CN110427563A - A kind of professional domain system cold start-up recommended method of knowledge based map - Google Patents
A kind of professional domain system cold start-up recommended method of knowledge based map Download PDFInfo
- Publication number
- CN110427563A CN110427563A CN201910811379.7A CN201910811379A CN110427563A CN 110427563 A CN110427563 A CN 110427563A CN 201910811379 A CN201910811379 A CN 201910811379A CN 110427563 A CN110427563 A CN 110427563A
- Authority
- CN
- China
- Prior art keywords
- user
- article
- node
- professional domain
- knowledge
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- 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
- 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
- G06F16/9538—Presentation of query results
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Mathematical Physics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
A kind of professional domain system cold start-up recommended method of knowledge based map, the following steps are included: (1) constructs professional domain knowledge mapping by the way of semi-automatic, it is primarily based on professional domain classified catalogue and manually initializes professional domain knowledge mapping, then extend professional domain knowledge mapping automatically using internet knowledge；(2) it extracts label respectively from the description text of the registration information of user and article, and is based on internet text training professional domain term vector；(3) entity link is carried out in knowledge mapping to user tag and article tag respectively first, the shortest path being then based between hinged node calculates user/article matching angle value；(4) recommend matching several highest articles of angle value to user.The present invention considers user's registration information and item contents information simultaneously, realizes system cold start-up；Based on professional domain knowledge architecture knowledge mapping, the accurate and quantization matching of user's registration information and item contents information is realized.
Description
Technical field
The present invention relates to natural language processings and data mining technology, and in particular to a kind of recommended method.
Background technique
Cold start-up recommendation problem is divided into user's cold start-up recommendation, article cold start-up recommendation and system cold start-up recommendation three greatly
Class.Wherein, user's cold start-up, which refers to do for new user, recommends, and article cold start-up, which refers to, recommends interested user for new article, is
System cold start-up refers to carries out personalized recommendation in system (i.e. absolutely not user's behavioral data) newly developed.Obviously, system is cold
Start available minimum data, realizes that difficulty is maximum.
The existing method for solving the problems, such as cold start-up recommendation recommends (such as recommendation mainly include the following types: (1) provides impersonal theory
Popular ranking list), but the article of this method recommendation is unable to satisfy the individual demand of user.(2) user's registration information is utilized,
User is allowed to fill in its all kinds of personal attribute information (such as gender, age, occupation) when registration, then according to statistical data
Recommend different articles to the user of different attribute.However, this method needs the historical statistical data of article, therefore only it is applicable in
User is cold-started recommendation problem.(3) utilize item contents information, i.e., based on natural language processing technique to item contents information into
Row analysis, carries out article similarity calculation on this basis, then recommends similar object to it if user likes certain articles
Product.However, this method needs the interesting data of user, therefore only it is applicable in article and is cold-started recommendation problem.
Recommendation problem is cold-started for system, user's registration information and item contents information can be utilized simultaneously.However, this
The maximum challenge of method is how to match the contents semantic of the personal attribute of user and article.For general domain,
The mode of artificial constructed rule can be used (for example, women prefers clothing product, male prefers electronic product).So
And this mode is not suitable for professional domain, and due to: (1) difficulty of professional domain artificial constructed rule and cost too big.For example,
Common people are difficult to judge books or the soft project side in terms of user's machine learning interested of a research artificial intelligence
The books in face.(2) mode of artificial constructed rule can only qualitative analysis, it is difficult to quantitative analysis.For example, i.e. enabled judge a use
Books in terms of family machine learning interested and natural language processing, but be difficult judgement it is interested which, degree is the more few.
Summary of the invention
It cannot achieve the deficiency that system is cold-started, accuracy is poor for existing recommended method, the invention proposes one
The professional domain system of kind knowledge based map is cold-started recommended method, while considering user's registration information and item contents letter
Breath realizes system cold start-up；Based on professional domain knowledge architecture knowledge mapping, user's registration information and item contents information are realized
It is accurate and quantization matching.
