CN109800307A - Analysis method, device, computer equipment and the storage medium of product evaluation - Google Patents
Analysis method, device, computer equipment and the storage medium of product evaluation Download PDFInfo
- Publication number
- CN109800307A CN109800307A CN201910046997.7A CN201910046997A CN109800307A CN 109800307 A CN109800307 A CN 109800307A CN 201910046997 A CN201910046997 A CN 201910046997A CN 109800307 A CN109800307 A CN 109800307A
- Authority
- CN
- China
- Prior art keywords
- vector
- term vector
- target
- data
- speech
- 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
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of analysis methods of product evaluation, device, computer equipment and storage medium, the described method includes: obtaining the data to be analyzed of application product, word segmentation processing is carried out to data to be analyzed, obtain target participle, and construct the target term vector of target participle, and target is segmented and carries out part-of-speech tagging, obtain the part of speech of target term vector, substatement vector sum sentence vector is constructed using target term vector, in the corresponding clustering cluster of every kind of part of speech in preset semantic analysis model, calculate the weight coefficient of each target term vector, according to the weight coefficient of target term vector, calculate the score of each substatement vector and the appraisal result of sentence vector, user preferences degree corresponding with appraisal result is inquired from score mapping table, so that it is determined that fancy grade of the user to application product.The embodiment of the present invention can rapidly and accurately obtain the degree of recognition of the comment to application product of user, improve the analysis accuracy rate of the comment to application product.
Description
Technical field
The present invention relates to data analysis technique field more particularly to a kind of analysis methods of product evaluation, device, computer
Equipment and storage medium.
Background technique
In the application market of each application software, user can carry out application software after downloading application software use
Evaluation, the content of user's evaluation can embody the advantage and disadvantage of application software, be the direct mode for obtaining the feedback information of user, raw
Business men can get the design advantage of application software from the comment of user, and need the shortcomings that improving.Currently, usually making
It is analyzed with comment content of the sentiment classification model to user, still, the analysis accuracy rate of sentiment classification model is not high, and
Sentiment analysis model can only judge that user is to hold front or negative attitude to application software, be unable to specific manifestation and go out user
Positive and passive degree, therefore, the effect analyzed at present to the user comment in application market is poor.
Summary of the invention
Analysis method, device, computer equipment and the storage medium of a kind of product evaluation are provided in the embodiment of the present invention, with
It solves the problems, such as currently low to the analysis accuracy rate of the comment of application software.
A kind of analysis method of product evaluation, comprising:
Obtain the data to be analyzed of application product;
Word segmentation processing is carried out to the data to be analyzed, obtains target participle;
The target term vector of the target participle is constructed, and the target is segmented and carries out part-of-speech tagging, obtains each institute
State the part of speech of target term vector;
Putting in order in the data to be analyzed is segmented according to the target, segments corresponding institute using the target
It states target term vector and constructs n substatement vector sum sentence vector, wherein n is positive integer;
According to the corresponding K clustering cluster of every kind of part of speech and each clustering cluster pair in preset semantic analysis model
The weight coefficient answered calculates the phase between the target term vector and the clustering cluster according to the part of speech of the target term vector
Like degree, and weight coefficient corresponding with the maximum clustering cluster of the similarity of the target term vector is obtained as the target
The weight coefficient of term vector, wherein K is positive integer;
According to the weight coefficient of the target term vector, the score of each substatement vector is calculated, and according to n institute
The score for stating substatement vector calculates the appraisal result of the sentence vector;
According to the appraisal result of the sentence vector, inquired from preset score mapping table corresponding with the appraisal result
User preferences degree, and using the user preferences degree as the analysis result of the data to be analyzed.
A kind of analytical equipment of product evaluation, comprising:
Data acquisition module, for obtaining the data to be analyzed of application product;
Word segmentation processing module obtains target participle for carrying out word segmentation processing to the data to be analyzed;
Term vector constructs module, for constructing the target term vector of the target participle, and segments and carries out to the target
Part-of-speech tagging obtains the part of speech of each target term vector;
Sentence vector constructs module, for segmenting putting in order in the data to be analyzed according to the target, makes
N substatement vector sum sentence vector is constructed with the corresponding target term vector of target participle, wherein n is positive integer;
Weight Acquisition module is used for according to the corresponding K clustering cluster of every kind of part of speech in preset semantic analysis model, and
The corresponding weight coefficient of each clustering cluster calculates the target term vector and institute according to the part of speech of the target term vector
It states the similarity between clustering cluster, and obtains weight corresponding with the maximum clustering cluster of the similarity of the target term vector
Weight coefficient of the coefficient as the target term vector, wherein K is positive integer;
Score computing module, for the weight coefficient according to the target term vector, calculates each substatement vector
Score, and calculate according to the score of n substatement vectors the appraisal result of the sentence vector;
As a result output module is inquired from preset score mapping table for the appraisal result according to the sentence vector
User preferences degree corresponding with the appraisal result, and using the user preferences degree as the analysis of the data to be analyzed
As a result.
A kind of computer equipment, including memory, processor and storage are in the memory and can be in the processing
The computer program run on device, the processor realize the analysis side of the said goods evaluation when executing the computer program
Method.
A kind of computer readable storage medium, the computer-readable recording medium storage have computer program, the meter
Calculation machine program realizes the analysis method of the said goods evaluation when being executed by processor.
The said goods evaluation analysis method, device, computer equipment and storage medium, by obtain application product to
Data are analyzed, word segmentation processing is carried out to data to be analyzed, obtain target participle, and construct the target term vector of target participle, with
And target is segmented and carries out part-of-speech tagging, the part of speech of each target term vector is obtained, constructs number to be analyzed using target term vector
According to substatement vector sum sentence vector, in the clustering cluster corresponding to each part of speech in preset semantic analysis model, meter
The weight coefficient of each target term vector is calculated, finally, calculating each substatement vector according to the weight coefficient of target term vector
The appraisal result of score and sentence vector inquires user preferences degree corresponding with appraisal result from score mapping table, thus
Determine user to the fancy grade of application product.It is calculated by using trained semantic analysis model every in data to be analyzed
A target segments corresponding weight coefficient, and segments the scoring knot that corresponding weight coefficient finds out data to be analyzed according to target
Fruit can quickly and accurately obtain the degree of recognition of the comment to application product of user, embody the passiveness of comment data
Degree or aggressiveness level improve the analysis efficiency and analysis accuracy rate of the comment to application product.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below by institute in the description to the embodiment of the present invention
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention
Example, for those of ordinary skill in the art, without any creative labor, can also be according to these attached drawings
Obtain other attached drawings.
Fig. 1 is an application environment schematic diagram of the analysis method of product evaluation in one embodiment of the invention;
Fig. 2 is a flow chart of the analysis method of product evaluation in one embodiment of the invention;
Fig. 3 be score in the analysis method of product evaluation data to be analyzed in one embodiment of the invention it is one specific
Flow chart;
Fig. 4 is the specific stream for storing data into database in one embodiment of the invention in the analysis method of product evaluation
Cheng Tu;
Fig. 5 is the detailed process for constructing semantic analysis model in one embodiment of the invention in the analysis method of product evaluation
Figure;
Fig. 6 is the detailed process clustered in the analysis method of product evaluation to term vector in one embodiment of the invention
Figure;
Fig. 7 is a functional block diagram of the analytical equipment of product evaluation in one embodiment of the invention;
Fig. 8 is a schematic diagram of computer equipment in one embodiment of the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are some of the embodiments of the present invention, instead of all the embodiments.Based on this hair
Embodiment in bright, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, shall fall within the protection scope of the present invention.
