methods 0.000 claims abstract description 13
- 238000013135 deep learning Methods 0.000 claims description 31
- 238000012545 processing Methods 0.000 claims description 22
- 238000004590 computer program Methods 0.000 claims description 9
- 230000015654 memory Effects 0.000 claims description 7
- 230000006399 behavior Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 2
- 238000013475 authorization Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0623—Item investigation
- G06Q30/0625—Directed, with specific intent or strategy
-
- 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
-
- 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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0641—Shopping interfaces
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/12—Hotels or restaurants
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Tourism & Hospitality (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Primary Health Care (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the application discloses a searching method, a searching device, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring user information of a current user; encoding the user information into a scene demand vector of a current user, wherein the dimension of the scene demand vector is a preset dimension; acquiring a preset number of scene vectors, wherein each scene vector is a vector of a preset dimension corresponding to a preset scene; respectively determining the similarity of the scene demand vectors and the preset number of scene vectors to obtain the preset number of similarities; and searching and sequencing a plurality of candidate targets according to the similarity of the preset number. According to the method and the device, the similarity vector is determined by adopting the scene demand vector of the current user and the existing scene vector of the preset scene, searching is carried out based on the similarity vector, and the candidate targets are sequenced, so that the final result is more consistent with the scene required by the current user, and the accuracy of the searching result is improved.
Description
Technical Field
The embodiment of the application relates to the technical field of internet, in particular to a searching method, a searching device, electronic equipment and a storage medium.
Background
When a user carries out online shopping or uses the online service, different scene requirements exist, for example, when a hotel search is used, the scene requirements can be that the user enters and stays nearby as soon as possible, goes on a business trip, travels, goes out by relatives and children, and the like. Different scenes have different optimization modes, and the scene division modes are complex and diverse.
In the prior art, in the field of hotel search, when scene division is performed, rule judgment is basically performed according to part of signals of users, for example, local and different places, new and old guests, android and iphone, parent-child search and the like can be performed, the divided scenes are easy to overlap, and the accuracy of search results is low due to the fact that the used user signals are simple.
Disclosure of Invention
The embodiment of the application provides a searching method, a searching device, electronic equipment and a storage medium, so as to improve the accuracy of a searching result.
In order to solve the above problem, in a first aspect, an embodiment of the present application provides a search method, including:
acquiring user information of a current user;
encoding the user information into a scene demand vector of a current user, wherein the dimension of the scene demand vector is a preset dimension;
acquiring a preset number of scene vectors, wherein each scene vector is a vector of a preset dimension corresponding to a preset scene;
respectively determining the similarity of the scene demand vectors and the preset number of scene vectors to obtain the preset number of similarities;
and searching and sequencing a plurality of candidate targets according to the similarity of the preset number.
Optionally, the encoding the user information into a scene requirement vector of the current user includes:
and processing the user information through a deep learning network to obtain a scene demand vector of the current user.
Optionally, the processing the user information through the deep learning network to obtain a scene demand vector of the current user includes:
encoding the user information into an input feature vector of a deep learning network;
and carrying out multilayer full-connection processing on the input characteristic vector through a deep learning network to obtain a scene demand vector of the current user.
Optionally, the sorting the multiple candidate targets according to the preset number of similarities includes:
splicing the similarity of the preset number into a similarity vector according to the arrangement sequence of the preset scenes;
searching according to the similarity vector to obtain a plurality of candidate targets;
inputting the similarity vectors and the characteristics of a plurality of candidate targets into a ranking model respectively to obtain ranking scores of the candidate targets;
and sorting the candidate targets according to the sorting scores of the candidate targets.
Optionally, after searching and sorting the multiple candidate targets according to the preset number of similarities, the method further includes:
returning the sorting result to the current user; or
And recommending the sequencing result to the current user.
Optionally, the user information includes at least one of user portrait information, user location information, historical payment information, historical click information, and historical browsing information.
In a second aspect, an embodiment of the present application provides a search apparatus, including:
the user signal acquisition module is used for acquiring the user information of the current user;
the user signal coding module is used for coding the user information into a scene demand vector of the current user, and the dimension of the scene demand vector is a preset dimension;
the scene vector acquisition module is used for acquiring a preset number of scene vectors, and each scene vector is a vector of a preset dimension corresponding to a preset scene;
the similarity determining module is used for respectively determining the similarity of the scene demand vectors and the preset number of scene vectors to obtain the preset number of similarities;
and the sorting module is used for searching and sorting the candidate targets according to the preset number of similarities.
