CN103559269B - A kind of knowledge recommendation method towards mobile news subscription - Google Patents
A kind of knowledge recommendation method towards mobile news subscription Download PDFInfo
- Publication number
- CN103559269B CN103559269B CN201310538587.7A CN201310538587A CN103559269B CN 103559269 B CN103559269 B CN 103559269B CN 201310538587 A CN201310538587 A CN 201310538587A CN 103559269 B CN103559269 B CN 103559269B
- Authority
- CN
- China
- Prior art keywords
- knowledge point
- knowledge
- module
- recommendation
- dictionary
- 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.)
- Expired - Fee Related
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
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2455—Query execution
- G06F16/24552—Database cache management
Abstract
The present invention relates to a kind of knowledge recommendation method towards mobile news subscription, methods described includes：（1）Extraction module extracts knowledge point successively from knowledge base；（2）Search module is stored in memory module from the title and use integration module generation knowledge point set of everyday words dictionary lookup knowledge point；（3）Update the knowledge point classification information and incidence relation of body in memory module；（4）All knowledge point titles that searching modul is found out from body in news documents form set A；（5）Analysis module forms set B to each element disambiguation in set A；（6）Analysis module filters to form recommendation list C to each element in set B；（7）Output module output recommendation list C.The knowledge point that the present invention is provided using mass-rent knowledge base is classified and relation information, and effectively the background knowledge by mobile news subscription recommends required high real-time, high availability, enhanced scalability and high availability to be melted into the system of a lightweight.
Description
Technical field
The invention belongs to search field, in particular to a kind of knowledge recommendation method towards mobile news subscription.
Background technology
The rapid popularization of the mobile intelligent terminal such as smart mobile phone and tablet personal computer is greatly expanded people and obtains information
Approach, especially for news category information, it is interested that people can obtain oneself whenever and wherever possible by way of keyword subscription
Content.But user is while news is read, it is likely that it is desirable that to involved the fact property key element of the media event (such as
Related person, organization, place etc.) background knowledge (being usually the page of similar encyclopaedia entry property) do further
Solution.At present, it disclosure satisfy that the technology of such demand predominantly recommends (or filtering) technology, recommended technology is broadly divided into based on cooperation
Recommendation, content-based recommendation and Knowledge based engineering recommend three types.
Recommendation based on cooperation：Its basic assumption be if there is similar preference with some other users before user,
Then in future, they can also have similar preference.It is general that the similarity degree-user of user preference (is pressed using arest neighbors
Based, or the similarity degree-item based that are had a preference for of article) or the method for Mining Association Rules predict active user new
Preference on article (new knowledge point), recommendation results are provided further according to the preference of prediction.But the application of such recommendation is the most
Extensively, but need to obtain user preference data, with the growth of number of users and knowledge base scale, the demand meeting to computing resource
Sharply increase, and the recommendation based on cooperation does not account for the content of knowledge point to be recommended.
Content-based recommendation：If its basic assumption is that certain article is close with the article of user preference, it may belong to
In the article that user likes.Recommendation of this method especially suitable for text.Vector space model of the generally use based on tfidf
To represent document, recommendation list is being provided using arest neighbors (or top k neighbours) method.But this type recommends to be mainly used in text
Recommend, and knowledge page usually contains the information of multiple format, or even the content including dynamic change, thus be difficult application or
Build unified document (knowledge point) method for expressing (such as tfidf).
Knowledge based engineering is recommended：The recommendation of this type is that its recommended requirements (forms the pact to commodity by user's explicit definition
Beam), similarity or the special rule of use between system-computed user's request and article are recommended.But the recommendation master of this type
The commodity (such as automobile, large electric appliances) that will not be frequently bought in a period of time are directed to, because evaluating letter in the case of this kind of
Breath is seldom and easily fails.The input and selection of mobile terminal user is difficulty all compared with pc, and allowing user oneself to set complicated recommendation needs
The complexity that can increase user's operation, the notice for shifting user for a long time are asked, while user is intended to ask remote to correlated knowledge point
It is intended to ask not as good as first purchase large scale commercial product, therefore Knowledge based engineering recommended technology is unsuitable for the recommendation to knowledge point.
