CN107577759A - User comment auto recommending method - Google Patents
User comment auto recommending method Download PDFInfo
- Publication number
- CN107577759A CN107577759A CN201710779830.2A CN201710779830A CN107577759A CN 107577759 A CN107577759 A CN 107577759A CN 201710779830 A CN201710779830 A CN 201710779830A CN 107577759 A CN107577759 A CN 107577759A
- Authority
- CN
- China
- Prior art keywords
- comment
- user
- information
- preference
- useless
- 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
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of user comment auto recommending method, methods described step includes：Step 1：Global serviceability analysis and research are carried out to comment according to the behavior of comment content and commentator, filter out a part of useless comment；Step 2：Comment after being filtered to step 1 carries out commenting on relevant information excavation；Step 3：User personalized information is captured, and user individual intention and preference are excavated；Step 4：On the basis of step 2 and step 3, user view preference is matched with corresponding comment, selects the comment for meeting user's request preference；Step 5：The comment for recommending user is ranked up, to reach the purpose that the useful comment of acquisition is recommended to user according to the customized information of user.
Description
Technical field
The present invention relates to network comment screening recommendation field, and in particular to a kind of user comment auto recommending method.
Background technology
User comment is a kind of unique linguistic form for stating people's viewpoint, while is also that people realize the important of interaction
Medium, such as, people are often desirable to, by comment, obtain the understanding to objective things indirectly, and so as to instructing sentencing for itself
Disconnected and behavior.By the development of internet, the quantity and class of user comment are increasingly enriched, and transmit and to share speed all with day
Increase, people can realize efficient thinking collision by the subjective texts on network, and can be quick just by other people knowledge
Objective world is appreciated and understood sharply.However, the information of " who can write, and who can read " that free net assigns is derivative and transmits
Characteristic, a large amount of low-quality comments are caused, while automatically identify, organize and supervise because being short of artificial or machine, people are difficult
While low quality, uncorrelated even " harmfulness " (as " instigating " property is commented on) comment is shielded, acquisition meets that its knowledge obtains
Take meaning the high quality user comment of figure.At present, the recommended technology research of domestic and international user oriented comment is not goed deep into still, especially exists
Explore based on subjectivity feature and its combine query intention feature comment value it is automatic assess, and related commentary identification,
Match somebody with somebody, organize with recommendation in terms of research still belong to blank.
The content of the invention
In order to solve the above technical problems, the present invention proposes user comment auto recommending method, to reach according to user's
Customized information recommends the useful comment of acquisition the purpose of user.
To reach above-mentioned purpose, technical scheme is as follows：
A kind of user comment auto recommending method, including：
Step 1：Global serviceability analysis and research are carried out to comment according to the behavior of comment content and commentator, filter out one
The useless comment in part；
Step 2：Comment after being filtered to step 1 carries out commenting on relevant information excavation；
Step 3：User personalized information is captured, and user individual intention and preference are excavated；
Step 4：On the basis of step 2 and step 3, user view preference is matched with corresponding comment, selected
Meet the comment of user's request preference；
Step 5：The comment for recommending user is ranked up.
As preferable, the method that filtering useless described in step 1 is commented on is to be known by commentator with comment interaction
Do not go out useless comment, concretely comprise the following steps：According to the feature of comment, it is modeled using machine learning method, by enough
Language material carries out model training, judges and finds out useless comment, and the author of the comment is found by the useless comment, is then passed through
Author finds more useless comments, then continues to look for author using these useless comments, so continuous iteration, finally to nothing
Comment carries out comprehensive and accurate filtering.
As preferable, the comment after filtering described in step 2 is divided into the comment to commodity and the comment to event, institute's commentary
Being excavated by relevant information includes：In comment in the extraction and comment of item property event and its association attributes extraction.
As preferable, the abstracting method of item property is in the comment：First, a comment is selected, identifies that this is commented
By for merchandise classification, secondly, other generic some commodity are found according to the merchandise classification, then excavated to this
The descriptive text of a little commodity, subsequently using the keyword abstraction method of more documents, extracts this kind of business from these texts
The object that would generally be come into question, the attribute of product, keyword set is formed according to the object of extraction and attribute, eventually passes back to comment on,
Simple match goes out while appears in the word in comment neutralization keyword set, and these words are the object and object as the comment
Attribute.
As preferable, the abstracting method of event and its association attributes is in the comment：Original comment is carried out first
Text Pretreatment, next extracts some trigger words in training set as original trigger word, manually collects some passes herein
In give opinion, the trigger word of attitude, these trigger words are extended with synonymicon, then, with reference to Text Pretreatment
Result and trigger word spreading result candidate events are extracted, and obtain candidate categories, composition training example, bluebeard compound
Method feature, contextual feature, dictionary feature describes candidate events from different angles, and candidate events are carried out using grader
Binary classification, so as to obtain the classification of candidate events, finally, investigation is identified as text sentence of the event i.e. containing trigger word, obtains
Its all entities, time point etc. is taken to be used as candidate's element, class label is incited somebody to action as the template representation corresponding to event category
Event argument recognition is converted to multivariate classification problem and obtains Event element and its corresponding role.