Optionally, the user signal encoding module is specifically configured to:
and processing the user information through a deep learning network to obtain a scene demand vector of the current user.
Optionally, the user signal encoding module includes:
the user signal coding unit is used for coding the user information into an input characteristic vector of the deep learning network;
and the scene requirement determining unit is used for carrying out multilayer full-connection processing on the input characteristic vector through a deep learning network to obtain a scene requirement vector of the current user.
Optionally, the sorting module includes:
a similarity vector determination unit, configured to splice the preset number of similarities into similarity vectors according to an arrangement order of preset scenes;
the searching unit is used for searching according to the similarity vector to obtain a plurality of candidate targets;
the ranking score determining unit is used for inputting the similarity vectors and the characteristics of the candidate targets into a ranking model respectively to obtain ranking scores of the candidate targets;
and the sorting unit is used for sorting the candidate targets according to the sorting scores of the candidate targets.
Optionally, the apparatus further comprises:
the search result returning module is used for returning the sequencing result to the current user; or
And the search result recommending module is used for recommending the sequencing result to the current user.
Optionally, the user information includes at least one of user portrait information, user location information, historical payment information, historical click information, and historical browsing information.
In a third aspect, an embodiment of the present application further provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the search method according to the embodiment of the present application when executing the computer program.
In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps of the search method disclosed in the present application.
According to the searching method, the searching device, the electronic equipment and the storage medium, the user information of the current user is obtained, the user information is coded into the scene demand vector of the current user, the preset number of scene vectors are obtained, the similarity between the scene demand vector and the preset number of scene vectors is respectively determined, the preset number of similarities are obtained, searching is conducted according to the preset number of similarities and a plurality of candidate targets are sequenced, the similarity vector is determined by adopting the scene demand vector of the current user and the existing scene vectors of the preset scene, searching is conducted based on the similarity vector and the candidate targets are sequenced, the final result is enabled to be more consistent with the scene required by the current user, and therefore the accuracy of the searching result is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.
Fig. 1 is a flowchart of a search method according to a first embodiment of the present application;
FIG. 2 is a schematic structural diagram of a scenarized coding layer in an embodiment of the present application;
FIG. 3 is a schematic structural diagram of a ranking model in an embodiment of the present application;
fig. 4 is a schematic structural diagram of a search apparatus according to a second embodiment of the present application;
fig. 5 is a schematic structural diagram of an electronic device according to a third embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Example one
As shown in fig. 1, the searching method provided by this embodiment includes: step 110 to step 140.
The user information is related information used for judging a scene required by a user, and the user information can be selected from at least one of user portrait information, user position information, historical payment information, historical click information and historical browsing information. The user portrait refers to abstracting each concrete information of the user into labels under the background of the big data era, and the labels are utilized to concretize the user image, so that targeted service is provided for the user. The user portrait information comprises information such as user age, gender, professional background, standing area and the like. The user location information includes current location information of a current user. The historical browsing information may be information of a preset number of commodities browsed recently by the current user, for example, information of a preset number of commodities browsed before the current time, and a scene meeting the current requirements of the current user can be summarized by selecting the information of the recently browsed commodities.
When a search request of a current user is received, a scene to which a target searched by the current user belongs can be determined, so that the target of a corresponding scene is recommended for the current user. In the embodiment of the present application, when determining the scene to which the target queried by the current user belongs, the scene is determined according to all user information of the current user related to the determination scene, and therefore, the user information of the current user needs to be acquired. The search request may be a request for actively searching by the current user, or may be a request for recommending to the current user.
User portrait information of a current user is acquired, and user Location information is acquired through Location Based Services (LBS) under the authorization of the current user, and the user portrait information and the user Location information are used as basic user information. The location-based service is to acquire the current location of the positioning device by using various types of positioning technologies, and provide information resources and basic services for the positioning device through the mobile internet.
And acquiring historical behavior information of the current user from a database storing historical behavior information of the user, such as historical payment prices of the current user, historical clicked commodity information and historical browsed commodity information.
