CN110413882A - Information-pushing method, device and equipment - Google Patents
Information-pushing method, device and equipment Download PDFInfo
- Publication number
- CN110413882A CN110413882A CN201910636124.1A CN201910636124A CN110413882A CN 110413882 A CN110413882 A CN 110413882A CN 201910636124 A CN201910636124 A CN 201910636124A CN 110413882 A CN110413882 A CN 110413882A
- Authority
- CN
- China
- Prior art keywords
- user
- phrase
- word
- behavior
- under
- 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 43
- 230000006399 behavior Effects 0.000 claims abstract description 136
- 230000003542 behavioural effect Effects 0.000 claims abstract description 35
- 238000013507 mapping Methods 0.000 claims abstract description 35
- 238000013135 deep learning Methods 0.000 claims abstract description 32
- 238000010304 firing Methods 0.000 claims abstract description 11
- 238000013527 convolutional neural network Methods 0.000 claims description 8
- 238000012216 screening Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 7
- 230000015572 biosynthetic process Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 238000005194 fractionation Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 235000013305 food Nutrition 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000010992 reflux Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/55—Push-based network services
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
This specification embodiment provides a kind of information-pushing method, device and equipment and obtains the user behavior data of user in information-pushing method.Based on user behavior data, the behavior scene of user is determined.Identify the user tag of user.The semanteme of the user tag reflects user's portrait of user.The label phrase firing table pre-established is inquired, the corresponding user's phrase of user tag under behavior scene is obtained.The semanteme of user's phrase reflects the behavioural habits of user.Label phrase mapping table is used to be recorded in the corresponding relationship under different behavior scenes between user tag and user's phrase.Wherein, the user tag under a kind of behavior scene and user's phrase are behavioural habits and deep learning the algorithm determination based on the user under behavior scene or other corelation behaviour scenes.Search the target information to match with user's phrase.Target information is pushed to user.
Description
Technical field
This specification one or more embodiment be related to field of computer technology more particularly to a kind of information-pushing method,
Device and equipment.
Background technique
It, may using that would generally be pushed to it when user uses or accesses application (APP) in order to promote user experience
Interested information.In traditional information push scheme, the historical behavior data of user are normally based on, it can to analyze user
The interested information of energy simultaneously realizes push.
However, the information type that traditional information-pushing method is pushed is relatively simple.Such as, it is currently answered when user only passes through
When with browsing skin-protection product, the push of skin-protection product can only be realized, and cannot achieve the push of other information.
Accordingly, it is desirable to provide a kind of more efficient, more flexible information-pushing method.
Summary of the invention
This specification one or more embodiment describes the method for pushing, device and equipment of a kind of information, to improve
Flexibility to user's pushed information.
In a first aspect, providing a kind of information-pushing method, comprising:
Obtain the user behavior data of user；
Based on the user behavior data, the behavior scene of the user is determined；
Identify the user tag of the user；The user tag is used to describe user's portrait of the user；
The label phrase firing table that pre-establishes is inquired, is obtained under the behavior scene, the corresponding use of the user tag
Family phrase；User's phrase is used to describe the behavioural habits of user；The label phrase mapping table is not gone together for being recorded in
For the corresponding relationship between user tag under scene and user's phrase；Wherein, the user tag under a kind of behavior scene and user
Phrase be based under behavior scene or other corelation behaviour scenes the behavioural habits and deep learning algorithm of user it is true
Fixed；
Search the target information to match with user's phrase；
The target information is pushed to the user.
Second aspect provides a kind of information push-delivery apparatus, comprising:
Acquiring unit, for obtaining the user behavior data of user；
Determination unit, the user behavior data for being obtained based on the acquiring unit, determines the row of the user
For scene；
Recognition unit, for identification user tag of the user；The user tag is used to describe the use of the user
Family portrait；
The acquiring unit is also used to inquire the label phrase firing table pre-established, obtains the institute under the behavior scene
State the corresponding user's phrase of user tag；User's phrase is used to describe the behavioural habits of user；The label phrase mapping
Table is used to be recorded in the corresponding relationship under different behavior scenes between user tag and user's phrase；Wherein, a kind of behavior scene
Under user tag and user's phrase be the behavioural habits based on the user under behavior scene or other corelation behaviour scenes
And the determination of deep learning algorithm；
Searching unit, the target information to match for searching the user's phrase obtained with the acquiring unit；
Push unit, for pushing the target information to the user.
The third aspect provides a kind of information pushing equipment, comprising:
Memory；
One or more processors；And
One or more programs wherein the storage of one or more of programs is in the memory, and are configured to
It is executed by one or more of processors, described program performs the steps of when being executed by the processor
Obtain the user behavior data of user；
Based on the user behavior data, the behavior scene of the user is determined；
Identify the user tag of the user；The user tag is used to describe user's portrait of the user；
The label phrase firing table that pre-establishes is inquired, is obtained under the behavior scene, the corresponding use of the user tag
Family phrase；User's phrase is used to describe the behavioural habits of user；The label phrase mapping table is not gone together for being recorded in
For the corresponding relationship between user tag under scene and user's phrase；Wherein, the user tag under a kind of behavior scene and user
Phrase be based under behavior scene or other corelation behaviour scenes the behavioural habits and deep learning algorithm of user it is true
Fixed；
Search the target information to match with user's phrase；
The target information is pushed to the user.
