CN105512298A - Interested content prediction method based on machine learning - Google Patents

Interested content prediction method based on machine learning Download PDF

Info

Publication number
CN105512298A
CN105512298A CN201510918123.8A CN201510918123A CN105512298A CN 105512298 A CN105512298 A CN 105512298A CN 201510918123 A CN201510918123 A CN 201510918123A CN 105512298 A CN105512298 A CN 105512298A
Authority
CN
China
Prior art keywords
retrieval
interest
result
word
user
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201510918123.8A
Other languages
Chinese (zh)
Inventor
董政
吴文杰
陈露
李学生
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chengdu Mo Yun Science And Technology Ltd
Original Assignee
Chengdu Mo Yun Science And Technology Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Chengdu Mo Yun Science And Technology Ltd filed Critical Chengdu Mo Yun Science And Technology Ltd
Priority to CN201510918123.8A priority Critical patent/CN105512298A/en
Publication of CN105512298A publication Critical patent/CN105512298A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/279—Recognition of textual entities
    • G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention provides an interested content prediction method based on machine learning. The method comprises steps as follows: search words input by a user are optimized and adjusted according to feature information of the user, and interested content search is performed based on the search words. According to the interested content prediction method based on the machine learning, the recognition accuracy and timeliness of features of internet users are effectively improved.

Description

Based on the content of interest Forecasting Methodology of machine learning
Technical field
The present invention relates to large data, particularly a kind of content of interest Forecasting Methodology based on machine learning.
Background technology
Along with the development of mobile Internet, Web content provide the user abundant information resources and services but on network, information quality is but uneven, a large amount of information is replicated, reprints, and various promotion message advertisement retrieval result, have impact on Consumer's Experience; If returning the same result for retrieval for the term input that all users are identical has not been probably that user wishes.Only adopt the mode of term coupling, and ignore the real demand of isolated user, namely in conjunction with user behavior (comprising user interest, user preference, user's query note) and term, this query intention of user is not made and judging accurately, the result meeting user's request cannot be provided.Existing technical scheme sets up interest characteristics vector by watch attentively history or the individual descriptor of user for user, recycle this vector carries out similarity calculating to the result for retrieval that retrieval returns, this information but not in vector often some users really needed forecloses.
Summary of the invention
For solving the problem existing for above-mentioned prior art, the present invention proposes a kind of content of interest Forecasting Methodology based on machine learning, comprising:
According to described user's characteristic information, adjustment is optimized to the term that user inputs, is carried out the retrieval of content of interest by term.
Preferably, describedly according to described user's characteristic information, adjustment is optimized to the term that user inputs, comprises further:
Term analysis extraction is carried out to the content that user inputs at interface, by segmenter, word segmentation processing is carried out to user search content; Obtain retrieval vector, each dimension of this retrieval vector, with a numerical value, represents the weights that term is corresponding; Watch behavior attentively according to user and differentiate object of interest, then carry out analysis structure interest model by object of interest; Concrete estimation formulas is:
Ip＝α×T b+β×U o
Wherein:
T b = 1 2 π δ exp ( - ( Δ t - t ) 2 2 δ 2 )
U o＝a×C copy+b×S save+G×R reply
Wherein Ip represents result for retrieval interest-degree score value, α and β is regulation coefficient, i.e. the proportion that accounts in formula of result for retrieval fixation time and result for retrieval interactive operation, wherein alpha+beta=1; T bbe the time dimension that user stops at result for retrieval, calculated by normal distribution, reflection user watches the result for retrieval time attentively; The degree of closeness of fixation time △ t and reference time t reflects interest-degree, t be according to Document Length determine reference time, t and result for retrieval length proportional; U 0the interactive operation of user on result for retrieval, C copyrepresent whether user carries out replicate run at result for retrieval, is that being worth is 1, and no value is 0; S saverepresent whether user carries out result for retrieval and preserve operation, is that being worth is 1, and no value is 0; R replyrepresenting and whether carry out feedback associative operation for result for retrieval, is that being worth is 1, and no value is 0; A, b and c are U 0regulation coefficient, according to different operations to being whether the significance level that object of interest is passed judgment on, respectively different values is arranged to coefficient
Preferably, comprise further:
Calculate the change of the weights of interest word according to Intersted word frequency of utilization, after Feature Words is determined as Intersted word, arranging its initialization weights is 1; Weights Distribution Calculation is carried out, the position that this weight computing occurs in the page according to word frequency and entry after being defined as Intersted word; Weight computing formula is expressed as:
w i = ( w 0 + ( 1 / n ) Σ i = 1 k w p i ) × F ( t ) + y N
Wherein, N is the number of times that Intersted word is updated, and y is that each interest word is used rear weights to increase coefficient; w i, represent the weights of Intersted word, w 0be weights initial values, the initial value arranged after being namely determined as Intersted word is 1, w pibe the average weight comprising Intersted word result for retrieval, n is the result for retrieval number comprising this Intersted word, and k is the sum that Intersted word occurs in all result for retrieval; w pithe weights of term in corresponding result for retrieval;
F (t) describes interest and ignores process: F (t)=e -log2/ (hSt)
Wherein, St is the time interval, and represent that term is updated to now for the last time, namely current time deducts number of days when upgrading for the last time; H is predetermined period, makes the value of F (t) after number of days h be initial value half.
The present invention compared to existing technology, has the following advantages:
The present invention proposes a kind of content of interest Forecasting Methodology based on machine learning, effectively improve the recognition accuracy of Internet user's feature, fully take into account the degree of correlation of result for retrieval and user's query contents, be widely used, it is convenient to realize.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of the content of interest Forecasting Methodology based on machine learning according to the embodiment of the present invention.
Embodiment
Detailed description to one or more embodiment of the present invention is hereafter provided together with the accompanying drawing of the diagram principle of the invention.Describe the present invention in conjunction with such embodiment, but the invention is not restricted to any embodiment.Scope of the present invention is only defined by the claims, and the present invention contain many substitute, amendment and equivalent.Set forth many details in the following description to provide thorough understanding of the present invention.These details are provided for exemplary purposes, and also can realize the present invention according to claims without some in these details or all details.
An aspect of of the present present invention provides a kind of content of interest Forecasting Methodology based on machine learning.Fig. 1 is the content of interest Forecasting Methodology process flow diagram based on machine learning according to the embodiment of the present invention.The solution of the present invention is adding users interest module on former general retrieval architecture basis, adopts inquiry to improve and arranges combination with weights.User adjusts Query Information by interest vector after inputting term, and sets different weights initial values according to user interest, puts in order returning results in list adjustment.
In retrieval architecture, information acquisition module is responsible for collecting user data, and comprise user search word, watch behavior attentively, webpage fixation time etc. can reflect the information of user characteristics, prepares for building user interest model.Then interest module according to this query contents in conjunction with user interest, optimize and revise user search term, simultaneously using interest vector and adjustment after query word as Parameter transfer to sequence formula in, finally through optimization filter result for retrieval list return to user side browser.
Information acquisition module comprises term and extracts and user's associative operation two parts: 1) term extracts, and is carry out term analysis extraction to user in the content that search interface inputs, carries out word segmentation processing by segmenter to user's query contents.The corresponding Term of each word, finally obtains query vector V (q)=(term 1, term 2, term 3term n) wherein n>1, wherein each dimension of query vector is with a numerical value, represents the weights that term is corresponding, is used for the significance level of identification retrieval word.2) user's associative operation information is obtained.User, may for no other reason than that title attracts and clicks the forward result for retrieval of rank in the process of watching the results list attentively, but result for retrieval content do not meet user's request.So first behavior will be watched attentively according to user differentiate object of interest, then carry out analysis structure interest model by object of interest.
Fixation time length, whether carry out content replication when watching result for retrieval attentively, the interactive operations such as collection all characterize the relation of object of interest.To sum up consider that watching result for retrieval attentively to user estimates thus show that whether result for retrieval is the object of interest of user, is used as the reference content building interest model.Concrete estimation formulas as:
Ip＝α×T b+β×U o
Wherein:
T b = 1 2 π δ exp ( - ( Δ t - t ) 2 2 δ 2 )
U o＝a×C copy+b×S save+G×R reply
Ip represents result for retrieval interest-degree score value, α and β is regulation coefficient, the proportion accounted in estimation formulas by different values reflection result for retrieval fixation time and result for retrieval interactive operation, wherein alpha+beta=1.T bbe the time dimension that user stops at result for retrieval, calculated by normal distribution, reflect that user's watches the result for retrieval time attentively.The degree of closeness of fixation time △ t and reference time t reflects interest-degree, and fixation time is long or too short all can reduce the score value of interest on result for retrieval fixation time, t according to Document Length determine, t and result for retrieval length proportional.U 0the interactive operation of user on result for retrieval, C copyrepresent whether user carries out replicate run at result for retrieval, is that being worth is 1, and no value is 0; S saverepresent whether user carries out result for retrieval and preserve operation, is that being worth is 1, and no value is 0; R replyrepresenting and whether carry out feedback associative operation for result for retrieval, is that being worth is 1, and no value is 0.A, b and c are U 0regulation coefficient, according to different operations to being whether the significance level that object of interest is passed judgment on, respectively different values is arranged to coefficient.
User interest is divided into common interest and special interests by the present invention, and common interest here does not belong to any one user, and it is that disengaging user is self-existent, can regard the tree construction that Feature Words is formed as.Special interests is then the node set of above-mentioned tree construction, has the interest node-type identifier of common interest, has different weights according to the different levels of interest node in interest tree construction.Interest model is made to be depart from user to rely on, in the use that index stage or off-line phase interest model are not restricted.Interest model of the present invention builds based on ODP classification, the corresponding interest term of each node of tree construction, for expanding coverage rate and the application in practice of interest model, also needs to call tree construction Feature Words and near synonym expand.
Intersted word has been made into the identifier of Feature Words in interest tree construction by special interests, Intersted word is utilized to be extended to the set of interest vocabulary, be embodied in the identifier of user interest model, in reduction user interest and application process, tree construction resolved and expands.Near synonym expansion on the one hand, on the other hand to there being the Feature Words of ambiguity or relation of inclusion to carry out semantic analysis and Intersted word differentiation.User interest is by vector representation, and the element in vector is a key-value pair, is identifier in interest model respectively and has weights.
The structure of general user's interest model needs first to realize through result for retrieval pre-service and result for retrieval classification again.The result for retrieval used first differentiates through object of interest.On interest is determined, set a threshold value, the Feature Words only reaching threshold value just can be identified as interest, and more the new stage carries out weights increase in interest afterwards, or weights reduce to this interest of cancellation.Certain filtration is carried out in the extraction of result for retrieval Feature Words, and result for retrieval carries out participle and after cancelling the respective handling such as stop words, adopts low frequency threshold value to screen result for retrieval Feature Words.
For preventing the impact extraction of Feature Words being caused to misleading, former result for retrieval Feature Words extracting rule being provided with high frequency limit, containing the impact that user interest is differentiated that term is piled up to a certain extent.The entry that item frequency has exceeded high frequency threshold value can not be identified as result for retrieval Feature Words equally, records the word frequency of this word and the positional information of appearance while Feature Words is determined, for the weight computing after being defined as Intersted word.After Intersted word fixes on the differentiation of result for retrieval feature vocabulary really, determine according to the number of times that all pages of result for retrieval Feature Words occur, the discrimination formula of Intersted word is expressed as follows:
I term＝(1/n)(T page+T search-d)+T submit
Wherein, I termbe interest level when differentiating Intersted word, if value is greater than 1, be determined as interest word.T pagerefer to the object of interest quantity comprising Intersted word; T searchit is the number of times of the retrieval term appearance that user manually inputs; T pageand T searchit is accumulation calculating.N is the count threshold meeting Intersted word condition, only has T page+ T searchthe value of-d is more than or equal to n and just can be identified as interest word.T submitbe then the interest word that user submits to, this value can only be 0 or 1.
The update strategy of user interest model is according to the change being Intersted word frequency of utilization, and the concrete weights embodying interest word that calculate change.After Feature Words is determined as Intersted word, needing to arrange its initialization weights is 1, and this value is the minimum weights of interest word, if weights be less than 1 so this word should cancel from interest vector.Also need to carry out a weights Distribution Calculation according to the importance of word after being defined as Intersted word, except the word frequency also position that occurs in the page of with good grounds entry, the importance of position relationship is identified by source file mark in result for retrieval.
Weight computing formula is expressed as:
w i = w 0 + ( 1 / n ) Σ i = 1 k w p i
Wherein w i, represent the weights of Intersted word, w 0be weights initial values, the initial value arranged after being namely determined as Intersted word is 1, w pibe the average weight comprising Intersted word result for retrieval, n is the result for retrieval number comprising this Intersted word, and k is the sum that Intersted word occurs in all result for retrieval.W pithe weights of term in corresponding result for retrieval, if the higher last calculating of frequency that in same document, term occurs is also larger.
Interest term is not used by user, is equivalent to user and is ignoring this interest word, and therefore interest being ignored process prescription is:
F(t)＝e -log2/(hSt)
Wherein, St is the time interval, and represent that term is updated to now for the last time, namely current time deducts number of days when upgrading for the last time.H is predetermined period, and after h days, the value of F (t) is initial value half.
The right value update computing formula finally obtained is:
w i＝w i×F(t)+yN
Wherein, N is the number of times that Intersted word is updated, and y is that each interest word is used rear weights to increase coefficient.
When user interest is formed, acquiescence is all short-term interest, along with the increase of N, represents that this word is often used, just this interest is determined as Long-term Interest when its value exceedes threshold value, and the N threshold value that the present invention uses is 100.
In sum, the present invention proposes a kind of content of interest Forecasting Methodology based on machine learning, effectively improve the recognition accuracy of Internet user's feature and ageing.
Obviously, it should be appreciated by those skilled in the art, above-mentioned of the present invention each module or each step can realize with general computing system, they can concentrate on single computing system, or be distributed on network that multiple computing system forms, alternatively, they can realize with the executable program code of computing system, thus, they can be stored and be performed by computing system within the storage system.Like this, the present invention is not restricted to any specific hardware and software combination.
Should be understood that, above-mentioned embodiment of the present invention only for exemplary illustration or explain principle of the present invention, and is not construed as limiting the invention.Therefore, any amendment made when without departing from the spirit and scope of the present invention, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.In addition, claims of the present invention be intended to contain fall into claims scope and border or this scope and border equivalents in whole change and modification.

Claims (3)

1., based on a content of interest Forecasting Methodology for machine learning, it is characterized in that, comprising:
According to described user's characteristic information, adjustment is optimized to the term that user inputs, is carried out the retrieval of content of interest by term.
2. method according to claim 1, is characterized in that, is describedly optimized adjustment according to described user's characteristic information to the term that user inputs, and comprises further:
Term analysis extraction is carried out to the content that user inputs at interface, by segmenter, word segmentation processing is carried out to user search content; Obtain retrieval vector, each dimension of this retrieval vector, with a numerical value, represents the weights that term is corresponding; Watch behavior attentively according to user and differentiate object of interest, then carry out analysis structure interest model by object of interest; Concrete estimation formulas is:
Ip＝α×T b+β×U o
Wherein:
U o＝a×C copy+b×S save+G×R reply
Wherein Ip represents result for retrieval interest-degree score value, α and β is regulation coefficient, i.e. the proportion that accounts in formula of result for retrieval fixation time and result for retrieval interactive operation, wherein alpha+beta=1; T bbe the time dimension that user stops at result for retrieval, calculated by normal distribution, reflection user watches the result for retrieval time attentively; The degree of closeness of fixation time △ t and reference time t reflects interest-degree, t be according to Document Length determine reference time, t and result for retrieval length proportional; U 0the interactive operation of user on result for retrieval, C copyrepresent whether user carries out replicate run at result for retrieval, is that being worth is 1, and no value is 0; S saverepresent whether user carries out result for retrieval and preserve operation, is that being worth is 1, and no value is 0; R replyrepresenting and whether carry out feedback associative operation for result for retrieval, is that being worth is 1, and no value is 0; A, b and c are U 0regulation coefficient, according to different operations to being whether the significance level that object of interest is passed judgment on, respectively different values is arranged to coefficient.
3. method according to claim 2, is characterized in that, comprises further:
Calculate the change of the weights of interest word according to Intersted word frequency of utilization, after Feature Words is determined as Intersted word, arranging its initialization weights is 1; Weights Distribution Calculation is carried out, the position that this weight computing occurs in the page according to word frequency and entry after being defined as Intersted word; Weight computing formula is expressed as:
Wherein, N is the number of times that Intersted word is updated, and y is that each interest word is used rear weights to increase coefficient; w i, represent the weights of Intersted word, w 0be weights initial values, the initial value arranged after being namely determined as Intersted word is 1, be the average weight comprising Intersted word result for retrieval, n is the result for retrieval number comprising this Intersted word, and k is the sum that Intersted word occurs in all result for retrieval; w pithe weights of term in corresponding result for retrieval;
F (t) describes interest and ignores process:
Wherein, St is the time interval, and represent that term is updated to now for the last time, namely current time deducts number of days when upgrading for the last time; H is predetermined period, makes the value of F (t) after number of days h be initial value half.
CN201510918123.8A 2015-12-10 2015-12-10 Interested content prediction method based on machine learning Pending CN105512298A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510918123.8A CN105512298A (en) 2015-12-10 2015-12-10 Interested content prediction method based on machine learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510918123.8A CN105512298A (en) 2015-12-10 2015-12-10 Interested content prediction method based on machine learning

Publications (1)

Publication Number Publication Date
CN105512298A true CN105512298A (en) 2016-04-20

Family

ID=55720279

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510918123.8A Pending CN105512298A (en) 2015-12-10 2015-12-10 Interested content prediction method based on machine learning

Country Status (1)

Country Link
CN (1) CN105512298A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107346336A (en) * 2017-06-29 2017-11-14 北京百度网讯科技有限公司 Information processing method and device based on artificial intelligence
CN108733684A (en) * 2017-04-17 2018-11-02 合信息技术（北京）有限公司 The recommendation method and device of multimedia resource
CN111222039A (en) * 2019-11-14 2020-06-02 电子科技大学 Session recommendation method and system based on long-term and short-term interest combination
CN112487313A (en) * 2020-11-30 2021-03-12 北京百度网讯科技有限公司 Interest point identification method and device, electronic equipment and storage medium

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101408897A (en) * 2008-10-10 2009-04-15 清华大学 Personalized enquiry expanding method based on collaboration filtering
CN101820592A (en) * 2009-02-27 2010-09-01 华为技术有限公司 Method and device for mobile search
CN103593425A (en) * 2013-11-08 2014-02-19 南方电网科学研究院有限责任公司 Preference-based intelligent retrieval method and system

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101408897A (en) * 2008-10-10 2009-04-15 清华大学 Personalized enquiry expanding method based on collaboration filtering
CN101820592A (en) * 2009-02-27 2010-09-01 华为技术有限公司 Method and device for mobile search
CN103593425A (en) * 2013-11-08 2014-02-19 南方电网科学研究院有限责任公司 Preference-based intelligent retrieval method and system

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
徐康: "基于用户兴趣模型的个性化搜索排序研究", 《中国优秀硕士学位论文全文数据库》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108733684A (en) * 2017-04-17 2018-11-02 合信息技术（北京）有限公司 The recommendation method and device of multimedia resource
CN107346336A (en) * 2017-06-29 2017-11-14 北京百度网讯科技有限公司 Information processing method and device based on artificial intelligence
US11620321B2 (en) 2017-06-29 2023-04-04 Beijing Baidu Netcom Science And Technology Co., Ltd. Artificial intelligence based method and apparatus for processing information
CN111222039A (en) * 2019-11-14 2020-06-02 电子科技大学 Session recommendation method and system based on long-term and short-term interest combination
CN111222039B (en) * 2019-11-14 2022-05-20 电子科技大学 Session recommendation method and system based on long-term and short-term interest combination
CN112487313A (en) * 2020-11-30 2021-03-12 北京百度网讯科技有限公司 Interest point identification method and device, electronic equipment and storage medium

Similar Documents

Publication Publication Date Title
CN109815308B (en) Method and device for determining intention recognition model and method and device for searching intention recognition
CN103207899B (en) Text recommends method and system
US9589277B2 (en) Search service advertisement selection
US20140195893A1 (en) Method and Apparatus for Generating Webpage Content
US9934293B2 (en) Generating search results
US20190012392A1 (en) Method and device for pushing information
CN103268348B (en) A kind of user's query intention recognition methods
US20140108417A1 (en) Method and apparatus for web ad matching
CN105786838B (en) A kind of information matches treating method and apparatus
CN106504011A (en) A kind of methods of exhibiting of business object and device
CN102902691A (en) Recommending method and recommending system
CN103365839A (en) Recommendation search method and device for search engines
CN103310003A (en) Method and system for predicting click rate of new advertisement based on click log
CN103064945A (en) Situation searching method based on body
CN103473354A (en) Insurance recommendation system framework and insurance recommendation method based on e-commerce platform
WO2007133334A1 (en) System and method for determining semantically related terms
CN102541971A (en) Mapping advertiser intents to keywords
CN103729359A (en) Method and system for recommending search terms
CN104268292A (en) Label word library update method of portrait system
CN106062743A (en) Systems and methods for keyword suggestion
CN103186574A (en) Method and device for generating searching result
CN103699603A (en) Information recommendation method and system based on user behaviors
CN103020049A (en) Searching method and searching system
US20120197732A1 (en) Action-aware intent-based behavior targeting
CN104376058A (en) User interest model updating method and device

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20160420

RJ01 Rejection of invention patent application after publication