As preferable, method that user personalized information is captured described in step 3 is heuristic information and rule to collect
The customized information of user, the customized information include demand and preference and user pair of the user for commodity and its attribute
The demand and preference of comment, the information for contributing to customized information to collect have：Comment that user oneself delivered, user are to other people
The historical information evaluated, buy commodity, the personal information that provides of user and other personal information and user browse
Behavioural information.
As preferable, user individual described in step 3 is intended to and preference carries out method for digging and included：Customized information
Processing is with organizing, the adaptive tracing of user personalized information and study, processing and the method for organizing of the customized information are：
User's request preference is designed to data structure, the data of the data structure include：ID as unique mark,<It is emerging
Interesting point：Interest-degree>, the comment write of this user and this user think helpful comment, the above-mentioned data of root square pass through engineering
Practise and data mining obtains user personalized information, adaptive tracing and the study of the user personalized information are to user
Customized information is regularly updated and real-time update.
As preferable, user view preference is included with the corresponding method matched of commenting on described in step 4：According to
The matching of content and the matching according to commentator and the relational network of reader are commented on, the matching process according to comment content is
Using machine learning method mapping relations model is established between the individuation data of user and recommended comment.So that system
Some comments matched with user are found, these comments are to be recommended to the comment of the user, described according to commentator and reading
The matching process of the relational network of person is to establish this cyberrelationship by machine learning method, can according to the commentator of comment
To evaluate the matching degree between user and comment.
The comment that user pair is recommended as preferable, described in step 5 is ranked up using RankSvm sort algorithms to choosing
Surely recommended comment is ranked up, and the feature being related in the sequence includes：What is obtained in global serviceability analysis comments
By mass fraction, user view and the feature used in corresponding comment and the authority of comment.
The invention has the advantages that：
(1) the present invention carries out going deep into detection to low quality comment, the global serviceability of comment is analyzed, by commenting
The evaluation method of theorist and comment interaction identifies useless comment and rejected, and has effectively been filtered by the above method useless
Comment.
(2) the present invention carries out the extraction of attribute to useful comment, recommends comment to provide effective support for user.
(3) the present invention carries out depth excavation to the customized information of user, analyzes the Demand perference of user, accurately
Provide the user valuable comment.
(4) the present invention matches user view preference with corresponding comment, then the comment after matching is ranked up, and obtains
The personalized comment of high quality is recommended.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
There is the required accompanying drawing used in technology description to be briefly described.
Fig. 1 is user comment auto recommending method flow chart disclosed in the embodiment of the present invention；
Fig. 2 is commentator disclosed in the embodiment of the present invention and the evaluation method flow chart of comment interaction；
Fig. 3 is right disclosed in the embodiment of the present invention<Point of interest：Interest-degree>The flow chart excavated；
Fig. 4 is object-oriented event extraction flow chart in the disclosed comment of the embodiment of the present invention.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation describes.
The invention provides a kind of its operation principle of user comment auto recommending method commented on and take out by filtering useless
The attribute of useful comment is taken, is that user recommends the personalized comment of high quality with reference to user personalized information, reaches and is accurately
The purpose of the personalized comment of lead referral high quality.
With reference to embodiment and embodiment, the present invention is further detailed explanation.
As Figure 1-Figure 4, a kind of user comment auto recommending method, including：
Step 1：Global serviceability analysis and research are carried out to comment according to the behavior of comment content and commentator, filter out one
The useless comment in part；
Step 2：Comment after being filtered to step 1 carries out commenting on relevant information excavation；
Step 3：User personalized information is captured, and user individual intention and preference are excavated；
Step 4：On the basis of step 2 and step 3, user view preference is matched with corresponding comment, selected
Meet the comment of user's request preference；
Step 5：The comment for recommending user is ranked up.
The method of step 1 filtering useless comment is identified using " commentator and the evaluation method of comment interaction "
Useless comment, its main thought are：Its author is found by useless comment, then found by author more useless
Comment, then continue to look for author using these useless comments, so continuous iteration, finally have to useless comment more comprehensive
Accurately filtering.
Specifically, we analyze useless commentator's manufacture, and corresponding to the feature of useless comment, they typically have this
A little features：
(A) tend to issue same section of text (or very much like text) in different places
(B) writing style is poor, not rigorous (corresponding to low-quality comment)
(C) comment of same person (or same class people), wherein having, quite a few is more extreme, and with the sight of main flow
Point tendency is runed counter to (such as deception, false comment producer).
One commentator some comments that he writes by correspondence, are expressed as：R=<author,reviews>.With reference to commentator
Comment serviceability appraisal procedure it is as shown in Figure 2.First, commentator is commented on it is corresponding, formed one with<author,
reviews>For the set S of element.In this set, the corresponding element of each commentator, its comment is this element
Attached attribute.Some comments in set have been previously identified as useless comment in addition.Then the element in set S is carried out
Cluster, feature used in cluster essentially from each element comment collection, including the content of comment, style, repeatability, whether
Useless comment etc. is had been labeled as, wherein whether being marked as useless this feature of comment has larger weight.To in set S
The cluster of element clusters according to the comment of commentator to commentator, and this process gathers the commentator with similar comment
Together, the classification of some commentators is obtained.Then, judge whether each commentator's classification is " useless comment producer " respectively
Classification, comment of the basis for estimation mainly in the category be marked as useless comment quantity and useless comment some are special
Sign.Class for being judged as useless comment producer, identifies and removes some most unlikely useless comments；For being determined
For the class of non-useless comment producer, the comment that wherein most probable is useless is identified, new is judged as result in formation of one
Useless comment set, using these mark increases to original corpus labeling, eventually pass back to starting point and restart.Whole process
It is the process of an iteration, when when operation result is stable or reaching iterations set in advance, you can stop iteration.
Step 2 comment is carried out relevant information excavate be divided into extraction and comment to item property in comment event and its
The extraction of association attributes, item property indicates the content topic that the comment is discussed in comment, is a kind of letter of more shallow-layer
They are identified with the following method by breath, the present invention：
(A) for a certain comment, the merchandise classification that this comment is directed to is identified first；
(B) other generic some commodity are found according to this merchandise classification；
(C) descriptive text to these commodity is excavated, these texts are often appeared in shopping website, it is assumed that dig
The text collection dug is D={ d1,d2,…,dn}。
(D) using the keyword abstraction technology of more documents, would generally being begged for for this kind of commodity is extracted from these texts
Object, the attribute of opinion.Assuming that being drawn into keyword set is combined into W={ w1,w2,…,wm}。
(E) comment is returned to, simple match goes out while appears in the word in comment neutralization keyword set, and these words are conduct
The object of the comment and the attribute of object, its weight are calculated using Boolean type weight or by TFIDF：
W (i)=TFi*log(N/DFi)
Wherein, W (i) represents weights of the attribute i in this comment；TFiRepresent the number that attribute i occurs in this comment；
DFiThe document frequency counted for attribute i in document sets D；N is word document sets number of documents.
Event and its association attributes extract being decomposed into following two aspects of the task in comment：
(A) identification of event trigger word and event category
(B) identification of Event element
Herein, event trigger word refers to the word for causing event to occur, and is the key character for determining event category；Event element
Refer to the participant of event, be the core of composition event, it constructs the framework of whole event with event trigger word；Event
Trigger word and Event element determine the classification and subclass of event.
Comment text has that some are different from general media event text, its distinguish essentially consist in occur in 1) text compared with
More emotion words, 2) can also give expression to some attitudes of commentator in event, 3) need to expression viewpoint, emotion this kind of thing
Part will be paid special attention to.Accordingly, the present invention intends solving above-mentioned two main tasks using machine learning algorithm, comprises the following steps that：
Text Pretreatment is carried out to original comment first, next extracts some trigger words in training set as original trigger word,
Manually collect herein some on giving opinion, the trigger word of attitude, these trigger words are extended with synonymicon, so
Afterwards, candidate events are extracted with reference to the result and trigger word spreading result of Text Pretreatment, and obtains candidate categories, group
Into training example, with reference to lexical characteristics, contextual feature, dictionary feature describes candidate events from different angles, and uses and divide
Class device carries out binary classification to candidate events, and so as to obtain the classification of candidate events, finally, investigation is identified as event and contained
The text sentence of trigger word, obtaining its all entities, time point etc. is used as candidate's element, and class label is as corresponding to event category
Template representation, therefore event argument recognition is converted into multivariate classification problem and obtains Event element and its corresponding role.
Step 3 captures user personalized information using heuristic information and rule to collect the customized information of user, this
In customized information include two aspect, be on the one hand demand and preference of the user for commodity and its attribute, be on the other hand
User indirectly and directly serves pushing away for personalized reviews respectively to the demand and preference of comment, the collection of the information of these two aspects
Recommend, and have similar aspect between them, be collected so these information are put together in the present invention, by people's
Web environment residing for the research and comment of cognitive law, it has been found that following information contributes to customized information to collect：
The information of direct exposition need preference：
(A) comment that user oneself delivered.In the case where low-quality comment is by filtering, it is clear that this class text is
Most accurate, directly reflection user is to the interest of comment and the information of preference.
(B) comment of the user to other people is evaluated.There is this in the comment of some shopping websites in short：" 59/61 people thinks
This comment is useful:", in the bottom of comment, there are two buttons to be used for allowing user to evaluate this serviceability commented on.It is right
In comment, there are 61 users to be evaluated, wherein having required for 59 users think that this comment is them, it is believed that
This comments on the demand and preference for directly reflecting this 59 users.
The information of indirect expression Demand perference：
(A) historical information of commodity is bought.Primarily with regard to the comment of commodity, user purchased for the comment that the present invention is directed to
The commodity history bought is also very important customized information.Using data mining technology by the commodity to having bought same
Feature mining is carried out in class commodity, different characteristic, the characteristic of purchased commodity commodity similar with other can be found, and these are believed
Breath reflects the demand and preference of user.
(B) personal information and other personal information that user provides.These information include sex, age, educational background, residence
(or by accessing the geographical position of IP acquisitions).These information are advantageous to understand the interest background of user, and for some specific
Field, user is clustered or classified using statistical information, comes the potential preference of auxiliary judgment user and intention.
(C) user browsing behavior information.These information are obtained by monitoring user in the behavior of Web page, are such as gathered
User a certain page residence time, the length of document, the URL addresses that user accesses, URL paths the data such as history, shape
Into journal file, the characteristic of user is summed up by analyzing the journal file.Research shows the web access of certain period of time
The stabilized interest of user is contained in daily record.This method is transparent to user, but the collection of user data generally require one section compared with
The long time.
User individual is intended to and preference is excavated and is divided into the processing of customized information and organizes and user personalized information
Adaptive tracing and study,
After it have collected various customized informations above, ensuing task is that these information are handled, and is excavated
Go out the demand and preference of user.
First, the present invention devises the data structure for representing user's request preference, and is stored in database：
Wherein, one user of " ID " unique mark；“<Point of interest:Interest-degree>" in, point of interest typicallys represent certain
The attribute of one commodity or commodity, characteristic, interest-degree represent interest level or desirability of this user to this point of interest；
The direct exposition need preference that " comment that this user is write " and " this user thinks helpful comment " is set forth above
Information.In this data structure, user "<Point of interest:Interest-degree>" need by means such as machine learning, data minings
Learnt on the described various customized informations of upper one section and excavation obtains.And for " comment that this user is write " and
" this user thinks helpful comment ", it is presently believed that wherein containing abundant information, user is not only expressed to commodity
Which attribute it is interested, also give expression to user to comment style, the preference of describing mode, therefore we retain original text
This, directly uses these texts in research below, various information is retained as far as possible.
The interest of user changes over time in general, and the demand of user show sometimes it is sudden,
In the present invention, intend being updated user individual data at following two aspects：
Regularly update：First, regular (such as every three days) collection user personalized information, legacy data collection is added,
Recalculate user data.Its weight of the more long raw information of history is smaller in the calculation, the interest-degree of corresponding point of interest
It can reduce therewith.The authority of historical information changes over time：
Wherein, W (inf) is current information inf authority；△TinfIt is information inf from being collected between current time
Every；α, β are relevant parameter, are adjusted in study, test.
In addition, the present invention is updated using some rules to the weight of some points of interest.Such as when user has bought one
Computer, then, after a period of time, he may can reduce to the degree of concern of computer, and the concern to computer peripheral equipment can
It is able to can improve.
Real-time update：The demand of user has sudden, jumping characteristic, especially searches, buys the situation of commodity in user
In.For example first day user has searched camera, but the second day information for having searched furniture.For such situation, the present invention exists
Using following solution：
(A) customized information of collection is classified according to theme, obtaining point of interest per one kind, (research above is
Realize)
(B) identify that active user is concerned about that class of things is other, this classification is one in the classification in previous step.
(C) interest-degree of point of interest corresponding to current class is accordingly improved, point of interest corresponding to remaining classification reduces.
Herein, the change of individuation data is just for current demand, interest, so above to individuation data
Change and do not need in write into Databasce.
Step 4 user view preference carries out matching process with corresponding comment and is divided into according to the matching of comment content and according to commenting
The matching of the relational network of theorist and reader, according to the matching process of comment content mainly using machine learning method in user
Individuation data and recommended comment between establish mapping relations model so that system finds matched with user some and commented
By these comments are to be recommended to the comment of the user, and problem is described as follows：
V=MatchF (user_personal_data, review)
Wherein user_personal_data represents the individuation data of a certain user；Review represent one it is to be matched
Comment, MathchF () are to assess this user to comment on the function for the degree that matches, it is necessary to pass through machine learning method with this
Training obtains；V is matching degree value, can be recommended which to be selected comment on given threshold.Mainly to search out herein
This function of MatchF, a kind of matching relationship model of the function representation.The problem is converted into binary classification problems by the present invention, is turned
Classification problem after changing is expressed as follows：
(A) classification sample：<user_personal_data,review>, i.e. the combination of user and comment.
(B) class categories：For each sample, its classification is " matching " or " non-matching ", and the classification of training examples is
It is known.
(C) feature that classification uses：Characteristic of division is formed by the combinations of features of three aspects, is respectively：
User_personal_data feature：Point of interest, related commentary (user writes or thought helpful comment)
Feature, such as comment on the event that occurs and its each attribute in the attribute referred in length, comment, comment.
Review feature：The event and its each attribute occurred in comment length, the attribute referred in commenting on, comment
Deng.
Relationship characteristic between User_personal_data and review：Similarity between related commentary and review
(text similarity, semantic similarity etc.).
Matching according to commentator and the relational network of reader is to establish this cyberrelationship, root by machine learning method
According to the commentator of comment, it can be estimated that the matching degree gone out between user and comment.
Step 5 is ranked up using RankSvm sort algorithms to the comment for recommending user to selecting recommended comment
It is ranked up.In the ranking, the feature being related to includes：
(A) the comment mass fraction obtained in global serviceability analysis；
(B) user view and some features used in corresponding comment；
(C) authority of comment, mainly obtained by the authority of commentator, the normalization of comment writing style.
Above-described is only the preferred embodiment of user comment auto recommending method disclosed in this invention, should be referred to
Go out, for the person of ordinary skill of the art, without departing from the concept of the premise of the invention, can also make some
Modification and improvement, these belong to protection scope of the present invention.
Claims (9)
1. a kind of user comment auto recommending method, it is characterised in that methods described step includes：
Step 1：Global serviceability analysis and research are carried out to comment according to the behavior of comment content and commentator, filter out a part
Useless comment；
Step 2：Comment after being filtered to step 1 carries out commenting on relevant information excavation；
Step 3：User personalized information is captured, and user individual intention and preference are excavated；
Step 4：On the basis of step 2 and step 3, user view preference is matched with corresponding comment, selects and meets
The comment of user's request preference；
Step 5：The comment for recommending user is ranked up.
2. user comment auto recommending method according to claim 1, it is characterised in that filtering useless is commented described in step 1
The method of opinion is to be interacted by commentator with comment to identify useless comment, is concretely comprised the following steps：According to the spy of comment
Sign, is modeled using machine learning method, is carried out model training by enough language materials, is judged and find out useless comment, leads to
The author that the comment is found in the useless comment is crossed, more useless comments are then found by author, then utilize these nothings
Continued to look for author with comment, so continuous iteration, comprehensive and accurate filtering finally is carried out to useless comment.
3. user comment auto recommending method according to claim 1, it is characterised in that commenting after filtering described in step 2
Comment by the comment being divided into commodity and to event, the comment relevant information, which is excavated, to be included：Item property takes out in comment
The extraction of event and its association attributes in taking and commenting on.
4. user comment auto recommending method according to claim 3, it is characterised in that item property in the comment
Abstracting method is：First, a comment is selected, identifies the merchandise classification that the comment is directed to, secondly, is looked for according to the merchandise classification
Some commodity generic to other, then excavate the descriptive text to these commodity, subsequently using the pass of more documents
Keyword abstracting method, the object that would generally be come into question, the attribute of this kind of commodity are extracted from these texts, according to pair of extraction
As forming keyword set with attribute, eventually pass back to comment on, simple match, which goes out while appears in comment, to be neutralized in keyword set
Word, these words i.e. as the object of the comment and the attribute of object.
5. user comment auto recommending method according to claim 3, it is characterised in that event and its phase in the comment
Close attribute abstracting method be：Text Pretreatment is carried out to original comment first, next extracts some triggerings in training set
Word as original trigger word, manually collect herein some on giving opinion, the trigger word of attitude, to these trigger words with same
Adopted word dictionary is extended, and then, candidate events are carried with reference to the result and trigger word spreading result of Text Pretreatment
Take, and obtain candidate categories, composition training example, with reference to lexical characteristics, contextual feature, dictionary feature is retouched from different angles
Candidate events are stated, and binary classification is carried out to candidate events using grader, so as to obtain the classification of candidate events, finally, are examined
Text sentence of the event of being identified as i.e. containing trigger word is examined, obtaining its all entities, time point etc. is used as candidate's element, classification
Event argument recognition is converted to multivariate classification problem and is obtained event member by label as the template representation corresponding to event category
Plain and its corresponding role.
6. user comment auto recommending method according to claim 1, it is characterised in that user is captured described in step 3
Property information method be that heuristic information and rule collect the customized information of user, the customized information includes use
Family, to the demand and preference of comment, contributes to what customized information was collected for the demand and preference of commodity and its attribute and user
Information has：The historical information being evaluated, buy commodity of the comment, user that user oneself delivered to other people, user
The personal information of offer and other personal information and user browsing behavior information.
7. user comment auto recommending method according to claim 1, it is characterised in that user individual described in step 3
It is intended to and preference carries out method for digging and included：The processing of customized information and tissue, the adaptive tracing of user personalized information
With study, processing and the method for organizing of the customized information are：User's request preference is designed to data structure, the data
The data of structure include：ID as unique mark,<Point of interest：Interest-degree>, the comment write of this user and this user
Think helpful comment, the above-mentioned data of root square obtain user personalized information, the use by machine learning and data mining
The adaptive tracing of family customized information and study are the customized information of user to be regularly updated and real-time update.
8. user comment auto recommending method according to claim 1, it is characterised in that by user view described in step 4
Preference includes with the corresponding method matched of commenting on：According to the matching of comment content and the relation according to commentator and reader
The matching of network, it is described according to comment content matching process be using machine learning method user individuation data and by
Mapping relations model is established between the comment of recommendation.So that system finds some comments matched with user, these comments are
The comment of the user is recommended to, the matching process according to commentator and the relational network of reader is by machine learning side
Method establishes this cyberrelationship, according to the commentator of comment, it can be estimated that the matching degree gone out between user and comment.
9. user comment auto recommending method according to claim 1, it is characterised in that pair recommend use described in step 5
The comment at family is ranked up to be ranked up using RankSvm sort algorithms to selecting recommended comment, is related in the sequence
And to feature include：The comment mass fraction that is obtained in the analysis of global serviceability, user view arrive with use in corresponding comment on
Feature and comment authority.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710779830.2A CN107577759B (en) | 2017-09-01 | 2017-09-01 | Automatic recommendation method for user comments |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710779830.2A CN107577759B (en) | 2017-09-01 | 2017-09-01 | Automatic recommendation method for user comments |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107577759A true CN107577759A (en) | 2018-01-12 |
| CN107577759B CN107577759B (en) | 2021-07-30 |
Family
ID=61030842
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710779830.2A Active CN107577759B (en) | 2017-09-01 | 2017-09-01 | Automatic recommendation method for user comments |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107577759B (en) |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108694647A (en) * | 2018-05-11 | 2018-10-23 | 北京三快在线科技有限公司 | A kind of method for digging and device of trade company's rationale for the recommendation, electronic equipment |
| CN109189892A (en) * | 2018-09-17 | 2019-01-11 | 北京点网聚科技有限公司 | A kind of recommended method and device based on article review |
| CN109189922A (en) * | 2018-08-07 | 2019-01-11 | 阿里巴巴集团控股有限公司 | Comment on the training method and device of assessment models |
| CN109344241A (en) * | 2018-09-25 | 2019-02-15 | Oppo广东移动通信有限公司 | Recommended method, device, terminal and the storage medium of information |
| CN109410011A (en) * | 2018-11-05 | 2019-03-01 | 深圳市鹏朗贸易有限责任公司 | Adopt the recommended method, terminal and computer readable storage medium of product |
| CN109753651A (en) * | 2018-12-14 | 2019-05-14 | 昆明理工大学 | It is a kind of to comment on method for digging for the APP software users for embodying user's intention |
| CN109933648A (en) * | 2019-02-28 | 2019-06-25 | 北京学之途网络科技有限公司 | A kind of differentiating method and discriminating device of real user comment |
| CN110061900A (en) * | 2018-01-18 | 2019-07-26 | 腾讯科技（深圳）有限公司 | Message display method, device, terminal and computer readable storage medium |
| CN110110084A (en) * | 2019-04-23 | 2019-08-09 | 北京科技大学 | The recognition methods of high quality user-generated content |
| CN110858378A (en) * | 2018-08-23 | 2020-03-03 | 阿里巴巴集团控股有限公司 | Evaluation information display method, order creation method, purchase adding method and purchase adding device |
| CN110929021A (en) * | 2018-08-31 | 2020-03-27 | 阿里巴巴集团控股有限公司 | Text information generating method and text information generating device |
| CN110990700A (en) * | 2019-11-29 | 2020-04-10 | 北京达佳互联信息技术有限公司 | Comment information display method, device, client, server and system |
| CN111666413A (en) * | 2020-06-09 | 2020-09-15 | 重庆邮电大学 | Commodity comment recommendation method based on reviewer reliability regression prediction |
| CN111930927A (en) * | 2020-08-21 | 2020-11-13 | 腾讯科技（深圳）有限公司 | Evaluation information display method and device, electronic equipment and readable storage medium |
| WO2021008025A1 (en) * | 2019-07-18 | 2021-01-21 | 平安科技（深圳）有限公司 | Speech recognition-based information analysis method and apparatus, and computer device |
| CN112528141A (en) * | 2020-12-03 | 2021-03-19 | 中山大学 | Personalized comment text recommendation method based on multitask learning |
| CN112801745A (en) * | 2021-02-02 | 2021-05-14 | 李海涛 | Big data platform based online comment validity recommendation method |
| CN112818082A (en) * | 2019-11-15 | 2021-05-18 | 北京沃东天骏信息技术有限公司 | Evaluation text pushing method and device |
| CN113240526A (en) * | 2020-12-14 | 2021-08-10 | 王玉华 | Service recommendation method applied to cloud computing and block chain finance and cloud computing system |
| CN113342221A (en) * | 2021-05-13 | 2021-09-03 | 北京字节跳动网络技术有限公司 | Comment information guiding method and device, storage medium and electronic equipment |
| CN113553421A (en) * | 2021-06-22 | 2021-10-26 | 北京百度网讯科技有限公司 | Comment text generation method and device, electronic equipment and storage medium |
| WO2022021400A1 (en) * | 2020-07-31 | 2022-02-03 | 深圳齐心集团股份有限公司 | E-commerce comment identification and marking system |
| CN110929021B (en) * | 2018-08-31 | 2024-05-31 | 阿里巴巴集团控股有限公司 | Text information generation method and text information generation device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110113041A1 (en) * | 2008-10-17 | 2011-05-12 | Louis Hawthorne | System and method for content identification and customization based on weighted recommendation scores |
| CN102467709A (en) * | 2010-11-17 | 2012-05-23 | 阿里巴巴集团控股有限公司 | Product information sending method and device |
| US20130332382A1 (en) * | 2003-08-04 | 2013-12-12 | Skill Survey, Inc. | System and method for evaluating job candidates |
| CN103778214A (en) * | 2014-01-16 | 2014-05-07 | 北京理工大学 | Commodity property clustering method based on user comments |
| CN103823893A (en) * | 2014-03-11 | 2014-05-28 | 北京大学 | User comment-based product search method and system |
| CN106126648A (en) * | 2016-06-23 | 2016-11-16 | 华南理工大学 | A kind of based on the distributed merchandise news reptile method redo log |
-
2017
- 2017-09-01 CN CN201710779830.2A patent/CN107577759B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130332382A1 (en) * | 2003-08-04 | 2013-12-12 | Skill Survey, Inc. | System and method for evaluating job candidates |
| US20110113041A1 (en) * | 2008-10-17 | 2011-05-12 | Louis Hawthorne | System and method for content identification and customization based on weighted recommendation scores |
| CN102467709A (en) * | 2010-11-17 | 2012-05-23 | 阿里巴巴集团控股有限公司 | Product information sending method and device |
| CN103778214A (en) * | 2014-01-16 | 2014-05-07 | 北京理工大学 | Commodity property clustering method based on user comments |
| CN103823893A (en) * | 2014-03-11 | 2014-05-28 | 北京大学 | User comment-based product search method and system |
| CN106126648A (en) * | 2016-06-23 | 2016-11-16 | 华南理工大学 | A kind of based on the distributed merchandise news reptile method redo log |
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110061900A (en) * | 2018-01-18 | 2019-07-26 | 腾讯科技（深圳）有限公司 | Message display method, device, terminal and computer readable storage medium |
| CN108694647A (en) * | 2018-05-11 | 2018-10-23 | 北京三快在线科技有限公司 | A kind of method for digging and device of trade company's rationale for the recommendation, electronic equipment |
| CN109189922B (en) * | 2018-08-07 | 2021-06-29 | 创新先进技术有限公司 | Comment evaluation model training method and device |
| CN109189922A (en) * | 2018-08-07 | 2019-01-11 | 阿里巴巴集团控股有限公司 | Comment on the training method and device of assessment models |
| CN110858378A (en) * | 2018-08-23 | 2020-03-03 | 阿里巴巴集团控股有限公司 | Evaluation information display method, order creation method, purchase adding method and purchase adding device |
| CN110858378B (en) * | 2018-08-23 | 2023-07-28 | 浙江天猫技术有限公司 | Evaluation information display method, order creation method, purchasing method and device |
| CN110929021B (en) * | 2018-08-31 | 2024-05-31 | 阿里巴巴集团控股有限公司 | Text information generation method and text information generation device |
| CN110929021A (en) * | 2018-08-31 | 2020-03-27 | 阿里巴巴集团控股有限公司 | Text information generating method and text information generating device |
| CN109189892A (en) * | 2018-09-17 | 2019-01-11 | 北京点网聚科技有限公司 | A kind of recommended method and device based on article review |
| CN109344241A (en) * | 2018-09-25 | 2019-02-15 | Oppo广东移动通信有限公司 | Recommended method, device, terminal and the storage medium of information |
| CN109410011A (en) * | 2018-11-05 | 2019-03-01 | 深圳市鹏朗贸易有限责任公司 | Adopt the recommended method, terminal and computer readable storage medium of product |
| CN109753651A (en) * | 2018-12-14 | 2019-05-14 | 昆明理工大学 | It is a kind of to comment on method for digging for the APP software users for embodying user's intention |
| CN109753651B (en) * | 2018-12-14 | 2022-05-17 | 昆明理工大学 | APP software user comment mining method for embodying user intention |
| CN109933648A (en) * | 2019-02-28 | 2019-06-25 | 北京学之途网络科技有限公司 | A kind of differentiating method and discriminating device of real user comment |
| CN110110084A (en) * | 2019-04-23 | 2019-08-09 | 北京科技大学 | The recognition methods of high quality user-generated content |
| WO2021008025A1 (en) * | 2019-07-18 | 2021-01-21 | 平安科技（深圳）有限公司 | Speech recognition-based information analysis method and apparatus, and computer device |
| CN112818082A (en) * | 2019-11-15 | 2021-05-18 | 北京沃东天骏信息技术有限公司 | Evaluation text pushing method and device |
| CN110990700A (en) * | 2019-11-29 | 2020-04-10 | 北京达佳互联信息技术有限公司 | Comment information display method, device, client, server and system |
| CN111666413A (en) * | 2020-06-09 | 2020-09-15 | 重庆邮电大学 | Commodity comment recommendation method based on reviewer reliability regression prediction |
| WO2022021400A1 (en) * | 2020-07-31 | 2022-02-03 | 深圳齐心集团股份有限公司 | E-commerce comment identification and marking system |
| CN111930927A (en) * | 2020-08-21 | 2020-11-13 | 腾讯科技（深圳）有限公司 | Evaluation information display method and device, electronic equipment and readable storage medium |
| CN111930927B (en) * | 2020-08-21 | 2024-03-12 | 腾讯科技（深圳）有限公司 | Evaluation information display method and device, electronic equipment and readable storage medium |
| CN112528141A (en) * | 2020-12-03 | 2021-03-19 | 中山大学 | Personalized comment text recommendation method based on multitask learning |
| CN113240526A (en) * | 2020-12-14 | 2021-08-10 | 王玉华 | Service recommendation method applied to cloud computing and block chain finance and cloud computing system |
| CN112801745A (en) * | 2021-02-02 | 2021-05-14 | 李海涛 | Big data platform based online comment validity recommendation method |
| CN113342221A (en) * | 2021-05-13 | 2021-09-03 | 北京字节跳动网络技术有限公司 | Comment information guiding method and device, storage medium and electronic equipment |
| CN113553421A (en) * | 2021-06-22 | 2021-10-26 | 北京百度网讯科技有限公司 | Comment text generation method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107577759B (en) | 2021-07-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107577759A (en) | User comment auto recommending method | |
| Kim et al. | A scientometric review of emerging trends and new developments in recommendation systems | |
| Ortega | Academic search engines: A quantitative outlook | |
| Mei et al. | A probabilistic approach to spatiotemporal theme pattern mining on weblogs | |
| Haustein | Multidimensional journal evaluation: Analyzing scientific periodicals beyond the impact factor | |
| US9659084B1 (en) | System, methods, and user interface for presenting information from unstructured data | |
| CN106339502A (en) | Modeling recommendation method based on user behavior data fragmentation cluster | |
| Song et al. | Identification of ambiguous queries in web search | |
| Hou et al. | Newsminer: Multifaceted news analysis for event search | |
| CN103853824A (en) | In-text advertisement releasing method and system based on deep semantic mining | |
| JP2010055618A (en) | Method and system for providing search based on topic | |
| CN110705288A (en) | Big data-based public opinion analysis system | |
| Godoy et al. | Interface agents personalizing Web-based tasks | |
| CN110263169A (en) | A kind of focus incident detection method based on convolutional neural networks and keyword clustering | |
| Gong et al. | Phrase-based hashtag recommendation for microblog posts. | |
| CN104809253B (en) | Internet data analysis system | |
| Rajkumar et al. | Users’ click and bookmark based personalization using modified agglomerative clustering for web search engine | |
| Hsu et al. | Mining various semantic relationships from unstructured user-generated web data | |
| Lalitha et al. | Potential Web Content Identification and Classification System using NLP and Machine Learning Techniques | |
| Nguyen et al. | Pagerank-based approach on ranking social events: a case study with flickr | |
| Nguyen et al. | Vnu-smm: A social media monitoring framework on vietnamese online news | |
| Abuoda et al. | Automatic Tag Recommendation for the UN Humanitarian Data Exchange. | |
| Said et al. | Hotel reviews analysis based on sentiment classification: Oman case study | |
| Christensen et al. | News Hunter: a semantic news aggregator | |
| Weiying et al. | Topic Modelling for Malay News Aggregator |
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 |