CN113127669B - Advertisement mapping method, device, equipment and storage medium - Google Patents
Advertisement mapping method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN113127669B CN113127669B CN202010041786.7A CN202010041786A CN113127669B CN 113127669 B CN113127669 B CN 113127669B CN 202010041786 A CN202010041786 A CN 202010041786A CN 113127669 B CN113127669 B CN 113127669B
- Authority
- CN
- China
- Prior art keywords
- node
- advertisement
- nodes
- picture
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/5866—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, manually generated location and time information
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- 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
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0276—Advertisement creation
Abstract
The embodiment of the application discloses a method, a device, equipment and a storage medium for advertisement mapping, and relates to the technical field of intelligent searching. The specific implementation scheme is as follows: searching according to the keyword nodes and the picture nodes in the heterograms to obtain sample node pairs comprising content similarity information and structure similarity information between the keywords and the pictures; determining node vector representations in the heterograms according to the sample node pairs to obtain keyword node vector representations and picture node vector representations; and determining similarity relations among nodes according to node vector representations in the heterograms, wherein the similarity relations are used for retrieving the target advertisement map of the advertisement to be mapped according to target keyword node vector representations of the advertisement to be mapped. The heterogeneous triggering of the advertisement to the distribution map is directly realized by mining the content similarity and the structure similarity of the keywords and the pictures through the heterogeneous map, the dependence on the picture description text is eliminated, and the accuracy of picture recall and the quality of advertisement distribution map are improved.
Description
Technical Field
The embodiment of the application relates to the technical field of computers, in particular to the technical field of intelligent searching, and particularly relates to an advertisement mapping method, device, equipment and storage medium.
Background
With the rapid development of the internet market, people have increasingly high requirements for advertisements. In order to enrich the advertisement presentation information, improve the user experience and the rendering efficiency of advertisements, it is necessary to map advertisements based on massive pictures in the internet. However, the picture recall mode based on text matching or semantic matching in the prior art has strong dependence on picture description text, and text semantics have diversity, so that the accuracy of picture recall is reduced.
Disclosure of Invention
The embodiment of the application provides an advertisement mapping method, device, equipment and storage medium, which can improve the accuracy of picture recall and advertisement mapping quality.
In a first aspect, an embodiment of the present application provides an advertisement mapping method, including:
searching according to the keyword nodes and the picture nodes in the heterograms to obtain sample node pairs comprising content similarity information and structure similarity information between the keywords and the pictures;
determining node vector representations in the heterograms according to the sample node pairs to obtain keyword node vector representations and picture node vector representations;
and determining similarity relations among nodes according to node vector representations in the heterograms, wherein the similarity relations are used for retrieving the target advertisement matching chart of the advertisement to be matched according to target keyword node vector representations of the advertisement to be matched.
One embodiment of the above application has the following advantages or benefits: the heterogeneous trigger of advertisement to picture allocation is directly realized by mining the content similarity and the structure similarity of keywords and pictures through the heterogeneous pictures, the dependence on picture description texts is eliminated, the defect of inaccurate picture allocation caused by semantic diversity is avoided, and the accuracy of picture recall and the quality of advertisement picture allocation are improved.
Optionally, before searching the keyword node and the picture node in the heterogram to obtain a sample node pair including content similarity information and structure similarity information between the keyword and the picture, the method further includes:
and constructing the heterograms according to the keywords and the pictures in the advertisement logs.
One embodiment of the above application has the following advantages or benefits: and mining advertisement keywords and advertisement pictures by using the advertisement logs, and constructing a heterogeneous graph consisting of the keywords and the pictures, so as to provide basis for mining similarity relations between the keywords and the pictures in terms of content and structure.
Optionally, the constructing the heterogeneous graph according to the keywords and the pictures in the advertisement log includes:
extracting keywords and pictures from the advertisement log, and determining keyword nodes and picture nodes in the heterograms;
Determining edges between keyword nodes and picture nodes in the iso-composition according to the keywords and advertising sources of the pictures;
and determining the weight of the edge in the abnormal composition according to the showing times and clicking times of the same advertisement to which the keyword and the picture belong.
One embodiment of the above application has the following advantages or benefits: through node determination, edge formation among nodes and edge weight calculation, a weighted undirected heterogeneous graph is constructed, a basis is provided for mining of similarity relations between keywords and pictures in content and structure, and good construction of different patterns is a foundation for mining of similarity relations among nodes.
Optionally, the searching according to the keyword node and the picture node in the heterogram to obtain a sample node pair including content similarity information and structure similarity information between the keyword and the picture includes:
based on a Node2vec algorithm, carrying out random walk according to the weight of the edge in the heterogram to generate a Node sequence containing content similarity information and structure similarity information between the keywords and the pictures;
sample node pairs are determined from the node sequence.
One embodiment of the above application has the following advantages or benefits: the node sequences fusing the content similarity information and the structure similarity information are obtained by random walk of the heterogeneous graphs, so that the similarity is conveyed, a sample node pair comprising the content similarity information and the structure similarity information between the keywords and the pictures is obtained, and a good basis is provided for the subsequent node vector representation learning.
Optionally, the generating a node sequence including content similarity information and structure similarity information between the keywords and the pictures by performing random walk according to the weight of the edges in the heterograms includes:
performing breadth-first search according to the weight of the edge in the heterogram to obtain content similarity nodes with content similarity;
performing depth-first search according to the weight of the edge in the heterogram to obtain a structural similarity node with structural similarity;
and generating a node sequence containing content similarity information and structure similarity information between the keywords and the pictures according to the content similarity nodes and the structure similarity nodes.
One embodiment of the above application has the following advantages or benefits: the node sequence integrating the content similarity information and the structure similarity information is obtained by performing breadth-first search and depth-first search on the heterogeneous graph, the node types are not distinguished in the node sequence, the adjacent nodes can be similar in content or structure, and the similarity relation between the nodes is fully mined.
Optionally, the determining the sample node pair according to the node sequence includes:
Traversing the node sequence by adopting a sliding window to obtain a positive sample node pair containing content similarity information and structure similarity information between the keywords and the pictures;
and extracting a negative sample node pair from the heterogeneous graph according to the positive sample node pair.
One embodiment of the above application has the following advantages or benefits: through traversing the node sequence by the sliding window, nodes in the window form positive sample node pairs, and similarity is continued to the positive sample node pairs, namely the nodes in the positive sample node pairs are similar in content or structure. And extracting the opposite negative sample node pair of the positive sample node pair from the heterogeneous graph, i.e. there is no similarity in content or structure between nodes in the negative sample node pair. Therefore, a sufficient basis is provided for the learning of the node vector representation.
Optionally, the determining a node vector representation in the heterogram according to the sample node pair includes:
and obtaining the node vector representation in the heterogeneous graph according to the sample node pairs based on a skip-gram algorithm.
One embodiment of the above application has the following advantages or benefits: according to the sample node pairs, the node vector representation is learned, and the node vector representation fused with the similarity relation between the nodes can be learned.
Optionally, the retrieving the target advertisement map for the advertisement to be mapped according to the target keyword node vector representation of the advertisement to be mapped includes:
determining target keywords of the advertisement to be matched according to the advertisement to be matched;
determining a target keyword node vector representation according to the heterogram and the target keyword;
and based on an approximate nearest neighbor algorithm, carrying out picture index retrieval according to the target keyword node vector representation to obtain the target advertisement map of the advertisement to be mapped.
One embodiment of the above application has the following advantages or benefits: based on the similarity relation among the nodes covered in the node vector representation, the target advertisement matching graph similar to the target keyword in content and structure can be retrieved from a plurality of pictures according to the target keyword node vector representation of the advertisement to be matching graph. The index is established for the picture nodes by adopting an approximate nearest neighbor algorithm, so that the efficiency of picture retrieval based on the picture index can be greatly improved. Therefore, the mode of searching the picture through the node vector representation gets rid of the dependence on the picture description text, avoids the defect of inaccurate picture allocation caused by semantic diversity, and improves the picture recall accuracy and the advertisement picture allocation quality.
In a second aspect, an embodiment of the present application provides an advertisement mapping apparatus, including:
the sample node pair determining module is used for searching according to the keyword nodes and the picture nodes in the heterograms to obtain sample node pairs comprising content similarity information and structure similarity information between the keywords and the pictures;
the node representation determining module is used for determining node vector representations in the heterograms according to the sample node pairs so as to obtain keyword node vector representations and picture node vector representations;
and the picture retrieval module is used for determining the similarity relation between the nodes according to the node vector representation in the heterogram and retrieving the target advertisement matching diagram of the advertisement to be matched according to the target keyword node vector representation of the advertisement to be matched.
In a third aspect, an embodiment of the present application provides an electronic device, including:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein, the liquid crystal display device comprises a liquid crystal display device,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the advertisement mapping method of any embodiment of the present application.
In a fourth aspect, embodiments of the present application provide a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the advertisement mapping method according to any of the embodiments of the present application.
One embodiment of the above application has the following advantages or benefits: and mining sample node pairs comprising content similarity information and structure similarity information between keywords and pictures based on heterogeneous pictures comprising keyword nodes and picture nodes, and determining each node vector representation according to the sample node pairs, so that when the pictures are matched with the advertisements to be matched, picture retrieval is carried out according to the node vector representations of the target keywords of the advertisements to be matched based on similarity relations among the nodes covered by the node vector representations, and the target advertisement matching pictures of the advertisements to be matched are obtained. According to the embodiment of the application, the content similarity and the structure similarity of the keywords and the pictures are mined through the different composition, the heterogeneous triggering of the advertisement to the distribution picture is directly realized, the dependence on the picture description text is eliminated, the defect of inaccurate distribution picture caused by semantic diversity is avoided, and the picture recall accuracy and the advertisement distribution picture quality are improved.
Other effects of the above alternative will be described below in connection with specific embodiments.
Drawings
The drawings are included to provide a better understanding of the present application and are not to be construed as limiting the application. Wherein:
FIG. 1 is a flow chart of an advertising scheme method according to a first embodiment of the application;
fig. 2 is an exemplary diagram of an iso-pattern according to a first embodiment of the present application;
FIG. 3 is a flow chart of an advertising scheme method according to a second embodiment of the present application;
FIG. 4 is a flow chart of an advertisement mapping method according to a third embodiment of the present application;
FIG. 5 is a schematic diagram of an advertisement mapping apparatus according to a fourth embodiment of the present application;
fig. 6 is a block diagram of an electronic device for implementing the advertising scheme method of an embodiment of the application.
Detailed Description
Exemplary embodiments of the present application will now be described with reference to the accompanying drawings, in which various details of the embodiments of the present application are included to facilitate understanding, and are to be considered 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 application. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
First embodiment
Fig. 1 is a flowchart of an advertisement mapping method according to a first embodiment of the present application, where the present embodiment is applicable to a situation of providing mapping for an advertisement according to keywords in advertisement content, so as to enrich the presentation content of the advertisement, and facilitate improving the rendering efficiency of the advertisement. The method may be performed by an advertising profiling apparatus implemented in software and/or hardware, preferably deployed in an electronic device, such as an advertising platform server. As shown in fig. 1, the method specifically includes the following steps:
and S110, searching according to the keyword nodes and the picture nodes in the heterograms to obtain sample node pairs comprising content similarity information and structure similarity information between the keywords and the pictures.
In particular embodiments of the present application, the ad campaigns are typically provided by advertisers, but there are situations where advertisers are not willing to provide the campaigns, and situations where the ad presentation is not rich enough to require further campaigns. Therefore, in the prior art, core word matching is usually carried out on advertisement text and candidate picture description text, or semantic matching is carried out on the basis of advertisement text semantics and picture description text semantics, so that pictures are recalled as advertisement map matching. However, the text matching usually has the condition of word ambiguity, for example, "apple" can represent fruit apple and mobile phone brands, thereby reducing the accuracy of text and semantic matching on recalled pictures, and having extremely high requirements on the accuracy of picture description text labeling.
Therefore, the embodiment breaks the dependence on the text in the picture recall process, adopts the heterogram to mine the similarity relation between the advertisement keywords and the picture, and improves the accuracy of the recall picture. The heterogeneous graph refers to a network graph formed by nodes and edges, wherein the nodes in the heterogeneous graph have different forms, and the relationships among the nodes in the heterogeneous graph also have different forms. Correspondingly, the embodiment adopts keywords and pictures as nodes in the heterograms. Wherein, the keywords refer to keywords in advertisements, such as auction words purchased by advertisers, etc.; the pictures refer to pictures which appear in advertisements and can be further expanded into a large number of pictures existing in the Internet.
Optionally, constructing an heterogram according to the keywords and the pictures in the advertisement log.
In this embodiment, a large number of advertisement logs may be collected from the advertisement platform in advance, which is used as a basis for constructing the heterograms. The advertisement log may include an advertisement presentation log and an advertisement click log. The advertisement log may include advertisement content data and advertisement picture data, and may also include other advertisement data within the page on which the advertisement is presented.
Specifically, keywords and pictures are extracted from the advertisement log, and keyword nodes and picture nodes in the heterograms are determined. And determining edges between the keyword nodes and the picture nodes in the heterograms according to the keywords and advertising sources of the pictures. And determining the weight of the edge in the heterogram according to the showing times and clicking times of the same advertisement to which the keyword and the picture belong.
For example, keywords and pictures related to advertisements in the advertisement log can be extracted as nodes in the heterograms. Edges between nodes can be formed between keywords in the same advertisement and between keywords and pictures. Accordingly, the same keyword may appear in a plurality of advertisements, and thus the keywords and pictures to which the same keyword is connected may include a plurality of. The weight of an edge can then be calculated using the following formula: w=α×show_num+ (1- α) ×click_num. The show_num represents the showing times of the same advertisement of the keyword and the picture or the same advertisement of the keyword and the keyword in the showing log; click_num represents the number of clicks of the same advertisement to which the keyword and the picture belong or the same advertisement to which the keyword and the keyword belong in the click log; alpha is a parameter with a value in the range of 0 to 1 for controlling the degree of influence of the number of presentations and clicks.
For another example, fig. 2 is an exemplary diagram of an iso-composition, as shown in fig. 2, a bidword represents a keyword node, and pic represents a picture node. The bid 1 and pic1 may be in the same advertisement at the same time, the bid 1 and pic3 may be in the same advertisement at the same time, but the bid 2 and pic3 may be in the same advertisement at the same time. The random walk is performed in the heterograph, so that nodes with content similarity, such as picture nodes directly connected with the bidword1, can be obtained, and nodes with structure similarity, such as bidword1 and bidword2, can be obtained. Thus, through the heterograms, the similarity relationship between nodes can be mined.
In this embodiment, the sample node pairs are obtained by iso-patterning for learning of the node vector representation. The sample node pairs include positive sample node pairs having content and/or structural similarity between nodes in the positive sample node pairs and negative sample node pairs having no similarity between nodes in the negative sample node pairs. By mining from the iso-graph, the similarity relationship between nodes in the iso-graph can be passed to the sample node pairs. Accordingly, the content similarity information refers to a sample node pair including a content similarity node, and the structure similarity information refers to a sample node pair including a structure similarity node.
Illustratively, random walk may be performed according to the weights of the edges in the iso-graph based on a Node2vec algorithm, node pairs with content similarity may be obtained through breadth first search (Breadth First Search, BFS), and Node pairs with structure similarity may be obtained through depth first search (Depth First Search, DFS). According to the connection relation of the edges between the nodes, the nodes under a complete path are arranged in sequence, and a node sequence of the content similarity node and the structure similarity node is obtained. And sampling the node sequence by adopting a sliding window to obtain positive sample node pairs which are similar in content or structure, and extracting nodes without any content or structure similarity from the heterogeneous graph as negative sample node pairs. The similarity relation among the nodes is transferred to the node sequence by the heterogeneous graph, and is transferred to the positive sample node pair by the node sequence.
S120, determining node vector representations in the heterograms according to the sample node pairs to obtain keyword node vector representations and picture node vector representations.
In a specific embodiment of the present application, the node types in the sample node pair are not limited, and the sample node pair may include nodes of the same type or may include nodes of different types. Because the positive sample node pairs contain content and/or structural similarity between nodes, and the negative sample node pairs do not contain any similarity, the similarity relationship is learned based on the sample node pairs to obtain a node vector representation, and then another node with the similarity relationship with the node can be directly searched through the node vector representation.
The present embodiment is not limited to the learning algorithm of the node vector representation, and any algorithm that can realize the node vector representation learning can be applied to the present embodiment. The node vector representation in the iso-graph is derived from the sample node pairs, e.g., based on skip-gram algorithm.
S130, according to node vector representation in the heterogram, similarity relation between nodes is determined, and the target advertisement matching diagram of the advertisement to be matched is retrieved according to the target keyword node vector representation of the advertisement to be matched.
In a specific embodiment of the application, in view of similarity relationships between nodes, the heterogeneous graph is transferred to a node sequence, the node sequence is transferred to a sample node pair, and the sample node pair is transferred to a node vector representation. Therefore, the node vector representation covers the similarity relation among the nodes, and the similarity relation among the nodes can be determined according to the node vector representation in the heterogram. Accordingly, the similarity relationship between nodes is used to retrieve content and/or structurally similar ad drawings for the ad to be drawn.
According to the technical scheme, based on the heterogeneous graph comprising the keyword nodes and the picture nodes, sample node pairs comprising content similarity information and structure similarity information between the keywords and the pictures are mined, and node vector representations are determined according to the sample node pairs, so that when the pictures are matched for the advertisements to be matched, picture retrieval is carried out according to the node vector representations of the target keywords of the advertisements to be matched, and the target advertisement matching pictures of the advertisements to be matched are obtained based on similarity relations among the nodes covered by the node vector representations. According to the embodiment of the application, the content similarity and the structure similarity of the keywords and the pictures are mined through the different composition, the heterogeneous triggering of the advertisement to the distribution picture is directly realized, the dependence on the picture description text is eliminated, the defect of inaccurate distribution picture caused by semantic diversity is avoided, and the picture recall accuracy and the advertisement distribution picture quality are improved.
Second embodiment
Fig. 3 is a flowchart of an advertisement mapping method according to a second embodiment of the present application, in which learning of node vector representations is further explained based on the first embodiment, a node sequence can be generated based on a heterogeneous graph, sample node pairs can be obtained from the node sequence, and node vector representations can be learned according to the sample nodes. As shown in fig. 3, the method specifically includes the following steps:
s310, based on a Node2vec algorithm, random walk is carried out according to the weight of the edges in the heterograms, and a Node sequence containing content similarity information and structure similarity information between the keywords and the pictures is generated.
In a specific embodiment of the present application, the Node2vec algorithm is a graph embedding (graph embedding) method that comprehensively considers the BFS neighborhood and the DFS neighborhood, and can be regarded as deep walk combining BFS and DFS random walk. The BFS can search to obtain direct neighbor nodes with content similarity, and the DFS can break through the neighbor node to search to obtain nodes with structure similarity. In the representation of the network, both content similarity and structural similarity are important, so that both similarities can be considered simultaneously when generating the node sequence, and there can be parameters, i.e. edge weights, to control one bias of both.
Optionally, performing breadth-first search according to the weight of the edge in the heterogram to obtain content similarity nodes with content similarity; performing depth-first search according to the weight of the edge in the heterogram to obtain a structural similarity node with structural similarity; and generating a node sequence containing content similarity information and structure similarity information between the keywords and the pictures according to the content similarity nodes and the structure similarity nodes.
Illustratively, taking the iso-graph in fig. 2 as an example, when BFS is adopted, random walk can be performed based on solid arrows, so as to obtain a direct neighbor node with content similarity to a bidword1 node, and at least the direct neighbor node can comprise a pic1 node, a pic2 node, a pic3 node and a pic4 node. When DFS is adopted, random walk can be performed based on a dotted arrow, so that a node with structural similarity with the bidword1 node can be obtained, and the node at least can comprise the bidword2 node. And wherein the weight of an edge determines how next a node goes, i.e. depends on its previous and next relationships. It is assumed that the node sequences pic2, bidword1, pic3, bidword2, pic6 are available.
S320, determining sample node pairs according to the node sequences.
In a specific embodiment of the present application, in view of the fusion of content similarity and structure similarity between nodes in a node sequence, node pairs with content and/or structure similarity relationships may be extracted from the node sequence as sample node pairs.
Alternatively, traversing the node sequence by adopting a sliding window to obtain a positive sample node pair containing content similarity information and structure similarity information between the keywords and the pictures; from the positive sample node pairs, negative sample node pairs are extracted from the heterograms.
Illustratively, in the above example, it is assumed that the sliding window length is 1, i.e., 1 node adjacent to the target node is extracted. For the node sequence { pic2, bidword1, pic3, bidword2, pic6}, a positive sample node pair may be obtained comprising: { pic2, bidword1}, { bidword1, pic3}, { pic3, bidword2} and { bidword2, pic6}. Wherein a positive sample node pair may represent that given keyword bidword1, the probability of recalling pictures pic2 and pic3 is greater than pic 6. And assuming that two negative sample node pairs are randomly extracted for each node, then the negative sampling algorithm is employed, assuming that the negative sample node pair for pic2 node is not { pic2, pic6} and { pic2, pic7} can be obtained.
S330, determining node vector representations in the heterograms according to the sample node pairs to obtain keyword node vector representations and picture node vector representations.
In a specific embodiment of the present application, optionally, based on a skip-gram algorithm, node vector representations in the heterograms are obtained according to the sample node pairs. The neural network is trained based on sample node pairs by adopting a skip-gram algorithm, and similarity relations and dissimilarity relations are learned to obtain node vector representation.
S340, according to node vector representation in the heterogram, similarity relation between nodes is determined, and the target advertisement matching diagram of the advertisement to be matched is retrieved according to the target keyword node vector representation of the advertisement to be matched.
According to the technical scheme, through constructing the heterogram comprising the keywords and the pictures, mining the content and/or structural similarity relation between the keywords and the pictures, transmitting the similarity relation between the nodes to the node sequence from the heterogram, transmitting the node sequence to the sample node pair from the node sequence, and transmitting the node pair to the sample node to represent the node vector, so that the target advertisement distribution diagram of the advertisement to be distributed can be retrieved according to the target keyword node vector representation of the advertisement to be distributed. The heterogeneous triggering of the advertisement to the distribution diagram is directly realized, the dependence on the picture description text is eliminated, the defect of inaccurate distribution diagram caused by semantic diversity is avoided, and the accuracy of picture recall and the quality of advertisement distribution diagram are improved.
Third embodiment
Fig. 4 is a flowchart of an advertisement matching method according to a third embodiment of the present application, where, based on the first embodiment, the retrieval of the advertisement matching based on the node vector representation is further explained, and the target advertisement matching can be retrieved according to the target keyword node vector representation of the advertisement to be matched based on the similarity relationship between the nodes included in the node vector representation. As shown in fig. 4, the method specifically includes the following steps:
S410, determining target keywords of the advertisement to be matched according to the advertisement to be matched.
In a specific embodiment of the present application, the advertisement to be mapped refers to an advertisement which is determined and is about to be displayed to a user when the user initiates a search. Whether the advertisement requires further mapping may be detected based on certain advertisement detection rules. The method for determining the advertisement of the map to be allocated is not limited in the embodiment, and any detection method which can meet the popularization requirement of the advertisement platform can be applied to the embodiment. For example, an advertisement that does not contain a map may be detected as a map-to-be-mapped advertisement, an advertisement that has a historical click rate lower than a click rate threshold may be detected as a map-to-be-mapped advertisement, and the like.