And 120, encoding the user information into a scene demand vector of the current user, wherein the dimension of the scene demand vector is a preset dimension.
Since the user information is related information capable of judging a scene required by the user, the scene demand vector of the current user can be obtained by processing the user information. Respectively coding the user basic information and the historical behavior information in the user information, splicing the coded user basic information and the coded historical behavior information into a vector, and processing the vector to map the vector into a vector with a preset dimension as a scene demand vector of the current user. When the scene demand vector of the current user is determined, all information related to the scene of the current user can be used instead of only part of information, so that the determined scene demand vector is more accurate, and the accuracy of a subsequent search result can be improved.
In an embodiment of the present application, the encoding the user information into a scene requirement vector of a current user includes: and processing the user information through a deep learning network to obtain a scene demand vector of the current user.
The deep learning network is a network model trained in advance, and network parameters are obtained through training and learning.
After the user information of the current user is acquired, the user information is input into a deep learning network, the user information is processed into a vector with preset dimensionality through the deep learning network, and the vector is used as a scene demand vector of the current user. The user information is processed through the deep learning network, and the obtained scene demand vector better meets the scene demand of the user.
In an embodiment of the present application, the processing the user information through the deep learning network to obtain a scene demand vector of a current user includes:
encoding the user information into an input feature vector of a deep learning network; and carrying out multilayer full-connection processing on the input characteristic vector through a deep learning network to obtain a scene demand vector of the current user.
User portrait information, user position information, historical payment information, historical click information and historical browsing information in the user information are respectively encoded, and the encoded information is spliced into a vector which is an input feature vector of the deep learning network. The deep learning network is composed of a plurality of fully-connected layers, input feature vectors are input into the deep learning network, the input feature vectors are subjected to multi-layer fully-connected processing by the fully-connected layers in the deep learning network, the input feature vectors are processed into vectors with preset dimensionality, the vectors are scene demand vectors of current users, namely, the input feature vectors are subjected to dimensionality reduction processing through the deep learning network, the scene demand vectors are obtained, and subsequent calculation is facilitated.
For example, the information such as gender, age and the like in the user portrait information is respectively encoded, the user position information is encoded, the information such as historical payment price and historically browsed commodity information in the historical behavior information is encoded, the encoded information is spliced to obtain a vector which is an input feature vector of the deep learning network, the input feature vector has a large dimension such as 100 dimensions, the input feature vector is processed through the deep learning network, the input feature vector is mapped to a vector with preset dimensions such as 100 dimensions, and the obtained vector is used as a scene demand vector of the current user.
Wherein the preset number is the total number of the preset scenes.
And respectively coding preset scenes with preset quantity to obtain an initial vector, and carrying out multilayer full-connection processing on the initial vector through a deep learning network to map the initial vector into a vector with preset dimensionality to obtain a scene vector corresponding to the preset scenes. The preset scene may be expanded, and when the preset scene is expanded, the preset scene may be mapped to a scene vector.
The preset scenes with preset quantity can be mapped into scene vectors in advance, the scene vectors corresponding to the preset scenes are stored, and when the current required scene of the current user needs to be obtained, the scene vector corresponding to each preset scene can be obtained from the storage position of the scene vectors.
The scene demand vectors are obtained according to the user information of the current user, the preset number of scene vectors are used for representing the preset number of scenes, and therefore the similarity between the scene demand vectors and the preset number of scene vectors is calculated, the preset number of similarity is the probability distribution of the current user belonging to the preset number of scenes, and the obtained preset number of similarity can be used for describing the scene required by the current user.
And respectively calculating the similarity between the scene demand vector and each of the preset number of scene vectors to obtain the preset number of similarities. The similarity can be cosine similarity, and the cosine similarity is obtained by calculating the cosine value of an included angle between two vectors to evaluate the similarity between the two vectors. When the similarity between a scene demand vector and a scene vector is calculated, the cosine value of the included angle between the scene demand vector and the scene vector can be calculated, and the cosine value of the included angle is used as the similarity between the scene demand vector and the scene vector. Of course, besides cosine similarity, other similarities may also be used, and are not limited herein.
The process of calculating the similarity may be used as a scene coding layer of a sequencing model, fig. 2 is a schematic structural diagram of the scene coding layer in the embodiment of the present application, and as shown in fig. 2, a user signal including user basic information and historical behavior information is processed to obtain a scene demand vector of a current user, and similarity calculation is performed on the scene demand vector and scene vectors corresponding to a preset number of preset scenes respectively to obtain similarity vectors corresponding to the preset number of preset scenes of the current user. In fig. 2, K is a preset number, i.e., the number of preset scenes.
The preset scene can be expanded as required, and when the preset scene is expanded, only the scene vectors corresponding to the preset scene need to be stored, so that when the similarity is determined, the scene vectors of all the preset scenes can be obtained from the storage positions of the scene vectors, the similarity between the scene demand vector and each scene vector is respectively calculated, a plurality of similarities can be obtained, and the obtained similarities can be used for representing the scene required by the current user. Therefore, the expansibility of scenes is improved, and more scenes can be represented by using a preset number of scenes through calculating the similarity, so that the scenes are maintained only by maintaining the preset number of scenes, the problem of poor maintainability caused by the fact that the scenes are divided more finely is solved, and the maintenance cost is reduced.
And 150, searching and sequencing a plurality of candidate targets according to the preset number of similarities.
When the search request is a request for actively searching by the current user, searching according to the keywords in the search request and in combination with the similarity of the preset number to obtain a plurality of candidate targets; and when the search request is a request for recommending to the current user, searching according to the similarity of the preset number to obtain a plurality of candidate targets. After the plurality of candidate targets are obtained through searching, the plurality of candidate targets are ranked, and therefore a ranking result is obtained. Wherein the candidate target may comprise a hotel service or a good.
For example, in the field of hotel searching, a preset number of preset scenes, such as local, remote, new, old, android, iphone, parent-child searching and the like, may be preset, each preset scene is encoded to obtain a scene vector corresponding to each preset scene, the scene vector corresponding to the preset scene is stored, when the current user searches for a hotel, user information of the current user may be encoded into one vector as a scene demand vector, and the similarity between the scene demand vector and each preset scene is calculated respectively, so as to obtain a plurality of similarities, which may represent the probability distribution of the user to each scene, may represent the current required scene of the user comprehensively, for example, when the user uses a large number of parent-child searching scenes historically, the similarity between the scene demand vector and the parent-child searching scene is large, so as to search and sort candidate hotel types based on the plurality of similarities, and the search result is more consistent with the scene required by the current user. In the field of commodity searching, if a commodity is a garment, a preset number of preset scenes can be preset as scenes such as men, women, mothers and babies, each preset scene is coded to obtain a scene vector corresponding to the preset scene, the scene vector corresponding to the preset scene is stored, when a current user searches the garment, the similarity can be determined based on user information and the scene vector corresponding to the preset scene, and the garment can be searched and sorted based on the similarity.
In an embodiment of the present application, the ranking the multiple candidate targets according to the preset number of similarities includes: splicing the similarity of the preset number into a similarity vector according to the arrangement sequence of the preset scenes; searching according to the similarity vector to obtain a plurality of candidate targets; inputting the similarity vectors and the characteristics of a plurality of candidate targets into a ranking model respectively to obtain ranking scores of the candidate targets; and sorting the candidate targets according to the sorting scores of the candidate targets.
The ranking model is a model for ranking the candidate targets, and may be used to calculate ranking scores corresponding to the candidate targets.
The preset scenes with the preset number have a certain arrangement sequence, the similarity with the preset number is arranged according to the arrangement sequence of the preset scenes, and the similarity is spliced into a similarity vector corresponding to the current user. And searching according to the similarity vector to obtain a plurality of candidate targets.
When a candidate target is recommended to a current user, after a similarity vector corresponding to the current user is determined, the similarity vector and the features of the candidate target are jointly used as input features of a ranking model, and a ranking score of the candidate target is calculated. Fig. 3 is a schematic structural diagram of a ranking model in an embodiment of the present application, and as shown in fig. 3, input features of the ranking model include similarity vectors and features of candidate targets, the features of the candidate targets include continuous features and discrete features, the similarity vectors, the continuous features, and the discrete features after cross processing are spliced into the input features of the ranking model, the input features are input into the ranking model, the input features are subjected to full connection processing through a multi-layer full connection layer of the ranking model, the features and the discrete features after the full connection processing are processed through a sigmoid function to obtain ranking scores of the candidate targets, after a ranking score is obtained by calculating each candidate target, the candidate targets can be ranked based on the ranking scores, and the candidate targets are recommended to a current user according to the ranking. The continuous features may be statistical features of the candidate targets, and may include, for example, the number of times that the candidate targets are placed in the latest preset time, the number of clicks of the candidate targets by the current user in the latest preset time, and the like, the discrete features may be identifiers of the candidate targets, features of the current time, and features of the current user, and the features of the current time may include, for example, date, day of the week, holiday information, and the like, and the features of the current user may include, for example, age, gender, address, occupation, and the like.
The similarity vector and the characteristics of the candidate targets are jointly used as the input characteristics of the ranking model, so that the ranking model can calculate the ranking scores of the candidate targets by combining the similarity vector, the calculated ranking scores are more consistent with the expectation of a current user, the ranking results of the candidate targets are more accurate, and the accuracy of the search results can be improved.
The searching method provided by the embodiment of the application is suitable for recommending candidate targets to a user according to different scenes, for example, the method can be suitable for the field of hotel searching, in the field of hotel searching, the preset scenes can comprise one or more of local/different places, new/old guests, android/iphone, parent-child searching, business trips, traveling and the like, different requirements can be met for users in different scenes, for example, the user generally requires the hotel to be located near a scenic spot during traveling, the hotel is generally required to be a hotel in a business area during traveling, two different scenes for parent-child searching and traveling can be overlapped, the searching method can be used for determining the scene needed by the user by integrating all the preset scenes, and even if the preset scenes are overlapped, the similarity vector which is more in line with the user requirements can be determined.
The searching method provided by the embodiment of the application comprises the steps of obtaining user information of a current user, coding the user information into scene demand vectors of the current user, obtaining a preset number of scene vectors, respectively determining the similarity between the scene demand vectors and the preset number of scene vectors to obtain the preset number of similarities, searching according to the preset number of similarities and sequencing a plurality of candidate targets, because the similarity vector is determined by adopting the scene demand vector of the current user and the scene vector of the existing preset scene, the search is carried out based on the similarity vector and the candidate targets are sequenced, the final result is more in line with the scene required by the current user, therefore, the accuracy of the search result is improved, the scenes required by the user and the scenes to which the candidate targets belong can be automatically found, and the difficulty in manually extracting the corresponding scenes of the commodities can be solved.
On the basis of the above technical solution, after searching and ranking the plurality of candidate targets according to the preset number of similarities, the method further includes: returning the sorting result to the current user; or recommending the sequencing result to the current user.
When the search request is a request for actively searching by the current user, after a plurality of candidate targets are searched and ranked, returning the ranking result to the current user, thereby displaying the ranking result in a client corresponding to the current user; and when the search request is a request for recommending the current user, after a plurality of candidate targets are searched and ranked, recommending the ranking result to the current user, and displaying the ranking result in a client corresponding to the current user. Therefore, based on the multiple similarity degrees obtained in the method of the embodiment of the application, the result can be obtained by searching according to the active search request of the current user, and the recommendation result can also be obtained by searching according to the recommendation request of the current user.
Example two
In the present embodiment, as shown in fig. 4, the search apparatus 400 includes:
a user signal obtaining module 410, configured to obtain user information of a current user;
a user signal encoding module 420, configured to encode the user information into a scene demand vector of a current user, where a dimension of the scene demand vector is a preset dimension;
a scene vector obtaining module 430, configured to obtain a preset number of scene vectors, where each scene vector is a vector of a preset dimension corresponding to a preset scene;
a similarity determining module 440, configured to determine similarities of the scene demand vectors and the preset number of scene vectors respectively, so as to obtain a preset number of similarities;
the sorting module 450 is configured to search and sort the multiple candidate targets according to the preset number of similarities.
Optionally, the user signal encoding module is specifically configured to:
and processing the user information through a deep learning network to obtain a scene demand vector of the current user.
Optionally, the user signal encoding module includes:
the user signal coding unit is used for coding the user information into an input characteristic vector of the deep learning network;
and the scene requirement determining unit is used for carrying out multilayer full-connection processing on the input characteristic vector through a deep learning network to obtain a scene requirement vector of the current user.
Optionally, the sorting module includes:
the similarity vector determining unit is used for splicing the preset number of similarities into a similarity vector of the current user according to the arrangement sequence of the preset scenes;
the searching unit is used for searching according to the similarity vector to obtain a plurality of candidate targets;
the ranking score determining unit is used for inputting the similarity vectors and the characteristics of the candidate targets into a ranking model respectively to obtain ranking scores of the candidate targets;
and the sorting unit is used for sorting the candidate targets according to the sorting scores of the candidate targets.
Optionally, the apparatus further comprises:
the search result returning module is used for returning the sequencing result to the current user; or
And the search result recommending module is used for recommending the sequencing result to the current user.
Optionally, the user information includes at least one of user portrait information, user location information, historical payment information, historical click information, and historical browsing information.
The search apparatus provided in the embodiment of the present application is configured to implement each step of the search method described in the first embodiment of the present application, and for specific implementation of each module of the apparatus, reference is made to the corresponding step, which is not described herein again.
According to the searching device provided by the embodiment of the application, the user signal obtaining module obtains the user information of the current user, the user signal coding module codes the user information into the scene demand vector of the current user, the scene vector obtaining module obtains the scene vectors of the preset number, the similarity determining module determines the similarity between the scene demand vector and the scene vectors of the preset number respectively to obtain the similarity of the preset number, the sorting module searches according to the similarity of the preset number and sorts a plurality of candidate targets, the similarity vector is determined by adopting the scene demand vector of the current user and the existing scene vectors of the preset scene, the search is carried out based on the similarity vector and the candidate targets are sorted, the final result is enabled to be more accordant with the scene required by the current user, and therefore the accuracy of the search result is improved.
EXAMPLE III
Embodiments of the present application also provide an electronic device, as shown in fig. 5, the electronic device 500 may include one or more processors 510 and one or more memories 520 connected to the processors 510. Electronic device 500 may also include input interface 530 and output interface 540 for communicating with another apparatus or system. Program code executed by processor 510 may be stored in memory 520.
The processor 510 in the electronic device 500 calls the program code stored in the memory 520 to perform the search method in the above-described embodiment.
The above elements in the above electronic device may be connected to each other by a bus, such as one of a data bus, an address bus, a control bus, an expansion bus, and a local bus, or any combination thereof.
Accordingly, the present application further provides a computer-readable storage medium, on which a computer program is stored, where the computer program is executed by a processor to implement the steps of the search method according to the first embodiment of the present application.
The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the device embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, refer to the partial description of the method embodiment.
The search method, the search device, the electronic device, and the storage medium provided in the embodiments of the present application are described in detail above, and a specific example is applied in the present application to explain the principle and the implementation of the present application, and the description of the above embodiments is only used to help understand the method and the core idea of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Claims (10)
1. A search method, comprising:
acquiring user information of a current user;
encoding the user information into a scene demand vector of a current user, wherein the dimension of the scene demand vector is a preset dimension;
acquiring a preset number of scene vectors, wherein each scene vector is a vector of a preset dimension corresponding to a preset scene;
respectively determining the similarity of the scene demand vectors and the preset number of scene vectors to obtain the preset number of similarities;
and searching and sequencing a plurality of candidate targets according to the similarity of the preset number.
2. The method of claim 1, the encoding the user information as a scene requirement vector for a current user, comprising:
and processing the user information through a deep learning network to obtain a scene demand vector of the current user.
3. The method of claim 2, wherein the processing the user information through the deep learning network to obtain a scene requirement vector of a current user comprises:
encoding the user information into an input feature vector of a deep learning network;
and carrying out multilayer full-connection processing on the input characteristic vector through a deep learning network to obtain a scene demand vector of the current user.
4. The method of claim 1, said ranking the plurality of candidate targets according to the preset number of similarities, comprising:
splicing the similarity of the preset number into a similarity vector according to the arrangement sequence of the preset scenes;
searching according to the similarity vector to obtain a plurality of candidate targets;
inputting the similarity vectors and the characteristics of a plurality of candidate targets into a ranking model respectively to obtain ranking scores of the candidate targets;
and sorting the candidate targets according to the sorting scores of the candidate targets.
5. The method according to any one of claims 1-4, further comprising, after searching and ranking the plurality of candidate objects according to the preset number of similarities:
returning the sorting result to the current user; or
And recommending the sequencing result to the current user.
6. The method of any of claims 1-4, the user information including at least one of user portrait information, user location information, historical payment information, historical click information, and historical browsing information.
7. A search apparatus, comprising:
the user signal acquisition module is used for acquiring the user information of the current user;
the user signal coding module is used for coding the user information into a scene demand vector of the current user, and the dimension of the scene demand vector is a preset dimension;
the scene vector acquisition module is used for acquiring a preset number of scene vectors, and each scene vector is a vector of a preset dimension corresponding to a preset scene;
the similarity determining module is used for respectively determining the similarity of the scene demand vectors and the preset number of scene vectors to obtain the preset number of similarities;
and the sorting module is used for searching and sorting the candidate targets according to the preset number of similarities.
8. The apparatus of claim 7, the user signal encoding module being specifically configured to:
and processing the user information through a deep learning network to obtain a scene demand vector of the current user.
9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the search method of any one of claims 1 to 6 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the search method of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010209482.7A CN111523962A (en) | 2020-03-23 | 2020-03-23 | Searching method, searching device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010209482.7A CN111523962A (en) | 2020-03-23 | 2020-03-23 | Searching method, searching device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111523962A true CN111523962A (en) | 2020-08-11 |
Family
ID=71900992
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010209482.7A Pending CN111523962A (en) | 2020-03-23 | 2020-03-23 | Searching method, searching device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111523962A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112684907A (en) * | 2020-12-24 | 2021-04-20 | 科大讯飞股份有限公司 | Text input method, device, equipment and storage medium |
| CN113569138A (en) * | 2021-07-08 | 2021-10-29 | 深圳Tcl新技术有限公司 | Intelligent device control method and device, electronic device and storage medium |
-
2020
- 2020-03-23 CN CN202010209482.7A patent/CN111523962A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112684907A (en) * | 2020-12-24 | 2021-04-20 | 科大讯飞股份有限公司 | Text input method, device, equipment and storage medium |
| CN112684907B (en) * | 2020-12-24 | 2024-04-26 | 科大讯飞股份有限公司 | Text input method, device, equipment and storage medium |
| CN113569138A (en) * | 2021-07-08 | 2021-10-29 | 深圳Tcl新技术有限公司 | Intelligent device control method and device, electronic device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111061946B (en) | Method, device, electronic equipment and storage medium for recommending scenerized content | |
| US10846534B1 (en) | Systems and methods for augmented reality navigation | |
| US11397772B2 (en) | Information search method, apparatus, and system | |
| CN104199896B (en) | The video similarity of feature based classification is determined and video recommendation method | |
| CN107924404A (en) | Eliminate the ambiguity of search inquiry | |
| CN108596695B (en) | Entity pushing method and system | |
| CN111695960A (en) | Object recommendation system, method, electronic device and storage medium | |
| KR101981136B1 (en) | Program, method and system for providing recommendation contents | |
| CN111523962A (en) | Searching method, searching device, electronic equipment and storage medium | |
| CN112488781A (en) | Search recommendation method and device, electronic equipment and readable storage medium | |
| KR20210032691A (en) | Method and apparatus of recommending goods based on network | |
| CN112364184A (en) | Method, device, server and storage medium for ordering multimedia data | |
| CN112036987B (en) | Method and device for determining recommended commodity | |
| CN114528474A (en) | Method and device for determining recommended object, electronic equipment and storage medium | |
| US20190318386A1 (en) | Information processing method, program, information processing system, and information processing apparatus | |
| CN112115354A (en) | Information processing method, information processing apparatus, server, and storage medium | |
| CN116611891A (en) | Content information recommendation method, device, server and storage medium | |
| CN112182126A (en) | Model training method and device for determining matching degree, electronic equipment and readable storage medium | |
| CN110309406B (en) | Click rate estimation method, device, equipment and storage medium | |
| CN116957128A (en) | Service index prediction method, device, equipment and storage medium | |
| CN111831686A (en) | Optimization method, device and system of sequencing model, electronic equipment and storage medium | |
| CN112035740B (en) | Project use time length prediction method, device, equipment and storage medium | |
| CN114399352A (en) | Information recommendation method and device, electronic equipment and storage medium | |
| CN110879863B (en) | Cross-domain search method and cross-domain search device | |
| CN113688295B (en) | Data determination method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20200811 |