The analysis method of product evaluation provided by the embodiments of the present application can be applicable in the application environment such as Fig. 1, the application
Environment includes server-side and client, wherein is attached between server-side and client by network, user passes through client
The evaluation to application product is inputted, the evaluation that server-side obtains user is used as band analysis data, and uses preset semantic analysis
Model scores to data to be analyzed, obtains user to the user preferences degree of application product.Client specifically can with but not
It is limited to be various personal computers, laptop, smart phone, tablet computer and portable wearable device, server-side tool
Body can be realized with the server cluster that independent server or multiple servers form.The embodiment of the present invention provides product and comments
The method of the analysis of valence is applied to server-side.
In one embodiment, Fig. 2 shows a flow chart of the analysis method of product evaluation in the present embodiment, this method applications
Server-side in Fig. 1 is analyzed for the data to be analyzed to application product, is obtained user and is liked to the user of application product
Good degree.As shown in Fig. 2, the analysis method of the product evaluation includes step S1 to step S7, details are as follows:
S1: the data to be analyzed of application product are obtained.
In the present embodiment, data to be analyzed are evaluation data of the user to application product, and user is in application product platform
After downloading application product, for example, the application products platform such as pea pods or Android market, user can be according to the downloading of application product
The application product to be commented on the case where use, the evaluation data that service provider need to obtain the application product are analyzed, with
Just further application product is improved.
Specifically, the data to be analyzed of application product, can be that user inputs after client downloads application product comments
Valence mumber evidence is also possible to the evaluation data that server-side is obtained from internet, and specific data source can be according to actual needs
It is configured, herein with no restrictions.
S2: carrying out word segmentation processing to data to be analyzed, obtains target participle.
Specifically, server-side be based on preset corpus, to data to be analyzed carry out word segmentation processing, by data to be analyzed into
At least one target participle is obtained after row cutting, which is the data preset for storing Chinese word
Library, wherein preset corpus includes the word of standard and the part of speech of word, and the content in preset corpus specifically can be from
It is obtained in the Chinese vocabulary banks such as " the huge Chinese wordbank of millions " of open source, " stammerer Chinese word segmentation library " or " new edition Beijing University mark collection "
It takes.
It is to be appreciated that being conducive to the discriminance analysis of machine mould after carrying out word segmentation processing to data to be analyzed, obtaining
Target participle granularity it is smaller so that the accuracy of analysis of machine mould is higher.
S3: building target participle target term vector, and to target segment carry out part-of-speech tagging, obtain each target word to
The part of speech of amount.
Wherein, term vector, which refers to, uses the describing mode of mathematics or the describing mode table of machine language for Chinese word
Show, so as to representation language in a computer, and computer program identifying processing can be allowed.Target term vector in the present embodiment is to use
The form of vector indicates target participle.
Specifically, server-side can be used term vector generation technique and target participle be converted to target term vector, and building is every
A target segments corresponding term vector, for example, the target term vector that target participle " product " building obtains is < 0.3,0.5,0.1,
0.1 >, which specifically can be one-hot coding techniques, be also possible to word2vec (word to
Vector, word converting vector) etc. natural language processing techniques.
According to the part of speech of word in preset corpus, each target is segmented and carries out part-of-speech tagging, so that each target
Participle carries corresponding part of speech, and the target term vector generated by target participle obtains each mesh with corresponding part of speech
Mark the part of speech of term vector, wherein part of speech refers to adverbial word, adjective, noun and pronoun etc., for example, the word of target participle " beauty "
Property be " adjective ".
By preset corpus, the target term vector of each target participle is constructed, it can not accurate understanding by computer
Text conversion easily identifies the term vector with operation at machine, and carries out part-of-speech tagging processing to target term vector, so as to
It is enough to be analyzed according to semanteme of the part of speech to word, the degree of recognition to application product for including in the data to be analyzed is obtained,
Accurately identify the thought expressed in data to be analyzed.
S4: putting in order in data to be analyzed is segmented according to target, segments corresponding target term vector using target
N substatement vector is constructed, and sentence vector is constructed according to n substatement vector, wherein n is positive integer.
In the present embodiment, data to be analyzed refer to the evaluation data that user comments on application product, evaluation
The corresponding substatement of each sentence in data is analysed to data according to the point number in the punctuation mark of data to be analyzed
It is divided into n substatement, wherein point number is intended to indicate that the punctuation mark that the sentence in data to be analyzed pauses, point number include
Fullstop (.), question mark (?), exclamation mark (!), comma (), branch (;) and colon (:), the specific value of n is equal to the evaluation of user
The quantity of sentence in data indicates that, comprising 2 sentences in the data to be analyzed, n is also possible to for example, n specifically can be 2
3, indicate that the data to be analyzed include 3 sentences.
Specifically, it puts in order according to word of the target participle in data to be analyzed, segments corresponding mesh using target
It marks term vector and constructs n substatement vector, obtained substatement vector can indicate are as follows: SubStenP={ [SubStenId]
[W1][W2]…[Wj], wherein SubStenP is substatement vector, and SubStenId is unique identification in data to be analyzed
The identifier of sentence, for distinguishing the substatement in data to be analyzed, WjIndicate j-th target segment corresponding target word to
Amount.
Put in order according to sentence of the n substatement in data to be analyzed, using the corresponding substatement of n substatement to
Amount constructs the corresponding sentence vector of data to be analyzed, and data to be analyzed are indicated by way of vector, and the data to be analyzed are corresponding
One sentence vector, also, the sentence vector can specifically indicate are as follows: SenP={ [SenPId] [SubStenP1]
[SubStenP2]…[SubStenPn], wherein SenP is sentence vector, and SenPId is the mark of unique identification data to be analyzed
Symbol, for the data to be analyzed of Division identification, SubStenPnFor the corresponding substatement of n-th of substatement in data to be analyzed to
Amount.
For example, in one embodiment, one got data to be analyzed are " well, useful!", it is to be analyzed to this
After data are segmented, available target participle is " good ", " fine " and " use ", according to target participle construct target word to
Amount, also, target segments corresponding target term vector and is respectively as follows: " good "=(0.2,0.5,1.1,0.7,1.3), and " fine "=
(1.2,2.5,1.8,0.8,0.3), " use "=(1,0.4,0.8,1.1,0.9).
According to the point number in data to be analyzed, data can be analysed to and be divided into 2 substatements, and segmented according to target
Word in data to be analyzed puts in order, and segments corresponding target term vector using target and constructs substatement vector, obtains
Substatement vector be respectively as follows: " good "=SubStenP1={ [SubSten1] (0.2,0.5,1.1,0.7,1.3) }, " very well
With "=SubStenP2={ [SubSten2] (1.2,2.5,1.8,0.8,0.3) (1,0.4,0.8,1.1,0.9) }.
It is put in order according to sentence of the substatement in data to be analyzed, the language constructed using 2 substatement vectors
Sentence vector are as follows: " it is good, it is useful!"=SenP={ [SenP1] [SubStenP1][SubStenP2]}。
S5: corresponding according to the corresponding K clustering cluster of every kind of part of speech in preset semantic analysis model and each clustering cluster
Weight coefficient calculate the similarity between target term vector and clustering cluster according to the part of speech of target term vector, and obtain and mesh
Mark weight coefficient of the corresponding weight coefficient of the maximum clustering cluster of similarity of term vector as target term vector, wherein K is positive
Integer.
In the present embodiment, preset semantic analysis model is preparatory trained machine learning model, preset semanteme
Analysis model includes that the corresponding K clustering cluster of every kind of part of speech and each clustering cluster set corresponding weight coefficient, wherein
Part of speech in preset semantic analysis model includes but is not limited to adverbial word, adjective, noun and pronoun etc., and K is positive integer, and K can
To be 5, it is also possible to 6, can be specifically configured according to the actual situation.
Specifically, according to the part of speech of target term vector, in the part of speech with target term vector of preset semantic analysis model
In corresponding clustering cluster, the similarity between target term vector each clustering cluster corresponding with the part of speech, acquisition and target are calculated
Weight coefficient of the corresponding weight coefficient of the maximum clustering cluster of the similarity of term vector as target term vector, obtains each target
The weight coefficient of term vector, if similarity is bigger, then it is assumed that the space length of two objects is closer, and the semantic of word gets over phase
Seemingly.
S6: according to the weight coefficient of target term vector, the score of each substatement vector is calculated, and according to n substatement
The appraisal result of the score computing statement vector of vector.
Specifically, according to the weight coefficient of the obtained target term vector of step S5, to the target in each substatement vector
The weight coefficient of term vector carries out quadrature calculation process, calculates the score of each substatement vector, obtains to n substatement vector
Divide and carries out summation operation processing, the appraisal result of computing statement vector, so as to be divided into granularity by being analysed to data
Smaller participle analyzes the semanteme of each participle, and identification obtains the feature of semanteme of each participle, then by all participles
The feature of semanteme combine, obtain the appraisal result of sentence vector.
Wherein, weight coefficient includes positive number and negative, and positive number indicates commendatory term, is the expression positive state of user in user comment
The word of degree, negative number representation derogatory term are the words that user's passive attitude is indicated in user comment.The appraisal result of sentence vector
It indicates that the feature of semanteme of data to be analyzed, appraisal result are positive number, indicates that the data to be analyzed are positive comment, appraisal result
For negative, indicate that the data to be analyzed are passive comments, the numerical value of appraisal result is bigger, indicates user to the product of application product
The degree of pole attitude or passive attitude is higher.
S7: according to the appraisal result of sentence vector, use corresponding with appraisal result is inquired from preset score mapping table
Family fancy grade, and using user preferences degree as the analysis result of data to be analyzed.
Specifically, preset score mapping table is the user preferences journey for storing appraisal result and user to application product
The tables of data of corresponding relationship between degree can be inquired according to the appraisal result of sentence vector from preset score mapping table
User is obtained to the user preferences degree of application product, and using user preferences degree as the analysis of data to be analyzed as a result, should
The mapping relations of preset score mapping table storage can be specifically configured according to the needs of actual conditions, not limited herein
System.
For example, fraction representation user of the appraisal result between (- 20, -10) is to application in preset score mapping table
Product is extremely dissatisfied, is badly in need of the defect of correction application product;Fraction representation user of the appraisal result between (- 10,0) is corresponding
Less satisfied with product, urgent need is improved;Print of fraction representation user of the appraisal result between (0,10) to application product
As good, need further perfect;Fraction representation user of the appraisal result between (10,20) feels to praise very much to application product,
Promotion efficiency can be increased;Fraction representation user of the appraisal result between (20,30) enjoys a lot application product, can be with
Continue to continue the outstanding design style of the application product.
In the corresponding embodiment of Fig. 2, by obtaining the data to be analyzed of application product, data to be analyzed are segmented
Processing obtains target participle, and constructs the target term vector of target participle, and segments to target and carry out part-of-speech tagging, obtains
The part of speech of each target term vector constructs the substatement vector sum sentence vector of data to be analyzed using target term vector, pre-
If semantic analysis model in each part of speech corresponding in clustering cluster, calculate the weight coefficient of each target term vector, most
Afterwards, according to the weight coefficient of target term vector, the score of each substatement vector and the appraisal result of sentence vector are calculated, from point
Inquiry user preferences degree corresponding with appraisal result in number mapping table, so that it is determined that fancy grade of the user to application product.
Each target in data to be analyzed is calculated by using trained semantic analysis model and segments corresponding weight coefficient, and root
The appraisal result that corresponding weight coefficient finds out data to be analyzed is segmented according to target, can quickly and accurately obtain user's
The degree of recognition to application product is commented on, the passive degree or aggressiveness level of comment data are embodied, is improved to application product
Comment analysis efficiency and analysis accuracy rate.
In one embodiment, the present embodiment calculates the weight coefficient according to target term vector mentioned in step S6
The score of each substatement vector, and the specific reality of the appraisal result according to the score computing statement vector of n substatement vector
Existing method is described in detail.
Referring to Fig. 3, Fig. 3 shows a specific flow chart of step S6, details are as follows:
S61: according to the weight coefficient of target term vector, the score of each substatement vector is calculated according to formula (1):
Wherein, DxFor the weight coefficient of x-th of target term vector, | Dx| it is the exhausted of the weight coefficient of x-th of target term vector
To value, Sx-1For DxThe score of the prefix sentence of corresponding target term vector, SxFor the score of substatement vector, x is just greater than 0
Integer, S0Equal to 1.
Specifically, formula is used since the first target term vector of the substatement vector for each substatement vector
(1) the sentence characteristic of entire substatement vector, target term vector D are successively calculatedxPrefix sentence refer to, in x-th of target word
Word combination before vector, using the score of the last one target term vector as the score of the substatement vector.
For example, substatement " I likes China " can be segmented with cutting for " I ", " love " and " China " three targets, and target
The weight coefficient for segmenting corresponding target term vector is followed successively by 2,2.5 and 1, wherein the prefix sentence of " love " is " I ", " China "
Prefix sentence be " I like ".The corresponding substatement vector of the substatement is calculated according to formula (1), available: And obtaining third target term vector
It is allocated as the score of the substatement vector, i.e. score of the corresponding score 25 of acquisition S3 as the substatement vector, therefore, sub- language
The corresponding substatement vector of sentence " I likes China " is scored at 25.
S62: according to the scoring of n substatement vector, according to the appraisal result of formula (2) computing statement vector:
Wherein, S is the appraisal result of data to be analyzed, and n is the number of substatement vector, SiFor i-th of substatement vector
Score, i is positive integer, i ∈ [1, n].
Specifically, the scoring of n substatement vector is added using formula (2), expressed by comprehensive each substatement vector
The feature of semanteme, obtain the appraisal result of sentence vector, which indicates the feature of semanteme of data to be analyzed, appraisal result
For positive number, indicate that the data to be analyzed are positive comments, appraisal result is negative, indicates that the data to be analyzed are passive comment
By the numerical value of appraisal result is bigger, indicates that user is higher to the positive attitude of application product or the degree of passive attitude.
In the corresponding embodiment of Fig. 3, the score of each substatement vector is calculated by using formula (1), and according to institute
The score of some substatement vectors enables server-side according to each using the appraisal result of formula (2) computing statement vector
The semanteme of participle, analysis obtains the feature of semanteme of each substatement, then the feature of semanteme of all substatements is combined, thus
The feature of semanteme of data to be analyzed is obtained, the accuracy of analysis of data to be analyzed is improved.
In one embodiment, as shown in figure 4, the mentioned appraisal result according to sentence vector in the step s 7, from pre-
If score mapping table in corresponding with the appraisal result user preferences degree of inquiry, and using user preferences degree as number to be analyzed
According to analysis result after, the data to be analyzed for obtaining appraisal result can also be saved in number by the analysis method of the product evaluation
According in library, details are as follows:
S7: if the appraisal result of data to be analyzed is greater than preset first threshold, which is saved in pre-
If first database in.
In the present embodiment, preset first threshold is the score value preset for obtaining positive evaluation data, if
The appraisal result of data to be analyzed is greater than preset first threshold, then the data to be analyzed belong to positive evaluation data, for example,
First threshold can be preset as 10 points, can also be preset as 15 points, and the data that can specifically obtain according to actual needs are set
It sets, herein with no restrictions.
The data to be analyzed that appraisal result is greater than preset first threshold are saved in preset first database by server-side
In, which is the database for storing the positive comment sentence of user, so that staff can be with
Evaluation data in first database are further analyzed.
Optionally, staff can assess application product according to the evaluation data in first database, obtain
The advantages of application product, so that it is determined that the scheme of later product design and iteration, continues the outstanding core concept of application product, and
And can also estimate to obtain the consumption hobby of user, it is applied in application product leaved for development, formulates relevant operation activity, mention
The usage amount of high application product.
S8: if the appraisal result of data to be analyzed is less than preset second threshold, which is saved in pre-
If the second database in, wherein second threshold be less than first threshold.
In the present embodiment, preset second threshold is the score value for obtaining passive evaluation data, if number to be analyzed
According to appraisal result be less than preset second threshold, then the data to be analyzed belong to passive evaluation data, for example, second threshold
It can be preset as 0 point, -5 points can also be preset as, the data that can specifically obtain according to actual needs are configured, herein not
It is limited.
The data to be analyzed that appraisal result is less than preset second threshold are saved in preset second database by server-side
In, which is the database for storing the passive comment sentence of user, so that staff can be with
Evaluation data in second database are further analyzed.
Optionally, staff can analyze application product according to the evaluation data in first database, retrospect
The shortcomings that source to go wrong, acquisition application product, application product is improved to formulate reasonable solution, is mentioned
The consumption figure of high application product.
It is understood that for appraisal result be less than preset first threshold and be greater than preset second threshold to
Data are analyzed, which can be able to maintain neutral attitude to application product, it is also possible to be filled in refresh download
The comment content unrelated with the application product, this partial data is screened out, and can be improved the working efficiency of staff.
In the corresponding embodiment of Fig. 4, the data to be analyzed by the way that appraisal result to be greater than to preset first threshold are saved
Into preset first database, and by appraisal result be less than preset second threshold data to be analyzed be saved in it is preset
In second database, so that staff can quickly and easily obtain the positive or passive evaluation data of application product
It is analyzed, and can application product be improved or be improved based on the analysis results, improve the usage amount of application product.
In one embodiment, as shown in figure 5, in step sl the data to be analyzed of mentioned acquisition application product it
Before, the analysis method of the product evaluation further includes the building of preset semantic analysis model, constructs preset semantic analysis model
Concrete methods of realizing details are as follows:
S101: the history evaluation data of application product are obtained.
In the present embodiment, application product refer to application product platform provide application software product, for example, pea pods or
The application products platforms such as person's Android market, history evaluation data are evaluation data of the user to application product, and user produces in application
After product platform downloads application product, can according to the downloading of application product and using the case where the application product is commented on.
Specifically, the data-interface that the platform that application software product specifically can be used in history evaluation data provides is adopted
Collection can also grab the evaluation data of application product, specific acquisition modes by network protocol in the platform of application software product
It can be configured according to the actual situation, herein without limitation.
Preferably, the history evaluation data of application product can be used reptile instrument and be crawled, for example, octopus crawler
Tool, Boston ivy reptile instrument or collection search objective reptile instrument etc., by URL (the Uniform Resource of application product platform
Locator, uniform resource locator) it is set as the acquisition address of history evaluation data, which uniquely marked in internet
The network address of knowledge, reptile instrument determine the webpage for crawling data according to the address URL, crawl comment corresponding with preset keyword
By data, which is keyword relevant to history evaluation data, such as " comment " or " evaluation ", crawler work
Keyword is spliced to chaining for the address URL by tool, is searched for and is downloaded history evaluation data corresponding with preset keyword.
S102: word segmentation processing is carried out to history evaluation data, obtains the corresponding participle of history evaluation data.
Specifically, server-side can segment tool or segmentation methods by third party, divide history evaluation data
Word processing, obtains the corresponding participle of history evaluation data.
Wherein, common third party participle tool includes but is not limited to: Stanford NLP segmenter, ICTClAS participle
System, ansj segment tool and HanLP Chinese word segmentation tool etc..
For example, carrying out word segmentation processing to history evaluation data " function of this product is very powerful ", can be with cutting
" this ", " money ", " product ", " ", " function ", the participle such as " very " and " powerful ".
S103: being based on preset corpus, constructs the term vector of participle, and carry out part-of-speech tagging to participle, obtains each
The part of speech of term vector.
Specifically, preset corpus is the database preset for storing Chinese word, wherein preset corpus
Library includes the word of standard and the part of speech of word, the content in preset corpus specifically can from open source " millions are huge
Chinese wordbank ", obtain in the Chinese vocabulary banks such as " stammerer Chinese word segmentation library " or " new edition Beijing University mark collection ".
Based on preset corpus, participle is converted to term vector using term vector generation technique by server-side, and building is each
Corresponding term vector is segmented, and according to the part of speech of word in preset corpus, part-of-speech tagging is carried out to each participle, so that often
A participle carries corresponding part of speech, and the target term vector generated by participle is with corresponding part of speech, obtain each word to
The part of speech of amount, wherein part of speech includes but is not limited to adverbial word, adjective, noun and pronoun etc., for example, the part of speech of participle " beauty "
For " adjective ".
S104: according to the part of speech of term vector, term vector is clustered using preset clustering algorithm, obtains every kind of part of speech
Corresponding K clustering cluster.
Specifically, by way of cluster, clustering is carried out to the identical term vector of part of speech, it is corresponding to obtain every kind of part of speech
K clustering cluster, wherein clustering is the statistical analysis method based on similitude, the data in the same clustering cluster
Between similitude it is higher than the similitude between the data in different clustering clusters, preset clustering algorithm specifically can be K-
Means algorithm is also possible to Clara algorithm, can specifically be configured according to the needs of practical application, herein with no restrictions.
Preferably, the present embodiment carries out clustering to term vector using K-Means algorithm, is used as similitude using distance
Evaluation index, that is, think that the distance between two term vectors are closer, similarity is bigger.Each clustering cluster is leaned on by distance
Close term vector composition, therefore, will obtain compact and independent K clustering cluster as the cluster result of term vector.
S105: the cluster central point in each clustering cluster is obtained, and distributes corresponding weight coefficient for each clustering cluster, is obtained
Preset semantic analysis model after cluster.
Specifically, in clustering cluster, if all term vectors to some term vector distance and it is minimum, which is
Cluster central point in the clustering cluster obtains the corresponding participle of cluster central point and is used as keyword, which represents the language of clustering cluster
Justice, and be that the participle distributes corresponding weight coefficient according to the part of speech feature of the participle, that is, corresponding for the setting of each clustering cluster
Weight coefficient, obtain the weight sets H=[H of each part of speech1, H2..., Hk]=[H11, H22..., Hkk], wherein part of speech H includes
K clustering cluster, HkIndicate the cluster central point of k-th clustering cluster, HkkIndicate the corresponding weight coefficient of k-th clustering cluster, thus
To the semantic analysis model of the K clustering cluster comprising each part of speech and the corresponding weight coefficient of each clustering cluster.
For example, being in adverbial word, according to the weight coefficient of the semantic setting of each cluster central point in part of speech are as follows: [pole, very,
Very, very, compare, a bit, general]=[2,1.6,1.5,1.3,1.1,1], it is in adjective, according in each cluster in part of speech
The weight coefficient of the semantic setting of heart point are as follows: [it is excellent, it is good, it is good, in, it is poor, bad]=[2,1.5,1.2,1, -1.5, -2], in part of speech
For in pronoun, the weight coefficient of each cluster central point can be disposed as 1.
It should be noted that weight coefficient includes positive number and negative, it is the expression in user comment that positive number, which indicates commendatory term,
The word of user's positive attitude, negative number representation derogatory term are the words of expression user's passive attitude in user comment.
In the corresponding embodiment of Fig. 5, by obtaining the historical review data of application product, historical review data are carried out
Word segmentation processing obtains the corresponding participle of historical review data, and is based on preset corpus, constructs the term vector of participle, and
Part-of-speech tagging is carried out to participle, the part of speech of each term vector is obtained, according to the part of speech of term vector, using preset clustering algorithm pair
Term vector is clustered, and the corresponding clustering cluster of every kind of part of speech is obtained, then obtains the cluster central point in each clustering cluster, is each poly-
Class cluster distributes corresponding weight coefficient, obtains semantic analysis model, so as to it is subsequent using semantic analysis model to application product
Comment carries out intellectual analysis, can quickly and accurately obtain the degree of recognition of the comment to application product of user, embody
The passive degree or aggressiveness level of data are evaluated, the analysis efficiency and analysis accuracy rate of the comment to application product are improved.
In one embodiment, for the present embodiment to the part of speech according to term vector mentioned in step S104, use is preset
Clustering algorithm clusters term vector, and the concrete methods of realizing for obtaining the corresponding K clustering cluster of every kind of part of speech carries out specifically
It is bright.
Referring to Fig. 6, Fig. 6 shows a specific flow chart of step S104, details are as follows:
S1041: sorting out term vector according to the part of speech of term vector, will store comprising the term vector of identical part of speech to same
One classification.
Specifically, according to the part of speech of term vector, the corresponding term vector of the participle of history evaluation data is carried out according to part of speech
Term vector comprising identical part of speech is classified as a kind of storage to the same classification, obtains describing word set, name word set, generation by classification
The classifications such as word set and adverbial word collection improve the efficiency of cluster so that the word to different parts of speech distinguishes.
S1042: for the term vector of each classification, using K-means clustering algorithm according to the every two word in the category
Euclidean distance between vector carries out clustering processing, obtains the corresponding K clustering cluster of each part of speech.
Specifically, for the term vector of each classification, K term vector is randomly choosed as initial cluster center, passes through public affairs
Formula (3) calculates the Euclidean distance of unchecked other term vectors and initial cluster center:
D=sqrt (∑ (p-q) ^2) formula (3)
Wherein, p is initial cluster center, and q is term vector, p=(X1, X2..., Xn), q=(Y1, Y2..., Yn), sqrt
It is the function of sqrt, D is the Euclidean distance between p and q.
In each unchecked other term vectors, according to the term vector and all initial cluster centers it is European away from
From the term vector is divided into and in the smallest initial cluster center of the Euclidean distance of the term vector, K interim clusters are formed
Cluster.
Calculate the mean value of each interim clustering cluster.According to the mean value of interim clustering cluster, each word in interim clustering cluster is calculated
Euclidean distance between the vector sum mean value chooses the corresponding term vector of minimum Eustachian distance as the new of the interim clustering cluster
Cluster centre, and by term vector be divided into in the smallest new cluster centre of the Euclidean distance of the term vector, after obtaining update
K interim clustering clusters.
The standard deviation of each updated interim clustering cluster is calculated according to formula (4):
Wherein, σ is standard deviation, AtFor t-th of term vector in updated clustering cluster, n is word in updated clustering cluster
The quantity of vector, μ are term vector AtThe mean value of the updated interim clustering cluster at place, t ∈ [1, m], and t and m are positive whole
Number.
If at least there is a standard deviation in the standard deviation of K updated interim clustering clusters more than or equal to preset
Standard deviation threshold method then returns to the step of calculating each temporarily mean value of clustering cluster, continues to update K interim clustering clusters, Zhi Daogeng
Until the standard deviation of K interim clustering clusters after new is respectively less than preset standard deviation threshold method.
If the standard deviation of K updated interim clustering clusters is respectively less than preset standard deviation threshold method, after this K is updated
Interim clustering cluster cluster centre as cluster analysis result, obtain the corresponding K clustering cluster of each part of speech.
In the corresponding embodiment of Fig. 6, stored by the term vector that will include identical part of speech to the same classification, and be directed to
The term vector of each classification, using K-means clustering algorithm according to the Euclidean distance between the every two term vector in the category
Clustering processing is carried out, the corresponding multiple clustering clusters of each part of speech are obtained, by the participle of semanteme identical in the evaluation to application product
Clustering processing is done, enabling machine to learn the feature of semanteme to such participle can in the analytic process to evaluation data
According to the knowledge learnt, identification obtains the semanteme of each participle in evaluation data, improves semantics recognition model to evaluation
The recognition accuracy of data.
It should be understood that the size of the serial number of each step is not meant that the order of the execution order in above-described embodiment, each process
Execution sequence should be determined by its function and internal logic, the implementation process without coping with the embodiment of the present invention constitutes any limit
It is fixed.
In one embodiment, a kind of analytical equipment of product evaluation, the analytical equipment of the product evaluation and above-mentioned reality are provided
The analysis method for applying product evaluation in example corresponds.As shown in fig. 7, the analytical equipment of the product evaluation includes: data acquisition
Module 71, word segmentation processing module 72, term vector building module 73, sentence vector building module 74, Weight Acquisition module 75, scoring
Computing module 76 and result output module 77.
Detailed description are as follows for each functional module:
Data acquisition module 71, for obtaining the data to be analyzed of application product;
Word segmentation processing module 72 obtains target participle for carrying out word segmentation processing to data to be analyzed;
Term vector constructs module 73, for constructing the target term vector of target participle, and segments to target and carries out part of speech mark
Note, obtains the part of speech of each target term vector;
Sentence vector constructs module 74, for segmenting putting in order in data to be analyzed according to target, uses target
It segments corresponding target term vector and constructs n substatement vector sum sentence vector, wherein n is positive integer;
Weight Acquisition module 75, for according to the corresponding K clustering cluster of every kind of part of speech in preset semantic analysis model, with
And the corresponding weight coefficient of each clustering cluster calculates between target term vector and clustering cluster according to the part of speech of target term vector
Similarity, and obtain power of the weight coefficient corresponding with the maximum clustering cluster of the similarity of target term vector as target term vector
Weight coefficient, wherein K is positive integer;
Scoring computing module 76 calculates the score of each substatement vector for the weight coefficient according to target term vector,
And the appraisal result of the score computing statement vector according to n substatement vector;
As a result output module 77, for the appraisal result according to sentence vector, from preset score mapping table inquiry with
The corresponding user preferences degree of appraisal result, and using user preferences degree as the analysis result of data to be analyzed.
Further, scoring computing module 76 includes:
First computing unit 761 calculates every sub- language according to following formula for the weight coefficient according to target term vector
The score of sentence vector:
Wherein, DxFor the weight coefficient of x-th of target term vector, | Dx| it is the exhausted of the weight coefficient of x-th of target term vector
To value, Sx-1For DxThe score of the prefix sentence of corresponding target term vector, SxFor the score of substatement vector, x is just greater than 0
Integer, S0Equal to 1;
Second computing unit 762, for the scoring according to n substatement vector, according to following formula computing statement vector
Appraisal result:
Wherein, S is the appraisal result of data to be analyzed, and n is the number of substatement vector, SiFor i-th of substatement vector
Score, i is positive integer, i ∈ [1, n].
Further, the analytical equipment of the product evaluation further include:
First memory module 78 waits for this if the appraisal result for data to be analyzed is greater than preset first threshold
Analysis data are saved in preset first database;
Second memory module 78 waits for this if the appraisal result for data to be analyzed is less than preset second threshold
Analysis data are saved in preset second database, wherein second threshold is less than first threshold.
Further, the analytical equipment of the product evaluation further include:
Data acquisition module 701, for obtaining the history evaluation data of application product;
It is corresponding to obtain history evaluation data for carrying out word segmentation processing to history evaluation data for data word segmentation module 702
Participle;
Part-of-speech tagging module 703 constructs the term vector of participle, and carry out word to participle for being based on preset corpus
Property mark, obtain the part of speech of each term vector;
Vector clusters module 704 gathers term vector using preset clustering algorithm for the part of speech according to term vector
Class obtains the corresponding K clustering cluster of every kind of part of speech;
Model construction module 705, for obtaining the cluster central point in each clustering cluster, and it is corresponding for the distribution of each clustering cluster
Weight coefficient, the preset semantic analysis model after being clustered.
Further, vector clusters module 704 includes:
Taxon 7041, for being sorted out according to the part of speech of term vector to term vector, by the word comprising identical part of speech
Vector is stored to the same classification;
Cluster cell 7042, for being directed to the term vector of each classification, using K-means clustering algorithm according in the category
Every two term vector between Euclidean distance carry out clustering processing, obtain the corresponding K clustering cluster of each part of speech.
The specific of analytical equipment about product evaluation limits the analysis method that may refer to above for product evaluation
Restriction, details are not described herein.Modules in the analytical equipment of the said goods evaluation can be fully or partially through software, hard
Part and combinations thereof is realized.Above-mentioned each module can be embedded in the form of hardware or independently of in the processor in computer equipment,
It can also be stored in a software form in the memory in computer equipment, execute the above modules in order to which processor calls
Corresponding operation.
In one embodiment, a kind of computer equipment is provided, which can be server, internal junction
Composition can be as shown in Figure 8.The computer equipment include by system bus connect processor, memory, network interface and
Database.Wherein, the processor of the computer equipment is for providing calculating and control ability.The memory packet of the computer equipment
Include non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system, computer program and data
Library.The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.The calculating
The network interface of machine equipment is used to communicate with external terminal by network connection.When the computer program is executed by processor with
Realize a kind of analysis method of product evaluation.
In one embodiment, a kind of computer equipment is provided, including memory, processor and storage are on a memory
And the computer program that can be run on a processor, processor realize product evaluation in above-described embodiment when executing computer program
Analysis method in step, such as step S1 shown in Fig. 2 to step S7, alternatively, processor execute it is real when computer program
The function of each module/unit of the analytical equipment of product evaluation in existing above-described embodiment, such as module 71 shown in Fig. 7 is to module 77
Function.To avoid repeating, which is not described herein again.
In one embodiment, a kind of computer readable storage medium is provided, computer program is stored thereon with, is calculated
Machine program realizes the step in above-described embodiment in the analysis method of product evaluation when being executed by processor, such as shown in Fig. 2
Step S1 to step S7, alternatively, processor realizes the analytical equipment of product evaluation in above-described embodiment when executing computer program
Each module/unit function, such as module 71 shown in Fig. 7 is to the function of module 77.To avoid repeating, which is not described herein again.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Instruct relevant hardware to complete by computer program, computer program to can be stored in a non-volatile computer readable
It takes in storage medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, this Shen
Please provided by any reference used in each embodiment to memory, storage, database or other media, may each comprise
Non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,
Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancing
Type SDRAM (ESDRAM), synchronization link (SyKchliKk) DRAM (SLDRAM), memory bus (Rambus) direct RAM
(RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
It is apparent to those skilled in the art that for convenience of description and succinctly, only with above-mentioned each function
Can unit, module division progress for example, in practical application, can according to need and by above-mentioned function distribution by different
Functional unit, module are completed, i.e., the internal structure of device are divided into different functional unit or module, to complete above description
All or part of function.
The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although with reference to the foregoing embodiments
Invention is explained in detail, those skilled in the art should understand that: it still can be to aforementioned each implementation
Technical solution documented by example is modified or equivalent replacement of some of the technical features;And these modification or
Replacement, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution should all include
Within protection scope of the present invention.
Claims (10)
1. a kind of analysis method of product evaluation, which is characterized in that the analysis method of the product evaluation includes:
Obtain the data to be analyzed of application product;
Word segmentation processing is carried out to the data to be analyzed, obtains target participle;
The target term vector of the target participle is constructed, and the target is segmented and carries out part-of-speech tagging, obtains each mesh
Mark the part of speech of term vector;
Putting in order in the data to be analyzed is segmented according to the target, segments the corresponding mesh using the target
It marks term vector and constructs n substatement vector, and construct sentence vector according to the n substatement vectors, wherein n is positive integer;
It is corresponding according to the corresponding K clustering cluster of every kind of part of speech and each clustering cluster in preset semantic analysis model
Weight coefficient calculates the similarity between the target term vector and the clustering cluster according to the part of speech of the target term vector,
And obtain weight coefficient corresponding with the maximum clustering cluster of the similarity of the target term vector as the target word to
The weight coefficient of amount, wherein K is positive integer;
According to the weight coefficient of the target term vector, the score of each substatement vector is calculated, and according to the n sons
The score of sentence vector calculates the appraisal result of the sentence vector;
According to the appraisal result of the sentence vector, use corresponding with the appraisal result is inquired from preset score mapping table
Family fancy grade, and using the user preferences degree as the analysis result of the data to be analyzed.
2. the analysis method of product evaluation as described in claim 1, which is characterized in that described according to the target term vector
Weight coefficient calculates the score of each substatement vector, and calculates the sentence according to the score of the substatement vector
The appraisal result of vector includes:
According to the weight coefficient of the target term vector, the score of each substatement vector is calculated according to following formula:
Wherein, DxFor the weight coefficient of x-th of target term vector, | Dx| it is the weight coefficient of x-th of target term vector
Absolute value, Sx-1For DxThe score of the prefix sentence of corresponding target term vector, SxFor the score of the substatement vector, x is
Positive integer greater than 0, S0Equal to 1;
According to the scoring of the n substatement vectors, the appraisal result of the sentence vector is calculated according to following formula:
Wherein, S is the appraisal result of the data to be analyzed, and n is the number of the substatement vector, SiFor i-th of sub- language
The score of sentence vector, i is positive integer, i ∈ [1, n].
3. the analysis method of product evaluation as described in claim 1, which is characterized in that described according to the data to be analyzed
Appraisal result, from preset score mapping table inquire user to the fancy grade of the application product, obtain the user
After fancy grade is as the analysis result, the analysis of the product evaluation further include:
If the appraisal result of the data to be analyzed is greater than preset first threshold, which is saved in preset
In first database;
If the appraisal result of the data to be analyzed is less than preset second threshold, which is saved in preset
In second database, wherein the second threshold is less than the first threshold.
4. the analysis method of product evaluation as described in claim 1, which is characterized in that it is described obtain data to be analyzed it
Before, the analysis method of the product evaluation further include:
Obtain the history evaluation data of the application product;
Word segmentation processing is carried out to the history evaluation data, obtains the corresponding participle of the history evaluation data;
Based on preset corpus, the term vector of the participle is constructed, and part-of-speech tagging is carried out to the participle, obtains each institute
The part of speech of predicate vector;
According to the part of speech of the term vector, term vector is clustered using preset clustering algorithm, obtains every kind of part of speech
The corresponding K clustering clusters;
The cluster central point in each clustering cluster is obtained, and distributes corresponding weight coefficient for each clustering cluster, is obtained
The preset semantic analysis model after cluster.
5. the analysis method of product evaluation as claimed in claim 4, which is characterized in that the word according to the term vector
Property, clusters term vector using preset clustering algorithm, obtains the corresponding K clustering cluster of every kind of part of speech and includes:
The term vector is sorted out according to the part of speech of the term vector, the term vector storage comprising identical part of speech is arrived
The same classification;
For the term vector of each classification, using K-means clustering algorithm according to every two institute's predicate in the category to
Euclidean distance between amount carries out clustering processing, obtains the corresponding K clustering cluster of each part of speech.
6. a kind of analytical equipment of product evaluation, which is characterized in that the analytical equipment of the product evaluation includes:
Data acquisition module, for obtaining the data to be analyzed of application product;
Word segmentation processing module obtains target participle for carrying out word segmentation processing to the data to be analyzed;
Term vector constructs module, for constructing the target term vector of the target participle, and segments to the target and carries out part of speech
Mark, obtains the part of speech of each target term vector;
Sentence vector constructs module, for segmenting putting in order in the data to be analyzed according to the target, uses institute
It states the corresponding target term vector of target participle and constructs n substatement vector sum sentence vector, wherein n is positive integer;
Weight Acquisition module is used for according to the corresponding K clustering cluster of every kind of part of speech in preset semantic analysis model, and each
The corresponding weight coefficient of the clustering cluster calculates the target term vector and described poly- according to the part of speech of the target term vector
Similarity between class cluster, and obtain weight coefficient corresponding with the maximum clustering cluster of the similarity of the target term vector
Weight coefficient as the target term vector, wherein K is positive integer;
Score computing module, for the weight coefficient according to the target term vector, calculates obtaining for each substatement vector
Divide, and calculates the appraisal result of the sentence vector according to the score of the n substatement vectors;
As a result output module, for the appraisal result according to the sentence vector, inquiry and institute from preset score mapping table
The corresponding user preferences degree of appraisal result is stated, and using the user preferences degree as the analysis knot of the data to be analyzed
Fruit.
7. the analytical equipment of product evaluation as claimed in claim 6, which is characterized in that the scoring computing module includes:
First computing unit calculates each son according to following formula for the weight coefficient according to the target term vector
The score of sentence vector:
Wherein, DxFor the weight coefficient of x-th of target term vector, | Dx| it is the weight coefficient of x-th of target term vector
Absolute value, Sx-1For DxThe score of the prefix sentence of corresponding target term vector, SxFor the score of the substatement vector, x is
Positive integer greater than 0, S0Equal to 1;
Second computing unit calculates the sentence vector according to following formula for the scoring according to the n substatement vectors
Appraisal result:
Wherein, S is the appraisal result of the data to be analyzed, and n is the number of the substatement vector, SiFor i-th of sub- language
The score of sentence vector, i is positive integer, i ∈ [1, n].
8. the analytical equipment of product evaluation as claimed in claim 6, which is characterized in that the analytical equipment of the product evaluation is also
Include:
Data acquisition module, for obtaining the history evaluation data of the application product;
It is corresponding to obtain the history evaluation data for carrying out word segmentation processing to the history evaluation data for data word segmentation module
Participle;
Part-of-speech tagging module constructs the term vector of the participle, and carry out word to the participle for being based on preset corpus
Property mark, obtain the part of speech of each term vector;
Vector clusters module clusters term vector using preset clustering algorithm for the part of speech according to the term vector,
Obtain the corresponding K clustering cluster of every kind of part of speech;
Model construction module distributes phase for obtaining the cluster central point in each clustering cluster, and for each clustering cluster
The weight coefficient answered, the preset semantic analysis model after being clustered.
9. a kind of computer equipment, including memory, processor and storage are in the memory and can be in the processor
The computer program of upper operation, which is characterized in that the processor realized when executing the computer program as claim 1 to
The analysis method of any one of 5 product evaluations.
10. a kind of computer readable storage medium, the computer-readable recording medium storage has computer program, and feature exists
In the analysis side of realization product evaluation as described in any one of claim 1 to 5 when the computer program is executed by processor
Method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910046997.7A CN109800307B (en) | 2019-01-18 | 2019-01-18 | Product evaluation analysis method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910046997.7A CN109800307B (en) | 2019-01-18 | 2019-01-18 | Product evaluation analysis method and device, computer equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109800307A true CN109800307A (en) | 2019-05-24 |
CN109800307B CN109800307B (en) | 2022-08-02 |
Family
ID=66559702
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910046997.7A Active CN109800307B (en) | 2019-01-18 | 2019-01-18 | Product evaluation analysis method and device, computer equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109800307B (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110377900A (en) * | 2019-06-17 | 2019-10-25 | 深圳壹账通智能科技有限公司 | Checking method, device, computer equipment and the storage medium of Web content publication |
CN111177402A (en) * | 2019-12-13 | 2020-05-19 | 中移(杭州)信息技术有限公司 | Evaluation method and device based on word segmentation processing, computer equipment and storage medium |
CN111241811A (en) * | 2020-01-06 | 2020-06-05 | 平安科技(深圳)有限公司 | Method, apparatus, computer device and storage medium for determining search term weight |
CN111506726A (en) * | 2020-03-18 | 2020-08-07 | 大箴(杭州)科技有限公司 | Short text clustering method and device based on part-of-speech coding and computer equipment |
CN111737374A (en) * | 2019-06-21 | 2020-10-02 | 腾讯科技(深圳)有限公司 | Position coordinate determination method and device, electronic equipment and storage medium |
CN111931497A (en) * | 2020-07-16 | 2020-11-13 | 中国汽车技术研究中心有限公司 | Optimization method for language of questionnaire for automobile consumer |
CN112785335A (en) * | 2021-01-21 | 2021-05-11 | 安徽商信政通信息技术股份有限公司 | Data processing method and system for electronic government affair performance assessment system |
CN113610559A (en) * | 2021-07-13 | 2021-11-05 | 广东丸美生物技术股份有限公司 | Method and device for evaluating cosmetics |
CN113761182A (en) * | 2020-06-17 | 2021-12-07 | 北京沃东天骏信息技术有限公司 | Method and device for determining service problem |
CN114707507A (en) * | 2022-04-02 | 2022-07-05 | 永道工程咨询有限公司 | List information detection method and device based on artificial intelligence algorithm |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9336192B1 (en) * | 2012-11-28 | 2016-05-10 | Lexalytics, Inc. | Methods for analyzing text |
CN106844330A (en) * | 2016-11-15 | 2017-06-13 | 平安科技(深圳)有限公司 | The analysis method and device of article emotion |
CN107229610A (en) * | 2017-03-17 | 2017-10-03 | 咪咕数字传媒有限公司 | The analysis method and device of a kind of affection data |
CN107391482A (en) * | 2017-07-12 | 2017-11-24 | 成都准星云学科技有限公司 | A kind of method that fuzzy matching and beta pruning are carried out based on sentence mould |
CN107704558A (en) * | 2017-09-28 | 2018-02-16 | 北京车慧互动广告有限公司 | A kind of consumers' opinions abstracting method and system |
US20180165554A1 (en) * | 2016-12-09 | 2018-06-14 | The Research Foundation For The State University Of New York | Semisupervised autoencoder for sentiment analysis |
CN108763326A (en) * | 2018-05-04 | 2018-11-06 | 南京邮电大学 | A kind of sentiment analysis model building method of the diversified convolutional neural networks of feature based |
CN108959383A (en) * | 2018-05-31 | 2018-12-07 | 平安科技(深圳)有限公司 | Analysis method, device and the computer readable storage medium of network public-opinion |
US20180357531A1 (en) * | 2015-11-27 | 2018-12-13 | Devanathan GIRIDHARI | Method for Text Classification and Feature Selection Using Class Vectors and the System Thereof |
-
2019
- 2019-01-18 CN CN201910046997.7A patent/CN109800307B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9336192B1 (en) * | 2012-11-28 | 2016-05-10 | Lexalytics, Inc. | Methods for analyzing text |
US20180357531A1 (en) * | 2015-11-27 | 2018-12-13 | Devanathan GIRIDHARI | Method for Text Classification and Feature Selection Using Class Vectors and the System Thereof |
CN106844330A (en) * | 2016-11-15 | 2017-06-13 | 平安科技(深圳)有限公司 | The analysis method and device of article emotion |
US20180165554A1 (en) * | 2016-12-09 | 2018-06-14 | The Research Foundation For The State University Of New York | Semisupervised autoencoder for sentiment analysis |
CN107229610A (en) * | 2017-03-17 | 2017-10-03 | 咪咕数字传媒有限公司 | The analysis method and device of a kind of affection data |
CN107391482A (en) * | 2017-07-12 | 2017-11-24 | 成都准星云学科技有限公司 | A kind of method that fuzzy matching and beta pruning are carried out based on sentence mould |
CN107704558A (en) * | 2017-09-28 | 2018-02-16 | 北京车慧互动广告有限公司 | A kind of consumers' opinions abstracting method and system |
CN108763326A (en) * | 2018-05-04 | 2018-11-06 | 南京邮电大学 | A kind of sentiment analysis model building method of the diversified convolutional neural networks of feature based |
CN108959383A (en) * | 2018-05-31 | 2018-12-07 | 平安科技(深圳)有限公司 | Analysis method, device and the computer readable storage medium of network public-opinion |
Non-Patent Citations (2)
Title |
---|
张焕成: "网购评论信息细粒度情感分析方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
段旭磊 等: "微博文本的句向量表示及相似度计算方法研究", 《计算机工程》 * |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110377900A (en) * | 2019-06-17 | 2019-10-25 | 深圳壹账通智能科技有限公司 | Checking method, device, computer equipment and the storage medium of Web content publication |
CN111737374A (en) * | 2019-06-21 | 2020-10-02 | 腾讯科技(深圳)有限公司 | Position coordinate determination method and device, electronic equipment and storage medium |
CN111737374B (en) * | 2019-06-21 | 2023-05-09 | 腾讯科技(深圳)有限公司 | Position coordinate determination method, device, electronic equipment and storage medium |
CN111177402B (en) * | 2019-12-13 | 2023-09-22 | 中移(杭州)信息技术有限公司 | Evaluation method, device, computer equipment and storage medium based on word segmentation processing |
CN111177402A (en) * | 2019-12-13 | 2020-05-19 | 中移(杭州)信息技术有限公司 | Evaluation method and device based on word segmentation processing, computer equipment and storage medium |
CN111241811A (en) * | 2020-01-06 | 2020-06-05 | 平安科技(深圳)有限公司 | Method, apparatus, computer device and storage medium for determining search term weight |
CN111241811B (en) * | 2020-01-06 | 2024-05-10 | 平安科技(深圳)有限公司 | Method, apparatus, computer device and storage medium for determining search term weight |
CN111506726A (en) * | 2020-03-18 | 2020-08-07 | 大箴(杭州)科技有限公司 | Short text clustering method and device based on part-of-speech coding and computer equipment |
CN111506726B (en) * | 2020-03-18 | 2023-09-22 | 大箴(杭州)科技有限公司 | Short text clustering method and device based on part-of-speech coding and computer equipment |
CN113761182A (en) * | 2020-06-17 | 2021-12-07 | 北京沃东天骏信息技术有限公司 | Method and device for determining service problem |
CN111931497A (en) * | 2020-07-16 | 2020-11-13 | 中国汽车技术研究中心有限公司 | Optimization method for language of questionnaire for automobile consumer |
CN112785335A (en) * | 2021-01-21 | 2021-05-11 | 安徽商信政通信息技术股份有限公司 | Data processing method and system for electronic government affair performance assessment system |
CN113610559A (en) * | 2021-07-13 | 2021-11-05 | 广东丸美生物技术股份有限公司 | Method and device for evaluating cosmetics |
CN114707507A (en) * | 2022-04-02 | 2022-07-05 | 永道工程咨询有限公司 | List information detection method and device based on artificial intelligence algorithm |
Also Published As
Publication number | Publication date |
---|---|
CN109800307B (en) | 2022-08-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109800307A (en) | Analysis method, device, computer equipment and the storage medium of product evaluation | |
CN110147551B (en) | Multi-category entity recognition model training, entity recognition method, server and terminal | |
CN108363790B (en) | Method, device, equipment and storage medium for evaluating comments | |
CN110046304B (en) | User recommendation method and device | |
CN106202010B (en) | Method and apparatus based on deep neural network building Law Text syntax tree | |
WO2021159632A1 (en) | Intelligent questioning and answering method and apparatus, computer device, and computer storage medium | |
CN116194912A (en) | Method and system for aspect-level emotion classification using graph diffusion transducers | |
CN107590133A (en) | The method and system that position vacant based on semanteme matches with job seeker resume | |
CN109271493A (en) | A kind of language text processing method, device and storage medium | |
CN110175227A (en) | A kind of dialogue auxiliary system based on form a team study and level reasoning | |
CN109857846B (en) | Method and device for matching user question and knowledge point | |
CN109684627A (en) | A kind of file classification method and device | |
US10685012B2 (en) | Generating feature embeddings from a co-occurrence matrix | |
CN108959531A (en) | Information search method, device, equipment and storage medium | |
CN111310440A (en) | Text error correction method, device and system | |
CN110598869B (en) | Classification method and device based on sequence model and electronic equipment | |
CN110309114A (en) | Processing method, device, storage medium and the electronic device of media information | |
CN106886579A (en) | Real-time streaming textual hierarchy monitoring method and device | |
CN112749556B (en) | Multi-language model training method and device, storage medium and electronic equipment | |
CN113761220A (en) | Information acquisition method, device, equipment and storage medium | |
CN113821527A (en) | Hash code generation method and device, computer equipment and storage medium | |
CN112131261A (en) | Community query method and device based on community network and computer equipment | |
CN109299459B (en) | Word vector training method and device for single semantic supervision | |
Korade et al. | Strengthening Sentence Similarity Identification Through OpenAI Embeddings and Deep Learning. | |
US20230351473A1 (en) | Apparatus and method for providing user's interior style analysis model on basis of sns text |
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 |