The technical solution adopted by the present invention to solve the technical problems is:
A kind of professional domain system cold start-up recommended method of knowledge based map, comprising the following steps:
(1) professional domain knowledge mapping constructs: constructing professional domain knowledge mapping, first base by the way of semi-automatic
Professional domain knowledge mapping is manually initialized in professional domain classified catalogue, then extends profession neck automatically using internet knowledge
Domain knowledge map；
(2) user/article portrait and term vector training: from being taken out respectively in the description text of the registration information of user and article
Label is taken, and is based on internet text training professional domain term vector；
(3) user/article matching degree scoring: user tag and article tag are carried out in knowledge mapping respectively first real
Body link, the shortest path being then based between hinged node calculate user/article matching angle value；
(4) personalized recommendation: recommend matching several highest articles of angle value to user.
Further, in the step (1), professional domain knowledge mapping is configured to offline task, and process is as follows:
The initialization of (1-1) professional domain knowledge mapping: particular professional field can usually find relevant classified catalogue, use
In the related notion and hyponymy that describe the professional domain roughly；Based on classified catalogue, the artificial constructed professional domain
Initial knowledge map G；Wherein, node on behalf concept, the side between node represent the hyponymy between node, the weight on side
It is 1, standard nodes will be known as based on the node that classified catalogue constructs, standard knot point set is denoted as SN；
(1-2) extends knowledge search: to standard nodes sd each in Gi, first in encyclopaedic knowledge library (such as wikipedia, hundred
Degree encyclopaedia) in search obtain corresponding sdiEntry seiWeb page text spi, and obtain spiIn all reference words with hyperlink
Item.Then, each TF-IDF value with reference to entry is calculated, and takes the reference entry of the highest α % of TF-IDF value as seiExpansion
Entry collection is opened up, EE is denoted asi；
(1-3) expanding node insertion: to standard nodes sd each in GiExtension entry collection EEiIn each extension entry
eeij, searched in G whether have existed corresponding ee firstijNode dij, a connection sd is then created if it existsiAnd dijSide
eij, otherwise create a node dijAnd create a connection sdiAnd dijSide eij, newly-built node is known as expanding node；So
Afterwards, side e is calculated according to following formulaijWeight wij, wherein tfidf (eeik) it is eeikIn spiIn TF-IDF value；
Further, in the step (2), user/article portrait and term vector are trained for offline task, and process is as follows:
(2-1) user tag extracts: it is required that user selects to divide belonging to its people from classified catalogue when registration
Category, and as the label of user；
(2-2) article tag extracts: firstly, describing to extract the highest K of score in text from article based on TF-IDF algorithm
The keyword of a noun part-of-speech；Then, this K keyword is searched in encyclopaedic knowledge library, retains the pass that can search entry
Keyword, and as the label of article；
The training of (2-3) term vector: it based on disclosed pre-training term vector model on internet, is based on
Word2vec algorithm is finely adjusted pre-training term vector model on all web page texts that step (1-2) obtains, and obtains most
Whole term vector model WV.
Further, in the step (3), user/article matching degree scoring is online task, gives user u and article
The tally set of p, u are UL (u), and the tally set of p is PL (p), and process is as follows:
The link of (3-1) user subject: to label UL (u) [i] each in UL (u), its corresponding node D (u) is searched in G
[i], since the label of user is the sorting item in classified catalogue, D (u) [i] is standard nodes, by what is searched for u
Node collection is denoted as D (u)；
(3-2) article entity link: to label PL (p) [i] each in PL (p), its corresponding node D (p) is searched in G
[i], since the label of article is keyword, D (p) [i] may be sky；If D (p) [i] is not sky, D (p) [i] is set
Confidence level coefficient c (p) [i] be 1；If D (p) [i] is sky, word-based vector model WV search is semantic similar to PL (p) [i]
Spend highest entry lejCorresponding node is as D (p) [i], and the confidence level coefficient c (p) [i] that D (p) [i] is arranged is PL (p)
[i] and lejSemantic similarity value.The node collection searched for p is denoted as D (p)；
(3-3) node matching degree calculates: the node D (p) [j] in node D (u) [i] and D (p) in given D (u), first
The shortest path SP that D (p) [j] arrives D (u) [i] is searched in Gji, then, D (u) [i] and D (p) [j] are calculated according to following formula
Node matching angle value；
(3-4) user/article matching degree calculates: the matching angle value of user u and article p, user u are calculated according to following formula
Matching angle value with article p is max { c (p) [1] × w24,c(p)[0]×w38×w35×w25}；
In the step (4), personalized recommendation is online task, and process is as follows: given user u is calculated needed first
Recommend the matching angle value of article and u, and by matching angle value descending arrangement；Then, several articles before the highest that sorts are recommended
To u.