In this embodiment, text keywords may be extracted from advertisement text of an advertisement to be configured, auction words of an advertiser to which the advertisement to be configured belongs may be extracted, and character recognition may be performed on pictures of the advertisement to be configured to extract picture keywords, so that at least one of the text keywords, the auction words, and the picture keywords may be used as target keywords of the advertisement to be configured.
S420, determining a target keyword node vector representation according to the heterograms and the target keywords.
In the embodiment of the application, the target keyword can be matched with the keyword node in the heterogeneous graph, and the node vector representation of the matched keyword node is determined as the target keyword node vector representation.
And S430, carrying out picture retrieval according to the target keyword node vector representation to obtain the target advertisement map of the advertisement to be mapped.
In the embodiment of the application, for each target keyword, according to the target keyword node vector representation, searching is performed by traversing all the picture node vector representations, and at least one picture with the shortest distance, namely the most similar picture, is calculated based on the matching of the vector representations. The different target keywords may search out a plurality of pictures, and then vote according to the plurality of pictures of the different target keywords in the advertisement to be matched, and select the pictures which have similarity relation with all the target keywords as the target advertisement matching picture, and combine the pictures with the target keywords to be matched to the advertisement to be displayed to the user.
Optionally, based on an approximate nearest neighbor algorithm, the picture index retrieval is performed according to the target keyword node vector representation.
In this embodiment, an approximate nearest neighbor algorithm (Approximate Nearest Neighbor, ANN) may be employed to find the approximate nearest neighbor vector. Specifically, a classical algorithm product quantization algorithm (Product Quantization, PQ) series under an ANN algorithm can be adopted, and indexes are established for the picture nodes through k-means clustering and subspace division under each class. Therefore, ANN retrieval is performed based on the picture node index, and through spatial positioning, all pictures do not need to be traversed, so that the retrieval efficiency of the pictures is greatly improved.
According to the technical scheme, through determining the node vector representation of the target keyword of the advertisement to be matched, searching of the picture nodes is carried out based on the node vector representation of the target keyword, and pictures with similarity relation with the target keyword in content and/or structure are searched out to serve as the target advertisement matching. The heterogeneous triggering of the advertisement to the distribution diagram is directly realized, the dependence on the picture description text is eliminated, the defect of inaccurate distribution diagram caused by semantic diversity is avoided, and the accuracy of picture recall and the quality of advertisement distribution diagram are improved.
Fourth embodiment
Fig. 5 is a schematic structural diagram of an advertisement mapping device according to a fourth embodiment of the present application, where the present embodiment is applicable to a situation of providing mapping for advertisements according to keywords in advertisement content, so as to enrich the presentation content of the advertisements, and facilitate improving the rendering efficiency of the advertisements. The device can realize the advertisement mapping method according to any embodiment of the application. The apparatus 500 specifically includes the following:
the sample node pair determining module 510 is configured to search for a keyword node and a picture node in the heterogram to obtain a sample node pair including content similarity information and structure similarity information between the keyword and the picture;
A node representation determining module 520, configured to determine a node vector representation in the iso-graph according to the sample node pair, so as to obtain a keyword node vector representation and a picture node vector representation;
the picture retrieval module 530 is configured to determine a similarity relationship between nodes according to the node vector representation in the heterogram, and retrieve a target advertisement matching graph of the advertisement to be matched according to the target keyword node vector representation of the advertisement to be matched.
Further, the apparatus 500 further includes an iso-composition creation module 540, specifically configured to:
and before searching according to the keyword nodes and the picture nodes in the heterograms to obtain sample node pairs comprising content similarity information and structure similarity information between the keywords and the pictures, constructing the heterograms according to the keywords and the pictures in the advertisement logs.
Optionally, the heterogeneous map creation module 540 is specifically configured to:
extracting keywords and pictures from the advertisement log, and determining keyword nodes and picture nodes in the heterograms;
determining edges between keyword nodes and picture nodes in the iso-composition according to the keywords and advertising sources of the pictures;
And determining the weight of the edge in the abnormal composition according to the showing times and clicking times of the same advertisement to which the keyword and the picture belong.
Optionally, the sample node pair determining module 510 is specifically configured to:
based on a Node2vec algorithm, carrying out random walk according to the weight of the edge in the heterogram to generate a Node sequence containing content similarity information and structure similarity information between the keywords and the pictures;
sample node pairs are determined from the node sequence.
Optionally, the sample node pair determining module 510 is specifically configured to:
performing breadth-first search according to the weight of the edge in the heterogram to obtain content similarity nodes with content similarity;
performing depth-first search according to the weight of the edge in the heterogram to obtain a structural similarity node with structural similarity;
and generating a node sequence containing content similarity information and structure similarity information between the keywords and the pictures according to the content similarity nodes and the structure similarity nodes.
Optionally, the sample node pair determining module 510 is specifically configured to:
traversing the node sequence by adopting a sliding window to obtain a positive sample node pair containing content similarity information and structure similarity information between the keywords and the pictures;
And extracting a negative sample node pair from the heterogeneous graph according to the positive sample node pair.
Optionally, the node representation determining module 520 is specifically configured to:
and obtaining the node vector representation in the heterogeneous graph according to the sample node pairs based on a skip-gram algorithm.
Optionally, the picture retrieval module 530 is specifically configured to:
determining target keywords of the advertisement to be matched according to the advertisement to be matched;
determining a target keyword node vector representation according to the heterogram and the target keyword;
and based on an approximate nearest neighbor algorithm, carrying out picture index retrieval according to the target keyword node vector representation to obtain the target advertisement map of the advertisement to be mapped.
According to the technical scheme, through mutual coordination among the functional modules, the functions of keyword node determination, picture node determination, iso-composition establishment, node sequence generation, sample node pair generation, node vector representation learning, picture index establishment, picture retrieval and the like are realized. According to the embodiment of the application, the content similarity and the structure similarity of the keywords and the pictures are mined through the different composition, the heterogeneous triggering of the advertisement to the distribution picture is directly realized, the dependence on the picture description text is eliminated, the defect of inaccurate distribution picture caused by semantic diversity is avoided, and the picture recall accuracy and the advertisement distribution picture quality are improved.
Fifth embodiment
According to an embodiment of the present application, the present application also provides an electronic device and a readable storage medium.
As shown in fig. 6, a block diagram of an electronic device of an advertisement mapping method according to an embodiment of the present application is shown. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the applications described and/or claimed herein.
As shown in fig. 6, the electronic device includes: one or more processors 601, memory 602, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The various components are interconnected using different buses and may be mounted on a common motherboard or in other manners as desired. The processor may process instructions executing within the electronic device, including instructions stored in or on memory to display graphical information of a graphical user interface (Graphical User Interface, GUI) on an external input/output device, such as a display device coupled to the interface. In other embodiments, multiple processors and/or multiple buses may be used, if desired, along with multiple memories and multiple memories. Also, multiple electronic devices may be connected, each providing a portion of the necessary operations, e.g., as a server array, a set of blade servers, or a multiprocessor system. One processor 601 is illustrated in fig. 6.
The memory 602 is a non-transitory computer readable storage medium provided by the present application. The memory stores instructions executable by the at least one processor to cause the at least one processor to perform the advertisement mapping method provided by the present application. The non-transitory computer readable storage medium of the present application stores computer instructions for causing a computer to execute the advertisement mapping method provided by the present application.
The memory 602 is used as a non-transitory computer readable storage medium for storing non-transitory software programs, non-transitory computer executable programs, and modules, such as program instructions/modules corresponding to the advertisement mapping method in the embodiment of the present application, for example, the sample node pair determining module 510, the node representation determining module 520, the picture retrieving module 530, and the heterogeneous map creating module 540 shown in fig. 5. The processor 601 executes various functional applications of the server and data processing by running non-transitory software programs, instructions and modules stored in the memory 602, i.e., implements the advertisement mapping method in the method embodiments described above.
The memory 602 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, at least one application program required for a function; the storage data area may store data created according to the use of the electronic device of the advertisement mapping method, etc. In addition, the memory 602 may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 602 may optionally include memory remotely located with respect to processor 601, which may be connected to the electronic device of the advertising scheme method via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The electronic device of the advertisement mapping method may further include: an input device 603 and an output device 604. The processor 601, memory 602, input device 603 and output device 604 may be connected by a bus or otherwise, for example in fig. 6.
The input device 603 may receive input numeric or character information and generate key signal inputs related to user settings and function control of the electronic device of the advertising mapping method, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, a pointer stick, one or more mouse buttons, a track ball, a joystick, and the like. The output means 604 may include a display device, auxiliary lighting means, such as light emitting diodes (Light Emitting Diode, LEDs), tactile feedback means, and the like; haptic feedback devices such as vibration motors and the like. The display device may include, but is not limited to, a liquid crystal display (Liquid Crystal Display, LCD), an LED display, and a plasma display. In some implementations, the display device may be a touch screen.
Various implementations of the systems and techniques described here can be implemented in digital electronic circuitry, integrated circuitry, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.
These computing programs, also referred to as programs, software applications, or code, include machine instructions for a programmable processor, and may be implemented in a high-level procedural and/or object-oriented programming language, and/or in assembly/machine language. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and/or device for providing machine instructions and/or data to a programmable processor, e.g., magnetic discs, optical disks, memory, programmable logic devices (Programmable Logic Device, PLD), including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal used to provide machine instructions and/or data to a programmable processor.
To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device for displaying information to a user, for example, a Cathode Ray Tube (CRT) or an LCD monitor; and a keyboard and pointing device, such as a mouse or trackball, by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user may be any form of sensory feedback, e.g., visual feedback, auditory feedback, or tactile feedback; and input from the user may be received in any form, including acoustic input, speech input, or tactile input.
The systems and techniques described here can be implemented in a computing system that includes a background component, e.g., a data server, or that includes a middleware component, e.g., an application server, or that includes a front-end component, e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here, or any combination of such background, middleware, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication, e.g., a communication network. Examples of communication networks include: local area network (Local Area Network, LAN), wide area network (Wide Area Network, WAN), the internet and blockchain networks.
The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.
According to the technical scheme provided by the embodiment of the application, the heterogeneous triggering of advertisement to the picture allocation is directly realized by mining the content similarity and the structure similarity of the keywords and the pictures through the heterogeneous pictures, the dependence on the picture description text is eliminated, the defect of inaccurate picture allocation caused by semantic diversity is avoided, and the accuracy of picture recall and the quality of advertisement picture allocation are improved.
In addition, advertisement logs are utilized to mine advertisement keywords and advertisement pictures, and heterogeneous pictures composed of the keywords and the pictures are constructed, so that basis is provided for mining of similarity relations between the keywords and the pictures in content and structure.
In addition, by node determination, edge formation among nodes and edge weight calculation, a weighted undirected heterogeneous graph is constructed, a basis is provided for mining of similarity relations between keywords and pictures in content and structure, and good construction of the heterogeneous graph is a foundation for mining of similarity relations among nodes.
In addition, through random walk of the heterogeneous graph, a node sequence integrating content similarity information and structure similarity information is obtained, so that similarity is conveyed, a sample node pair comprising the content similarity information and the structure similarity information between the keywords and the pictures is obtained, and a good basis is provided for learning of subsequent node vector representation.
In addition, by performing breadth-first search and depth-first search on the heterogeneous graph, a node sequence integrating the content similarity information and the structure similarity information is obtained, the node types are not distinguished in the node sequence, the adjacent nodes can be similar in content or structure, and the similarity relation between the nodes is fully mined.
In addition, by traversing the node sequence through the sliding window, the nodes in the window form positive sample node pairs, and similarity is continued to the positive sample node pairs, namely the nodes in the positive sample node pairs can be similar in content or similar in structure. And extracting the opposite negative sample node pair of the positive sample node pair from the heterogeneous graph, i.e. there is no similarity in content or structure between nodes in the negative sample node pair. Therefore, a sufficient basis is provided for the learning of the node vector representation.
In addition, according to the learning of the node vector representation by the sample node pairs, the node vector representation fused with the similarity relation between the nodes can be learned.
In addition, based on the similarity relation among the nodes covered in the node vector representation, according to the node vector representation of the target keyword of the advertisement to be matched, the target advertisement matching graph similar to the target keyword in content and structure can be retrieved from a plurality of pictures. The index is established for the picture nodes by adopting an approximate nearest neighbor algorithm, so that the efficiency of picture retrieval based on the picture index can be greatly improved. Therefore, the mode of searching the picture through the node vector representation gets rid of the dependence on the picture description text, avoids the defect of inaccurate picture allocation caused by semantic diversity, and improves the picture recall accuracy and the advertisement picture allocation quality.
It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps described in the present application may be performed in parallel, sequentially, or in a different order, so long as the desired results of the technical solution disclosed in the present application can be achieved, and are not limited herein.
The above embodiments do not limit the scope of the present application. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the scope of the present application.
Claims (8)
1. An advertising mapping method, comprising:
searching according to the keyword nodes and the picture nodes in the heterograms to obtain sample node pairs comprising content similarity information and structure similarity information between the keywords and the pictures;
determining node vector representations in the heterograms according to the sample node pairs to obtain keyword node vector representations and picture node vector representations;
Determining a similarity relation between nodes according to node vector representation in the heterogram, wherein the similarity relation is used for retrieving a target advertisement matching chart of the advertisement to be matched according to target keyword node vector representation of the advertisement to be matched;
the searching according to the keyword node and the picture node in the heterogram to obtain a sample node pair comprising content similarity information and structure similarity information between the keyword and the picture comprises the following steps:
based on a Node2vec algorithm, breadth-first searching is carried out according to the weight of the edge in the heterogram, so that content similarity nodes with content similarity are obtained; performing depth-first search according to the weight of the edge in the heterogram to obtain a structural similarity node with structural similarity;
generating a node sequence containing content similarity information and structure similarity information between keywords and pictures according to the content similarity nodes and the structure similarity nodes;
traversing the node sequence by adopting a sliding window to obtain a positive sample node pair containing content similarity information and structure similarity information between the keywords and the pictures;
and extracting a negative sample node pair from the heterogeneous graph according to the positive sample node pair.
2. The method according to claim 1, further comprising, before searching for the keyword node and the picture node in the heterogram to obtain a sample node pair including content similarity information and structure similarity information between the keyword and the picture:
and constructing the heterograms according to the keywords and the pictures in the advertisement logs.
3. The method of claim 2, wherein constructing the heterogeneous map from keywords and pictures in an ad log comprises:
extracting keywords and pictures from the advertisement log, and determining keyword nodes and picture nodes in the heterograms;
determining edges between keyword nodes and picture nodes in the iso-composition according to the keywords and advertising sources of the pictures;
and determining the weight of the edge in the abnormal composition according to the showing times and clicking times of the same advertisement to which the keyword and the picture belong.
4. The method of claim 1, wherein said determining a node vector representation in the iso-graph from the sample node pairs comprises:
and obtaining the node vector representation in the heterogeneous graph according to the sample node pairs based on a skip-gram algorithm.
5. The method of claim 1, wherein retrieving the target ad campaigns for the ad campaigns from the target keyword node vector representation for the ad campaigns comprises:
determining target keywords of the advertisement to be matched according to the advertisement to be matched;
determining a target keyword node vector representation according to the heterogram and the target keyword;
and based on an approximate nearest neighbor algorithm, carrying out picture index retrieval according to the target keyword node vector representation to obtain the target advertisement map of the advertisement to be mapped.
6. An advertising graphics device, comprising:
the sample node pair determining module is used for searching according to the keyword nodes and the picture nodes in the heterograms to obtain sample node pairs comprising content similarity information and structure similarity information between the keywords and the pictures;
the node representation determining module is used for determining node vector representations in the heterograms according to the sample node pairs so as to obtain keyword node vector representations and picture node vector representations;
the picture retrieval module is used for determining similarity relation among nodes according to node vector representation in the heterogram and retrieving target advertisement matching images of the advertisement to be matched according to target keyword node vector representation of the advertisement to be matched;
The sample node pair determining module is specifically configured to:
based on a Node2vec algorithm, breadth-first searching is carried out according to the weight of the edge in the heterogram, so that content similarity nodes with content similarity are obtained; performing depth-first search according to the weight of the edge in the heterogram to obtain a structural similarity node with structural similarity;
generating a node sequence containing content similarity information and structure similarity information between keywords and pictures according to the content similarity nodes and the structure similarity nodes;
traversing the node sequence by adopting a sliding window to obtain a positive sample node pair containing content similarity information and structure similarity information between the keywords and the pictures;
and extracting a negative sample node pair from the heterogeneous graph according to the positive sample node pair.
7. An electronic device, comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein, the liquid crystal display device comprises a liquid crystal display device,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the advertising mapping method of any one of claims 1-5.
8. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the advertising mapping method of any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010041786.7A CN113127669B (en) | 2020-01-15 | 2020-01-15 | Advertisement mapping method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010041786.7A CN113127669B (en) | 2020-01-15 | 2020-01-15 | Advertisement mapping method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113127669A CN113127669A (en) | 2021-07-16 |
| CN113127669B true CN113127669B (en) | 2023-10-10 |
Family
ID=76771363
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010041786.7A Active CN113127669B (en) | 2020-01-15 | 2020-01-15 | Advertisement mapping method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113127669B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114329016B (en) * | 2022-01-04 | 2023-04-25 | 北京百度网讯科技有限公司 | Picture label generating method and text mapping method |
| CN114528589B (en) * | 2022-01-26 | 2022-11-08 | 广东南方新视界传媒科技有限公司 | Outdoor media ecological management system and method based on block chain |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101321190A (en) * | 2008-07-04 | 2008-12-10 | 清华大学 | Recommend method and recommend system of heterogeneous network |
| US7716229B1 (en) * | 2006-03-31 | 2010-05-11 | Microsoft Corporation | Generating misspells from query log context usage |
| CN102129431A (en) * | 2010-01-13 | 2011-07-20 | 阿里巴巴集团控股有限公司 | Search method and system applied to online trading platform |
| CN105654345A (en) * | 2015-12-30 | 2016-06-08 | 上海珍岛信息技术有限公司 | Bayesian network-based advertisement click-through rate prediction method and device |
| CN106709747A (en) * | 2015-11-17 | 2017-05-24 | 北京奇虎科技有限公司 | Method and device for recalling ad |
| KR20170089142A (en) * | 2016-01-26 | 2017-08-03 | 경북대학교 산학협력단 | Generating method and system for triple data |
| CN107491518A (en) * | 2017-08-15 | 2017-12-19 | 北京百度网讯科技有限公司 | Method and apparatus, server, storage medium are recalled in one kind search |
| WO2018122238A1 (en) * | 2016-12-30 | 2018-07-05 | Robert Bosch Gmbh | Method and system for fuzzy keyword search over encrypted data |
| CN108415961A (en) * | 2018-02-06 | 2018-08-17 | 厦门集微科技有限公司 | A kind of advertising pictures recommendation method and device |
| CN110032606A (en) * | 2019-03-29 | 2019-07-19 | 阿里巴巴集团控股有限公司 | A kind of sample clustering method and device |
| CN110569437A (en) * | 2019-09-05 | 2019-12-13 | 腾讯科技（深圳）有限公司 | click probability prediction and page content recommendation methods and devices |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9256670B2 (en) * | 2013-10-10 | 2016-02-09 | International Business Machines Corporation | Visualizing conflicts in online messages |
| US10146875B2 (en) * | 2014-12-19 | 2018-12-04 | International Business Machines Corporation | Information propagation via weighted semantic and social graphs |
-
2020
- 2020-01-15 CN CN202010041786.7A patent/CN113127669B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7716229B1 (en) * | 2006-03-31 | 2010-05-11 | Microsoft Corporation | Generating misspells from query log context usage |
| CN101321190A (en) * | 2008-07-04 | 2008-12-10 | 清华大学 | Recommend method and recommend system of heterogeneous network |
| CN102129431A (en) * | 2010-01-13 | 2011-07-20 | 阿里巴巴集团控股有限公司 | Search method and system applied to online trading platform |
| CN106709747A (en) * | 2015-11-17 | 2017-05-24 | 北京奇虎科技有限公司 | Method and device for recalling ad |
| CN105654345A (en) * | 2015-12-30 | 2016-06-08 | 上海珍岛信息技术有限公司 | Bayesian network-based advertisement click-through rate prediction method and device |
| KR20170089142A (en) * | 2016-01-26 | 2017-08-03 | 경북대학교 산학협력단 | Generating method and system for triple data |
| WO2018122238A1 (en) * | 2016-12-30 | 2018-07-05 | Robert Bosch Gmbh | Method and system for fuzzy keyword search over encrypted data |
| CN107491518A (en) * | 2017-08-15 | 2017-12-19 | 北京百度网讯科技有限公司 | Method and apparatus, server, storage medium are recalled in one kind search |
| CN108415961A (en) * | 2018-02-06 | 2018-08-17 | 厦门集微科技有限公司 | A kind of advertising pictures recommendation method and device |
| CN110032606A (en) * | 2019-03-29 | 2019-07-19 | 阿里巴巴集团控股有限公司 | A kind of sample clustering method and device |
| CN110569437A (en) * | 2019-09-05 | 2019-12-13 | 腾讯科技（深圳）有限公司 | click probability prediction and page content recommendation methods and devices |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113127669A (en) | 2021-07-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111967256B (en) | Event relation generation method and device, electronic equipment and storage medium | |
| CN111522967B (en) | Knowledge graph construction method, device, equipment and storage medium | |
| KR101911999B1 (en) | Feature-based candidate selection | |
| CN111522994A (en) | Method and apparatus for generating information | |
| CN111488740B (en) | Causal relationship judging method and device, electronic equipment and storage medium | |
| KR20210038467A (en) | Method and apparatus for generating an event theme, device and storage medium | |
| CN112650907A (en) | Search word recommendation method, target model training method, device and equipment | |
| CN111949814A (en) | Searching method, searching device, electronic equipment and storage medium | |
| CN112541362B (en) | Generalization processing method, device, equipment and computer storage medium | |
| CN113127669B (en) | Advertisement mapping method, device, equipment and storage medium | |
| CN112380847B (en) | Point-of-interest processing method and device, electronic equipment and storage medium | |
| CN111984825A (en) | Method and apparatus for searching video | |
| CN110532404B (en) | Source multimedia determining method, device, equipment and storage medium | |
| CN111984775A (en) | Question and answer quality determination method, device, equipment and storage medium | |
| CN111523007A (en) | User interest information determination method, device, equipment and storage medium | |
| CN112001265A (en) | Video event identification method and device, electronic equipment and storage medium | |
| CN111737966A (en) | Document repetition degree detection method, device, equipment and readable storage medium | |
| CN111241242A (en) | Method, device and equipment for determining target content and computer readable storage medium | |
| CN113111216B (en) | Advertisement recommendation method, device, equipment and storage medium | |
| US20210216710A1 (en) | Method and apparatus for performing word segmentation on text, device, and medium | |
| CN111125445A (en) | Community theme generation method and device, electronic equipment and storage medium | |
| CN114491232B (en) | Information query method and device, electronic equipment and storage medium | |
| CN112052402B (en) | Information recommendation method and device, electronic equipment and storage medium | |
| CN111080369B (en) | Advertisement information display method, device, equipment and readable storage medium | |
| CN111125362B (en) | Abnormal text determination method and device, electronic equipment and medium |
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 |