CN102103641A - Method for adding banner advertisement into user-browsed network image - Google Patents

Method for adding banner advertisement into user-browsed network image Download PDF

Info

Publication number
CN102103641A
CN102103641A CN2011100549918A CN201110054991A CN102103641A CN 102103641 A CN102103641 A CN 102103641A CN 2011100549918 A CN2011100549918 A CN 2011100549918A CN 201110054991 A CN201110054991 A CN 201110054991A CN 102103641 A CN102103641 A CN 102103641A
Authority
CN
China
Prior art keywords
image
user
advertisement
unit
visual
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
Application number
CN2011100549918A
Other languages
Chinese (zh)
Inventor
钱学明
汪欢
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Jiaotong University
Original Assignee
Xian Jiaotong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian Jiaotong University filed Critical Xian Jiaotong University
Priority to CN2011100549918A priority Critical patent/CN102103641A/en
Publication of CN102103641A publication Critical patent/CN102103641A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

本发明公开了一种在用户浏览网络图像中添加图标广告的方法，包括下述步骤：首先对用户浏览的网络图像单元10执行视觉相似图像检索单元20，将用户浏览的网络图像单元10中的图像画面100进行相似图像确定；然后执行用户兴趣描述词排序单元30，按照不同时间约束信息来获取用户兴趣信息；接下来执行广告排序方法与选择单元40；接下来执行广告位置选择及链接单元50，按照广告排序的结果，对每个图标广告与当前图像的视觉相似性进行计算，确定插入图标广告的位置，并在相应的广告插入位置添加有关描述该广告更详细内容的超链接；最终的检索结果在显示插入广告效果图单元60中进行显示。

Figure 201110054991

The invention discloses a method for adding an icon advertisement in a network image browsed by a user, which includes the following steps: first, execute a visually similar image retrieval unit 20 on the network image unit 10 browsed by the user, and retrieve the image in the network image unit 10 browsed by the user The image screen 100 performs similar image determination; then executes the user interest descriptor sorting unit 30 to obtain user interest information according to different time constraint information; then executes the advertisement sorting method and selection unit 40; next executes the advertisement position selection and linking unit 50 , calculate the visual similarity between each icon ad and the current image according to the results of the ad sorting, determine the position to insert the icon ad, and add a hyperlink describing more detailed content of the ad at the corresponding ad insertion position; the final The retrieval result is displayed in the display insertion advertisement rendering unit 60 .

Figure 201110054991

Description

In user's browse network image, add the method for banner advertisement
Technical field
The present invention relates to a kind of method of in user's browse network image, adding banner advertisement.
Background technology
Add banner advertisement in the image that user on the network is browsed, not only directly perceived but also the very big market space arranged.Take in nearly 20,000,000,000 dollars at the web advertisement in the first half of the year in 2010.There is following several problem in existing network icon advertisement recommend method: 1) user's specific aim deficiency of advertisement, and the advertisement insertion method on the internet is not distinguished well to the user group; (2) the correlativity deficiency of advertisement icon content and browse graph picture; (user's affinity deficiency of 3 advertisements, these advertisements all are irrelevant with user's interest under a lot of situations.
A kind of method that the web advertisement is sorted is disclosed in Chinese patent ZL200710117607.8.Adopt the keyword in advertisement keyword and the webpage to do coupling in advertisement is obtained, this order ads method based on the keyword coupling does not have user's specific aim, therefore can not reach to attract on the network user to the effect of advertisement notice.Therefore propose among the present invention the user on the internet is added the banner advertisement relevant with user interest in the image that it is browsed.This advertising method can attract user's notice effectively.
Summary of the invention
The objective of the invention is to overcome existing method and adds banner advertisement do not have user's deficiency targetedly in the network image that the user browses, proposing a kind of is the banner advertisement adding method of guiding with user and the network image content browsed thereof.
For reaching above purpose, the present invention adopts following technical scheme to be achieved:
A kind of method of adding banner advertisement in user's browse network image comprises the steps:
At first vision retrieving similar images unit 20 is carried out in the network image unit 10 that the user is browsed, wherein, the network image unit 10 that the user browses comprises image frame 100, image text 110 and user ID information 120, image frame 100 in the network image unit 10 that the user is browsed in vision retrieving similar images unit 20 carries out similar image to be determined, obtains user ID associated picture visual similarity ordering 220; Carry out user interest descriptor sequencing unit 30 according to user ID associated picture visual similarity ordering 220 then, obtain user interest information according to the different time constraint information; Next carry out order ads method and selected cell 40, promptly the result according to the ordering of user interest descriptor carries out order ads and advertisement selection to the advertisement in the banner advertisement storehouse according to correlativity; Next carrying out location advertising selects and link unit 50, result according to the previous step order ads, visual similarity to each banner advertisement and present image calculates, determine to insert the position of banner advertisement, and add the relevant more hyperlink of detailed content of this advertisement of describing in the correspondent advertisement insertion position; Final result for retrieval shows in showing insertion advertising results figure unit 60.
In the such scheme, described vision retrieving similar images unit 20 comprises following concrete steps: at first image frame 100 is carried out Visual Feature Retrieval Process step 101, extract color, texture and edge feature in the image, next carry out visual signature quantization step 102, after Visual Feature Retrieval Process, corresponding color characteristic, textural characteristics and edge feature are quantized with the method for K-mean cluster respectively; When carrying out Visual Feature Retrieval Process step 101 and visual signature quantization step 102, image and the image text execution index in the image text 200 downloaded on the network are set up, generate network image text message index database 201 based on user ID information; Then image and the image in the image text 200 downloaded on the network are carried out Visual Feature Retrieval Process step 101 and characteristic quantification step 102 successively, obtain network image visual signature index database 202 based on user ID information; Then to the visual signature quantitative information of characteristic quantification step 102 gained and the visual similarity metrology step of carrying out based on the Image Visual Feature index of all these users in the network image visual signature index database 202 of user ID information based on TF-IDF 210, carrying out similarity calculates, obtain based on the visual similarity score between each image and the user images picture 100 in the network image visual signature index database 202 of user ID information, at last above-mentioned visual similarity score is sorted, obtain user ID associated picture visual similarity ordering 220.
In color in the described extraction image, texture and the edge feature step, the extraction of color characteristic is 25 image blocks that wait size that original image are divided into 5x5, extracts the color moment feature of 9 dimensions in each piece respectively, and the dimension of color characteristic is 225; Scalable wavelet bag textural characteristics describing method is adopted in the extraction of textural characteristics, and the basis function of wavelet package transforms is ' DB2 ', the image divided mode be 2x2 and one placed in the middle etc. the sized images piece, the dimension of textural characteristics is 170.Edge Gradient Feature adopts the marginal distribution histogram based on 128 dimensions, and direction number is 16, and the quantification technique progression of gradient is 8.
Described user interest descriptor sequencing unit 30 comprises following concrete steps: according to user ID associated picture visual similarity ordering 220, carry out and extract user images text step, obtain vision similar image text message 310 with current browse graph picture, next execution in step 320, obtain user interest information according to the different time constraint information, step 320 comprises that the user interest of overall time-constrain obtains method 321, the user interest of time-constrain obtains method 322 recently, and both select one; Carry out the user interest ordered steps 330 based on the visual similarity weighting at last, this step is exactly to describe user interest according to the text message of image relevant in step 321 or the step 322.
The method of adding banner advertisement in the network image that the user browses that is provided among the present invention is compared with existing network icon adding method, and its beneficial effect shows that the advertisement of interpolation has user's specific aim.
Description of drawings
The present invention is described in further detail below in conjunction with the drawings and the specific embodiments.
Fig. 1 is the general steps synoptic diagram of the inventive method.
Fig. 2 is the concrete steps process flow diagram of visual similarity image retrieval unit 20 among Fig. 1.
Fig. 3 is the concrete steps process flow diagram of user interest descriptor sequencing unit 30 among Fig. 1.
Embodiment
Fig. 1 has provided the general steps synoptic diagram that adds the method for banner advertisement among the present invention in user's browse graph picture.Wherein comprise the network image unit 10 that the Internet user browses; Vision retrieving similar images unit 20 is carried out in the network image unit 10 that the user browses; Carry out user interest descriptor sequencing unit 30 then; Next carry out order ads method and selected cell 40; Next carrying out location advertising selects and link unit unit 50; Final result for retrieval shows in showing insertion advertising results figure unit 60.
The network image unit 10 that the user browses among the present invention comprises image frame 100, image text 110, user ID information 120.Image frame 100 in the network image unit 10 that the user is browsed in vision retrieving similar images unit 20 among the present invention carries out similar image and determines.Provide to example the network image unit 10 that the user is browsed among Fig. 2 and carried out the FB(flow block) that the vision similar image detects.At first image frame 100 is carried out Visual Feature Retrieval Process step 101, extract color, texture and edge feature in the image.Wherein the extraction of color characteristic is 25 image blocks that wait size that original image are divided into 5x5, extracts the color moment feature of 9 dimensions in each piece respectively, and the dimension of color characteristic is 225.Wherein scalable wavelet bag textural characteristics describing method is adopted in the description of textural characteristics, the basis function of wavelet package transforms is ' DB2 ', the image divided mode be 2x2 and one placed in the middle etc. the sized images piece, the dimension of textural characteristics is that 170 (correlation technique sees the paper of publishing: X.Qian for details, G.Liu, D.Guo, Z.Li, Z.Wang, and H.Wang, " Object Categorization using Hierarchical Wavelet Packet Texture Descriptors, " inProc.ISM 2009, pp.44-51.).Wherein edge feature adopts the marginal distribution histogram (direction number is 16, and the quantification progression of gradient is 8) based on 128 dimensions.
Next carry out visual signature quantization step 102, after feature extraction, corresponding color moment feature, wavelet packet textural characteristics and edge feature are quantized with the method for K-mean cluster respectively, the code book number that quantizes is respectively: 50000,10000 and 50000.Can change the code book number as required in practice.Suggestion code book number is more than 10000 among the present invention.
The similarity image comes from the image downloaded on the network and image text unit 200 thereof (this unit from website Bing, Flickr, the view data that download websites such as Google and the label text information of every width of cloth image) among the present invention.Then the image text execution index in the unit 200 is set up, generated network image text message index database 201 based on user ID information.Then the image in the unit 200 is carried out Visual Feature Retrieval Process step 101 and characteristic quantification step 102 successively, to obtain network image visual signature index database 202 based on user ID information.Then to visual signature quantitative information 102 and the visual similarity metrology step of carrying out based on the Image Visual Feature index of relevant all these users in the network image visual signature index database 202 of user ID information based on TF-IDF 210, calculate to carry out similarity, obtain the visual similarity score between each image and user images picture 100 in 202, suppose that active user's picture number is N, so wherein the visual similarity of any one image i must be divided into S (i), i=1～N, S (i) ∈ [0,1].In similarity is calculated, adopt and carry out (the TF-IDF method is a kind of known method) in this area based on the criterion of TF-IDF.At last above-mentioned visual similarity score is sorted (image being arranged by score order from high to low), obtain user ID associated picture visual similarity ordering 220.
In the user interest descriptor sequencing unit 30 of Fig. 1, realize user's interest is sorted.The pairing concrete steps of user interest sort method as shown in Figure 3.Comprising according to similar image ranking results 220 in the unit 20, carry out and extract user images text step, with the vision similar image text message 310 of acquisition with active user's browse graph picture, next execution in step 320, obtain user interest information according to the different time constraint information.Step 320 comprises that the user interest of overall time-constrain obtains method 321, the user interest of time-constrain obtains one of method 322 recently.The method of step 321 is the text message in all relevant images of user to be used for interest obtain; The method of step 322 is that the interest with the user is limited in the current slot.Carry out user interest ordered steps 330 at last, describe user interest according to the text message of image relevant in step 321 or the step 322 exactly based on the visual similarity weighting.Suppose that wherein picture number is M, the corresponding visual similarity of each image must be divided into S (i), i=1～M, and S (i) ∈ [0,1], the descriptive text vocabulary that is comprised among this image i has Z iIndividual.Suppose to include K vocabulary in these images, be designated as t respectively 1～t K, vocabulary t wherein kThe number of times that occurs is that the score of c in respective image is respectively s 1～s c, then final pairing user interest degree I kFor:
I k = Σ i = 1 c s i 2 × c p , p∈[0，1]
The final user interest degree of describing adopts normalized interest-degree:
I k = I k / Σ i = 1 k I i 2 , p∈[0，1]
Result according to the user interest descriptor ordering that is drawn in the unit 30 in order ads in Fig. 1 and the advertisement selection unit 40 carries out order ads and advertisement selection to the advertisement in the banner advertisement storehouse according to correlativity.Method for measuring similarity in the advertisement coupling adopts existing open source literature (T.Mei, X.-S.Hua, and S.Li, Contextual in-image advertising, in Proc.ACM Multimedia, Vancouver, Canada, 2008, the pp.439-448.) method in.After calculating, can draw each advertisement a iCorrelativity score U (a with the user i).
Location advertising in Fig. 1 select and link unit 50 in according to the result of the order ads that is drawn in the unit 40, the visual similarity of each banner advertisement and present image is calculated.During similarity is calculated with the color correlation of image as measurement criterion.The execution in step of insertion position system of selection is as follows, at first image division is become 5*5 piece that waits size, then each piece is carried out the texture complexity and the content importance degree is divided, with the position P that finds out the most suitable interpolation icon (x, y, z), x wherein, y denotation coordination, z are represented corresponding Color Channel number (for coloured image port number z=3, for gray level image port number z=1).Concrete grammar can adopt existing document (T.Mei, X.-S.Hua, and S.Li, Contextual in-image advertising, in Proc.ACM Multimedia, Vancouver, Canada, 2008, the pp.439-448.) method of publishing in.After determining to insert the position of advertisement, the criterion with the advertisement icon and the color distortion of corresponding insertion position are measured as visual similarity can draw each advertisement a after calculating iWith the current visual similarity score V (a that browses image frame 100 of user i).
V(a i)＝exp(-D(a i))
D (a wherein i) expression advertisement a iWith current local location P (x, y, vision difference z), the D (a that browses the most suitable interpolation advertisement in the image frame of user i) can be expressed as:
D ( a i ) = 1 N Σ x , y , z ( a i ( x , y , z ) - P ( x , y , z ) ) 2
In addition also with advertisement a iSimilarity T (a with the current browse network image text 110 of user i) also consider among tolerance T (a wherein i) computing method and U (a i) identical, do not do tired stating at this.
Score F (a of final advertisement selection i) be user's correlativity score U (a that weighting is considered i) and visual similarity score V (a i) and user version correlativity T (a i) and:
F (a i)=α * U (a i)+β * V (a i)+γ * T (a i), { α, beta, gamma } ∈ [0,1] is α wherein, and beta, gamma is a weighting coefficient,
α+β+γ＝1
α＝0.7，β＝0.1，γ＝0.2
Add the relevant more hyperlink of detailed content of this advertisement of describing then with the highest the picking out of advertisement score, and in the correspondent advertisement insertion position as final insertion advertisement.
In order ads and advertisement selection unit 40, can suitably select several alternative advertisements in the present invention and select and link the input of determining unit 50, like this, can effectively reduce the complexity of system handles as the advertisement insertion position.
The final advertising effect image that inserts shows in unit 60.

Claims (4)

1.一种在用户浏览网络图像中添加图标广告的方法，其特征在于，包括下述步骤：1. A method for adding icon advertisements in user browsing network images, is characterized in that, comprises the following steps: 首先对用户浏览的网络图像单元(10)执行视觉相似图像检索单元(20)，其中，用户浏览的网络图像单元(10)包括图像画面(100)、图像文本(110)和用户ID信息(120)，视觉相似图像检索单元(20)将用户浏览的网络图像单元(10)中的图像画面(100)进行相似图像确定，获得用户ID相关图像视觉相似性排序(220)；然后按照用户ID相关图像视觉相似性排序(220)执行用户兴趣描述词排序单元(30)，按照不同时间约束信息来获取用户兴趣信息；接下来执行广告排序方法与选择单元(40)，即根据用户兴趣描述词排序的结果对图标广告库中的广告按照相关性进行广告排序和广告选择；接下来执行广告位置选择及链接单元(50)，按照前一步骤广告排序的结果，对每个图标广告与当前图像的视觉相似性进行计算，确定插入图标广告的位置，并在相应的广告插入位置添加有关描述该广告更详细内容的超链接；最终的检索结果在显示插入广告效果图单元(60)中进行显示。Firstly, the visual similarity image retrieval unit (20) is executed on the network image unit (10) browsed by the user, wherein the network image unit (10) browsed by the user includes image frame (100), image text (110) and user ID information (120) ), the visual similarity image retrieval unit (20) carries out similar image determination to the image frame (100) in the network image unit (10) that the user browses, and obtains the user ID-related image visual similarity sorting (220); then according to the user ID correlation Image visual similarity sorting (220) executes user interest descriptor sorting unit (30), obtains user interest information according to different time constraint information; then executes advertisement sorting method and selection unit (40), namely sorts according to user interest descriptor Carry out advertisement sorting and advertisement selection according to the relevance of the advertisement in the icon advertisement storehouse according to the result of the advertisement in the icon advertisement storehouse; Next, carry out advertisement position selection and link unit (50), according to the result of advertisement sorting in the previous step, to each icon advertisement and the current image Calculate the visual similarity, determine the position for inserting the icon advertisement, and add a hyperlink about describing the more detailed content of the advertisement at the corresponding advertisement insertion position; the final retrieval result is displayed in the display insertion advertisement rendering unit (60). 2.如权利要求1所述的在用户浏览网络图像中添加图标广告的方法，其特征在于，所述视觉相似图像检索单元(20)包括下述具体步骤：首先对图像画面(100)执行视觉特征提取步骤(101)，提取图像中的颜色、纹理以及边缘特征，接下来执行视觉特征量化步骤(102)，在视觉特征提取之后对相应的颜色特征、纹理特征以及边缘特征分别用K-均值聚类的方法进行量化；在进行视觉特征提取步骤(101)和视觉特征量化步骤(102)的同时，对网络上下载的图像及图像文本(200)中的图像文本执行索引建立，生成基于用户ID信息的网络图像文本信息索引库(201)；然后对网络上下载的图像及图像文本(200)中的图像依次执行视觉特征提取步骤(101)和特征量化步骤(102)，获得基于用户ID信息的网络图像视觉特征索引库(202)；然后对特征量化步骤(102)所得的视觉特征量化信息与基于用户ID信息的网络图像视觉特征索引库(202)中所有该用户的图像视觉特征索引执行基于TF-IDF的视觉相似性度量步骤(210)，进行相似性计算，得到基于用户ID信息的网络图像视觉特征索引库(202)中每个图像与用户图像画面(100)之间的视觉相似性得分，最后对上述的视觉相似性得分进行排序，得到用户ID相关图像视觉相似性排序(220)。2. The method for adding icon advertisements in user browsing network images as claimed in claim 1, characterized in that said visually similar image retrieval unit (20) comprises the following specific steps: first performing visual inspection on the image frame (100) The feature extraction step (101) extracts the color, texture and edge features in the image, and then performs the visual feature quantization step (102). After the visual feature extraction, the corresponding color features, texture features and edge features are respectively used K-means The method for clustering is quantified; while performing the visual feature extraction step (101) and the visual feature quantization step (102), the image text in the image downloaded on the network and the image text (200) is indexed and built, and generated based on the user The network image text information index storehouse (201) of ID information; Then carry out visual feature extraction step (101) and feature quantization step (102) successively to the image in the image downloaded on the network and image text (200), obtain based on user ID The network image visual feature index storehouse (202) of information; Then the image visual feature index of all this user in the visual feature quantification information of feature quantization step (102) gained and the network image visual feature index storehouse (202) based on user ID information Execute the visual similarity measurement step (210) based on TF-IDF, perform similarity calculation, and obtain the visual distance between each image and the user image picture (100) in the network image visual feature index library (202) based on user ID information. Similarity score, and finally sort the above visual similarity scores to obtain the visual similarity ranking of user ID-related images (220). 3.如权利要求2所述的在用户浏览网络图像中添加图标广告的方法，其特征在于，所述提取图像中的颜色、纹理以及边缘特征步骤中，颜色特征的提取是将原始图像划分成5x5的25个等大小的图像块，每个块中分别提取9维的颜色矩特征，颜色特征的维数为225；纹理特征的提取采用可分级小波包纹理特征描述方法，小波包变换的基函数为‘DB2’，图像分开形式为2x2和一个居中的等大小图像块，纹理特征的维数为170；边缘特征提取采用基于128维的边缘分布直方图，方向数为16，梯度的量化级数为8。3. the method for adding icon advertisement in user's browsing network image as claimed in claim 2, it is characterized in that, in the color, texture and edge characteristic step in the described extraction image, the extraction of color characteristic is to divide original image into 25 equal-sized image blocks of 5x5, extract 9-dimensional color moment features from each block, and the dimension of color features is 225; The function is 'DB2', the image separation form is 2x2 and a centered image block of equal size, the dimension of the texture feature is 170; the edge feature extraction uses a 128-dimensional edge distribution histogram, the number of directions is 16, and the quantization level of the gradient The number is 8. 4.如权利要求1所述的在用户浏览网络图像中添加图标广告的方法，其特征在于，所述用户兴趣描述词排序单元(30)包括下述具体步骤：按照用户ID相关图像视觉相似性排序(220)，执行提取用户图像文本步骤，获得与当前浏览图像的视觉相似图像文本信息(310)，接下来执行步骤(320)，按照不同时间约束信息来获取用户兴趣信息，步骤(320)包括总体时间约束的用户兴趣获取法(321)、最近时间约束的用户兴趣获取法(322)，两者选其一；最后执行基于视觉相似性加权的用户兴趣排序步骤(330)，该步骤就是根据步骤(321)或者步骤(322)中相关的图像的文本信息来描述用户兴趣。4. the method for adding icon advertisement in user's browsing network image as claimed in claim 1, is characterized in that, described user interest descriptor ordering unit (30) comprises following concrete steps: according to user ID relevant image visual similarity Sorting (220), performing the step of extracting user image text, obtaining visually similar image text information (310) with the currently browsed image, and then performing step (320), obtaining user interest information according to different time constraint information, step (320) Including the overall time-constrained user interest acquisition method (321), the most recent time-constrained user interest acquisition method (322), choose one of the two; finally perform the user interest sorting step based on visual similarity weighting (330), this step is Describe the user's interest according to the text information of the relevant image in step (321) or step (322).
CN2011100549918A 2011-03-08 2011-03-08 Method for adding banner advertisement into user-browsed network image Pending CN102103641A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2011100549918A CN102103641A (en) 2011-03-08 2011-03-08 Method for adding banner advertisement into user-browsed network image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2011100549918A CN102103641A (en) 2011-03-08 2011-03-08 Method for adding banner advertisement into user-browsed network image

Publications (1)

Publication Number Publication Date
CN102103641A true CN102103641A (en) 2011-06-22

Family

ID=44156411

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2011100549918A Pending CN102103641A (en) 2011-03-08 2011-03-08 Method for adding banner advertisement into user-browsed network image

Country Status (1)

Country Link
CN (1) CN102103641A (en)

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2013063740A1 (en) * 2011-10-31 2013-05-10 Google Inc. Selecting images based on textual description
CN103428539A (en) * 2012-05-15 2013-12-04 腾讯科技（深圳）有限公司 Pushed information publishing method and device
CN104809251A (en) * 2015-05-19 2015-07-29 北京理工大学 Rapid optimal automatic data specification icon arranging method
CN105956878A (en) * 2016-04-25 2016-09-21 广州出益信息科技有限公司 Network advertisement pushing method and network advertisement pushing device
CN106156063A (en) * 2015-03-30 2016-11-23 阿里巴巴集团控股有限公司 Correlation technique and device for object picture search results ranking
CN107784061A (en) * 2016-08-24 2018-03-09 百度（美国）有限责任公司 It is determined that the method and system and machine readable media of the content genres based on image
CN108111897A (en) * 2017-12-12 2018-06-01 北京奇艺世纪科技有限公司 A kind of method and device for showing displaying information in video
CN109241374A (en) * 2018-06-07 2019-01-18 广东数相智能科技有限公司 A kind of book information library update method and books in libraries localization method
CN109389440A (en) * 2017-08-02 2019-02-26 阿里巴巴集团控股有限公司 The method, apparatus and electronic equipment of data object information are provided
CN113676775A (en) * 2021-08-27 2021-11-19 苏州因塞德信息科技有限公司 Method for implanting advertisement in video and game by using artificial intelligence

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2013063740A1 (en) * 2011-10-31 2013-05-10 Google Inc. Selecting images based on textual description
CN103428539A (en) * 2012-05-15 2013-12-04 腾讯科技（深圳）有限公司 Pushed information publishing method and device
CN106156063A (en) * 2015-03-30 2016-11-23 阿里巴巴集团控股有限公司 Correlation technique and device for object picture search results ranking
CN104809251A (en) * 2015-05-19 2015-07-29 北京理工大学 Rapid optimal automatic data specification icon arranging method
CN104809251B (en) * 2015-05-19 2017-11-28 北京理工大学 A kind of quickly optimal auto arranging method of data type specification icon
CN105956878A (en) * 2016-04-25 2016-09-21 广州出益信息科技有限公司 Network advertisement pushing method and network advertisement pushing device
CN107784061A (en) * 2016-08-24 2018-03-09 百度（美国）有限责任公司 It is determined that the method and system and machine readable media of the content genres based on image
CN109389440A (en) * 2017-08-02 2019-02-26 阿里巴巴集团控股有限公司 The method, apparatus and electronic equipment of data object information are provided
CN109389440B (en) * 2017-08-02 2022-05-24 阿里巴巴集团控股有限公司 Method and device for providing data object information and electronic equipment
CN108111897A (en) * 2017-12-12 2018-06-01 北京奇艺世纪科技有限公司 A kind of method and device for showing displaying information in video
CN109241374A (en) * 2018-06-07 2019-01-18 广东数相智能科技有限公司 A kind of book information library update method and books in libraries localization method
CN113676775A (en) * 2021-08-27 2021-11-19 苏州因塞德信息科技有限公司 Method for implanting advertisement in video and game by using artificial intelligence

Similar Documents

Publication Publication Date Title
CN102103641A (en) Method for adding banner advertisement into user-browsed network image
CN103559191B (en) Based on latent space study and Bidirectional sort study across media sort method
CN107679960B (en) A recommendation method for personalized clothing based on bimodal content analysis of clothing images and label text
CN102799591B (en) Method and device for providing recommended word
Zhan et al. Gather customer concerns from online product reviews–A text summarization approach
CN104572651B (en) Picture sort method and device
TWI631474B (en) Method and device for product identification label and method for product navigation
CN103544216B (en) The information recommendation method and system of a kind of combination picture material and keyword
US20240330361A1 (en) Training Image and Text Embedding Models
US7596552B2 (en) Method and system for extracting web data
US9607327B2 (en) Object search and navigation method and system
US9489400B1 (en) Interactive item filtering using images
CN101996191B (en) Method and system for searching for two-dimensional cross-media element
CN103577478B (en) Web page push method and system
CN106202362A (en) Image recommendation method and image recommendation device
CN101634996A (en) Individualized video sequencing method based on comprehensive consideration
CN104298749A (en) Commodity retrieval method based on image visual and textual semantic integration
CN103310003A (en) Method and system for predicting click rate of new advertisement based on click log
CN111309936A (en) A method of constructing movie user portraits
CN107577682A (en) Users' Interests Mining and user based on social picture recommend method and system
CN111291191B (en) Broadcast television knowledge graph construction method and device
WO2015066891A1 (en) Systems and methods for extracting and generating images for display content
CN109460519B (en) Browsing object recommendation method and device, storage medium and server
CN112948575B (en) Text data processing method, apparatus and computer readable storage medium
CN117829911B (en) AI-driven advertisement creative optimization method and system

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20110622