Beneficial effects of the present invention are mainly manifested in: (1) while considering user's registration information and item contents information, realize
System cold start-up.(2) it is based on professional domain knowledge architecture knowledge mapping, realizes the essence of user's registration information and item contents information
Quasi- and quantization matching.
Detailed description of the invention
Fig. 1 is that a kind of professional domain of knowledge based map is cold-started recommended method flow chart；
Fig. 2 is the implementation example figure of artificial intelligence field knowledge mapping building；
Fig. 3 is the implementation example figure that a user/article matching degree calculates.
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings.
Referring to Fig.1~Fig. 3, a kind of professional domain cold start-up recommended method of knowledge based map, comprising the following steps:
(1) professional domain knowledge mapping constructs: constructing professional domain knowledge mapping, first base by the way of semi-automatic
Professional domain knowledge mapping is manually initialized in professional domain classified catalogue, then extends profession neck automatically using internet knowledge
Domain knowledge map；
(2) user/article portrait and term vector training: from being taken out respectively in the description text of the registration information of user and article
Label is taken, and is based on internet text training professional domain term vector；
(3) user/article matching degree scoring: user tag and article tag are carried out in knowledge mapping respectively first real
Body link, the shortest path being then based between hinged node calculate user/article matching angle value；
(4) personalized recommendation: recommend matching several highest articles of angle value to user.
Further, in the step (1), professional domain knowledge mapping is configured to offline task, and process is as follows:
The initialization of (1-1) professional domain knowledge mapping: particular professional field can usually find relevant classified catalogue, use
In the related notion and hyponymy that describe the professional domain roughly；Based on classified catalogue, the artificial constructed professional domain
Initial knowledge map G；Wherein, node on behalf concept, the side between node represent the hyponymy between node, the weight on side
It is 1.It will be known as standard nodes based on the node that classified catalogue constructs, standard knot point set is denoted as SN；
(1-2) extends knowledge search: to standard nodes sd each in Gi, first in encyclopaedic knowledge library (such as wikipedia, hundred
Degree encyclopaedia) in search obtain corresponding sdiEntry seiWeb page text spi, and obtain spiIn all reference words with hyperlink
Item；Then, each TF-IDF value with reference to entry is calculated, and takes the reference entry of the highest α % of TF-IDF value as seiExpansion
Entry collection is opened up, EE is denoted asi；
(1-3) expanding node insertion: to standard nodes sd each in GiExtension entry collection EEiIn each extension entry
eeij, searched in G whether have existed corresponding ee firstijNode dij, a connection sd is then created if it existsiAnd dijSide
eij, otherwise create a node dijAnd create a connection sdiAnd dijSide eij, newly-built node is known as expanding node；So
Afterwards, side e is calculated according to following formulaijWeight wij, wherein tfidf (eeik) it is eeikIn spiIn TF-IDF value, for example, figure
2 give the embodiment of an artificial intelligence field knowledge mapping；
Further, in the step (2), user/article portrait and term vector are trained for offline task, and process is as follows:
(2-1) user tag extracts: it is required that user selects to divide belonging to its people from classified catalogue when registration
Category, and as the label of user；
(2-2) article tag extracts: firstly, describing to extract the highest K of score in text from article based on TF-IDF algorithm
The keyword of a noun part-of-speech；Then, this K keyword is searched in encyclopaedic knowledge library, retains the pass that can search entry
Keyword, and as the label of article；
The training of (2-3) term vector: it based on disclosed pre-training term vector model on internet, is based on
Word2vec algorithm is finely adjusted pre-training term vector model on all web page texts that step (1-2) obtains, and obtains most
Whole term vector model WV.
In the step (3), user/article matching degree scoring is online task.The label of given user u and article p, u
Integrate as UL (u), the tally set of p is PL (p), and process is as follows:
The link of (3-1) user subject: to label UL (u) [i] each in UL (u), its corresponding node D (u) is searched in G
[i], since the label of user is the sorting item in classified catalogue, D (u) [i] is standard nodes, by what is searched for u
Node collection is denoted as D (u)；
(3-2) article entity link: to label PL (p) [i] each in PL (p), its corresponding node D (p) is searched in G
[i], since the label of article is keyword, D (p) [i] may be sky；If D (p) [i] is not sky, D (p) [i] is set
Confidence level coefficient c (p) [i] be 1；If D (p) [i] is sky, word-based vector model WV search is semantic similar to PL (p) [i]
Spend highest entry lejCorresponding node is as D (p) [i], and the confidence level coefficient c (p) [i] that D (p) [i] is arranged is PL (p)
[i] and lejSemantic similarity value, the node collection searched for p is denoted as D (p)；
(3-3) node matching degree calculates: the node D (p) [j] in node D (u) [i] and D (p) in given D (u), first
The shortest path SP that D (p) [j] arrives D (u) [i] is searched in Gji；Then, D (u) [i] and D (p) [j] are calculated according to following formula
Node matching angle value；
(3-4) user/article matching degree calculates: the matching angle value of user u and article p is calculated according to following formula.For example,
Fig. 3 gives the embodiment that a user/article matching degree calculates, and wherein the matching angle value of user u and article p is max { c (p)
[1]×w24,c(p)[0]×w38×w35×w25}；
In the step (4), personalized recommendation is online task, and process is as follows: given user u is calculated needed first
Recommend the matching angle value of article and u, and by matching angle value descending arrangement；Then, several articles before the highest that sorts are recommended
To u.
Claims (5)
1. the professional domain system of knowledge based map a kind of is cold-started recommended method, which is characterized in that the method includes with
Lower step:
(1) professional domain knowledge mapping constructs: constructing professional domain knowledge mapping by the way of semi-automatic, is primarily based on specially
Industry domain classification catalogue manually initializes professional domain knowledge mapping, then extends professional domain automatically using internet knowledge and knows
Know map；
(2) user/article portrait and term vector training: from extracting mark respectively in the description text of the registration information of user and article
Label, and it is based on internet text training professional domain term vector；
(3) chain of entities user/article matching degree scoring: is carried out in knowledge mapping to user tag and article tag respectively first
It connects, the shortest path being then based between hinged node calculates user/article matching angle value；
(4) personalized recommendation: recommend matching several highest articles of angle value to user.
2. a kind of professional domain system of knowledge based map as described in claim 1 is cold-started recommended method, feature exists
In in the step (1), professional domain knowledge mapping is configured to offline task, and process is as follows:
The initialization of (1-1) professional domain knowledge mapping: particular professional field can usually find relevant classified catalogue, for thick
The related notion and hyponymy of the professional domain are slightly described；Based on classified catalogue, the artificial constructed professional domain it is initial
Knowledge mapping G；Wherein, node on behalf concept, the side between node represent the hyponymy between node, and the weight on side is 1,
It will be known as standard nodes based on the node that classified catalogue constructs, standard knot point set is denoted as SN；
(1-2) extends knowledge search: to standard nodes sd each in Gi, search obtains corresponding sd first in encyclopaedic knowledge libraryi's
Entry seiWeb page text spi, and obtain spiIn all reference entries with hyperlink；Then, it calculates each with reference to entry
TF-IDF value, and take the reference entry of the highest α % of TF-IDF value as seiExtension entry collection, be denoted as EEi；
(1-3) expanding node insertion: to standard nodes sd each in GiExtension entry collection EEiIn each extension entry eeij,
It is searched in G first and whether has existed corresponding eeijNode dij, a connection sd is then created if it existsiAnd dijSide eij,
Otherwise a node d is createdijAnd create a connection sdiAnd dijSide eij, newly-built node is known as expanding node；Then,
Side e is calculated according to following formulaijWeight wij, wherein tfidf (eeik) it is eeikIn spiIn TF-IDF value；
3. a kind of professional domain system of knowledge based map as claimed in claim 1 or 2 is cold-started recommended method, feature
It is, in the step (2), user/article portrait and term vector are trained for offline task, and process is as follows:
(2-1) user tag extracts: it is required that user selects sorting item belonging to its people when registration from classified catalogue,
And as the label of user；
(2-2) article tag extracts: firstly, describing to extract the highest K name of score in text from article based on TF-IDF algorithm
The keyword of word part of speech；Then, this K keyword is searched in encyclopaedic knowledge library, retains the keyword that can search entry,
And as the label of article；
The training of (2-3) term vector: it based on disclosed pre-training term vector model on internet, is calculated based on word2vec
Method is finely adjusted pre-training term vector model on all web page texts that step (1-2) obtains, and obtains final term vector
Model WV.
4. a kind of professional domain system of knowledge based map as claimed in claim 1 or 2 is cold-started recommended method, feature
It is, in the step (3), user/article matching degree scoring is online task, and the tally set of given user u and article p, u are
The tally set of UL (u), p are PL (p), and process is as follows:
The link of (3-1) user subject: to label UL (u) [i] each in UL (u), searching for its corresponding node D (u) [i] in G,
Since the label of user is the sorting item in classified catalogue, D (u) [i] is standard nodes, the node that will be searched for u
Collection is denoted as D (u)；
(3-2) article entity link: to label PL (p) [i] each in PL (p), searching for its corresponding node D (p) [i] in G,
Since the label of article is keyword, D (p) [i] may be sky；If D (p) [i] is not sky, setting for D (p) [i] is set
Coefficient of reliability c (p) [i] is 1；If D (p) [i] is sky, word-based vector model WV search and PL (p) [i] semantic similarity are most
High entry lejCorresponding node as D (p) [i], and be arranged D (p) [i] confidence level coefficient c (p) [i] be PL (p) [i] and
lejSemantic similarity value, the node collection searched for p is denoted as D (p)；
(3-3) node matching degree calculates: the node D (p) [j] in node D (u) [i] and D (p) in given D (u), first in G
Middle search D (p) [j] arrives the shortest path SP of D (u) [i]ji, then, the section of D (u) [i] and D (p) [j] are calculated according to following formula
Point matching angle value；
(3-4) user/article matching degree calculates: the matching angle value of user u and article p, user u and object are calculated according to following formula
The matching angle value of product p is max { c (p) [1] × w24,c(p)[0]×w38×w35×w25}；
5. a kind of professional domain system of knowledge based map as claimed in claim 1 or 2 is cold-started recommended method, feature
Be, in the step (4), personalized recommendation is online task, and process is as follows: given user u is calculated all to be recommended first
The matching angle value of article and u, and by matching angle value descending arrangement；Then, several articles before the highest that sorts are recommended into u.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910811379.7A CN110427563B (en) | 2019-08-30 | 2019-08-30 | Professional field system cold start recommendation method based on knowledge graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910811379.7A CN110427563B (en) | 2019-08-30 | 2019-08-30 | Professional field system cold start recommendation method based on knowledge graph |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110427563A true CN110427563A (en) | 2019-11-08 |
| CN110427563B CN110427563B (en) | 2023-02-28 |
Family
ID=68416644
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910811379.7A Active CN110427563B (en) | 2019-08-30 | 2019-08-30 | Professional field system cold start recommendation method based on knowledge graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110427563B (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111310051A (en) * | 2019-12-13 | 2020-06-19 | 北京邮电大学 | Cold start solution based on knowledge graph |
| CN111522994A (en) * | 2020-04-15 | 2020-08-11 | 北京百度网讯科技有限公司 | Method and apparatus for generating information |
| CN111737587A (en) * | 2020-08-21 | 2020-10-02 | 北京每日优鲜电子商务有限公司 | Device operation method, device, electronic device and computer readable medium |
| CN112069423A (en) * | 2020-08-14 | 2020-12-11 | 京华信息科技股份有限公司 | Information recommendation method and device, storage medium and computer equipment |
| CN112069388A (en) * | 2020-09-02 | 2020-12-11 | 上海风秩科技有限公司 | Entity recommendation method, system, computer device and computer-readable storage medium |
| CN112070586A (en) * | 2020-09-09 | 2020-12-11 | 腾讯科技（深圳）有限公司 | Article recommendation method and device based on semantic recognition, computer equipment and medium |
| CN112287218A (en) * | 2020-10-26 | 2021-01-29 | 安徽工业大学 | Knowledge graph-based non-coal mine literature association recommendation method |
| CN112364180A (en) * | 2020-11-26 | 2021-02-12 | 上海天旦网络科技发展有限公司 | Label definition and calculation method and system based on knowledge graph |
| CN112364181A (en) * | 2020-11-27 | 2021-02-12 | 深圳市慧择时代科技有限公司 | Insurance product matching degree determination method and device |
| CN112507132A (en) * | 2020-12-14 | 2021-03-16 | 天津大学 | Recommendation method based on knowledge graph and attention mechanism |
| CN112528146A (en) * | 2020-12-09 | 2021-03-19 | 北京百度网讯科技有限公司 | Content resource recommendation method and device, electronic equipment and storage medium |
| CN112992362A (en) * | 2021-04-19 | 2021-06-18 | 北京妙医佳健康科技集团有限公司 | Body state evaluation device, electronic equipment and storage medium |
| CN112989215A (en) * | 2019-12-13 | 2021-06-18 | 复旦大学 | Knowledge graph enhanced recommendation system based on sparse user behavior data |
| WO2023179689A1 (en) * | 2022-03-24 | 2023-09-28 | 重庆邮电大学工业互联网研究院 | Knowledge graph-based recommendation method for internet of things |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103593792A (en) * | 2013-11-13 | 2014-02-19 | 复旦大学 | Individual recommendation method and system based on Chinese knowledge mapping |
| CN109255033A (en) * | 2018-11-05 | 2019-01-22 | 桂林电子科技大学 | A kind of recommended method of the knowledge mapping based on location-based service field |
| CN110046240A (en) * | 2019-04-16 | 2019-07-23 | 浙江爱闻格环保科技有限公司 | In conjunction with the target domain question and answer method for pushing of keyword retrieval and twin neural network |
-
2019
- 2019-08-30 CN CN201910811379.7A patent/CN110427563B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103593792A (en) * | 2013-11-13 | 2014-02-19 | 复旦大学 | Individual recommendation method and system based on Chinese knowledge mapping |
| CN109255033A (en) * | 2018-11-05 | 2019-01-22 | 桂林电子科技大学 | A kind of recommended method of the knowledge mapping based on location-based service field |
| CN110046240A (en) * | 2019-04-16 | 2019-07-23 | 浙江爱闻格环保科技有限公司 | In conjunction with the target domain question and answer method for pushing of keyword retrieval and twin neural network |
Non-Patent Citations (1)
| Title |
|---|
| 徐静等: "基于领域知识个性化推荐在微课平台中的应用", 《黑龙江工业学院学报(综合版)》 * |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112989215A (en) * | 2019-12-13 | 2021-06-18 | 复旦大学 | Knowledge graph enhanced recommendation system based on sparse user behavior data |
| CN111310051A (en) * | 2019-12-13 | 2020-06-19 | 北京邮电大学 | Cold start solution based on knowledge graph |
| CN112989215B (en) * | 2019-12-13 | 2022-06-14 | 复旦大学 | Sparse user behavior data-based knowledge graph enhanced recommendation system |
| CN111522994A (en) * | 2020-04-15 | 2020-08-11 | 北京百度网讯科技有限公司 | Method and apparatus for generating information |
| US11847164B2 (en) | 2020-04-15 | 2023-12-19 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Method, electronic device and storage medium for generating information |
| CN111522994B (en) * | 2020-04-15 | 2023-08-01 | 北京百度网讯科技有限公司 | Method and device for generating information |
| CN112069423A (en) * | 2020-08-14 | 2020-12-11 | 京华信息科技股份有限公司 | Information recommendation method and device, storage medium and computer equipment |
| CN111737587A (en) * | 2020-08-21 | 2020-10-02 | 北京每日优鲜电子商务有限公司 | Device operation method, device, electronic device and computer readable medium |
| CN112069388A (en) * | 2020-09-02 | 2020-12-11 | 上海风秩科技有限公司 | Entity recommendation method, system, computer device and computer-readable storage medium |
| CN112069388B (en) * | 2020-09-02 | 2023-07-21 | 上海风秩科技有限公司 | Entity recommendation method, system, computer device and computer readable storage medium |
| CN112070586A (en) * | 2020-09-09 | 2020-12-11 | 腾讯科技（深圳）有限公司 | Article recommendation method and device based on semantic recognition, computer equipment and medium |
| CN112070586B (en) * | 2020-09-09 | 2023-11-28 | 腾讯科技（深圳）有限公司 | Item recommendation method and device based on semantic recognition, computer equipment and medium |
| CN112287218A (en) * | 2020-10-26 | 2021-01-29 | 安徽工业大学 | Knowledge graph-based non-coal mine literature association recommendation method |
| CN112364180A (en) * | 2020-11-26 | 2021-02-12 | 上海天旦网络科技发展有限公司 | Label definition and calculation method and system based on knowledge graph |
| CN112364181A (en) * | 2020-11-27 | 2021-02-12 | 深圳市慧择时代科技有限公司 | Insurance product matching degree determination method and device |
| CN112364181B (en) * | 2020-11-27 | 2024-05-28 | 深圳市慧择时代科技有限公司 | Insurance product matching degree determining method and apparatus |
| CN112528146A (en) * | 2020-12-09 | 2021-03-19 | 北京百度网讯科技有限公司 | Content resource recommendation method and device, electronic equipment and storage medium |
| CN112528146B (en) * | 2020-12-09 | 2022-05-27 | 北京百度网讯科技有限公司 | Content resource recommendation method and device, electronic equipment and storage medium |
| CN112507132A (en) * | 2020-12-14 | 2021-03-16 | 天津大学 | Recommendation method based on knowledge graph and attention mechanism |
| CN112992362A (en) * | 2021-04-19 | 2021-06-18 | 北京妙医佳健康科技集团有限公司 | Body state evaluation device, electronic equipment and storage medium |
| WO2023179689A1 (en) * | 2022-03-24 | 2023-09-28 | 重庆邮电大学工业互联网研究院 | Knowledge graph-based recommendation method for internet of things |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110427563B (en) | 2023-02-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110427563A (en) | A kind of professional domain system cold start-up recommended method of knowledge based map | |
| CN106776711B (en) | Chinese medical knowledge map construction method based on deep learning | |
| CN106709040B (en) | Application search method and server | |
| CN106599032B (en) | Text event extraction method combining sparse coding and structure sensing machine | |
| CN107590133A (en) | The method and system that position vacant based on semanteme matches with job seeker resume | |
| JP6381775B2 (en) | Information processing system and information processing method | |
| CN103593474B (en) | Image retrieval sort method based on deep learning | |
| CN105824933A (en) | Automatic question answering system based on main statement position and implementation method thereof | |
| CN103577534B (en) | Searching method and search engine | |
| CN103995853A (en) | Multi-language emotional data processing and classifying method and system based on key sentences | |
| CN113553429A (en) | Normalized label system construction and text automatic labeling method | |
| CN104778283B (en) | A kind of user's occupational classification method and system based on microblogging | |
| CN109582788A (en) | Comment spam training, recognition methods, device, equipment and readable storage medium storing program for executing | |
| CN109325120A (en) | A kind of text sentiment classification method separating user and product attention mechanism | |
| KR101319413B1 (en) | Summary Information Generating System and Method for Review of Product and Service | |
| Korpusik et al. | Distributional semantics for understanding spoken meal descriptions | |
| CN106886565A (en) | A kind of basic house type auto-polymerization method | |
| CN111914157A (en) | Travel scheme generation method and system based on user preference | |
| CN109325115A (en) | A kind of role analysis method and analysis system | |
| CN114255096A (en) | Data requirement matching method and device, electronic equipment and storage medium | |
| Korpusik et al. | Data collection and language understanding of food descriptions | |
| CN110781300B (en) | Tourism resource culture characteristic scoring algorithm based on Baidu encyclopedia knowledge graph | |
| CN112015907A (en) | Method and device for quickly constructing discipline knowledge graph and storage medium | |
| CN108804416B (en) | Training method for film evaluation emotion tendency analysis based on machine learning | |
| CN111191045A (en) | Entity alignment method and system applied to knowledge graph |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |