CN105205046A - System and method for on-line user recommendation based on semantic analysis - Google Patents
System and method for on-line user recommendation based on semantic analysis Download PDFInfo
- Publication number
- CN105205046A CN105205046A CN201510624936.6A CN201510624936A CN105205046A CN 105205046 A CN105205046 A CN 105205046A CN 201510624936 A CN201510624936 A CN 201510624936A CN 105205046 A CN105205046 A CN 105205046A
- Authority
- CN
- China
- Prior art keywords
- semantic analysis
- user
- information
- memory device
- 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.)
- Pending
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to an on-line user recommendation technology based on semantic analysis and belongs to the technical field of mass data processing and information safety processing. The system comprises a user group, a media information source and a semantic analysis and storage device. The method mainly comprises the steps that user behaviors of the user group are uploaded after being allowed and undergoes semantic and united semantic analysis together with the coarsely screened media information source, extracting precise information and releasing the precise information to the user group; users make a judgment on the precise information, the user behaviors are fed back to be processed, rational precise information is fed back again, and on-line user recommendation is achieved. The process of semantic analysis can comprise information form transformation and kernel keyword extraction. Methods for semantic and united semantic analysis comprise but are not limited to a hidden Markov model, a recurrent nerve network method, a deep learning method and a statistical learning method. The problems of shallowness in recommended content, formalization and lack of an on-line updating recommending strategy in a user recommendation technology are effectively solved.
Description
Technical field
The present invention relates to a kind of user's recommended technology, be specifically related to a kind of online user's recommended technology based on semantic analysis, belong to the technical fields such as magnanimity information processing, information security and Intelligent treatment.
Background technology
In decades, quantity of information is exponentially exploded in internet development, how in magnanimity information, to obtain the part personal relevant to oneself, has become the most urgent demand of user, the business of Ye Shi ISP primary study and deployment.
Amazon Company of the U.S. is famous with electric business business, it is said that its recommendation business helps it to improve the sales volume of about 30%.The commending system of effective practicality is by four parts: user profile logging modle, user preference processing module, proposed algorithm module and user feedback module.For external Amazon, Eachnet and domestic Taobao, bean cotyledon etc., their recommendation business mainly contains three kinds of forms:
1) recommendation is made based on user behavior (comprising residence time, search, feedback);
2) recommendation is made based on target information (as collection history, shopping cart information);
3) according to the relation (if history member is to the fancy grade of some commodity) of historic user and target information, recommendation mechanisms is utilized to do membership promotion.
Can see, for this group binary relation of user behavior-target information, a lot of service provider is studied.
Under above-mentioned three kinds of forms, common recommendation mechanisms algorithm comprises based on the algorithm (AssociationRule_basedRecommendation) of correlation rule, content-based synergetic (Content-basedRecommendation), Collaborative Filtering Recommendation Algorithm (CollaborativeFilteringRecommendation).
Patent " a kind of RSS information proposed algorithm based on collaborative filtering " (application number: 200910256457.8) from calculating user similarity square formation, new user is carried out.I.e. A, party B-subscriber has similar hobby, if A with the addition of new information source, then also recommends this information source for B.This method, based on cluster, simply effectively, but if new information source does not have user to bother about, just cannot recommendedly be gone to more user; That is this proposed algorithm have impact on the fairness of information recommendation.
Patent " information recommendation method based on potential community " (application number: 201210012345.X), patent " user's recommend method and device based on SNS " (application number: 201310337086.2) realize information by customer relationship cluster, user recommends.
Above-mentioned proposed algorithm ubiquity defect---the above-mentioned conventional recommendation algorithm based on similarity, relevance, the Deep Semantics that user behavior or target information contain behind is not analyzed, only depend on the shallow-layer implication of user action, cause the recommendation based on similarity can only rest in form.User is real it is desirable that one can be understood oneself, and the information about oneself individual provides pipeline.
And carry out semantic analysis, one can extract the Analysis of Deep Implications that user behavior contains behind, and two can the storage of brief user and target information relationship type database, and three fairly can recommend multiple possible information source.
Patent " a kind of user's recommend method and device " (application number: 201510089046.X) quantizes user behavior, sets some fixing formula to realize user's recommendation each other.
Patent " user's recommend method and user's commending system " (application number: the network of personal connections that 200910038601.0) with only user, those friend recommendations.Do not consider other outside authentic communication any.
Especially there is another one shortcoming in above-mentioned commending system, does not namely realize the online updating of Generalization bounds: rationally effective user's quantitatively evaluating, should make full use of other public data of user, go along with time continuous more Generalization bounds.
Summary of the invention
Object of the present invention is in order to the content recommendation shallow-layer, the formalization that solve user's recommended technology and exist and lack the problem of online updating Generalization bounds, proposes a kind of online user's commending system based on semantic analysis and method.
Based on online user's commending system of semantic analysis, it is characterized in that: comprise groups of users 100, media information source 300, semantic analysis and memory device 200; Semantic analysis and memory device 200 receive user behavior that groups of users 100 sends and the information that media information source 300 is submitted to, after carrying out semantic analysis coupling respectively, for groups of users 100 provides accurate recommendation information.
Further, described media information source 300 includes but not limited to merchandise news, multimedia messages, individual service information and good friend's recommendation information.
Further, described user behavior includes but not limited to that web page browsing or refusal record, goods browse or refusal record, good friend browses or refuses record, browsing time.
Based on online user's recommend method of semantic analysis, it is characterized in that: comprise the steps:
1) behavior of groups of users 100 upload user is in semantic analysis and memory device 200, and semantic analysis and memory device 200 pairs of user behaviors carry out semantic analysis;
2) information is supplied to semantic analysis and memory device 200 by media information source 300, and semantic analysis and memory device 200 pairs of information carry out semantic analysis;
3) semantic analysis and memory device 200 carry out combination semantic analysis in conjunction with user behavior and extensive information;
4) the comprehensively the above-mentioned 1st) to the 3rd) result of step, semantic analysis and memory device 200 extract precise information and throw in groups of users 100;
5) groups of users 100 pairs of precise informations judge, and feedback user behavior, circulation step 1)-3), again feed back precise information, realize online user and recommend.
Further, described step 1) in semantic analysis and memory device 200 pairs of user behaviors method of carrying out semantic analysis include but not limited to Hidden Markov Model (HMM) method, recurrent neural networks method, degree of deep learning method, statistical learning method.
Further, described step 2) in semantic analysis and memory device 200 pairs of information method of carrying out semantic analysis include but not limited to Hidden Markov Model (HMM) method, recurrent neural networks method, degree of deep learning method, statistical learning method.
Further, described step 3) in the method for combination semantic analysis include but not limited to Hidden Markov Model (HMM) method, recurrent neural networks method, degree of deep learning method, statistical learning method.
Beneficial effect
Adopt a kind of online user's recommended technology based on semantic analysis of the present invention, there is following beneficial effect:
1. semantic analysis is carried out to the information source in user behavior and database simultaneously, analyze data and user behavior Analysis of Deep Implications behind;
2. do not need user behavior to carry out auxiliary just independently to analyze information, and obtain information semantic, this technology when not having a large number of users behavioral data, can provide user's recommendation more accurately;
3. can carry out real-time study to the behavior of user, and feed back to user recommend in go.
Accompanying drawing explanation
Fig. 1 is the system construction drawing needed for a kind of online user's recommended technology based on semantic analysis of the present invention;
Fig. 2 is the realization flow figure of a kind of online user's recommended technology based on semantic analysis of the present invention.
Embodiment
Below in conjunction with drawings and Examples, the present invention will be further described.
Embodiment 1
The present embodiment specifically describes a kind of realization flow of the online user's recommended technology based on semantic analysis.
As depicted in figs. 1 and 2, flow process mainly comprises:
1 system boot, the groups of users 100 in accompanying drawing 1, semantic analysis and the relevant hardware devices such as memory device 200, media information source 300 are reached the standard grade;
The extensive information that 2 semantic analyses and memory device 200 reading media information source 300 provide, processed offline is carried out to information:
2.1 pairs of information carry out pre-service, as message forms such as unified voice, video, image, words;
2.2 select semantic analysis, as keyword extraction, information association, information characteristics analysis etc.;
2.3 the original extensive information material realized based on semantic analysis is analyzed.
Whether 3 are realizing while 2, investigate and have the user behavior from groups of users 100 to access in semantic analysis and memory device 200, if had, start the semantic analysis based on user behavior:
3.1 pairs of user behaviors and extensive information carry out pre-service, both integration, carry out the semantic analysis of combining;
3.2 can use the semantic analysis to extensive information carried out in advance in the 2nd step in combination semantic analysis;
3.3 combination semantic analyses have known that user's depth psychology is movable, thought, also obtain the abstract representation of extensive information, and in abstract aspect, the two is more prone to coupling, and the recommendation business being also is more accurately simple.
4., in the process realizing 2 and 3, after completing semantic analysis to a certain degree, system allows activated user recommendation request, and flow process is as follows:
4.1 investigate targeted customer, obtain this user behavior semantic analysis result;
The semantic analysis result of 4.2 investigation relevant informations;
4.3 select specific recommendations rule, to the information obtained based on semantic analysis, can adopt traditional user's proposed algorithm, including but not limited to neighbour's principle, and Support Vector Machine, collaborative filterings etc., realize user and recommend;
4.4 users obtain recommendation results and feed back, and feedack becomes new user behavior, realize process, and wait for new user's recommendation request information in feed-in the 3rd step.
Embodiment 2
The present embodiment Case-based Reasoning describes the advantage of user's recommended technology on information integration based on semantic analysis.
For traditional recommended technology, spotting information must be carried out by user behavior, namely provide auxiliary calibration by the cluster of user behavior, division, association to target information, when raw information is not when any user browses, have no idea to recommend it.
To an E-News center, morning one day is original, not proven news (not having column label etc.) submitted upper electronics press center in a large number, press center needs to recommend suitable news to user, as to party a subscriber, likes reading defense economy news information.Certain conventional recommendation method needs party a subscriber to read a certain amount of news, then finds some also to see user's second third fourth of related news, then the news that second third fourth has been seen is pushed to first.But those just send because of news, and little user has read news; Therefore use conventional methods and cannot recommend.
Use the method for semantic analysis described in this patent, semantic analysis can be carried out to news information, namely by keyword extraction and abstract, news semanteme can be analyzed, then for news tags.Finally the result of analysis is mated for the characteristic of user's first in database, and realize pushing accurately.
Embodiment 3
The present embodiment Case-based Reasoning describe based on semantic analysis user's recommended technology recommendation alleviate advantage.
To an electronic emporium, as slightly expensive commodity A deducts by party a subscriber from shopping cart, and add similar slightly cheap commodity B.Adopt conventional recommendation method, like product can be recommended to user, but but not look after the real demand of website and user.
Use semantic analysis described in this patent, the behavior that can analyze user is semantic: party a subscriber may have rigid demand to such commodity, but also has hesitation in price; If then website is considered for objective unit price, the like product that price is higher more can be recommended, if considered from rate of adoption, can the cheaper product of recommended price.And can financial service etc. be recommended.
The above is preferred embodiment of the present invention, and the present invention should not be confined to the content disclosed in this embodiment and accompanying drawing.Every do not depart from spirit disclosed in this invention under the equivalence that completes or amendment, all fall into the scope of protection of the invention.
Claims (7)
1. based on online user's commending system of semantic analysis, it is characterized in that: comprise groups of users 100, media information source 300, semantic analysis and memory device 200; Semantic analysis and memory device 200 receive user behavior that groups of users 100 sends and the information that media information source 300 is submitted to, after carrying out semantic analysis coupling respectively, for groups of users 100 provides accurate recommendation information.
2., as claimed in claim 1 based on online user's commending system of semantic analysis, it is characterized in that: described media information source 300 includes but not limited to merchandise news, multimedia messages, individual service information and good friend's recommendation information.
3. as claimed in claim 1 based on online user's commending system of semantic analysis, it is characterized in that: described user behavior includes but not limited to that web page browsing or refusal record, goods browse or refusal record, good friend browses or refuses record, browsing time.
4., based on online user's recommend method of semantic analysis, it is characterized in that: comprise the steps:
1) behavior of groups of users 100 upload user is in semantic analysis and memory device 200, and semantic analysis and memory device 200 pairs of user behaviors carry out semantic analysis;
2) information is supplied to semantic analysis and memory device 200 by media information source 300, and semantic analysis and memory device 200 pairs of information carry out semantic analysis;
3) semantic analysis and memory device 200 carry out combination semantic analysis in conjunction with user behavior and extensive information;
4) the comprehensively the above-mentioned 1st) to the 3rd) result of step, semantic analysis and memory device 200 extract precise information and throw in groups of users 100;
5) groups of users 100 pairs of precise informations judge, and feedback user behavior, circulation step 1)-3), again feed back precise information, realize online user and recommend.
5., as claimed in claim 4 based on online user's recommend method of semantic analysis, it is characterized in that: described step 1) in semantic analysis and memory device 200 pairs of user behaviors method of carrying out semantic analysis include but not limited to Hidden Markov Model (HMM) method, recurrent neural networks method, degree of deep learning method, statistical learning method.
6., as claimed in claim 4 based on online user's recommend method of semantic analysis, it is characterized in that: described step 2) in semantic analysis and memory device 200 pairs of information method of carrying out semantic analysis include but not limited to Hidden Markov Model (HMM) method, recurrent neural networks method, degree of deep learning method, statistical learning method.
7., as claimed in claim 4 based on online user's recommend method of semantic analysis, it is characterized in that: described step 3) in the method for combination semantic analysis include but not limited to Hidden Markov Model (HMM) method, recurrent neural networks method, degree of deep learning method, statistical learning method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510624936.6A CN105205046A (en) | 2015-09-25 | 2015-09-25 | System and method for on-line user recommendation based on semantic analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510624936.6A CN105205046A (en) | 2015-09-25 | 2015-09-25 | System and method for on-line user recommendation based on semantic analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105205046A true CN105205046A (en) | 2015-12-30 |
Family
ID=54952739
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510624936.6A Pending CN105205046A (en) | 2015-09-25 | 2015-09-25 | System and method for on-line user recommendation based on semantic analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105205046A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105678567A (en) * | 2015-12-31 | 2016-06-15 | 宁波领视信息科技有限公司 | Accurate prediction system based on big data deep learning |
| CN106227792A (en) * | 2016-07-20 | 2016-12-14 | 百度在线网络技术（北京）有限公司 | Method and apparatus for pushed information |
| CN106815356A (en) * | 2017-01-20 | 2017-06-09 | 泰华智慧产业集团股份有限公司 | Precision target user message method for pushing and system based on semantic analysis |
| WO2017118328A1 (en) * | 2016-01-04 | 2017-07-13 | 腾讯科技（深圳）有限公司 | Push information rough selection sorting method, device and computer storage medium |
| CN108776914A (en) * | 2018-05-22 | 2018-11-09 | 广州品唯软件有限公司 | A kind of user individual goods browse method and device based on neural network |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103703789A (en) * | 2013-06-28 | 2014-04-02 | 华为技术有限公司 | Method, terminal and system of data presentation |
| CN103778260A (en) * | 2014-03-03 | 2014-05-07 | 哈尔滨工业大学 | Individualized microblog information recommending system and method |
| CN104268130A (en) * | 2014-09-24 | 2015-01-07 | 南开大学 | Social advertising facing Twitter feasibility analysis method |
| CN104573109A (en) * | 2015-01-30 | 2015-04-29 | 深圳市中兴移动通信有限公司 | System, terminal and method for automatic recommendation based on group relation |
| CN104750856A (en) * | 2015-04-16 | 2015-07-01 | 天天艾米（北京）网络科技有限公司 | System and method for multi-dimensional synergic recommendation |
| CN104883358A (en) * | 2015-04-28 | 2015-09-02 | 腾讯科技（北京）有限公司 | Interaction method and device based on recommended content |
-
2015
- 2015-09-25 CN CN201510624936.6A patent/CN105205046A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103703789A (en) * | 2013-06-28 | 2014-04-02 | 华为技术有限公司 | Method, terminal and system of data presentation |
| CN103778260A (en) * | 2014-03-03 | 2014-05-07 | 哈尔滨工业大学 | Individualized microblog information recommending system and method |
| CN104268130A (en) * | 2014-09-24 | 2015-01-07 | 南开大学 | Social advertising facing Twitter feasibility analysis method |
| CN104573109A (en) * | 2015-01-30 | 2015-04-29 | 深圳市中兴移动通信有限公司 | System, terminal and method for automatic recommendation based on group relation |
| CN104750856A (en) * | 2015-04-16 | 2015-07-01 | 天天艾米（北京）网络科技有限公司 | System and method for multi-dimensional synergic recommendation |
| CN104883358A (en) * | 2015-04-28 | 2015-09-02 | 腾讯科技（北京）有限公司 | Interaction method and device based on recommended content |
Non-Patent Citations (2)
| Title |
|---|
| 刘蓉 等: "个性化网页推荐中基于神经网络的自适应用户模型研究", 《电子测量技术》 * |
| 周由 等: "语义分析与TF-IDF方法相结合的新闻推荐技术", 《计算机科学》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105678567A (en) * | 2015-12-31 | 2016-06-15 | 宁波领视信息科技有限公司 | Accurate prediction system based on big data deep learning |
| WO2017118328A1 (en) * | 2016-01-04 | 2017-07-13 | 腾讯科技（深圳）有限公司 | Push information rough selection sorting method, device and computer storage medium |
| CN106227792A (en) * | 2016-07-20 | 2016-12-14 | 百度在线网络技术（北京）有限公司 | Method and apparatus for pushed information |
| CN106227792B (en) * | 2016-07-20 | 2019-10-15 | 百度在线网络技术（北京）有限公司 | Method and apparatus for pushed information |
| CN106815356A (en) * | 2017-01-20 | 2017-06-09 | 泰华智慧产业集团股份有限公司 | Precision target user message method for pushing and system based on semantic analysis |
| CN106815356B (en) * | 2017-01-20 | 2019-04-26 | 泰华智慧产业集团股份有限公司 | Precision target user message method for pushing and system based on semantic analysis |
| CN108776914A (en) * | 2018-05-22 | 2018-11-09 | 广州品唯软件有限公司 | A kind of user individual goods browse method and device based on neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sivapalan et al. | Recommender systems in e-commerce | |
| Seo et al. | Personalized recommender system based on friendship strength in social network services | |
| Zhang et al. | Predicting purchase behaviors from social media | |
| CN103246980B (en) | Information output method and server | |
| US20090006469A1 (en) | Clustering users using contextual object interactions | |
| Forouzandeh et al. | Content marketing through data mining on Facebook social network. | |
| CN101454771A (en) | System and method of segmenting and tagging entities based on profile matching using a multi-media survey | |
| CN105205046A (en) | System and method for on-line user recommendation based on semantic analysis | |
| CN101916274A (en) | Method and system for polymerization display of LinkUGC (User Generated Content) | |
| CN108614832A (en) | A kind of user individual commercial articles searching implementation method and device | |
| Cherqi et al. | Analysis of hacking related trade in the darkweb | |
| Krämer et al. | The role of data for digital markets contestability: case studies and data access remedies | |
| Liang et al. | Collaborative filtering based on information-theoretic co-clustering | |
| Au Yeung et al. | Capturing implicit user influence in online social sharing | |
| Ren et al. | A survey of recommendation techniques based on offline data processing | |
| CN105590234A (en) | Method and system for recommending commodities to target users | |
| Huang et al. | Towards next-generation business intelligence: an integrated framework based on DME and KID fusion engine | |
| Kardan et al. | Targeted advertisement in social networks using recommender systems | |
| Gelli et al. | Learning visual elements of images for discovery of brand posts | |
| Mohamed et al. | Customer Profiling Using Internet of Things Based Recommendations | |
| Chew et al. | Recommender System for Retail Domain: An Insight on Techniques and Evaluations | |
| Chen et al. | Pattern filtering and classification for market basket analysis with profit‐based measures | |
| US20200111130A1 (en) | Systems and methods for automatic processing of marketing documents | |
| Wen | Development of personalized online systems for web search, recommendations, and e-commerce | |
| Zhu | Research on multi‐source mobile commerce service recommendation model of data fusion based on tree network |
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: 20151230 |
|
| RJ01 | Rejection of invention patent application after publication |