The content of the invention
In view of the shortcomings of the prior art, the present invention provides a kind of knowledge recommendation method towards mobile news subscription.For
Fast response time that mobile news subscription requires in itself, the features such as visit capacity is big, content update is rapid, devise and known based on mass-rent
Know the knowledge recommendation method of storehouse and body, lightweight can be realized using this method, there is high real-time, enhanced scalability, height
The backgrounding content recommendation system of availability.
The purpose of the present invention is realized using following technical proposals：
A kind of knowledge recommendation method towards mobile news subscription, it is theed improvement is that, methods described includes：
(1) extraction module extracts knowledge point successively from knowledge base；
(2) search module is deposited from the title and use integration module generation knowledge point set of everyday words dictionary lookup knowledge point
It is stored in memory module；
(3) the knowledge point classification information and incidence relation of body in memory module are updated；
(4) all knowledge point titles that searching modul is found out from body in news documents form set A；
(5) analysis module forms set B to each element disambiguation in set A；
(6) analysis module filters to form recommendation list C to each element in set B；
(7) output module output recommendation list C.
Preferably, the step (2) includes passing through the title that knowledge point is searched in everyday words dictionary, if everyday words, then
Continue to take next knowledge point；If it is not everyday words, integration module generation knowledge point set is used to be stored in storage
In module, for recommending.
Preferably, a knowledge point title may be corresponding with the polysemy of multiple knowledge points, knowledge point title with
Mapping relations between the id of knowledge point are safeguarded by memory module.
Preferably, the step (3) is included according to three knowledge point set, classificating word dictionary and relative dictionary dictionary lifes
Into and renewal classification tree and knowledge point graph of a relation renewal memory module in body.
Further, the body includes classification tree, graph of a relation, classificating word dictionary and relative dictionary；Knowledge point ID positions
In the leaf node of ontology classification tree, the intermediate node of classification tree is classifier ID, and classificating word dictionary includes classifier and its ID
Between mapping relations；Graph of a relation includes target sparse matrix under the ID ranks of knowledge point, and matrix element is relative ID list, is closed
Copula dictionary includes the mapping relations between relative and its ID.
Preferably, the step (5) includes
If corresponding knowledge point a, unambiguously, is added into Candidate Recommendation knowledge point set B；With
B, if ambiguity, then selection subscribes to knowledge point with minimum public ancestors with user from corresponding all knowledge points
That knowledge point, be added into set B.
Preferably, the step (6) includes
If a) title occurrence number in news documents is more than or equal to threshold value δ, it is added into and recommends knowledge point list C
In；
B) if title occurrence number is less than threshold value, but the knowledge point and user subscribe to knowledge point have dependency relation (, then by it
Add recommendation list C；With
If c) title occurrence number is less than threshold value, and subscribes to knowledge point without dependency relation with user, then the knowledge point is abandoned.
Further, the news documents are shorter, and to reduce empty recommendation, then threshold value δ takes 0.
Compared with the prior art, beneficial effects of the present invention are：
1) inventive algorithm is simple, it is easy to accomplish, efficiency high is higher to requirement of real-time suitable for mobile news subscription etc.
Environment.
2) present system magnitude is light, without handling the user preference data of magnanimity, takes calculating and storage resource is less.
3) scalability of the present invention is strong, and scale is not for main back-end data (everyday words dictionary, body, knowledge point set)
Greatly, and the uniformity of data need not be ensured
4) availability of the present invention is high, and everyday words dictionary, body, knowledge point set etc. can online updatings.
5) knowledge base of present invention structure body is based on mass-rent, both ensure that higher quality, and without using complexity
Natural language understanding technology.
6) present invention takes full advantage of the classification of high quality knowledge point and the relation information of mass-rent knowledge place offer, avoids
Acquisition, processing and the analysis of large-scale consumer preference information, it also avoid building the complexity nature that Opening field body is commonly used
Language processing techniques, high real-time, high availability, Gao Ke required by effectively the background knowledge of mobile news subscription is recommended
Autgmentability and high availability are melted into the system of a lightweight.
Brief description of the drawings
Fig. 1 is a kind of knowledge recommendation method structured flowchart towards mobile news subscription provided by the invention.
Fig. 2 is the flow of knowledge processing in a kind of knowledge recommendation method towards mobile news subscription provided by the invention
Figure.
Fig. 3 is the flow for recommending part in a kind of knowledge recommendation method towards mobile news subscription provided by the invention
Figure.
Embodiment
The embodiment of the present invention is described in further detail below in conjunction with the accompanying drawings.
As shown in figure 1, a kind of knowledge recommendation method towards mobile news subscription of the present invention is the knowledge base based on mass-rent
On the basis of build body and knowledge point dictionary, with carry out knowledge recommendation.System includes knowledge processing and recommends (to include content again
Analysis and two main modulars of disambiguation/filtering) two parts.
Knowledge processing part using everyday words dictionary (can be with manual maintenance) extracted from knowledge base useful knowledge point (with it is normal
The knowledge point of word matching is by as noise), and the classification in knowledge base and relation information structure and renewal body；
Recommend part to be used to handle request (pick out from by the document presented for user suitable recommend word), it
All available knowledge points are found out in news documents as candidate recommendation project, then utilize the body and mistake constructed by knowledge processing
Filter rule, disambiguation and filtering are carried out to candidate recommendation project, to improve the precision recommended.
Body includes classification tree, graph of a relation, classificating word dictionary and relative dictionary.Knowledge point id is located at ontology classification tree
Leaf node, the intermediate node of classification tree is classifier id, and the mapping relations between classifier and its id are deposited by classificating word dictionary；
Graph of a relation can be stored by making under ranks target sparse matrix with knowledge point id, and matrix element is relative id list, relation
Mapping relations between word and its id are deposited by relative dictionary.
Due to the presence of polysemy, a knowledge point title may be corresponding with multiple knowledge points, and (they have not
Same knowledge point id), the mapping relations between knowledge point title and knowledge point id can be safeguarded by knowledge point set.
As shown in Fig. 2 the flow key step of knowledge processing includes：
1) each knowledge point is taken out successively from knowledge base；
2) title of the knowledge point is searched in everyday words dictionary, if it is everyday words, continues to take next knowledge point.
If it is not everyday words, knowledge point set is added into, for recommending；
3) to the knowledge point of addition knowledge point set, its classification information and incidence relation are taken, renewal body is (by knowledge point
Set, the three dictionary generations of classificating word dictionary and relative dictionary and renewal classification tree and knowledge point graph of a relation).
As shown in figure 3, the flow key step recommended includes as follows：
1) knowledge point set is utilized, finds out all knowledge point titles in news documents, is taken in set A.
2) disambiguation operation is carried out to each element (knowledge point title) in set A；
If a) it is unambiguously (a corresponding knowledge point id), its corresponding knowledge point is added into Candidate Recommendation knowledge point set
Close B.
B) if ambiguity (corresponding multiple knowledge point id), then (classification of body is utilized from all knowledge points corresponding to it
Tree) that knowledge point of selection and user subscription knowledge point with minimum public ancestors, it is added into set B.
3) filter operation is carried out to each element (knowledge point) in set B；
If a) its title occurrence number in news documents is more than or equal to threshold value δ (such as taking 3), it is added into knowledge point and pushes away
Recommend in list C and (for newsflash document, recommend to reduce sky, then 0) threshold value δ takes；
If b) its title occurrence number is less than threshold value, but the knowledge point subscribes to knowledge point with user and has dependency relation (to use
The graph of a relation of body determines), then it is added into recommendation list C；
If c) title occurrence number is less than threshold value, and subscribes to knowledge point without dependency relation with user, then the knowledge point is abandoned.
4) recommendation list C is exported.
Embodiment
User a is interested in certain class news described by keyword w, and his (she) can only mobile device (such as intelligent hand at it
Machine) news subscription software S in have subscribed the keyword.S basic function is to be presented to the newest news t related to w
a.Existing S wishes to increase background knowledge recommendation function, while t is presented into a, the related of some words/phrase p in t
Structural knowledge (may include basic introduction, the hyperlink to other related pages and the various content of multimedia of text formatting
Link etc.) also recommend a.
Implementation method：Using mass-rent mode construction knowledge base K (p set P, each p page containing classification information,
The related content of various forms and dependency relation-such as related person with other p, and each p page elements and layout can
Can be different).General term dictionary D is built using artificial or semi-artificial mode.Using the knowledge process flow (utilizing D) shown in Fig. 2
Knowledge point set Z and body O are constructed from K.For news t, the recommendation stream shown in Fig. 3 can be used before a is presented to
Journey (utilizing Z and D), which provides, recommends knowledge point list, is presented to a together with t afterwards.Specific presentation mode can be used in t
Presenting as Anchor Text is occurred in the first time of word in recommendation list, can also be arranged near t presentation region (such as lower section)
Go out to recommend the mode of word to present.
Finally it should be noted that：The above embodiments are merely illustrative of the technical scheme of the present invention and are not intended to be limiting thereof, to the greatest extent
The present invention is described in detail with reference to above-described embodiment for pipe, those of ordinary skills in the art should understand that：Still
The embodiment of the present invention can be modified or equivalent substitution, and without departing from any of spirit and scope of the invention
Modification or equivalent substitution, it all should cover among scope of the presently claimed invention.
Claims (5)
- A kind of 1. knowledge recommendation method towards mobile news subscription, it is characterised in that methods described includes：(1) extraction module extracts knowledge point successively from knowledge base；(2) search module is stored in from the title and use integration module generation knowledge point set of everyday words dictionary lookup knowledge point In memory module；(3) the knowledge point classification information and incidence relation of body in memory module are updated；(4) all knowledge point titles that searching modul is found out from body in news documents form set A；(5) analysis module forms set B to each element disambiguation in set A；(6) analysis module filters to form recommendation list C to each element in set B；(7) output module output recommendation list C；One knowledge point title may be corresponding with the polysemy of multiple knowledge points, between knowledge point title and knowledge point ID Mapping relations safeguarded by memory module；The step (2) includes passing through the title that knowledge point is searched in everyday words dictionary, if everyday words, then continues to take next Knowledge point；If it is not everyday words, integration module generation knowledge point set is used to be stored in memory module, for recommending Use；The step (3) is included according to knowledge point set, the three dictionary generations of classificating word dictionary and relative dictionary and renewal point Class tree and knowledge point graph of a relation, update body in memory module.
- A kind of 2. knowledge recommendation method towards mobile news subscription as claimed in claim 1, it is characterised in that the body Including classification tree, graph of a relation, classificating word dictionary and relative dictionary；Knowledge point ID is located at the leaf node of Ontology tree, classification The intermediate node of tree is classifier ID, and classificating word dictionary includes the mapping relations between classifier and its ID；Graph of a relation includes knowledge Target sparse matrix under point ID ranks, matrix element are relative ID list, and relative dictionary is included between relative and its ID Mapping relations.
- A kind of 3. knowledge recommendation method towards mobile news subscription as claimed in claim 1, it is characterised in that the step (5) includeIf corresponding knowledge point a, unambiguously, is added into Candidate Recommendation knowledge point set B；WithB, if ambiguity, then selection and user's subscription knowledge point have that of minimum public ancestors from corresponding all knowledge points Individual knowledge point, it is added into set B.
- A kind of 4. knowledge recommendation method towards mobile news subscription as claimed in claim 1, it is characterised in that the step (6) includeIf a) title occurrence number in news documents is more than or equal to threshold value δ, it is added into and recommends in knowledge point list C；If b) title occurrence number is less than threshold value, but there is dependency relation the knowledge point with user's subscription knowledge point, then is added into Recommendation list C；WithIf c) title occurrence number is less than threshold value, and subscribes to knowledge point without dependency relation with user, then the knowledge point is abandoned.
- A kind of 5. knowledge recommendation method towards mobile news subscription as claimed in claim 4, it is characterised in that the news Document is shorter, and to reduce empty recommendation, then threshold value δ takes 0.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310538587.7A CN103559269B (en) | 2013-11-04 | 2013-11-04 | A kind of knowledge recommendation method towards mobile news subscription |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310538587.7A CN103559269B (en) | 2013-11-04 | 2013-11-04 | A kind of knowledge recommendation method towards mobile news subscription |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103559269A CN103559269A (en) | 2014-02-05 |
| CN103559269B true CN103559269B (en) | 2018-02-06 |
Family
ID=50013515
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310538587.7A Expired - Fee Related CN103559269B (en) | 2013-11-04 | 2013-11-04 | A kind of knowledge recommendation method towards mobile news subscription |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103559269B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105302897B (en) * | 2015-10-21 | 2018-11-20 | 无锡天脉聚源传媒科技有限公司 | A kind of acquisition methods and device of search result |
| CN107967254B (en) * | 2017-10-31 | 2021-05-04 | 科大讯飞股份有限公司 | Knowledge point prediction method and device, storage medium and electronic equipment |
| CN108536872B (en) * | 2018-04-27 | 2020-12-25 | 湖北时代万新国际教育发展有限公司 | Method and device for optimizing knowledge base structure |
| CN111737407B (en) * | 2020-08-25 | 2020-11-10 | 成都数联铭品科技有限公司 | Event unique ID construction method based on event disambiguation |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1998020405A2 (en) * | 1996-11-06 | 1998-05-14 | U.C. Media, Inc. | Method and system for screen text capturing |
| CN102611785B (en) * | 2011-01-20 | 2014-04-02 | 北京邮电大学 | Personalized active news recommending service system and method for mobile phone user |
-
2013
- 2013-11-04 CN CN201310538587.7A patent/CN103559269B/en not_active Expired - Fee Related
Non-Patent Citations (3)
| Title |
|---|
| 基于本体知识库的自动语义标注;戚欣 等;《计算机应用研究》;20110531;第28卷(第5期);第1742-1744页、第1747页 * |
| 基于本体论的知识检索研究;艾丹祥;《中国优秀博硕士学位论文全文数据库（硕士） 信息科技辑》;20041215;第11-21页 * |
| 社交媒体的语义标注应用现状研究;刘振;《图书与情报》;20131020(第5期);第95-99页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103559269A (en) | 2014-02-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7201730B2 (en) | Intention recommendation method, device, equipment and storage medium | |
| Wang et al. | A sentiment-enhanced hybrid recommender system for movie recommendation: a big data analytics framework | |
| CN110321482B (en) | Information recommendation method, device and equipment | |
| Deveaud et al. | Accurate and effective latent concept modeling for ad hoc information retrieval | |
| Deshpande et al. | Building, maintaining, and using knowledge bases: a report from the trenches | |
| Konstas et al. | On social networks and collaborative recommendation | |
| KR101166130B1 (en) | Method and apparatus for constructing user profile using contents tag, and method for contents recommendation using the constructed user profile | |
| JP2010067175A (en) | Hybrid content recommendation server, recommendation system, and recommendation method | |
| Martín et al. | Using semi-structured data for assessing research paper similarity | |
| Hu et al. | Enhancing accessibility of microblogging messages using semantic knowledge | |
| CN103559269B (en) | A kind of knowledge recommendation method towards mobile news subscription | |
| CN103384883A (en) | Semantic enrichment by exploiting Top-K processing | |
| Yang et al. | A topic model for co-occurring normal documents and short texts | |
| CN109582869A (en) | A kind of data processing method, device and the device for data processing | |
| CN101770291B (en) | Semantic analysis data hashing storage and analysis methods for input system | |
| Celikyilmaz et al. | Leveraging web query logs to learn user intent via bayesian latent variable model | |
| Spitz et al. | Exploring entity-centric networks in entangled news streams | |
| CN112487161A (en) | Enterprise demand oriented expert recommendation method, device, medium and equipment | |
| CN111625718A (en) | User portrait construction method based on user search keyword data | |
| Bouras et al. | Assisting cluster coherency via n-grams and clustering as a tool to deal with the new user problem | |
| Liu et al. | Recent advances in personal recommender systems | |
| Tatli et al. | A tag-based hybrid music recommendation system using semantic relations and multi-domain information | |
| CN109145089B (en) | Hierarchical thematic attribute extraction method based on natural language processing | |
| CN104376034B (en) | Information processing equipment, information processing method and program | |
| WO2017135889A1 (en) | Ontology determination methods and ontology determination devices |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C41 | Transfer of patent application or patent right or utility model | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20160412 Address after: 100086, No. 2, building 43, No. 5 West Third Ring Road, Haidian District, Beijing, 01-03A Applicant after: Beijing Wyatt Network Technology Co.,Ltd. Address before: Shou Heng Technology Building No. 51 Beijing 100191 Haidian District Xueyuan Road room 0902 Applicant before: BEIJING ZHONGSOU NETWORK TECHNOLOGY Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180206 Termination date: 20211104 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |