CN110348907A - A kind of orientation method and device of advertisement crowd - Google Patents
A kind of orientation method and device of advertisement crowd Download PDFInfo
- Publication number
- CN110348907A CN110348907A CN201910633443.7A CN201910633443A CN110348907A CN 110348907 A CN110348907 A CN 110348907A CN 201910633443 A CN201910633443 A CN 201910633443A CN 110348907 A CN110348907 A CN 110348907A
- Authority
- CN
- China
- Prior art keywords
- user
- label
- tag
- target
- users
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000012549 training Methods 0.000 claims description 58
- 239000012141 concentrate Substances 0.000 claims description 17
- 238000013528 artificial neural network Methods 0.000 claims description 7
- 238000012216 screening Methods 0.000 claims description 4
- 238000009792 diffusion process Methods 0.000 abstract description 12
- 239000000284 extract Substances 0.000 abstract description 9
- 238000012545 processing Methods 0.000 abstract description 6
- 238000010586 diagram Methods 0.000 description 9
- 238000003780 insertion Methods 0.000 description 6
- 230000037431 insertion Effects 0.000 description 6
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 238000004590 computer program Methods 0.000 description 4
- 238000013523 data management Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000006399 behavior Effects 0.000 description 3
- 238000000586 desensitisation Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000003066 decision tree Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000007477 logistic regression Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000004579 scanning voltage microscopy Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000000151 deposition Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06Q30/0251—Targeted advertisements
- G06Q30/0269—Targeted advertisements based on user profile or attribute
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Data Mining & Analysis (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Marketing (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Business, Economics & Management (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses the orientation methods and device of a kind of advertisement crowd, belong to technical field of data processing, while to realize tagging user diffusion when advertisement crowd extracts, take into account algorithm complexity and accuracy.This method comprises: obtaining the target user's label for being used to extract advertisement crowd of user's input；In the user tag of each description user characteristics obtained in advance, the similarity value between target user's label and other each user tags is determined respectively；According to the similarity value between target user's label and other each user tags, the determining and maximum K similar users label of target user's label similarity, K is natural number；Using the user comprising each similar users label as the advertisement crowd being directed to, or using the user comprising each similar users label and the user comprising target user's label as the advertisement crowd being directed to.
Description
Technical field
The present invention relates to technical field of data processing, in particular to a kind of the orientation method and device of advertisement crowd.
Background technique
Label crowd extraction is that one in data management platform (Data Management Platform, DMP) is important
Component, user give sequence of user label as stereotactic conditions, advertisement crowd are extracted by association user portrait, that is, is passed through
The combination of the user tags such as interest abundant, behavior, draws a circle to approve required user group.
Extract generally, based on the advertisement crowd of user tag and can only extract the user comprising user tag, namely comprising with
The user of family label how many just extract how much, to extract more massive user group or extracting the user of specific scale
Group then needs to carry out tagging user diffusion.
The tagging user diffusion way of mainstream is at present: use the original user comprising user tag as positive sample, with
Machine chooses other users as negative sample (the methods of PU-Learning can be used in the acquisition of negative sample), in positive sample and bears
80% is randomly selected in sample as training data, 20% and is used as test data, passes through decision tree, logistic
Regression, SVM scheduling algorithm carry out model training, and model is finally applied to all users, obtains classification results, by maximum
Score obtains the diffusion result of the user tag.
On the one hand above-mentioned tagging user broadcast algorithm needs to construct each user tag one classifier, training
Very big with the cost of maintenance, algorithm complexity is high；On the other hand, for some long-tail unexpected winner user tags, due to including user
The original user of label is less, and the accuracy of classifier is influenced when training.
Summary of the invention
The embodiment of the present invention provides the orientation method and device of a kind of advertisement crowd, to realize when advertisement crowd extracts
While tagging user is spread, algorithm complexity and accuracy are taken into account.
On the one hand, the orientation method of advertisement crowd a kind of is provided, method includes:
Obtain the target user's label for being used to extract advertisement crowd of user's input；
In the user tag of each description user characteristics obtained in advance, target user's label and other each is determined respectively
Similarity value between a user tag；
According to the similarity value between target user's label and other each user tags, determination and target user's label phase
Like maximum K similar users label is spent, K is natural number；
It using the user comprising each similar users label as the advertisement crowd being directed to, or will include each similar use
The user of family label and user comprising target user's label are as the advertisement crowd being directed to.
On the one hand, the orienting device of advertisement crowd a kind of is provided, device includes:
Label acquiring unit, for obtaining the target user's label for being used to extract advertisement crowd of user's input；
Similarity determining unit, for being determined in the user tag of each description user characteristics obtained in advance respectively
Similarity value between target user's label and other each user tags；
Similar tags determination unit, for according to the similarity between target user's label and other each user tags
Value, the determining and maximum K similar users label of target user's label similarity, K is natural number；
Directed element, for will include the user of each similar users label as the advertisement crowd being directed to, or general
User comprising each similar users label and the user comprising target user's label are as the advertisement crowd being directed to.
Optionally, device further include:
Number obtainment unit, for obtaining directional user's quantity of user's input；
User selection unit is greater than or equal to directional user's quantity for the quantity in the advertisement crowd being directed to
When, the user of directional user's quantity is selected from the advertisement crowd being directed to；And the number in the advertisement crowd being directed to
When amount is less than directional user's quantity, increases K value, redirect advertisement crowd, and big in the quantity of the advertisement crowd redirected
When directional user's quantity, the user of directional user's quantity is selected from the advertisement crowd redirected.
Optionally, user selection unit is specifically used for:
In directed element using the user comprising each similar users label as when the advertisement crowd being directed to, for K
Similar users label determines that the user comprising each similar users label collects, and obtains K user's collection；
For each user that K user concentrates, is collected according to the collection of user belonging to each user and each user and corresponded to
Score value, determine the score value of each user, it is that each user collects corresponding similar use that each user, which collects corresponding score value,
The similarity value of family label and target user's label；
Based on the score value of each user, according to preset rules, directional user's number is selected from the advertisement crowd being directed to
The user of amount.
Optionally, user selection unit is specifically used for:
In directed element using the user comprising each similar users label and the user comprising target user's label as fixed
When to the advertisement crowd arrived, for K similar users label, determines that the user comprising each similar users label collects, obtain K
User's collection；
Each user that each user concentrated for K user and the user comprising target user's label concentrate, root
Collect corresponding score value according to the collection of user belonging to each user and each user, determines the score value of each user, Mei Geyong
Collection corresponding score value in family is the similarity value that each user collects corresponding similar users label and target user's label；
Based on the score value of each user, according to preset rules, directional user's number is selected from the advertisement crowd being directed to
The user of amount.
Optionally, similarity determining unit is specifically used for: determining that target user marks using term vector model trained in advance
Label term vector corresponding with each other user tags；Based on target user's label word corresponding with other each user tags to
Amount determines the similarity value between target user's label and other each user tags respectively.
Optionally, similarity determining unit generates term vector model using following steps training: obtaining multiple users and use
The corresponding relationship of family tally set is as training sample data；Based on training sample data and pre-generated training label, use
Deep neural network training generates term vector model, training label be in training sample data the corresponding word of each user tag to
Amount.
Optionally, similarity determining unit is specifically used for: in the corresponding relationship of user gathered in advance and user tag collection
In, the corresponding relationship of multiple users and user tag collection is randomly selected as training sample data.
Optionally, similarity determining unit is specifically used for: in the corresponding relationship of user gathered in advance and user tag collection
In, screening user tag concentrates user tag quantity greater than the corresponding relationship of the first preset threshold；In the corresponding relationship filtered out
In, the corresponding relationship of multiple users and user tag collection is chosen as training sample data.
Optionally, similarity determining unit is specifically used for: in the corresponding relationship of user gathered in advance and user tag collection
In, statistics includes the number of users of each user tag；For each popular user tag, from the use comprising popular user tag
The user of the second pre-set threshold numbers is extracted in family, popular user tag is that the number of users comprising user tag is greater than second in advance
If the user tag of threshold value；In the user extracted and user comprising non-popular user tag, multiple target users are chosen,
And the corresponding relationship of multiple target users and user tag collection is used as to training sample data, non-hot topic user tag for comprising with
The number of users of family label is less than or equal to the user tag of the second preset threshold.
On the one hand, a kind of computer equipment is provided, including memory, processor and storage on a memory and can handled
The computer program run on device, the processor realize that method described in above-mentioned aspect walks when executing the computer program
Suddenly.
On the one hand, a kind of computer readable storage medium is provided, the computer-readable recording medium storage has computer
Instruction, when the computer instruction is run on computers, enables a computer to execute method described in above-mentioned aspect.
In the embodiment of the present invention, by obtaining the target user's label for being used to extract advertisement crowd of user's input, pre-
In the user tag of each description user characteristics first obtained, determine respectively target user's label and other each user tags it
Between similarity value, it is determining to be used with target and according to the similarity value between target user's label and other each user tags
The maximum K similar users label of family label similarity, the user comprising each similar users label is wide as what is be directed to
Announcement crowd, or the user comprising each similar users label and the user comprising target user's label is wide as what is be directed to
Announcement crowd, to realize the label crowd diffusion comprising target user's label, while without to each user tag structure
A classifier is built, algorithm complexity is reduced, and for long-tail unexpected winner user tag, also can determine its similar users label,
Improve accuracy.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Inventive embodiments for those of ordinary skill in the art without creative efforts, can also be according to mentioning
The attached drawing of confession obtains other attached drawings.
Fig. 1 is the flow diagram of the orientation method of advertisement crowd provided in an embodiment of the present invention；
Fig. 2 is the interface schematic diagram that creation user tag provided in an embodiment of the present invention is extracted；
Fig. 3 is the interface schematic diagram of request user's selection target user tag provided in an embodiment of the present invention；
Fig. 4 is the interface schematic diagram provided in an embodiment of the present invention for showing the number of users comprising user tag；
Fig. 5 is the schematic diagram in term vector space provided in an embodiment of the present invention；
Fig. 6 is the flow diagram of the detailed process of the orientation method of advertisement crowd provided in an embodiment of the present invention；
Fig. 7 is the structural schematic diagram of the orienting device of advertisement crowd provided in an embodiment of the present invention；
Fig. 8 is a kind of structural schematic diagram of computer equipment provided in an embodiment of the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction in the embodiment of the present invention
Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only
It is a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people
Member's every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.?
In the case where not conflicting, the feature in embodiment and embodiment in the present invention can mutual any combination.Although also, flowing
Logical order is shown in journey figure, but in some cases, it can be to be different from shown or described by sequence execution herein
The step of.
Technical solution provided in an embodiment of the present invention for ease of understanding, some passes that first embodiment of the present invention is used here
Key name word explains:
User tag: for describing user characteristics, for example, user tag can describe the interest, hobby, personality of user
Deng.
Tagging user diffusion: according to user's portrait available user list comprising user tag of label, according to each
The primary object list of user tag carries out object diffusion to it, meets user's focal need, that is, is not appreciably affecting practical effect
In the case where fruit (for example, advertising results), expand tagging user scale.
In addition, the terms "and/or", only a kind of incidence relation for describing affiliated partner, indicates may exist
Three kinds of relationships, for example, A and/or B, can indicate: individualism A exists simultaneously A and B, these three situations of individualism B.Separately
Outside, character "/" herein typicallys represent the relationship that forward-backward correlation object is a kind of "or" in the case where not illustrating.
Label crowd extraction is that one in data management platform (Data Management Platform, DMP) is important
Component, user give sequence of user label as stereotactic conditions, advertisement crowd are extracted by association user portrait, that is, is passed through
The combination of the user tags such as interest abundant, behavior, draws a circle to approve required user group.
Extract generally, based on the advertisement crowd of user tag and can only extract the user comprising user tag, namely comprising with
The user of family label how many just extract how much, to extract more massive user group or extracting the user of specific scale
Group then needs to carry out tagging user diffusion.
The tagging user diffusion way of mainstream is at present: use the original user comprising user tag as positive sample, with
Machine chooses other users as negative sample (the methods of PU-Learning can be used in the acquisition of negative sample), in positive sample and bears
80% is randomly selected in sample as training data, 20% and is used as test data, passes through decision tree, logistic
Regression, SVM scheduling algorithm carry out model training, and model is finally applied to all users, obtains classification results, by maximum
Score obtains the diffusion result of the user tag.
On the one hand above-mentioned tagging user broadcast algorithm needs to construct each user tag one classifier, training
Very big with the cost of maintenance, algorithm complexity is high；On the other hand, for some long-tail unexpected winner user tags, due to including user
The original user of label is less, and the accuracy of classifier is influenced when training.
In consideration of it, the embodiment of the invention provides the orientation methods of advertisement crowd a kind of, in the method, used by obtaining
The target user's label for being used to extract advertisement crowd of family input, in the user tag of each description user characteristics obtained in advance
In, determine the similarity value between target user's label and other each user tags respectively, and according to target user's label with
Similarity value between other each user tags, the determining and maximum K similar users label of target user's label similarity,
Using the user comprising each similar users label as the advertisement crowd being directed to, or each similar users label will be included
User and user comprising target user's label are as the advertisement crowd being directed to, to realize comprising target user's label
Label crowd diffusion, while without constructing a classifier to each user tag, reduce algorithm complexity, and for
Long-tail unexpected winner user tag also can determine its similar users label, improve accuracy.
Also, the embodiment of the present invention is for given directional user's quantity, in the quantity for the advertisement crowd being directed to
When more than or equal to directional user's quantity, the user of directional user's quantity can be selected directly from the advertisement crowd being directed to,
And when the quantity for the advertisement crowd being directed to is less than directional user's quantity, increase K value, redirects advertisement crowd, and
When the quantity of the advertisement crowd redirected is greater than or equal to directional user's quantity, selected from the advertisement crowd redirected
The user of directional user's quantity meets the requirement for being differently directed number of users.
After having introduced the design philosophy of the embodiment of the present invention, the technical solution of the embodiment of the present invention can be fitted below
Application scenarios do some simple introductions, it should be noted that application scenarios introduced below are merely to illustrate of the invention real
Apply example and non-limiting.In the specific implementation process, skill provided in an embodiment of the present invention can be neatly applied according to actual needs
Art scheme.
In a kind of scene, by taking application of the technical solution in inventive embodiments in advertisement orientation dispensing field as an example,
In the scene, advertisement is launched on single page application program (Single Page web Application, SPA) by advertisement dispensing side
When, advertisement is if desired delivered to the user comprising user tag " audiophile ", and it is desirable that be delivered to 5,000,000 users.
If there was only 1,000,000 comprising the user of " audiophile " label in user tag in SPA DMP system, at this time in order to
The dispensing requirement for meeting advertisement dispensing side searches out and is best suitable for this user tag characteristic of audiophile while meeting scale
User group can obtain 5,000,000 using the orientation method of advertisement crowd provided in an embodiment of the present invention and include " music hair
The user of burning friend " label.
In practical application, using the method for the embodiment of the present invention, advertisement dispensing side, which can choose, launches the rule that advertisement needs
Mould, and do not have to be limited to the original user quantity comprising user tag.
In another scene, by taking application of the technical solution in inventive embodiments in data desensitization field as an example, at this
Jing Zhong, for example, which label user has, make sures if the initial data comprising user tag is more sensitive to non-advertising
With when, in order to desensitize to data, original tag crowd can be diffused, in this way spread after result can play
The effect of desensitization, so that it may which the business for being supplied to some non-commercial papers uses.
Specifically, for the interest tags system constructed exclusively for SPA advertising business, be supplied to if necessary it is external its
Its non-advertising business uses, and can encounter data sensitive problem.For example, for user interest label, it can be in SPA interest tags
It on the basis of system, is diffused using the orientation method of advertisement crowd provided in an embodiment of the present invention, for example, being diffused into original
5 times of label crowd size, the crowd after diffusion can be supplied to other business and use, since which other business can not know
User is original true tag crowd, therefore can play desensitization effect.
Certainly, method provided in an embodiment of the present invention and unlimited above-mentioned two application scenarios, can be also used for other needs
The scene of user group of the orientation comprising user tag, the embodiment of the present invention are simultaneously not limited.
It referring to Figure 1, is the flow diagram of the orientation method of advertisement crowd provided in an embodiment of the present invention, party's rule
It can such as be executed with the orienting device of advertisement crowd provided in an embodiment of the present invention, certainly, in actual application, the device
It can also be realized by having the computer equipment of corresponding computing capability, such as personal computer (Personal can be passed through
Computer, PC), server or computer cluster realize.The process of this method is described as follows.
Step 101: obtaining the target user's label for being used to extract advertisement crowd of user's input.
It, can be defeated with direct request user when the specific target user's label for being used to extract advertisement crowd for obtaining user's input
Enter, such as: input frame is shown in terminal interface, request user inputs target user's label for extracting advertisement crowd；?
User tag can be selected as defeated from the user tag of displaying to user's exposition or whole user tags, request user
The target user's label entered.
Certainly, it should be noted that user's exposition or whole user tags, request user of the user from displaying
When selecting user tag target user's label as input in label, target user's label is chosen for the convenience of the user, it can will
Some or all of displaying user tag is sorted out, and is shown in the form of multistage list.
In one example, as shown in Fig. 2, triggering is used based on user tag if user clicks the Create button
Family is extracted, then user is requested to input or select target user's label for extracting user.Specifically refer to Fig. 3, Yong Huke
To select target user's label for extracting user in a kind of label, two class labels, three classes label, for example, being shown in Fig. 3
Target user's label for selecting of user be " tourism ".
Step 102: in the user tag of each description user characteristics obtained in advance, determining target user's label respectively
With the similarity value between other each user tags.
It should be noted that the user tag of each description user characteristics, can mark in user gathered in advance and user
It is obtained in the corresponding relationship of label, for example, fetching portion or all unduplicated user tag.Wherein, user and user mark
The corresponding relationship of label can be collected and recorded in real time by user tag server.
For example, user A Adds User when label a, label a is added the corresponding user of user A and marked by user tag server
Label are concentrated；For another example the user tag b of user B by the user tag addition user delete when, user tag server with
B corresponding user tag in family, which is concentrated, deletes user tag b.
It, can be in a manner of traversal from user gathered in advance when specific fetching portion or all unduplicated user tag
It, can also be first from use gathered in advance with fetching portion in the corresponding relationship of user tag collection or whole unduplicated user tags
User tag is obtained in the corresponding relationship of family and user tag collection, then progress duplicate removal processing obtains partly or entirely unduplicated
User tag.
When it is implemented, can be incited somebody to action when determining the similarity value between target user's label and other each user tags
User tag is as a word, the directly semantic similarity between calculating target user's label and other each user tags
Value, first can also be converted to target user's label with other each user tags corresponding using user tag as a kind of behavior
Term vector, then calculate the similarity value between target user's label term vector corresponding with other each user tags.
Specifically, the similarity value between target user's label term vector corresponding with other each user tags is calculated
When, the cosine similarity between target user's label term vector corresponding with other each user tags can be calculated, specifically may be used
To be calculated using following formula:
Wherein, similarity value of the sim (A, B) between user tag A and user tag B, cosine (VA, VB) it is user
The corresponding term vector V of label AATerm vector V corresponding with user tag BBBetween cosine similarity.
Step 103: determining to be used with target according to the similarity value between target user's label and other each user tags
The maximum K similar users label of family label similarity, K is natural number.
When it is implemented, according to the similarity value between target user's label and other each user tags, determining and mesh
The maximum K similar users label of user tag similarity is marked, it can first will be between target user's label and other user tags
Similarity value descending arrangement, then by the corresponding user tag of K similarity value preceding in ranking results, be determined as and target use
The maximum K similar users label of family label similarity.Wherein, K can be set based on experience value, for example, the value of K is 10.
Step 104: using the user comprising each similar users label as the advertisement crowd being directed to, or will be comprising each
The user of a similar users label and user comprising target user's label are as the advertisement crowd being directed to.
When it is implemented, determining and the maximum K similar users label of target user's label similarity in step 103
It afterwards, can also will be able to include each similar using the user comprising each similar users label as the advertisement crowd being directed to
The user of user tag and user comprising target user's label realize and use comprising target as the advertisement crowd being directed to
The label crowd of family label is spread.
In concrete application, using the user comprising each similar users label as the mode for the advertisement crowd being directed to, phase
Compared with will include the user of each similar users label and user comprising target user's label as the advertisement crowd being directed to
Mode can preferably desensitize in the more sensitive scene of data for data.
Fig. 4 is referred to, the embodiment of the present invention can show to include target user after being directed to advertisement crowd to user
The number of users of label.Wherein, the number of users comprising target user's label is shown to user, can be and mark comprising target user
The sum of the number of users of the number of users of label and the similar users label comprising target user's label is also possible to use comprising target
The number of users of the similar users label of family label.
It should be noted that be illustrated for target user's label in above-described embodiment, in practical application,
If desired advertisement crowd orientation is carried out to multiple target user's labels, then it can be for each mesh in multiple target user's labels
User tag is marked, aforesaid way is all made of and is oriented, the embodiment of the present invention repeats no more this.
In practical application, user can may also have the advertisement crowd's quantity being directed to when carrying out advertisement crowd orientation
Certain requirement in this way, the present invention can also obtain directional user's quantity of user's input when implementing, and is being directed to
When the quantity of advertisement crowd is greater than or equal to directional user's quantity, directional user's quantity is selected from the advertisement crowd being directed to
User, and when the quantity for the advertisement crowd being directed to is less than directional user's quantity, increase K value, using above-mentioned steps
103- step 104 redirects advertisement crowd, and is greater than or equal to directional user's number in the quantity of the advertisement crowd redirected
When amount, the user of directional user's quantity is selected from the advertisement crowd redirected.
For example, advertisement dispensing side needs to be delivered to advertisement comprising user tag " sound when launching advertisement on SPA
The user of happy enthusiast ", and it is desirable that 5,000,000 users namely target user's label are delivered to as " audiophile ", orientation is used
Amount amount is 5,000,000.If the original user in SPA data system comprising " audiophile " user tag has 1,000,000, fixed
It, can be when the quantity for the advertisement crowd being directed to be greater than 5,000,000, from orientation when to advertisement crowd comprising " audiophile "
To advertisement crowd in select 5,000,000 users as the advertisement crowd being directed to.Certainly, if the number for the advertisement crowd being directed to
Amount is equal to 5,000,000, then without selection, the advertisement crowd being directed to directly can be supplied to user.
In practical application, the advertisement crowd and directional user's quantity being directed to are often unequal, and therefore, it is necessary to from being directed to
Advertisement crowd in select the object of directional user's quantity, when specific choice, according to targeted ads crowd in above-mentioned steps 104
Mode it is different, select the user of directional user's quantity also slightly different from the advertisement crowd being directed to.Specifically, may be used
To be divided into following two embodiment.
Embodiment one
If using the user comprising each similar users label as the advertisement crowd being directed in step 104, from being directed to
Advertisement crowd in when selecting the user of directional user's quantity, for K similar users label, determine to include each similar use
The user of family label collects, and K user's collection is obtained, for each user that K user concentrates, according to user belonging to each user
Collection and each user collect corresponding score value, determine the score value of each user, it is every that each user, which collects corresponding score value,
A user collects the similarity value of corresponding similar users label and target user's label, based on the score value of each user, foundation
Preset rules select the user of directional user's quantity from the advertisement crowd being directed to.
Embodiment two
If using the user comprising each similar users label and the user comprising target user's label as fixed in step 104
To the advertisement crowd arrived, when selecting the user of directional user's quantity from the advertisement crowd being directed to, for K similar users
Label determines that the user comprising each similar users label collects, and obtains K user's collection, each user concentrated for K user
And each user that the user comprising target user's label concentrates, according to the collection of user belonging to each user and each user
Collect corresponding score value, determines the score value of each user, it is that each user collects corresponding that each user, which collects corresponding score value,
The similarity value of similar users label and target user's label, based on the score value of each user, according to preset rules, from orientation
To advertisement crowd in select directional user's quantity user.
Wherein, preset rules can be the rule of score value from high to low, is also possible to score value and is greater than default scoring threshold
The rule of value, it is not limited in the embodiment of the present invention.
In an example it is assumed that K similar users label of target user's label is respectively user tag A, Yong Hubiao
B, user tag C are signed, if the similarity value of user tag A and target user's label is 0.9, user tag B and target user are marked
The similarity value of label is 0.8, and the similarity value of user tag C and target user's label is 0.7, and then includes user tag A
It includes user 1, user 2, user 3 and user 4 that user, which concentrates, and it includes user 1, user that the user comprising user tag B, which concentrates,
3, user 5 and user 7, it includes user 1, user 2, user 6 and user 9 that the user comprising user tag C, which concentrates,.
When selecting the user of directional user's quantity from the advertisement crowd being directed to, the score value of each user is calculated, it is first
First determining that the score value that the user comprising user tag A collects is 0.9, the score value of user's collection comprising user tag B is 0.8,
The score value of user's collection comprising user tag C is 0.7, then since user 1 is included in user's collection, the Yong Hubiao of user tag A
The user of the user's collection and user tag C of signing B concentrates, then the score value that user 1 can be calculated is 2.4, since user 2 wraps
The user of the user's collection and user tag C that are contained in user tag A concentrates, then the score value that user 2 can be calculated is 1.6,
And so on, the score value that user 3 can be calculated is 1.7, and the score value of user 4 is 0.9, and the score value of user 5 is
0.8, the score value of user 6 is 0.7, and the score value of user 7 is 0.8, and the score value of user 9 is 0.7.
If preset rules are the rule of score value from high to low, directional user's quantity is 3, then from the advertisement crowd being directed to
When selecting the user of directional user's quantity in (user comprising user tag A, user tag B and user tag C), chooses and use
Family 1, user 2 and user 3.
Step 102 specifically calculates the similarity value between user tag in above-described embodiment, and user tag is converted to pair
When the term vector answered, it can use term vector model trained in advance and determine the corresponding term vector of each user tag.Wherein, in advance
First training generates trained term vector model in the following way: obtaining the corresponding relationship conduct of multiple users Yu user tag collection
Training sample data generate word using deep neural network training based on training sample data and pre-generated training label
Vector model, training label are the corresponding term vector of user tag each in training sample data.
Specifically, by taking the word2vec in word insertion embedding as an example, word2vec is inputted using neural network analysis
Training sample data (user tag), each user tag is generated and represents the vector of this user tag: a dimension
It is embedded into the much lower vector row space of a dimension for the higher dimensional space of all user tag quantity, each user tag
The vector being mapped as in real number field.If two different user tags frequently appear in similar context, it is believed that
Using any of two user tags as input, neural network will export very similar predicted value.Two user's marks
The number for checking out present similar situation is more, their coordinate will be closer, i.e., so that semantic close word after insertion
Space in position also can be close.
In one example, as shown in figure 5, word " Obama " in document 1 (document1) and document 2
(document2) number that the word " President " in appears in identical context is more, they are empty in the vector of word insertion
Between middle position be closer to；The word in word " speaks " and document 2 (document2) in document 1 (document1)
The number that " greets " is appeared in identical context is more, they are closer to position in the vector space of word insertion；Document 1
(document1) word " press " in word " media " and document 2 (document2) in appears in identical context
Number is more, they are closer to position in the vector space of word insertion；Word in document 1 (document1)
" Illinois " and the number that the word " Chicago " in document 2 (document2) appears in identical context are more, they
Position is closer in the vector space of word insertion.
It should be noted that when obtaining the corresponding relationship of multiple users and user tag collection as training sample data,
It, can be in the corresponding relationship of user gathered in advance and user tag collection for quick obtaining in a kind of possible embodiment
In, the corresponding relationship of multiple users and user tag collection is randomly selected as training sample data.
In another possible embodiment, in order to increase the user tag quantity got, can also first it be adopted in advance
In the user of collection and the corresponding relationship of user tag collection, screening user tag concentrates user tag quantity to be greater than the first preset threshold
Corresponding relationship choose the corresponding relationship of multiple users and user tag collection as instructing then in the corresponding relationship filtered out
Practice sample data.Wherein, the first preset threshold can be set based on experience value, for example, the value of the first preset threshold is
3。
In another possible embodiment, for the frequency that user tag occurs in equalizing training sample data, pre-
In the corresponding relationship of the user and user tag collection that first acquire, the corresponding relationship of multiple users and user tag collection is obtained as instruction
When practicing sample data, the number of users comprising each user tag can be first counted, and for each popular user tag, from packet
The user of the second pre-set threshold numbers is extracted in user containing popular user tag, popular user tag is to include user tag
Number of users is greater than the user tag of the second preset threshold, then in the user extracted and use comprising non-popular user tag
In family, multiple target users are chosen, and using the corresponding relationship of multiple target users and user tag collection as training sample data,
Non- hot topic user tag is the user tag that the number of users comprising user tag is less than or equal to the second preset threshold.Wherein,
Second preset threshold can be set based on experience value, can also be set according to the number of users comprising each user tag, for example,
Second preset threshold is set as the median of the number of users comprising each user tag.
Certainly it should be noted that obtaining the corresponding relationship of multiple users and user tag collection as training sample data
When, above-mentioned three kinds of embodiments can be used alone, and can also be used in combination, and present invention implementation does not limit this.
It is provided in an embodiment of the present invention in conjunction with the orientation method of above-mentioned advertisement crowd and the training method of term vector model
The detailed process of the orientation method of advertisement crowd, as shown in Figure 6, comprising:
Step 601: obtaining the corresponding relationship of multiple users and user tag collection as training sample data, based on training sample
Notebook data and pre-generated training label generate term vector model using deep neural network training.
Step 602: obtaining the target user's label for being used to extract advertisement crowd of user's input.
Step 603: obtaining the corresponding term vector of target user's label using term vector model and obtain in advance each
The corresponding term vector of user tag.
Step 604: the corresponding term vector of each user tag obtained according to target user's label and in advance word to
Distance in quantity space, the determining and maximum K similar users label of target user's label similarity.
It should be noted that in term vector space, the term vector closer word of distance corresponding with target user's label to
Amount, corresponding user tag and the similarity of target user's label are bigger.
Step 605: using the user comprising each similar users label as the advertisement crowd being directed to, or will be comprising each
The user of a similar users label and user comprising target user's label are as the advertisement crowd being directed to.
Fig. 7 is referred to, based on the same inventive concept, the embodiment of the invention also provides the orienting devices of advertisement crowd a kind of
70, comprising:
Label acquiring unit 701, for obtaining the target user's label for being used to extract advertisement crowd of user's input.
Similarity determining unit 702, in the user tag of each description user characteristics obtained in advance, difference to be true
Similarity value between the user tag that sets the goal and other each user tags.
Similar tags determination unit 703, for according to similar between target user's label and other each user tags
Angle value, the determining and maximum K similar users label of target user's label similarity, K is natural number.
Directed element 704, for using include each similar users label user as the advertisement crowd being directed to, or
Using the user comprising each similar users label and the user comprising target user's label as the advertisement crowd being directed to.
Optionally, device further include: number obtainment unit 705, for obtaining directional user's quantity of user's input；User
Selecting unit 706, when being greater than or equal to directional user's quantity for the quantity in the advertisement crowd being directed to, from being directed to
Advertisement crowd in select directional user's quantity user；And it is less than orientation use in the quantity for the advertisement crowd being directed to
When amount amount, increase K value, redirects advertisement crowd, and be greater than or equal to orientation in the quantity of the advertisement crowd redirected
When number of users, the user of directional user's quantity is selected from the advertisement crowd redirected.
Optionally, user selection unit 706 are specifically used for: in directed element by the use comprising each similar users label
When family is as the advertisement crowd being directed to, for K similar users label, the user comprising each similar users label is determined
Collection obtains K user's collection；For each user that K user concentrates, according to the collection of user belonging to each user and each use
Collection corresponding score value in family determines the score value of each user, and it is that each user collects correspondence that each user, which collects corresponding score value,
Similar users label and target user's label similarity value；Based on the score value of each user, according to preset rules, from calmly
To advertisement crowd in select directional user's quantity user.
Optionally, user selection unit 706 are specifically used for: in directed element by the use comprising each similar users label
When family and user comprising target user's label are as the advertisement crowd being directed to, for K similar users label, determination includes
The user of each similar users label collects, and obtains K user's collection；Each user for concentrating for K user and include target
Each user that the user of user tag concentrates collects corresponding scoring according to the collection of user belonging to each user and each user
Value determines the score value of each user, and it is that each user collects corresponding similar users label that each user, which collects corresponding score value,
With the similarity value of target user's label；Based on the score value of each user, according to preset rules, from the advertisement crowd being directed to
The user of middle selection directional user's quantity.
Optionally, similarity determining unit 702 are specifically used for: determining that target is used using term vector model trained in advance
Family label and the corresponding term vector of each other user tags；It is corresponding with other each user tags based on target user's label
Term vector determines the similarity value between target user's label and other each user tags respectively.
Optionally, similarity determining unit 702 generate term vector model using following steps training: obtaining multiple users
Corresponding relationship with user tag collection is as training sample data；Based on training sample data and pre-generated training label,
Term vector model is generated using deep neural network training, training label is that each user tag is corresponding in training sample data
Term vector.
Optionally, similarity determining unit 702 are specifically used for: corresponding with user tag collection in user gathered in advance
In relationship, the corresponding relationship of multiple users and user tag collection is randomly selected as training sample data.
Optionally, similarity determining unit 702 are specifically used for: corresponding with user tag collection in user gathered in advance
In relationship, screening user tag concentrates user tag quantity greater than the corresponding relationship of the first preset threshold；In the correspondence filtered out
In relationship, the corresponding relationship of multiple users and user tag collection is chosen as training sample data.
Optionally, similarity determining unit 702 are specifically used for: corresponding with user tag collection in user gathered in advance
In relationship, statistics includes the number of users of each user tag；For each popular user tag, from including popular user tag
User in extract the users of the second pre-set threshold numbers, popular user tag is that the number of users comprising user tag is greater than the
The user tag of two preset thresholds；In the user extracted and user comprising non-popular user tag, multiple targets are chosen
User, and using the corresponding relationship of multiple target users and user tag collection as training sample data, non-hot topic user tag is
Number of users comprising user tag is less than or equal to the user tag of the second preset threshold.
Fig. 8 is referred to, same technical concept is based on, it, can be with the embodiment of the invention also provides a kind of computer equipment 80
Including memory 801 and processor 802.
The memory 801, the computer program executed for storage processor 802.Memory 801 can mainly include depositing
Store up program area and storage data area, wherein storing program area can application program needed for storage program area, at least one function
Deng；Storage data area, which can be stored, uses created data etc. according to computer equipment.Processor 802 can be in one
Central Processing Unit (central processing unit, CPU), or be digital processing element etc..In the embodiment of the present invention
The specific connection medium between above-mentioned memory 801 and processor 802 is not limited.The embodiment of the present invention is in fig. 8 with memory
It is connected between 801 and processor 802 by bus 803, bus 803 is indicated in fig. 8 with thick line, the connection between other components
Mode is only to be schematically illustrated, does not regard it as and be limited.The bus 803 can be divided into address bus, data/address bus, control
Bus processed etc..Only to be indicated with a thick line in Fig. 8, it is not intended that an only bus or a type of convenient for indicating
Bus.
Memory 801 can be volatile memory (volatile memory), such as random access memory
(random-access memory, RAM)；Memory 801 is also possible to nonvolatile memory (non-volatile
Memory), such as read-only memory, flash memory (flash memory), hard disk (hard disk drive, HDD) or solid
State hard disk (solid-state drive, SSD) or memory 801 can be used for carrying or storing have instruction or data
The desired program code of structure type and can by any other medium of computer access, but not limited to this.Memory 801
It can be the combination of above-mentioned memory.
Processor 802 executes as shown in Figure 7 when for calling the computer program stored in the memory 801
Method performed by equipment in embodiment.
In some possible embodiments, the various aspects of method provided by the invention are also implemented as a kind of program
The form of product comprising program code, when described program product is run on a computing device, said program code is used for
Execute the computer equipment in the method for illustrative embodiments various according to the present invention of this specification foregoing description
Step, for example, the computer equipment can execute method performed by equipment in as in the embodiment depicted in figure 7.
Described program product can be using any combination of one or more readable mediums.Readable medium can be readable letter
Number medium or readable storage medium storing program for executing.Readable storage medium storing program for executing for example may be-but not limited to-electricity, magnetic, optical, electromagnetic, red
The system of outside line or semiconductor, device or device, or any above combination.The more specific example of readable storage medium storing program for executing
(non exhaustive list) includes: the electrical connection with one or more conducting wires, portable disc, hard disk, random access memory
(RAM), read-only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc
Read memory (CD-ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.
Although preferred embodiments of the present invention have been described, it is created once a person skilled in the art knows basic
Property concept, then additional changes and modifications can be made to these embodiments.So it includes excellent that the following claims are intended to be interpreted as
It selects embodiment and falls into all change and modification of the scope of the invention.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art
Mind and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to include these modifications and variations.
Claims (10)
1. a kind of orientation method of advertisement crowd, which is characterized in that the described method includes:
Obtain the target user's label for being used to extract advertisement crowd of user's input；
In the user tag of each description user characteristics obtained in advance, target user's label and other each is determined respectively
Similarity value between a user tag；
It is determining to be marked with the target user according to the similarity value between target user's label and other each user tags
The maximum K similar users label of similarity is signed, K is natural number；
It using the user comprising each similar users label as the advertisement crowd being directed to, or will include each similar users mark
The user of label and user comprising target user's label are as the advertisement crowd being directed to.
2. the method according to claim 1, wherein the method also includes:
Obtain directional user's quantity of user's input；
When the quantity for the advertisement crowd being directed to described in the determination is greater than or equal to directional user's quantity, it is directed to from described
Advertisement crowd in select the user of directional user's quantity；
When the quantity for the advertisement crowd being directed to described in the determination is less than directional user's quantity, increase the K value, it is again fixed
To advertisement crowd, and when the quantity of the advertisement crowd redirected is greater than or equal to directional user's quantity, from described heavy
The user of directional user's quantity is selected in the advertisement crowd newly oriented.
3. according to the method described in claim 2, it is characterized in that, using the user comprising each similar users label as orientation
When the advertisement crowd arrived, the user that directional user's quantity is selected from the advertisement crowd being directed to, comprising:
It for the K similar users label, determines that the user comprising each similar users label collects, obtains K user's collection；
For each user that the K user concentrates, is collected according to the collection of user belonging to each user and each user and corresponded to
Score value, determine the score value of each user, it is that each user collects and corresponds to that each user, which collects corresponding score value,
Similar users label and target user's label similarity value；
Based on the score value of each user, according to preset rules, it is described fixed to select from the advertisement crowd being directed to
To the user of number of users.
4. according to the method described in claim 2, it is characterized in that, by the user comprising each similar users label and including institute
It is described to select institute from the advertisement crowd being directed to when stating the user of target user's label as the advertisement crowd being directed to
State the user of directional user's quantity, comprising:
It for the K similar users label, determines that the user comprising each similar users label collects, obtains K user's collection；
Each use that each user concentrated for the K user and the user comprising target user's label concentrate
Family collects corresponding score value according to the collection of user belonging to each user and each user, determines the score value of each user, institute
Stating each user to collect corresponding score value is that each user collects corresponding similar users label and target user's label
Similarity value；
Based on the score value of each user, according to preset rules, it is described fixed to select from the advertisement crowd being directed to
To the user of number of users.
5. the method according to claim 1, wherein the target user's label and other each of determining respectively
Similarity value between a user tag, comprising:
Target user's label and the corresponding term vector of each other user tags are determined using term vector model trained in advance；
Based on target user's label term vector corresponding with other each user tags, target user's mark is determined respectively
Similarity value between label and other each user tags.
6. according to the method described in claim 5, it is characterized in that, the term vector model trained in advance in the following way
Training generates:
The corresponding relationship of multiple users and user tag collection is obtained as training sample data；
Based on the training sample data and pre-generated training label, using deep neural network training generate institute's predicate to
Model is measured, the trained label is the corresponding term vector of user tag each in the training sample data.
7. according to the method described in claim 6, it is characterized in that, described obtain multiple users pass corresponding with user tag collection
System is used as training sample data, comprising:
In the corresponding relationship of user gathered in advance and user tag collection, pair of multiple users Yu user tag collection are randomly selected
It should be related to as training sample data.
8. the method according to the description of claim 7 is characterized in that described obtain multiple users pass corresponding with user tag collection
System is used as training sample data, comprising:
In the corresponding relationship of user gathered in advance and user tag collection, screening user tag concentrates user tag quantity to be greater than
The corresponding relationship of first preset threshold；
In the corresponding relationship filtered out, the corresponding relationship of multiple users and user tag collection is chosen as training sample data.
9. method according to claim 7 or 8, which is characterized in that pair for obtaining multiple users and user tag collection
It should be related to as training sample data, comprising:
In the corresponding relationship of user gathered in advance and user tag collection, statistics includes the number of users of each user tag；
For each popular user tag, the second pre-set threshold numbers are extracted from the user comprising the popular user tag
User, the hot topic user tag are the user tag that the number of users comprising user tag is greater than second preset threshold；
In the user extracted and user comprising non-popular user tag, multiple target users are chosen, and by multiple targets
For the corresponding relationship of user and user tag collection as training sample data, the non-popular user tag is to include user tag
Number of users is less than or equal to the user tag of second preset threshold.
10. a kind of orienting device of advertisement crowd, which is characterized in that described device includes:
Label acquiring unit, for obtaining the target user's label for being used to extract advertisement crowd of user's input；
Similarity determining unit, described in being determined in the user tag of each description user characteristics obtained in advance respectively
Similarity value between target user's label and other each user tags；
Similar tags determination unit, for according to the similarity between target user's label and other each user tags
Value, the determining and maximum K similar users label of target user's label similarity, K is natural number；
Directed element, for using include each similar users label user as the advertisement crowd being directed to, or will include
The user of each similar users label and user comprising target user's label are as the advertisement crowd being directed to.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910633443.7A CN110348907B (en) | 2019-07-12 | 2019-07-12 | Advertisement crowd orientation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910633443.7A CN110348907B (en) | 2019-07-12 | 2019-07-12 | Advertisement crowd orientation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110348907A true CN110348907A (en) | 2019-10-18 |
| CN110348907B CN110348907B (en) | 2024-05-28 |
Family
ID=68176253
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910633443.7A Active CN110348907B (en) | 2019-07-12 | 2019-07-12 | Advertisement crowd orientation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110348907B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111080359A (en) * | 2019-12-13 | 2020-04-28 | 北京搜狐新媒体信息技术有限公司 | Label algorithm determination method, device, server and storage medium |
| CN111598630A (en) * | 2020-06-30 | 2020-08-28 | 成都新潮传媒集团有限公司 | Cell portrait construction method and device and storage medium |
| CN112288461A (en) * | 2020-09-25 | 2021-01-29 | 北京沃东天骏信息技术有限公司 | Advertisement information pushing method, device, medium and electronic equipment |
| CN112488859A (en) * | 2020-11-26 | 2021-03-12 | 泰康保险集团股份有限公司 | Data processing method, device, equipment and storage medium |
| CN112734502A (en) * | 2021-03-29 | 2021-04-30 | 腾讯科技（深圳）有限公司 | Testing method and device for multimedia information directional delivery and electronic equipment |
| CN113674008A (en) * | 2020-05-14 | 2021-11-19 | 北京达佳互联信息技术有限公司 | Directional label recommendation method, device, server and storage medium |
| CN113836903A (en) * | 2021-08-17 | 2021-12-24 | 淮阴工学院 | Method and device for extracting enterprise portrait label based on situation embedding and knowledge distillation |
| CN113919877A (en) * | 2021-10-15 | 2022-01-11 | 深圳市酷开网络科技股份有限公司 | Method and device for processing human-circled task progress based on DMP platform and readable storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105825396A (en) * | 2016-03-11 | 2016-08-03 | 合网络技术（北京）有限公司 | Co-occurrence-based advertisement label clustering method and system |
| CN108876470A (en) * | 2018-06-29 | 2018-11-23 | 腾讯科技（深圳）有限公司 | Tagging user extended method, computer equipment and storage medium |
| CN109308332A (en) * | 2018-08-07 | 2019-02-05 | 腾讯科技（深圳）有限公司 | A kind of target user's acquisition methods, device and server |
| US20190213462A1 (en) * | 2017-07-20 | 2019-07-11 | Laava Id Pty Ltd | Systems and methods for generating secure tags |
-
2019
- 2019-07-12 CN CN201910633443.7A patent/CN110348907B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105825396A (en) * | 2016-03-11 | 2016-08-03 | 合网络技术（北京）有限公司 | Co-occurrence-based advertisement label clustering method and system |
| US20190213462A1 (en) * | 2017-07-20 | 2019-07-11 | Laava Id Pty Ltd | Systems and methods for generating secure tags |
| CN108876470A (en) * | 2018-06-29 | 2018-11-23 | 腾讯科技（深圳）有限公司 | Tagging user extended method, computer equipment and storage medium |
| CN109308332A (en) * | 2018-08-07 | 2019-02-05 | 腾讯科技（深圳）有限公司 | A kind of target user's acquisition methods, device and server |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111080359A (en) * | 2019-12-13 | 2020-04-28 | 北京搜狐新媒体信息技术有限公司 | Label algorithm determination method, device, server and storage medium |
| CN113674008A (en) * | 2020-05-14 | 2021-11-19 | 北京达佳互联信息技术有限公司 | Directional label recommendation method, device, server and storage medium |
| CN111598630A (en) * | 2020-06-30 | 2020-08-28 | 成都新潮传媒集团有限公司 | Cell portrait construction method and device and storage medium |
| CN112288461A (en) * | 2020-09-25 | 2021-01-29 | 北京沃东天骏信息技术有限公司 | Advertisement information pushing method, device, medium and electronic equipment |
| CN112488859A (en) * | 2020-11-26 | 2021-03-12 | 泰康保险集团股份有限公司 | Data processing method, device, equipment and storage medium |
| CN112734502A (en) * | 2021-03-29 | 2021-04-30 | 腾讯科技（深圳）有限公司 | Testing method and device for multimedia information directional delivery and electronic equipment |
| CN112734502B (en) * | 2021-03-29 | 2021-06-18 | 腾讯科技（深圳）有限公司 | Testing method and device for multimedia information directional delivery and electronic equipment |
| CN113836903A (en) * | 2021-08-17 | 2021-12-24 | 淮阴工学院 | Method and device for extracting enterprise portrait label based on situation embedding and knowledge distillation |
| CN113836903B (en) * | 2021-08-17 | 2023-07-18 | 淮阴工学院 | Enterprise portrait tag extraction method and device based on situation embedding and knowledge distillation |
| CN113919877A (en) * | 2021-10-15 | 2022-01-11 | 深圳市酷开网络科技股份有限公司 | Method and device for processing human-circled task progress based on DMP platform and readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110348907B (en) | 2024-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110348907A (en) | A kind of orientation method and device of advertisement crowd | |
| JP6986527B2 (en) | How and equipment to process video | |
| Zhu et al. | Multimodal c4: An open, billion-scale corpus of images interleaved with text | |
| Dubey et al. | Deep learning the city: Quantifying urban perception at a global scale | |
| EP3473016B1 (en) | Method and system for automatically producing video highlights | |
| Du et al. | Personalized video recommendation using rich contents from videos | |
| KR20160040633A (en) | Systems and methods for image classification by correlating contextual cues with images | |
| CN113590854B (en) | Data processing method, data processing equipment and computer readable storage medium | |
| CN106528894B (en) | The method and device of label information is set | |
| CN108376164B (en) | Display method and device of potential anchor | |
| CN107368856A (en) | Clustering method and device, the computer installation and readable storage medium storing program for executing of Malware | |
| CN106919575A (en) | application program searching method and device | |
| CN110119477A (en) | A kind of information-pushing method, device and storage medium | |
| Zhu et al. | Feature engineering for place category classification | |
| US20230410222A1 (en) | Information processing apparatus, control method, and program | |
| CN112231468A (en) | Information generation method and device, electronic equipment and storage medium | |
| CN104281641A (en) | Method for enriching a multimedia content, and corresponding device | |
| Karunanayake et al. | A multi-modal neural embeddings approach for detecting mobile counterfeit apps: A case study on Google Play store | |
| CN111552865A (en) | User interest portrait method and related equipment | |
| CN108563713A (en) | Keyword rule generating method and device and electronic equipment | |
| CN104462151B (en) | Assess the method and relevant apparatus of Homepage Publishing time | |
| EP2835748A1 (en) | Systems and methods for image classification by correlating contextual cues with images | |
| CN109361929B (en) | Method for determining live broadcast room label and related equipment | |
| CN109033078B (en) | The recognition methods of sentence classification and device, storage medium, processor | |
| Rao et al. | Deep learning-based image retrieval system with clustering on attention-based representations |
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 |