Information-pushing method, device and the equipment that this specification one or more embodiment provides, obtain the user of user
Behavioral data.Based on user behavior data, the behavior scene of user is determined.Identify the user tag of user.The user tag
Semanteme reflects user's portrait of user.The label phrase firing table pre-established is inquired, the Yong Hubiao under behavior scene is obtained
Sign corresponding user's phrase.The semanteme of user's phrase reflects the behavioural habits of user.Label phrase mapping table is for recording
Corresponding relationship under different behavior scenes between user tag and user's phrase.Wherein, user's mark under a kind of behavior scene
Label and user's phrase are behavioural habits and depth based on the user under behavior scene or other corelation behaviour scenes
Practise what algorithm determined.Search the target information to match with user's phrase.Target information is pushed to user.Namely this specification mentions
In the scheme of confession, after capturing the behavior of user, the user tag of the user can be first identified, be based on marking with user later
Corresponding user's phrase is signed, to search the possible interested information of user and push, pushes letter thus, it is possible to greatly promote
The flexibility of breath.
Detailed description of the invention
In order to illustrate more clearly of the technical solution of this specification embodiment, will make below to required in embodiment description
Attached drawing is briefly described, it should be apparent that, the accompanying drawings in the following description is only some embodiments of this specification, right
For those of ordinary skill in the art, without creative efforts, it can also be obtained according to these attached drawings
Its attached drawing.
Fig. 1 is the information transmission system schematic diagram that this specification provides；
Fig. 2 is the method for building up flow chart for the label phrase mapping table that this specification provides；
Fig. 3 is the correlation word determination process schematic diagram that this specification provides；
Fig. 4 is the establishment process schematic diagram for the label phrase mapping table that this specification provides；
Fig. 5 is the information-pushing method flow chart that this specification one embodiment provides；
Fig. 6 is the information push-delivery apparatus schematic diagram that this specification one embodiment provides；
Fig. 7 is the information pushing equipment schematic diagram that this specification one embodiment provides.
Specific embodiment
With reference to the accompanying drawing, the scheme provided this specification is described.
Fig. 1 is the information transmission system schematic diagram that this specification provides.As shown in Figure 1, the system may include data work
Journey layer 102, model training layer 104, algorithm layer 106 and gas producing formation 108.Data engineering layer 102: majority is according to depth mode
(schema) change support, the filtering of feature extraction OP, repeated exposure configurationization.Model training layer 104: feature selecting reconstruct, just amount
Variation, optimizer tune ginseng and Topological expansion, hour grade model-aided tuning.Algorithm layer 106:DMP is carried out based on Ray frame
Deep learning (Deep learning) in intention assessment and enhances natural language sequence to sequence (Sequence to
Sequence, seq2seq) mode carries out multiprecision arithmetic output.Gas producing formation 108: for the mould of the Deep learning of DMP
Type is got through with gas producing formation, is formed the official documents and correspondence based on intention assessment in conjunction with special scenes (such as Spring Festival) and is intelligently split combined scene
Change the intelligent meaning of one's words and combine operation tool, and got through with the pending class product of project in intelligent reminding, forms entirety and user is closed
Note degree clicks wish and clicks the operation tool of the frequency (viscosity).
In addition, above system can also include: infrastructure layer 110, model validation layer 112 and model service layer 114.Its
In, infrastructure layer 110: system bubble reminder application programming interface (Application Programming
Interface, API) call association and the primary interaction page of H5 to conduct in real time.Model validation layer 112: prolong after load off-line model
Slow export, horse racing mechanism, real time monitoring.Model service layer 114: sufficiently pressure, which is surveyed, guarantees service stability, balance effect and pressure
Power monitoring.
In Fig. 1, deep learning algorithm of the gas producing formation 110 based on Ray can preferably utilize higher order relationship, especially with
Under family behavioral data missing and the insufficient situation of corpus, still operation scene can be combined to carry out by high-precision intention assessment
High compactness and accuracy user push.
It should be noted that the information-pushing method that this specification provides is realized based on label phrase mapping table,
Therefore, before the scheme that description this specification provides, first the establishment process of label phrase mapping table is illustrated.
Fig. 2 is the method for building up flow chart for the label phrase mapping table that this specification provides.As shown in Fig. 2, the foundation
Journey may include steps of:
Step 202, a plurality of business corpus relevant to multiple users is collected.
Such as, it can be and collect a plurality of business corpus relevant to multiple users from business corpus.Wherein, every business
Corpus can have corresponding initial labels.
Specifically, it can be based on specific behavior scene (e.g., the new year) and want keyword (e.g., five happinesses red packet outstanding
Deng) carry out business corpus collection and arrangement.In one example, collected business corpus for example can be with are as follows: " exquisite shirt
Dedicate exquisite you to ", " in short give you to strive: never stopping believing oneself good morning " etc..Wherein, with it is " exquisite
Shirt dedicates exquisite you to " corresponding initial labels can be with are as follows: and " the beautiful younger sister of fashion " and " in short gives you to strive: never
Stopping believing oneself good morning " corresponding initial labels can be with are as follows: " workplace office ".
It should be noted that after being collected into business corpus dependency rule can be based on, there are End-user relevances for filtering
The business corpus of difference, to realize recalling for related corpus.
Step 204, for business corpus relevant to each user, corresponding keyword is extracted from the business corpus.
The semanteme of above-mentioned keyword reflects behavioural habits of the user under specifies behavior scene.
In one implementation, corresponding pass can be extracted from business corpus based on context_service algorithm
Keyword.By business corpus: for for " exquisite shirt dedicate to exquisite you ", the keyword extracted for example can be with are as follows: " exquisite
You ".Again by business corpus: for for " in short give you to strive: never stopping believing oneself good morning ", being taken out
The keyword taken for example can be with are as follows: " strive you ".Here, the semanteme of " exquisite you " and " strive you " reflects user
It is the behavioural habits under scene of doing shopping.
It should be noted that can be carried out based on the rule such as part of speech word frequency, correlation for the keyword that said extracted arrives
Screening.
Step 206, correlation word is determined using deep learning algorithm based on keyword.
Here, used deep learning algorithm can be the DMP algorithm based on Ray frame.
Determine that the process of correlation word is referred to Fig. 3 using the DMP algorithm based on Ray frame.Specifically, based on figure
When DMP algorithm shown in 3 determines correlation word, keyword can be inputted in the position that top side is finally distributed.It later, can be right
Keyword is split, to obtain the first sub- word and the second sub- word.Wherein, the semanteme of the first sub- word reflects user's
Lifestyle.The personage that the semanteme of second sub- word reflects user claims in generation.By keyword: for for " exquisite you ", institute
The sub- word of first obtained can be with are as follows: " exquisiteness ", the second sub- word can be with are as follows: " you ".Again by keyword: for " strive you "
For, obtained first sub- word can be with are as follows: " striving ", the second sub- word can be with are as follows: " you ".For the son after above-mentioned fractionation
Word can screen it using convolutional neural networks algorithm (e.g., Text_CNN), to remove unreasonable fractionation word
Language.
In Fig. 3, determining correlation word includes at least the first correlation word.The output position of first correlation word can be with
For the position of bottom right side section abstract.The semanteme of first correlation word reflects other rows of the user under specifies behavior scene
For habit and/or user's portrait.For by taking user draws a portrait as an example, such as can be with are as follows: " gender ", " occupation ", " region ", " work
Make field " and " hobby " etc..In addition, identified correlation word can also include the second correlation word.Second related term
The output position of language is the position that lower left side carrys out source text.The semanteme of second correlation word reflects user in other behavior fields
Behavioural habits and user's portrait under scape.For by taking aforementioned " exquisite you " and " strive you " as an example, second here is related
Word can be user's portrait under professional scene: " white collar " and " student " etc..
From figure 3, it can be seen that determining that the process of correlation word experienced following process: decoding hidden state -> vocabulary point
Cloth -> final distribution, and the relationship between the three as shown in figure 3, do not repeat again herein.
In another example, pointer occupier model can also be used, determines correlation word.The model is in base
The mechanism and coverage mechanism that k-means has been incorporated on plinth DMP model, protect user in the precision of semantics recognition
Card.
This specification is above-mentioned when determining correlation word, be adjusted on BERT distributed strategy multiple branches jointly into
Row search is finally carried out preferably and is exported to the result of generation, so as to greatly promote the flexibility of determined word.This
Outside, above-mentioned determining method is to derive to obtain correlation word based on the intention of user, it is possible thereby to realize the intelligence of keyword
It splits.
It returns in Fig. 2, Fig. 2 can also include the following steps:
Step 208, it is based on keyword and correlation word, generates the corresponding user's phrase of user under specifies behavior scene
And user tag.
Herein, before executing above-mentioned generation step, first above-mentioned keyword and correlation word can be normalized
Mapping relations matching.Specifically, the identical keyword of meaning and/or correlation word can be normalized to identical word, and,
Keyword and/or correlation word are mapped as to embody the word (being referred to as intended to word) of user's intention.
It is understood that after executing above-mentioned normalized mapping relationship match, so that it may generate user's phrase and
User tag.Specifically, initial labels corresponding with business corpus, Lai Shengcheng user tag can be based on.Can also based on from
Correlation word determined by the keyword extracted in the corresponding business corpus of the initial labels, Lai Shengcheng user tag.Citing comes
It says, can be by initial labels above-mentioned: " the beautiful younger sister of fashion " be used as user tag." white collar " can also be used as user tag.
It, can be based on the keyword and the first correlation under behavior scene corresponding to relative users label for user's phrase
Word generates.For by taking " the beautiful younger sister of fashion " as an example, corresponding user's phrase can be with are as follows: " exquisite you ".I.e. to fractionation after
Keyword is reconfigured.
When above-mentioned correlation word further includes the second correlation word, user tag generated can be with are as follows: and " white collar ", and it is right
That answers can be with for phrase are as follows: " cause is more and more prosperous ".In this example embodiment, user's phrase can be based on dependency rule to the second phase
It closes word and carries out automatic combination producing.That is, can also be generated other when correlation word further includes the second correlation word
The corresponding user tag of user and user's phrase under behavior scene.
In the present specification, for the user's phrase and user tag of generation, convolutional neural networks algorithm can be used
(e.g., Text_CNN) screens it, to screen the higher user's phrase of mass.
Step 210, mark is at least established based on the corresponding user's phrase of user and user tag under specifies behavior scene
Sign phrase mapping table.
It is understood that in practical applications, above-mentioned steps 204 and step 208, which can be, to be repeated, until raw
At the corresponding user tag of each user and user's phrase.
When generating user tag corresponding with multiple users and user's phrase, step 210 be can be based on multiple use
The user tag and user's phrase at family, Lai Jianli label phrase mapping table.
It should be noted that the user tag and user's phrase in the label phrase mapping table finally established can be used as newly
The completion of business corpus into business corpus, it is possible thereby to achieve the purpose that information flow back.By the reflux of information, can make
Business corpus in business corpus is more and more abundant.
To sum up, this specification embodiment can be come based on behavioural habits data, intention word and the completion information with user
Establish above-mentioned label phrase mapping table.
Fig. 4 is the establishment process schematic diagram for the label phrase mapping table that this specification provides.In Fig. 4, firstly, from corpus
It is middle to collect a plurality of business corpus relevant to multiple users, and corresponding keyword is extracted from each corpus.Here
A plurality of business corpus is corresponding with specific behavior scene.Later, it is learned based on the keyword extracted using deep learning algorithm
Practise other words relevant to above-mentioned specific behavior scene；And study other words relevant to other behavior scenes.For
The other words learnt can be filtered it using convolutional neural networks algorithm.Finally, being based on keyword and filtering
Other words afterwards generate the corresponding relationship under different behavior scenes between user's phrase and user tag, and right based on this
It should be related to, establish label phrase mapping table.
It is understood that being the explanation to the establishment process of label phrase mapping table above, below to based on the mapping
The information-pushing method of table is illustrated.
Fig. 5 is the information-pushing method flow chart that this specification one embodiment provides.As shown in figure 5, the method can
To include the following steps:
Step 502, the user behavior data of user is obtained.
Here user behavior data can refer to user in the upper execution browsing behavior of application, click behavior or consumption row
Generated data when to wait business conducts.
Step 504, it is based on user behavior data, determines the behavior scene of user.
What can be here pre-defined, it can include but is not limited to the new year scene, shopping scene and occupation
Scene etc..
Step 506, the user tag of user is identified.
Here it is possible to be to identify the user tag of user based on user behavior data.The user tag is for describing user
User portrait.
Under scene of doing shopping, the user tag of user for example can be with are as follows: " the beautiful younger sister of fashion ", " housewife " and " workplace
Office " etc..Under professional scene, the user tag of user for example can be with are as follows: " white collar ", " university student " etc..
Step 508, the label phrase firing table pre-established is inquired, is obtained under behavior scene, user tag is corresponding
User's phrase.
User's phrase is used to describe the behavioural habits of user.
For by taking label phrase mapping table shown in Fig. 4 as an example, when the behavior scene for determining user are as follows: shopping scene, and
The user tag of user are as follows: when " the beautiful younger sister of fashion ", user's phrase of acquisition are as follows: " exquisite you ".
Step 510, the target information to match with user's phrase is searched.
Step 512, target information is pushed to user.
Here it is possible to be based on preset matching relationship, to search the target information to match.As an example it is assumed that having
Following matching relationship:So can to active user push digital product or with cook
It prepares food relevant article.
To sum up, the information-pushing method that this specification embodiment provides can first be known after capturing the behavior of user
The not user tag of the user is based on user's phrase corresponding with user tag later, may be interested to search user
Information simultaneously pushes, thus, it is possible to realize the accurate push of information.Further, it is also possible to greatly promote the flexibility of institute's pushed information.
Finally, relying on Ray as Distributed Computing Platform in this specification scheme, being accurate, the efficient hair of deep learning
It waves and plays important function.Realize that deep learning operation is efficient to final transformation model from data source header acquisition user behavior data
Energy intention assessment, End2End is within a frame always.Realize data engineering configurationization, model training real time implementation, model
Verify onlineization, the key properties such as minute grade model modification.
Accordingly with above- mentioned information method for pushing, a kind of information push-delivery apparatus that this specification one embodiment also provides,
As shown in fig. 6, the apparatus may include:
Acquiring unit 602, for obtaining the user behavior data of user.
Determination unit 604, the user behavior data for being obtained based on acquiring unit 602, determines the behavior scene of user.
Recognition unit 606, the user tag of user, the user tag are used to describe user's portrait of user for identification.
Acquiring unit 602 is also used to inquire the label phrase firing table pre-established, obtain under current behavior scene, uses
The corresponding user's phrase of family label.User's phrase is used to describe the behavioural habits of user.Here label phrase mapping table is used for
It is recorded in the corresponding relationship under different behavior scenes between user tag and user's phrase.Wherein, the use under a kind of behavior scene
Family label and user's phrase are behavioural habits and depth based on the user under behavior scene or other corelation behaviour scenes
Spend what learning algorithm determined.
Here deep learning algorithm for example can be the DMP algorithm based on Ray frame.
Searching unit 608, the target information to match for searching the user's phrase obtained with acquiring unit 602.
Push unit 610, for pushing the target information to user.
Optionally, which can also include:
Unit (not shown) is established, for collecting a plurality of business corpus relevant to multiple users.
For business corpus relevant to arbitrary first user, corresponding keyword is extracted from business corpus, the pass
The semanteme of keyword reflects behavioural habits of first user under specifies behavior scene.
Correlation word is determined using deep learning algorithm based on keyword.The correlation word includes at least the first related term
Language, the semanteme of the first correlation word reflect other behavioural habits and/or user of first user under specifies behavior scene and draw
Picture.
Based on keyword and correlation word, generate under specifies behavior scene the corresponding user's phrase of the first user and
User tag.
At least based on the corresponding user's phrase of the first user and user tag under specifies behavior scene, it is short to establish label
Language mapping table.
Optionally, above-mentioned correlation word can also include the second correlation word, and the semanteme of the second correlation word reflect the
Behavioural habits and user portrait of one user under other behavior scenes.
Determination unit 604 is also used to determine that the first user is corresponding under other behavior scenes based on the second correlation word
User's phrase and user tag.
Establishing unit specifically can be used for:
At least based on the corresponding user's phrase of the first user and user tag under specifies behavior scene, and other
The corresponding user's phrase of first user and user tag, establish label phrase mapping table under behavior scene.
Establishing unit also specifically can be used for:
Keyword is split, to obtain the first sub- word and the second sub- word.The semanteme of first sub- word reflects
The lifestyle of first user, the personage that the semanteme of the second sub- word reflects the first user claim in generation.
Correlation word is determined using deep learning algorithm based on the first sub- word and the second sub- word.
Optionally, which can also include:
Screening unit (not shown), for being based on convolutional neural networks algorithm, to the first sub- word and the second sub- word
Language is screened.
Establishing unit also specifically can be used for:
Based on after screening the first sub- word and the second sub- word correlation word determined using deep learning algorithm.
The function of each functional module of this specification above-described embodiment device can pass through each step of above method embodiment
Rapid to realize, therefore, the specific work process for the device that this specification one embodiment provides does not repeat again herein.
The information push-delivery apparatus that this specification one embodiment provides, acquiring unit 602 obtain the user behavior number of user
According to.User behavior data of the determination unit 604 based on acquisition determines the behavior scene of user.Recognition unit 606 identifies user's
User tag, the user tag are used to describe user's portrait of user.Acquiring unit 602 is inquired the label phrase pre-established and is penetrated
Table obtains the corresponding user's phrase of user tag under current behavior scene.The behavior that user's phrase is used to describe user is practised
It is used.Here label phrase mapping table is used to be recorded under different behavior scenes the corresponding pass between user tag and user's phrase
System.Wherein, the user tag under a kind of behavior scene and user's phrase are based in behavior scene or other corelation behaviours
The behavioural habits of user and deep learning algorithm determine under scene.Searching unit 608 searches user's phrase phase with acquisition
Matched target information.Push unit 610 pushes target information to user.Thus, it is possible to realize the accurate push of information.
Accordingly with above- mentioned information method for pushing, this specification embodiment additionally provides a kind of information pushing equipment, such as Fig. 7
Shown, which may include: memory 702, one or more processors 704 and one or more programs.Wherein, this one
A or multiple programs are stored in memory 702, and are configured to be executed by one or more processors 704, the program quilt
Processor 704 performs the steps of when executing
Obtain the user behavior data of user.
Based on user behavior data, the behavior scene of user is determined.
Identify that the user tag of user, the user tag are used to describe user's portrait of user.
The label phrase firing table pre-established is inquired, the corresponding user of user tag under determining behavior scene is obtained
Phrase.User's phrase is used to describe the behavioural habits of user.Label phrase mapping table is for being recorded under different behavior scenes
Corresponding relationship between user tag and user's phrase.Wherein, the user tag under a kind of behavior scene and user's phrase are bases
It is determined in the behavioural habits and deep learning algorithm of the user under behavior scene or other corelation behaviour scenes.
Search the target information to match with user's phrase.
Target information is pushed to user.
The information pushing equipment that this specification one embodiment provides, may be implemented the accurate push of information.
All the embodiments in this specification are described in a progressive manner, same and similar portion between each embodiment
Dividing may refer to each other, and each embodiment focuses on the differences from other embodiments.Especially for equipment reality
For applying example, since it is substantially similar to the method embodiment, so being described relatively simple, related place is referring to embodiment of the method
Part explanation.
The step of method in conjunction with described in this disclosure content or algorithm can realize in a manner of hardware,
It can be and the mode of software instruction is executed by processor to realize.Software instruction can be made of corresponding software module, software
Module can be stored on RAM memory, flash memory, ROM memory, eprom memory, eeprom memory, register, hard
Disk, mobile hard disk, CD-ROM or any other form well known in the art storage medium in.A kind of illustrative storage Jie
Matter is coupled to processor, to enable a processor to from the read information, and information can be written to the storage medium.
Certainly, storage medium is also possible to the component part of processor.Pocessor and storage media can be located in ASIC.In addition, should
ASIC can be located in server.Certainly, pocessor and storage media can also be used as discrete assembly and be present in server.
Those skilled in the art are it will be appreciated that in said one or multiple examples, function described in the invention
It can be realized with hardware, software, firmware or their any combination.It when implemented in software, can be by these functions
Storage in computer-readable medium or as on computer-readable medium one or more instructions or code transmitted.
Computer-readable medium includes computer storage media and communication media, and wherein communication media includes convenient for from a place to another
Any medium of one place transmission computer program.Storage medium can be general or specialized computer can access it is any
Usable medium.
It is above-mentioned that this specification specific embodiment is described.Other embodiments are in the scope of the appended claims
It is interior.In some cases, the movement recorded in detail in the claims or step can be come according to the sequence being different from embodiment
It executes and desired result still may be implemented.In addition, process depicted in the drawing not necessarily require show it is specific suitable
Sequence or consecutive order are just able to achieve desired result.In some embodiments, multitasking and parallel processing be also can
With or may be advantageous.
Above-described specific embodiment has carried out into one the purpose of this specification, technical scheme and beneficial effects
Step is described in detail, it should be understood that being not used to limit this foregoing is merely the specific embodiment of this specification
The protection scope of specification, all any modifications on the basis of the technical solution of this specification, made, change equivalent replacement
Into etc., it should all include within the protection scope of this specification.
Claims (13)
1. a kind of information-pushing method, comprising:
Obtain the user behavior data of user；
Based on the user behavior data, the behavior scene of the user is determined；
Identify the user tag of the user；The user tag is used to describe user's portrait of the user；
The label phrase firing table that pre-establishes is inquired, is obtained under the behavior scene, the corresponding user of the user tag is short
Language；User's phrase is used to describe the behavioural habits of user；The label phrase mapping table is for being recorded in different behavior fields
Corresponding relationship under scape between user tag and user's phrase；Wherein, the user tag under a kind of behavior scene and user's phrase
It is behavioural habits and deep learning the algorithm determination based on the user under behavior scene or other corelation behaviour scenes；
Search the target information to match with user's phrase；
The target information is pushed to the user.
2. according to the method described in claim 1, the label phrase mapping table is established by following steps:
Collect a plurality of business corpus relevant to multiple users；
For business corpus relevant to arbitrary first user, corresponding keyword is extracted from the business corpus；It is described
The semanteme of keyword reflects behavioural habits of first user under specifies behavior scene；
Correlation word is determined using deep learning algorithm based on the keyword；It is related that the correlation word includes at least first
Word；The semanteme of first correlation word reflects other behaviors of first user under the specifies behavior scene and practises
Used and/or user portrait；
Based on the keyword and the correlation word, it is corresponding to generate first user under the specifies behavior scene
User's phrase and user tag；
At least based on the corresponding user's phrase of the first user described under the specifies behavior scene and user tag, institute is established
State label phrase mapping table.
3. according to the method described in claim 2, the correlation word further includes the second correlation word；Second correlation word
Semanteme reflect behavioural habits of first user under other behavior scenes and user's portrait；The method also includes:
Based on second correlation word, determine the corresponding user's phrase of the first user described under other behavior scenes with
And user tag；
It is described to establish the label phrase mapping table and include:
At least based on the corresponding user's phrase of the first user described under the specifies behavior scene and user tag, Yi Ji
The corresponding user's phrase of first user and user tag under other behavior scenes establish the label phrase mapping
Table.
4. according to the method described in claim 2, described determine related term using deep learning algorithm based on the keyword
Language, comprising:
The keyword is split, to obtain the first sub- word and the second sub- word；The semanteme of the first sub- word is anti-
The lifestyle of first user is reflected；The personage that the semanteme of the second sub- word reflects first user claims in generation；
Correlation word is determined using deep learning algorithm based on the described first sub- word and the second sub- word.
5. according to the method described in claim 4, the described first sub- word and the second sub- word are based on described, using depth
Learning algorithm is spent, before determining correlation word, further includes:
Based on convolutional neural networks algorithm, the described first sub- word and the second sub- word are screened；
It is described that correlation word is determined using deep learning algorithm based on the described first sub- word and the second sub- word, it wraps
It includes:
Based on after screening the described first sub- word and the second sub- word correlation word determined using deep learning algorithm.
6. according to the method described in claim 1, the deep learning algorithm is the DMP algorithm based on Ray frame.
7. a kind of information push-delivery apparatus, comprising:
Acquiring unit, for obtaining the user behavior data of user；
Determination unit, the user behavior data for being obtained based on the acquiring unit, determines the behavior field of the user
Scape；
Recognition unit, for identification user tag of the user；The user that the user tag is used to describe the user draws
Picture；
The acquiring unit is also used to inquire the label phrase firing table pre-established, obtains the use under the behavior scene
The corresponding user's phrase of family label；User's phrase is used to describe the behavioural habits of user；The label phrase mapping table is used
In the corresponding relationship being recorded under different behavior scenes between user tag and user's phrase；Wherein, under a kind of behavior scene
User tag and user's phrase be based on the behavioural habits of user under behavior scene or other corelation behaviour scenes and
What deep learning algorithm determined；
Searching unit, the target information to match for searching the user's phrase obtained with the acquiring unit；
Push unit, for pushing the target information to the user.
8. device according to claim 7, further includes:
Unit is established, for collecting a plurality of business corpus relevant to multiple users；
For business corpus relevant to arbitrary first user, corresponding keyword is extracted from the business corpus；It is described
The semanteme of keyword reflects behavioural habits of first user under specifies behavior scene；
Correlation word is determined using deep learning algorithm based on the keyword；It is related that the correlation word includes at least first
Word；The semanteme of first correlation word reflects other behaviors of first user under the specifies behavior scene and practises
Used and/or user portrait；
Based on the keyword and the correlation word, it is corresponding to generate first user under the specifies behavior scene
User's phrase and user tag；
At least based on the corresponding user's phrase of the first user described under the specifies behavior scene and user tag, institute is established
State label phrase mapping table.
9. device according to claim 8, the correlation word further includes the second correlation word；Second correlation word
Semanteme reflect behavioural habits of first user under other behavior scenes and user's portrait；
The determination unit is also used to determine described first under other behavior scenes based on second correlation word
The corresponding user's phrase of user and user tag；
The unit of establishing is specifically used for:
At least based on the corresponding user's phrase of the first user described under the specifies behavior scene and user tag, Yi Ji
The corresponding user's phrase of first user and user tag under other behavior scenes establish the label phrase mapping
Table.
10. device according to claim 8, it is described establish unit also particularly useful for:
The keyword is split, to obtain the first sub- word and the second sub- word；The semanteme of the first sub- word is anti-
The lifestyle of first user is reflected；The personage that the semanteme of the second sub- word reflects first user claims in generation；
Correlation word is determined using deep learning algorithm based on the described first sub- word and the second sub- word.
11. device according to claim 10, further includes:
Screening unit sieves the described first sub- word and the second sub- word for being based on convolutional neural networks algorithm
Choosing；
It is described establish unit also particularly useful for:
Based on after screening the described first sub- word and the second sub- word correlation word determined using deep learning algorithm.
12. device according to claim 7, the deep learning algorithm is the DMP algorithm based on Ray frame.
13. a kind of information pushing equipment, comprising:
Memory；
One or more processors；And
One or more programs wherein the storage of one or more of programs is in the memory, and are configured to by institute
One or more processors execution is stated, described program performs the steps of when being executed by the processor
Obtain the user behavior data of user；
Based on the user behavior data, the behavior scene of the user is determined；
Identify the user tag of the user；The user tag is used to describe user's portrait of the user；
The label phrase firing table that pre-establishes is inquired, is obtained under the behavior scene, the corresponding user of the user tag is short
Language；User's phrase is used to describe the behavioural habits of user；The label phrase mapping table is for being recorded in different behavior fields
Corresponding relationship under scape between user tag and user's phrase；Wherein, the user tag under a kind of behavior scene and user's phrase
It is behavioural habits and deep learning the algorithm determination based on the user under behavior scene or other corelation behaviour scenes；
Search the target information to match with user's phrase；
The target information is pushed to the user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910636124.1A CN110413882B (en) | 2019-07-15 | 2019-07-15 | Information pushing method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910636124.1A CN110413882B (en) | 2019-07-15 | 2019-07-15 | Information pushing method, device and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110413882A true CN110413882A (en) | 2019-11-05 |
| CN110413882B CN110413882B (en) | 2023-10-31 |
Family
ID=68361509
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910636124.1A Active CN110413882B (en) | 2019-07-15 | 2019-07-15 | Information pushing method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110413882B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111684441A (en) * | 2019-11-15 | 2020-09-18 | 深圳海付移通科技有限公司 | Message pushing method and device based on image data and computer storage medium |
| CN112818023A (en) * | 2021-01-26 | 2021-05-18 | 龚世燕 | Big data analysis method and cloud computing server in associated cloud service scene |
| CN114706945A (en) * | 2022-03-24 | 2022-07-05 | 马上消费金融股份有限公司 | Intention recognition method and device, electronic equipment and storage medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010288024A (en) * | 2009-06-10 | 2010-12-24 | Univ Of Electro-Communications | Moving picture recommendation apparatus |
| US20110246496A1 (en) * | 2008-12-11 | 2011-10-06 | Chung Hee Sung | Information search method and information provision method based on user's intention |
| CN104199872A (en) * | 2014-08-19 | 2014-12-10 | 北京搜狗科技发展有限公司 | Information recommendation method and device |
| CN105824912A (en) * | 2016-03-15 | 2016-08-03 | 平安科技（深圳）有限公司 | Personalized recommending method and device based on user portrait |
| CN106844732A (en) * | 2017-02-13 | 2017-06-13 | 长沙军鸽软件有限公司 | The method that automatic acquisition is carried out for the session context label that cannot directly gather |
| CN107424043A (en) * | 2017-06-15 | 2017-12-01 | 北京三快在线科技有限公司 | A kind of Products Show method and device, electronic equipment |
| CN107911448A (en) * | 2017-11-14 | 2018-04-13 | 北京金山安全软件有限公司 | Content pushing method and device |
| CN108268617A (en) * | 2018-01-05 | 2018-07-10 | 阿里巴巴集团控股有限公司 | User view determines method and device |
| CN109829107A (en) * | 2019-01-23 | 2019-05-31 | 华为技术有限公司 | A kind of recommended method and electronic equipment based on user movement state |
-
2019
- 2019-07-15 CN CN201910636124.1A patent/CN110413882B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110246496A1 (en) * | 2008-12-11 | 2011-10-06 | Chung Hee Sung | Information search method and information provision method based on user's intention |
| JP2010288024A (en) * | 2009-06-10 | 2010-12-24 | Univ Of Electro-Communications | Moving picture recommendation apparatus |
| CN104199872A (en) * | 2014-08-19 | 2014-12-10 | 北京搜狗科技发展有限公司 | Information recommendation method and device |
| CN105824912A (en) * | 2016-03-15 | 2016-08-03 | 平安科技（深圳）有限公司 | Personalized recommending method and device based on user portrait |
| CN106844732A (en) * | 2017-02-13 | 2017-06-13 | 长沙军鸽软件有限公司 | The method that automatic acquisition is carried out for the session context label that cannot directly gather |
| CN107424043A (en) * | 2017-06-15 | 2017-12-01 | 北京三快在线科技有限公司 | A kind of Products Show method and device, electronic equipment |
| CN107911448A (en) * | 2017-11-14 | 2018-04-13 | 北京金山安全软件有限公司 | Content pushing method and device |
| CN108268617A (en) * | 2018-01-05 | 2018-07-10 | 阿里巴巴集团控股有限公司 | User view determines method and device |
| CN109829107A (en) * | 2019-01-23 | 2019-05-31 | 华为技术有限公司 | A kind of recommended method and electronic equipment based on user movement state |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111684441A (en) * | 2019-11-15 | 2020-09-18 | 深圳海付移通科技有限公司 | Message pushing method and device based on image data and computer storage medium |
| CN112818023A (en) * | 2021-01-26 | 2021-05-18 | 龚世燕 | Big data analysis method and cloud computing server in associated cloud service scene |
| CN112818023B (en) * | 2021-01-26 | 2022-03-18 | 广州市番禺融合小额贷款股份有限公司 | Big data analysis method and cloud computing server in associated cloud service scene |
| CN114706945A (en) * | 2022-03-24 | 2022-07-05 | 马上消费金融股份有限公司 | Intention recognition method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110413882B (en) | 2023-10-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102693309B (en) | Candidate phrase querying method and aided translation system for computer aided translation | |
| CN109997124A (en) | System and method for measuring the semantic dependency of keyword | |
| CN110413882A (en) | Information-pushing method, device and equipment | |
| CN110941698B (en) | Service discovery method based on convolutional neural network under BERT | |
| CN109446328A (en) | A kind of text recognition method, device and its storage medium | |
| CN106227714A (en) | A kind of method and apparatus obtaining the key word generating poem based on artificial intelligence | |
| Nguyen et al. | LifeSeeker 3.0: An Interactive Lifelog Search Engine for LSC'21 | |
| CN107193882A (en) | Why not query answer methods based on figure matching on RDF data | |
| CN102855317A (en) | Multimode indexing method and system based on demonstration video | |
| CN113723853A (en) | Method and device for processing post competence demand data | |
| CN112860850A (en) | Man-machine interaction method, device, equipment and storage medium | |
| CN114647719A (en) | Question-answering method and device based on knowledge graph | |
| Golshani et al. | A multimedia information repository for cross cultural dance studies | |
| CN112199958A (en) | Concept word sequence generation method and device, computer equipment and storage medium | |
| CN116127066A (en) | Text clustering method, text clustering device, electronic equipment and storage medium | |
| CN109785818A (en) | A kind of music music method and system based on deep learning | |
| CN113032567B (en) | Position embedding interpretation method and device, computer equipment and storage medium | |
| Alaofi et al. | Personalisation of Generic Library Search Results Using Student Enrolment Information. | |
| KR20190082453A (en) | Method, apparatus and computer program for analyzing new learning contents for machine learning modeling | |
| CN113468306A (en) | Voice conversation method, device, electronic equipment and storage medium | |
| Sarkar et al. | Representing Tasks with a Graph-Based Method for Supporting Users in Complex Search Tasks | |
| Kyriakakis et al. | Enabling ontology-based search: a case study in the bioinformatics domain | |
| Müller et al. | Multimodal music processing (dagstuhl seminar 11041) | |
| Matsuda et al. | Benchmark for Personalized Federated Learning | |
| Carta et al. | CulturAI: Semantic Enrichment of Cultural Data Leveraging Artificial Intelligence |
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 | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40016772 Country of ref document: HK |
|
| TA01 | Transfer of patent application right |
Effective date of registration: 20201010 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Innovative advanced technology Co.,Ltd. Address before: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant before: Advanced innovation technology Co.,Ltd. Effective date of registration: 20201010 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Advanced innovation technology Co.,Ltd. Address before: A four-storey 847 mailbox in Grand Cayman Capital Building, British Cayman Islands Applicant before: Alibaba Group Holding Ltd. |
|
| TA01 | Transfer of patent application right | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |