CN107526721A - A kind of disambiguation method and device to electric business product review vocabulary - Google Patents
A kind of disambiguation method and device to electric business product review vocabulary Download PDFInfo
- Publication number
- CN107526721A CN107526721A CN201710473766.5A CN201710473766A CN107526721A CN 107526721 A CN107526721 A CN 107526721A CN 201710473766 A CN201710473766 A CN 201710473766A CN 107526721 A CN107526721 A CN 107526721A
- Authority
- CN
- China
- Prior art keywords
- word
- history
- frequent
- emotion
- combination
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/253—Grammatical analysis; Style critique
Abstract
The disclosure discloses a kind of Word sense disambiguation method and device to electric business product review, and this method includes：Historical review text is obtained, therefrom extracts history feature word and the history emotion word arranged in pairs or groups therewith；According to the cooccurrence relation between the corresponding each history emotion word of history feature word, the history feature word and history emotion word combination of most frequent appearance are filtered out；Marked according to the senses of a dictionary entry of the history feature word of most frequent appearance and history emotion word combination producing history feature word；New comment text is obtained, therefrom extracts new feature word and corresponding new emotion word combination；According to new feature word and new emotion word combination, inquiry and new feature word and the history feature word and history emotion word combination of new emotion word combinations matches, the meaning of a word using the senses of a dictionary entry mark of the history feature word of matching as new feature word.The technical scheme realizes to be determined to the meaning of a word analysis of product review vocabulary and the senses of a dictionary entry, so as to which accurate judgement appears in the accurate lexical or textual analysis of the product feature word of the linguistic context in different comment contexts.
Description
Technical field
This disclosure relates to natural language processing technique field, more particularly to a kind of ambiguity elimination side to product review vocabulary
Method and device.
Background technology
Polysemy is the intrinsic feature of natural language.When active computer does natural language processing, this ambiguity is given
Analysis result brings very big influence.Research finds that the specific connotation of lexical item depends primarily on context, i.e., the linguistic context that word occurs
It is the key element for determining the meaning of a word.
Word sense disambiguation technology based on statistical learning is each from mark or un-annotated data learning using corpus as knowledge source
The different meaning of a word feature of kind.By calculating the probability right of vocabulary within a context in given text, selection has maximum probability
The meaning of a word of weight exports as optimum, such as Bayes classifier, maximum entropy classification.Statistical method is divided into again guidance
With guideless two class.The Model of Word Sense Disambiguation for having guidance needs to carry out word sense tagging to training corpus in advance, and guideless
Method is without this requirement.The machine learning method of current main-stream, such as decision tree (Decision Tree), SVMs
(Support Vector Machine, SVM), maximum entropy (Maximum Entropy, ME) may be used to the statistics meaning of a word and disappear
Discrimination.Illustrated below by taking Bayes as an example：
Bayes's disambiguation method regards the context of sequence of terms as one without structure word set, by contextual window
In the integration of numerous lexical informations carry out disambiguation.Comprise the concrete steps that：Firstly the need of a corpus, in this sample training collection
In the appearance of each ambiguity word marked its correctly semantic, the example of a statistical classification is provided for disambiguation；Then structure
Grader is built, based on context new ambiguity word is classified, algorithm is：If the window size centered on word w is n, this
Word in individual window can be expressed as " w1,w2,...w1/2,...wn-1" form；Setting ambiguity word w has L semantic item simultaneously "
S1...Sl...SL", then Bayesian Method, which is chosen, makes P (w/s1w1...wn-1) (l=1...L) meaning of a word S when taking maximuml(l=
1...L it is) the final semanteme of ambiguity word.
The comment and analysis of electric business platform is the public domain of the own speech of consumers in general, has speech lack of standardization, random
Property big, theme diverging the features such as.This nonstandard language environment carrys out new challenge to meaning of a word analytic band so that tradition is based on upper
Hereafter related statistical machine learning method can not accurately carry out meaning of a word analysis.
The content of the invention
Big in order to solve to exist in correlation technique the randomness of electric business product review vocabulary, the meaning of a word judges not accurate enough ask
Topic, present disclose provides a kind of Word sense disambiguation method to product review vocabulary.
On the one hand, present disclose provides a kind of disambiguation method to product review vocabulary, this method to include：
The historical review text for appointed product is obtained, product history feature word is extracted from the historical review text
With corresponding history emotion word；
According to the cooccurrence relation between the history feature word and corresponding each history emotion word, most frequent appearance is filtered out
History feature word and history emotion word combination；
According to the senses of a dictionary entry mark of history feature word described in the history feature word of most frequent appearance and history emotion word combination producing
Note；
Obtain for appointed product input new comment text, from the new comment text extract product new feature word and
Corresponding new emotion word combination；
According to the new feature word and new emotion word combination, inquiry and the new feature word and new emotion word combinations matches
History feature word and history emotion word combination, the word using the senses of a dictionary entry mark of the history feature word of matching as the new feature word
Justice.
On the other hand, the disclosure additionally provides a kind of ambiguity cancellation element to product review vocabulary, and the device includes：
Bilingual lexicon acquisition module, for obtaining the historical review text for appointed product, from the historical review text
Extract product history feature word and corresponding history emotion word；
Word combination module, for being closed according to the co-occurrence between the history feature word and corresponding each history emotion word
System, filter out the history feature word and history emotion word combination of most frequent appearance；
Word sense tagging module, for being gone through described in the history feature word according to most frequent appearance and history emotion word combination producing
The senses of a dictionary entry mark of history Feature Words；
Word retrieval module, for obtaining the new comment text for appointed product input, from the new comment text
Extract product new feature word and corresponding new emotion word combination；
Match disambiguation module, for according to the new feature word and new emotion word combination, inquiry and the new feature word and
The history feature word and history emotion word combination of new emotion word combinations matches, using the senses of a dictionary entry mark of the history feature word of matching as
The meaning of a word of the new feature word.
The technical scheme provided by this disclosed embodiment can include the following benefits：
The disambiguation method and device to product review vocabulary that the disclosure provides, pass through history in historical review text
Cooccurrence relation between Feature Words and corresponding each history emotion word, filter out the history feature word and history feelings of most frequent appearance
Feel word combination, senses of a dictionary entry mark is then carried out to history feature word according to combination, for the new feature word in new comment text and newly
Emotion word combination, is matched by Feature Words, the history feature word and history emotion word combination of matching can be filtered out, so as to match
History feature word the senses of a dictionary entry mark can be as the meaning of a word of new feature word, so as to eliminate the ambiguity of new feature word.The program
Collocation most often to occur realizes and the meaning of a word analysis of product evaluation vocabulary and the senses of a dictionary entry is determined, introduce emotion word as linguistic context
With the mode of Feature Words collocation, screening is further made to semantic selection from syntactic structure, so as to above and below different comments
Accurate judgement appears in the accurate lexical or textual analysis of the product feature word in the linguistic context in text.
It should be appreciated that the general description and following detailed description of the above are only exemplary, this can not be limited
It is open.
Brief description of the drawings
Accompanying drawing herein is merged in specification and forms the part of this specification, shows the implementation for meeting the present invention
Example, and in specification together for explaining principle of the invention.
Fig. 1 is the schematic diagram of the implementation environment according to involved by the disclosure；
Fig. 2 is a kind of block diagram of server according to an exemplary embodiment；
Fig. 3 is a kind of flow of disambiguation method to product review vocabulary according to an exemplary embodiment
Figure；
Fig. 4 is the flow chart of the details for the step 310 that Fig. 3 corresponds to embodiment；
Fig. 5 is the flow chart of the details for the step 320 that Fig. 3 corresponds to embodiment；
Fig. 6 is the flow chart of the details for the step 322 that Fig. 5 corresponds to embodiment；
Fig. 7 is a kind of block diagram of ambiguity cancellation element to product review vocabulary according to an exemplary embodiment；
Fig. 8 is the details block diagram for the word combination module that Fig. 7 corresponds to embodiment；
Fig. 9 is the details block diagram for the association mining unit that Fig. 7 corresponds to embodiment.
Embodiment
Here explanation will be performed to exemplary embodiment in detail, its example is illustrated in the accompanying drawings.Following description is related to
During accompanying drawing, unless otherwise indicated, the same numbers in different accompanying drawings represent same or analogous key element.Following exemplary embodiment
Described in embodiment do not represent and the consistent all embodiments of the present invention.On the contrary, they be only with it is such as appended
The example of the consistent apparatus and method of some aspects being described in detail in claims, of the invention.
Fig. 1 is the schematic diagram of the implementation environment according to involved by the disclosure.The implementation environment includes：It is at least one mobile whole
End 110 and server 120.
Interrelational form between mobile terminal 110 and server 120, include the network associate mode and/or agreement of hardware,
And the data correlation mode come and gone therebetween.Mobile terminal 110 can install shopping software APP, and mobile terminal 110 calls
Comment word to product can be inputted for user in product review region by doing shopping after software APP, and by the comment text of input
Upload onto the server 120.Server 120 has data storage and disposal ability, server 120 can be right after receiving comment text
Comment text is analyzed and processed, and can specifically use the disambiguation method to product review vocabulary that the disclosure provides to commenting
Paper is originally analyzed and processed, and then obtains the meaning of a word of the comment vocabulary of appointed product.
Referring to Fig. 2, Fig. 2 is a kind of server architecture schematic diagram that the embodiment of the present disclosure provides.The server 200 can because with
Put or performance is different and produce bigger difference, one or more central processing units (central can be included
Processing units, CPU) 222 (for example, one or more processors) and memory 232, one or more
Store the storage medium 230 (such as one or more mass memory units) of application program 242 or data 244.Wherein, deposit
Reservoir 232 and storage medium 230 can be of short duration storage or persistently storage.Being stored in the program of storage medium 230 can include
One or more modules (diagram is not shown), each module can include operating the series of instructions in server.More
Further, central processing unit 222 could be arranged to communicate with storage medium 230, perform storage medium on server 200
Series of instructions operation in 230.Server 200 can also include one or more power supplys 226, one or more
Wired or wireless network interface 250, one or more input/output interfaces 258, and/or, one or more operations
System 241, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM etc..Following Fig. 3-
The Fig. 2 can be based on as the disambiguation method to product review vocabulary performed by server described in 6 illustrated embodiments
Shown server architecture.
Fig. 3 is a kind of flow of disambiguation method to product review vocabulary according to an exemplary embodiment
Figure.The scope of application and executive agent for the disambiguation method to product review vocabulary that disclosure exemplary embodiment provides,
For example, this method is used for the server 120 of implementation environment shown in Fig. 1.As shown in figure 3, the disambiguation method, can be by servicing
Device 120 performs, and may comprise steps of.
In the step 310, the historical review text for appointed product is obtained, production is extracted from the historical review text
Product history feature word and corresponding history emotion word；
Specifically, server 120 can dispose crawler system, orientation crawl is from the upload of mobile terminal 110 for specifying
The historical product comment text of product input.For historical review text is new comment text relatively hereafter, refer to new
Comment text comment text received before producing.With mobile phone for example, comment text can be battery life it is very long,
It is unintelligible etc. that battery temperature is high, screen resolution is not high, camera pixel is very low, camera is taken pictures.
Optionally, as shown in figure 4, step 310 specifically includes following steps：
In step 311, participle operation is carried out to the historical review text, obtains some history feature words and history feelings
Feel word；
Wherein, participle operation first can be carried out to historical review text after historical review text is obtained, specifically can be with
Using existing segmentation methods, the sentence of historical review text is splitted into vocabulary one by one.Segmentation methods are included based on dictionary, word
The segmenting method of storehouse matching；The segmenting method that the segmenting method and knowledge based of word-based frequency statisticses understand, it is no longer superfluous herein
State.After participle, product feature word can be extracted by the method for feature extraction, for example, " battery ", " life-span ", " screen
Curtain ", " resolution ratio " etc. are exactly product feature word, and " length ", " short " " height ", " low ", " unintelligible " etc. are exactly emotion word.History is special
Sign word refers to the Feature Words that occur in historical review text, history emotion word refer in historical review text with history feature phrase
Close the emotion word occurred.Afterwards, can also carry out duplicate removal, denoising, part-of-speech tagging (noun, adjectival mark), Subject Clustering,
The text-processings such as polarity Judgment by emotion work.
In step 312, each history feature word and each history emotion word are counted in the historical review text
Occurrence number, calculate the support of each history feature word and each history emotion word in the historical review text；
It is to be understood that many vocabulary occurrence numbers are few, or even only occurred once, in order to reduce sample vocabulary
Quantity, the number that can occur by counting each history feature word and history emotion word, filters out the relatively low history of the frequency of occurrences
Feature Words and history emotion word.
Wherein, support refers to history feature word or the probability that history emotion word occurs.By statistical history Feature Words and
The number that history emotion word occurs, history feature word and the probability of history emotion word appearance can be calculated respectively.
In step 313, it is the history feature word and history emotion word is frequent by support descending progress arrangement form
Item list, according to the minimum support threshold value of setting, filter out the history feature that support is more than or equal to minimum support threshold value
Word and history emotion word, are respectively written into Feature Words transaction database and emotion word transaction database.
Wherein it is possible to minimum support threshold value is manually set, respectively to history feature word and history emotion word according to support
Degree carry out descending arrangement, formed frequent episode table, filter out support more than or equal to minimum support threshold value history feature word and
History emotion word, filter out history feature word and history emotion word that support is less than minimum support threshold value.After screening
Feature Words write-in characteristic word transaction database, the emotion word after screening is write into emotion word transaction database.
In step 320, according to the cooccurrence relation between the history feature word and corresponding each history emotion word, screening
Go out the history feature word and history emotion word combination of most frequent appearance；
Wherein, can according to the Feature Words and the cooccurrence relation of emotion word occurred in the same sentence of historical review text
With history of forming Feature Words and history emotion word combination, and most frequent (most normal) appearance can be filtered out by building frequent item set
History feature word and history emotion word collocation, from grammatical relation determine vocabulary the meaning of a word.
Specifically, as shown in figure 5, step 320 may comprise steps of：
In step 321, the Feature Words transaction database is scanned, the history feature word is generated into Feature Words frequent one
Item collection；The emotion word transaction database is scanned, the history emotion word is generated into the frequent item collection of emotion word；By the history
Feature Words and the history emotion word combination producing " history feature word-history emotion word " frequent two item collection；
It should be noted that after above-mentioned pretreatment, can be by history feature word write-in characteristic word transaction database
And write history emotion word in emotion word transaction database, so as to pass through scanning feature word transaction database and emotion word affairs
Database can generate the frequent item collection of Feature Words and the frequent item collection of emotion word respectively, by by frequent the one of history feature word
Frequent two item collection of frequent item collection combination producing of item collection and history emotion word.In other words, in the form of " Feature Words-emotion word "
The frequent 2- item collections of combination producing.One product feature word is referred to as a feature frequent episode, and an emotion word is referred to as an emotion frequency
Numerous item.Wherein, " product feature word " is prerequisite, and " emotion word " is corresponding association results, for representing to imply in data
Relevance.In the presence of a Feature Words and an emotion word combination, a Feature Words and multiple emotion word combinations, multiple Feature Words
Three kinds of situations are respectively combined with multiple emotion words, generate frequent 2- item collections according to Matching Relation.
In step 322, each " history feature word-history feelings is concentrated according to the frequent binomial of the historical review text
Feel word " number occurred is combined, the history feature word and history feelings of most frequent appearance are filtered out using association rules mining algorithm
Feel word combination.
Wherein, as shown in fig. 6, step 322 specifically includes following steps：
In step 3221, " history emotion word-history feature word " frequent two item collection calculates as candidate by described in
Frequent binomial concentrates the support each combined；
For the ease of description, history feature word is represented with X, history emotion word is represented with Y, support refers to all
In combination, possibility that { X, Y } occurs, i.e., the probability in all item collections simultaneously containing X and Y.In other words, as certain combination
The frequency of appearance, that is, " appearance rate ".More than occurrence number, support is big；Occurrence number is few, and support is small.Pass through
Support threshold is set, rejects the combination that support is less than threshold value, simplifies follow-up calculating.
In step 3222, by each combination that frequent binomial is concentrated according to support descending sort, it will be greater than being equal to most
The combination of small support threshold forms frequent binomial table；
Specifically, the number that each " history emotion word-history feature word " combination occurs can be concentrated according to frequent binomial,
The support each combined is calculated, arranges each combination according to support descending, i.e., before the big combination row of support, support is small
Combination row after, " the history emotion word-history feature word " for filtering out support less than threshold value combines, and will be greater than the group equal to threshold value
Close and form frequent binomial table L.
In step 3223, FP-Growth root vertex null are created, successively will be each according to the frequently binomial table
" history emotion word-history feature word " combination extracted out as an affairs from table, combined with the root node, composition one from
Root node null->Feature Words->The path of emotion word, passage path encode to the affairs, and by all sections on path
The frequency count of point is 1, generates some paths, forms frequent mode FP-Growth trees；
Specifically, according to frequent binomial table L, the root node of frequent pattern tree (fp tree) (FP- trees) is created, it is marked with " null ".
For each affairs in transaction database D (each " history emotion word-history feature word " combination), perform：Select in affairs
Frequent episode pair, and by the order sequence in L.If the frequent episode table after sequence is [p | P], wherein, p is first element, and P is
The table of surplus element.Calling insert_tree ([p | P], T).The process implementation status is as follows.If T has children, N causes
N.item-name=p.item-name, then N counting increase by 1；Otherwise one new node N of establishment, which is counted, is arranged to 1, chain
Its father node T is connected to, and the node with identical item-name is linked to by node chain structure.If P is non-
Sky, recursively call insert_tree (P, N).So that by the database compressing of frequent 2- item collections to a frequent pattern tree (fp tree), but
Still retain item collection related information.
In step 3224, according to the frequent mode FP-Growth trees, correlation rule is generated by frequent episode table, obtained
The most frequent co-occurrence of " history feature word-history emotion word " that the history emotion word associated by the history feature word collectively forms
Relation.
Optionally, step 3224 specifically includes：
Calculate the confidence level that the frequently binomial concentrates each " history emotion word-history feature word " combination；
It should be noted that the relatively low combination of confidence level can be filtered out, after simplifying by calculating the confidence level each combined
Continuous calculating.Confidence level expression is under conditions of prerequisite " Feature Words " generation, the probability of association results " emotion word " generation,
This is a threshold for generating Strong association rule, has weighed reliability of the correlation rule investigated in " matter ".Wherein it is possible to
Minimum threshold is set to confidence level to realize further screening, rejects the combination that confidence level is less than minimum threshold.
The support and confidence calculations result combined according to each " history emotion word-history feature word ", " goes through to each
History emotion word-history feature word " combination structure conditional pattern base and condition FP-Growth trees, whole frequent two item collection is traveled through, directly
To FP-Growth trees a single path is only included for sky, or FP-Growth trees；
Each combination in the combination of all subpaths in the path will be generated as a frequent mode, i.e. " history
The most frequent cooccurrence relation of Feature Words-history emotion word ".
Specifically, the condition pattern of frequent two item collection is built according to frequent two item collection after confidence level and support screening
Base, the condition frequent pattern tree (fp tree) of frequent two item collection is built to conditional pattern base；It is right according to the establishment process of condition frequent pattern tree (fp tree)
Each condition frequent pattern tree (fp tree) iteration said process newly created；
According to the generating process of the condition frequent pattern tree (fp tree) iteration, when condition frequent pattern tree (fp tree) result is space-time, or
When the condition frequent pattern tree (fp tree) only includes a single generation path, the path generates the condition frequent modes by all
The subpath of tree is combined, and each combination is a frequent mode.
It should be noted that after FP- trees are built up, it is possible to carry out the excavation of frequent item set, mining algorithm can be FP-
Growth (Frequent Pattern Growth) algorithm, is excavated since last of gauge outfit, by that analogy.Call
It is as follows that FP_growth (FP_tree, null) realizes that the process of frequent item set mining is realized：
FP_growth(Tree,α)
(1) if Tree contain single path P then
(2) in for path Ps node each combination (being denoted as β)
(3) minimum support of node in pattern β ∪ α, its support support=β is produced；
(4) else for each ai Tree head (being sequentially scanned from low to high according to support)
(5) pattern β=ai ∪ α, its support support=ai.support are produced；
(6) β conditional pattern base is constructed, then constructs β condition FP- tree Tree β；
(8) FP_growth (Tree β, β) is called；}
It is further alternative, after frequent item set mining is carried out, further the item in frequent item set can be screened.
Specifically can by calculating the degree of correlation in frequent item set between history feature word and each history emotion word, if the degree of correlation compared with
It is low, the appearance of history feature word is represented independently of history emotion word, otherwise claims to rely between history feature word and history emotion word,
The relatively low emotion word of the degree of correlation and feature word combination can be filtered out in this way.It is possible thereby to generate dependency rule shaped like
{i1,i2,...,im, wherein, item { i1,i2,...,imAppearance be related.
Further, the interest-degree of each frequent episode in frequent item set to frequent item set can be calculated, if interest-degree is high,
Represent that frequent item set can promote the presence of frequent episode, if interest-degree is negative value, and frequent item set can suppress the presence of frequent episode；If
Interest-degree is 0, then frequent item set on frequent episode without too big influence.Wherein, confidence level refers to frequent item set F and a certain j union
The ratio of the support of (i.e. F U { j }) and frequent item set F support.And interest-degree refer to F U { j } confidence level with comprising
Difference between the set ratio of { j }.The relatively low frequent episode of interest-degree can be removed by the calculating of interest-degree.Pass through interest-degree
Calculating with the degree of correlation can exclude uncommon history feature word and the collocation of history emotion word, so as to obtain and history feature word
The history emotion word of association, beneficial to semantic judgement is subsequently correctly made, make accuracy higher.
In a step 330, history is special according to the history feature word of most frequent appearance and history emotion word combination producing
Levy the senses of a dictionary entry mark of word.
Wherein, senses of a dictionary entry mark refers to be labeled the multinomial meaning of a word of history feature word.In other words, according to history feature
The different meaning of a word of history feature word are labeled by word and the final commonly-encountered collocations of history emotion word combination.
It is to be understood that history spy can be generated according to the history feature word of most frequent appearance and history emotion word combination
Levy the adopted class of word, collocation knowledge and statistics using the adopted class as word sense disambiguation；According to the justice corresponding to history feature word
Class generative semantics dictionary, distinguished so as to statically describe the senses of a dictionary entry of history feature word；According to history corresponding to history feature word
The collocation knowledge and statistics generative semantics tagged corpus (i.e. the senses of a dictionary entry marks) of emotion word, dynamically present not synonymity
Behaviour in service in true text up and down.
In step 340, the new comment text for appointed product input is obtained, production is extracted from the new comment text
Product new feature word and corresponding new emotion word combination；
For the new comment text inputted after historical review text, Feature Words and feelings can be extracted from new comment text
Feel word, in order to be made a distinction with the Feature Words in historical review text and emotion word, Feature Words and emotion in new comment text
Word respectively becomes new feature word and new emotion word.New feature word and the extracting mode of new emotion word are referred to above-mentioned history feature
The extracting mode of word and history emotion word, will not be repeated here.
In step 350, according to the new feature word and new emotion word combination, inquire about and the new feature word and new emotion
The history feature word and history emotion word combination of word combination matching, using the senses of a dictionary entry mark of the history feature word of matching as described new
The meaning of a word of Feature Words.
Specifically, historical review can be calculated according to the excavation of history feature word and the most normal cooccurrence relation of history emotion word
Method generative semantics dictionary, by the new feature word in the new comment text of acquisition and new emotion word, searched for, looked into by similarity mode
Semantic dictionary is ask, lookup is combined with the history feature word and history emotion device of new feature word and new emotion word combinations matches.Wherein,
In the history feature word and history emotion word combination of the matching found, the meaning of a word of the meaning of a word of history feature word as new feature word
Mark.
Found by the data analysis to domestic main flow electric business platform, obvious emotion is carried for the text of comment on commodity
Color, and product feature word compares focusing, obvious long tail effect is presented in the Feature Words distribution often occurred in comment.Pass through
Discovery is further analyzed to comment text, the qualifier near these Feature Words is appeared in and also concentrates in a scope.By
The constraint for grammer stipulations normal form of originally being experienced in language, syntactic structure can often confine the meaning of a word in certain limit, realization pair
The preliminary screening of polysemant.The qualifier (emotion word) occurred in historical review text is subjected to word frequency statisticses generation frequent item set
Close, which semanteme the lexical or textual analysis for observing the product feature word in these item collections is all concentrated under, regard the semanteme most often occurred as this
The final semanteme of product feature word.
In summary, the disambiguation method to product review vocabulary that the disclosure provides, by historical review text
Cooccurrence relation between history feature word and corresponding each history emotion word, filter out the history feature word of most frequent appearance and go through
History emotion word combination, senses of a dictionary entry mark is then carried out to history feature word according to combination, for the new feature word in new comment text
With new emotion word combination, matched by Feature Words, the history feature word and emotion word combination of matching can be filtered out, so as to match
History feature word the senses of a dictionary entry mark can be as the meaning of a word of new feature word, so as to eliminate the ambiguity of new feature word.The program
Collocation most often to occur realizes and the meaning of a word analysis of product evaluation vocabulary and the senses of a dictionary entry is determined, introduce emotion word as linguistic context
With the mode of Feature Words collocation, screening is further made to semantic selection from syntactic structure, so as to above and below different comments
Accurate judgement appears in the accurate lexical or textual analysis of the product feature word in the linguistic context in text.
Following is embodiment of the present disclosure, can be used for performing the commenting product of the above-mentioned execution of server 120 of the disclosure
By the disambiguation method embodiment of vocabulary.For the details not disclosed in embodiment of the present disclosure, the disclosure pair refer to
The disambiguation method embodiment of product review vocabulary.
Fig. 7 is a kind of block diagram of ambiguity cancellation element to product review vocabulary according to an exemplary embodiment,
The ambiguity cancellation element to product review vocabulary can be used in the server 120 of implementation environment shown in Fig. 1, perform Fig. 3-figure
The all or part of step of 6 any shown disambiguation methods to product review vocabulary.As shown in fig. 7, the judgement fills
Put including but not limited to：Bilingual lexicon acquisition module 710, word combination module 720, word sense tagging module 730, word retrieval module
740 and matching disambiguation module 750.
Wherein, bilingual lexicon acquisition module 710, for obtaining the historical review text for appointed product, commented from the history
Product history feature word and corresponding history emotion word are extracted in paper sheet；
Word combination module 720, for according to the co-occurrence between the history feature word and corresponding each history emotion word
Relation, filter out the history feature word and history emotion word combination of most frequent appearance；
Word sense tagging module 730, for the history feature word according to most frequent appearance and history emotion word combination producing institute
State the senses of a dictionary entry mark of history feature word；
Word retrieval module 740, for obtaining the new comment text for appointed product input, from the new comment text
Middle extraction product new feature word and corresponding new emotion word combination；
Disambiguation module 750 is matched, for according to the new feature word and new emotion word combination, inquiring about and the new feature word
With the history feature word and history emotion word combination of new emotion word combinations matches, the senses of a dictionary entry of the history feature word of matching is marked and made
For the meaning of a word of the new feature word.
The function of modules and the implementation process of effect specifically refer to above-mentioned to product review vocabulary in said apparatus
The implementation process of step is corresponded in disambiguation method, will not be repeated here.
Bilingual lexicon acquisition module 710 such as can be some physical arrangement input/output interface 258 in Fig. 2.
Word combination module 720, word sense tagging module 730 and matching disambiguation module 750 can also be functional modules, use
In performing the corresponding step in the above-mentioned disambiguation method to product review vocabulary.It is appreciated that these modules can pass through
Hardware, software, or a combination of both realize.When realizing in hardware, these modules may be embodied as one or more hard
Part module, such as one or more application specific integrated circuits.When being realized with software mode, these modules may be embodied as at one
Or the one or more computer programs performed on multiple processors, such as being stored in performed by Fig. 2 central processing unit 222
Program in memory 232.
Optionally, the bilingual lexicon acquisition module 710 includes：
Participle unit, for carrying out participle operation to the historical review text, obtain some history feature words and history
Emotion word；
Support computing unit, for counting each history feature word and each history emotion word in historical review text
Occurrence number in this, calculates the support of each history feature word and each history emotion word in the historical review text
Degree；
Sort screening unit, for the history feature word and history emotion word to be carried out into arrangement form by support descending
Frequent episode list, according to the minimum support threshold value of setting, filter out the history that support is more than or equal to minimum support threshold value
Feature Words and history emotion word, are respectively written into Feature Words transaction database and emotion word transaction database.
Optionally, as shown in figure 8, the word combination module 720 includes：
Word combination unit 721, for the Feature Words transaction database will to be scanned, the history feature word is generated special
Levy the numerous item collection of word frequency；The emotion word transaction database is scanned, the history emotion word is generated into the frequent item collection of emotion word；
By the history feature word and the history emotion word combination producing " history feature word-history emotion word " frequent two item collection；
Association mining unit 722, for concentrating each " history feature according to the frequent binomial of the historical review text
The number that word-history emotion word " combination occurs, the history feature word of most frequent appearance is filtered out using association rules mining algorithm
With history emotion word combination.
Optionally, as shown in figure 9, the association mining unit 722 includes：
Computation subunit 7221, for " history emotion word-history feature word " frequent two item collection described in general as candidate item
Collection, calculate frequent binomial and concentrate the support each combined；
Subelement 7222 is filtered, each combination for frequent binomial to be concentrated will be greater than according to support descending sort
Combination equal to minimum support threshold value forms frequent binomial table；
Tree structure subelement 7223, for creating FP-Growth root vertex null, according to the frequently binomial table, according to
It is secondary extract each " history emotion word-history feature word " combination from table as an affairs out, combined with the root node, structure
Into one from root node null->Feature Words->The path of emotion word, passage path encode to the affairs, and by path
The frequency count of upper all nodes is 1, generates some paths, forms frequent mode FP-Growth trees；
Association mining subelement 7224, for according to the frequent mode FP-Growth trees, being generated and being associated by frequent episode table
Rule, obtain " history feature word-history emotion word " that the history emotion word associated by the history feature word collectively forms
Most frequent cooccurrence relation.
Optionally, the association mining subelement 7224 includes：
Confidence calculations block, each " history emotion word-history feature word " combination is concentrated for calculating the frequently binomial
Confidence level；
Condition tree structure block, for the support and confidence level according to each " history emotion word-history feature word " combination
Result of calculation, structure conditional pattern base and condition FP-Growth trees are combined to each " history emotion word-history feature word ", time
Whole frequent two item collection is gone through, until FP-Growth trees only include a single path for sky, or FP-Growth trees；
Mode combinations block, for using each combination in the combination for generating all subpaths in the path as a frequency
The most frequent cooccurrence relation of numerous pattern, i.e. " history feature word-history emotion word ".
Optionally, the disclosure also provides a kind of electronic equipment, and the electronic equipment can be used for the clothes of implementation environment shown in Fig. 1
It is engaged in device 120, performs all or part of step of any shown disambiguation methods to product review vocabulary of Fig. 3-6.
The electronic equipment includes：
Processor；
For storing the memory of processor-executable instruction；
Wherein, the processor is configured as performing the ambiguity elimination side to product review vocabulary described in above-described embodiment
Method.Such as including：
The historical review text for appointed product is obtained, product history feature word is extracted from the historical review text
With corresponding history emotion word；
According to the cooccurrence relation between the history feature word and corresponding each history emotion word, most frequent appearance is filtered out
History feature word and history emotion word combination；
According to the senses of a dictionary entry mark of history feature word described in the history feature word of most frequent appearance and history emotion word combination producing
Note；
Obtain for appointed product input new comment text, from the new comment text extract product new feature word and
Corresponding new emotion word combination；
According to the new feature word and new emotion word combination, inquiry and the new feature word and new emotion word combinations matches
History feature word and history emotion word combination, the word using the senses of a dictionary entry mark of the history feature word of matching as the new feature word
Justice.
In the embodiment electronic equipment computing device operation concrete mode it is relevant this to product review word
Detailed description is performed in the embodiment of the disambiguation method of remittance, explanation will be not set forth in detail herein.
In the exemplary embodiment, a kind of storage medium is additionally provided, the storage medium is computer-readable recording medium,
Such as can be the provisional and non-transitorycomputer readable storage medium for including instruction.The storage medium is for example including instruction
Memory 232, above-mentioned instruction can be performed above-mentioned to product review vocabulary to complete by the central processing unit 218 of server 200
Disambiguation method.
It should be appreciated that the invention is not limited in the precision architecture for being described above and being shown in the drawings, and
And various modifications and changes can be being performed without departing from the scope.The scope of the present invention is only limited by appended claim.
Claims (10)
- A kind of 1. disambiguation method to product review vocabulary, it is characterised in that including：The historical review text for appointed product is obtained, product history feature word and right is extracted from the historical review text The history emotion word answered；According to the cooccurrence relation between the history feature word and corresponding each history emotion word, going through for most frequent appearance is filtered out History Feature Words and history emotion word combination；Marked according to the senses of a dictionary entry of history feature word described in the history feature word of most frequent appearance and history emotion word combination producing；The new comment text for appointed product input is obtained, product new feature word and correspondingly is extracted from the new comment text New emotion word combination；According to the new feature word and new emotion word combination, inquiry and the new feature word and the history of new emotion word combinations matches Feature Words and history emotion word combination, the meaning of a word using the senses of a dictionary entry mark of the history feature word of matching as the new feature word.
- 2. according to the method for claim 1, it is characterised in that the historical review text obtained for appointed product, Product history feature word and corresponding history emotion word are extracted from the historical review text, including：Participle operation is carried out to the historical review text, obtains some history feature words and history emotion word；The occurrence number of each history feature word and each history emotion word in the historical review text is counted, is calculated each The support of history feature word and each history emotion word in the historical review text；The history feature word and history emotion word are subjected to arrangement form frequent episode list by support descending, according to setting Minimum support threshold value, history feature word and history emotion word that support is more than or equal to minimum support threshold value are filtered out, point Other write-in characteristic word transaction database and emotion word transaction database.
- 3. according to the method for claim 2, it is characterised in that described according to history feature word and corresponding each history emotion Cooccurrence relation between word, the history feature word and history emotion word combination of most frequent appearance are filtered out, including：The Feature Words transaction database is scanned, the history feature word is generated into the frequent item collection of Feature Words；Scan the feelings Feel word transaction database, the history emotion word is generated into the frequent item collection of emotion word；The history feature word is gone through with described History emotion word combination producing " history feature word-history emotion word " frequent two item collection；Each " history feature word-history emotion word " combination is concentrated to occur according to the frequent binomial of the historical review text secondary Number, the history feature word and history emotion word combination of most frequent appearance are filtered out using association rules mining algorithm.
- 4. according to the method for claim 3, it is characterised in that frequent two item collection according to the historical review text In the number that occurs of each " history feature word-history emotion word " combination, filtered out using association rules mining algorithm most frequent The history feature word and history emotion word combination of appearance, including：" history emotion word-history feature word " frequent two item collection calculates frequent binomial and concentrated each as candidate by described in The support of combination；By each combination that frequent binomial is concentrated according to support descending sort, the combination equal to minimum support threshold value will be greater than Form frequent binomial table；FP-Growth root vertex null are created, successively will each " history emotion word-history spy according to the frequently binomial table Sign word " combination is extracted out from table as an affairs, combined with the root node, forms one from root node null->Feature Word->The path of emotion word, passage path encode to the affairs, and are 1 by the frequency count of all nodes on path, Some paths are generated, form frequent mode FP-Growth trees；According to the frequent mode FP-Growth trees, correlation rule is generated by frequent episode table, the history feature word is obtained and is closed The most frequent cooccurrence relation of " history feature word-history emotion word " that the history emotion word of connection collectively forms.
- 5. according to the method for claim 4, it is characterised in that it is described according to frequent mode FP-Growth trees, by frequent episode Table generates correlation rule, obtains " the history feature word-history that the history emotion word associated by the history feature word collectively forms The most frequent cooccurrence relation of emotion word ", including：Calculate the confidence level that the frequently binomial concentrates each " history emotion word-history feature word " combination；The support and confidence calculations result combined according to each " history emotion word-history feature word ", to each " history feelings Feel word-history feature word " structure conditional pattern base and condition FP-Growth trees are combined, whole frequent two item collection is traveled through, until FP-Growth trees only include a single path for sky, or FP-Growth trees；Each combination in the combination of all subpaths in the path will be generated as a frequent mode, i.e. " history feature The most frequent cooccurrence relation of word-history emotion word ".
- A kind of 6. ambiguity cancellation element to product review vocabulary, it is characterised in that including：Bilingual lexicon acquisition module, for obtaining the historical review text for appointed product, extracted from the historical review text Product history feature word and corresponding history emotion word；Word combination module, for according to the cooccurrence relation between the history feature word and corresponding each history emotion word, sieve Select the history feature word and history emotion word combination of most frequent appearance；Word sense tagging module, it is special for history described in the history feature word according to most frequent appearance and history emotion word combination producing Levy the senses of a dictionary entry mark of word；Word retrieval module, for obtaining the new comment text for appointed product input, extracted from the new comment text Product new feature word and corresponding new emotion word combination；Disambiguation module is matched, for according to the new feature word and new emotion word combination, inquiring about and the new feature word and new feelings Feel the history feature word and history emotion word combination of word combination matching, the senses of a dictionary entry of the history feature word of matching is marked as described in The meaning of a word of new feature word.
- 7. device according to claim 6, it is characterised in that the bilingual lexicon acquisition module includes：Participle unit, for carrying out participle operation to the historical review text, obtain some history feature words and history emotion Word；Support computing unit, for counting each history feature word and each history emotion word in the historical review text Occurrence number, calculate the support of each history feature word and each history emotion word in the historical review text；Sort screening unit, frequent for the history feature word and history emotion word to be carried out into arrangement form by support descending Item list, according to the minimum support threshold value of setting, filter out the history feature that support is more than or equal to minimum support threshold value Word and history emotion word, are respectively written into Feature Words transaction database and emotion word transaction database.
- 8. device according to claim 7, it is characterised in that the word combination module includes：Word combination unit, for scanning the Feature Words transaction database, it is frequent that the history feature word is generated into Feature Words One item collection；The emotion word transaction database is scanned, the history emotion word is generated into the frequent item collection of emotion word；Gone through described History Feature Words and the history emotion word combination producing " history feature word-history emotion word " frequent two item collection；Association mining unit, for concentrating each " history feature word-history feelings according to the frequent binomial of the historical review text Feel word " number occurred is combined, the history feature word and history feelings of most frequent appearance are filtered out using association rules mining algorithm Feel word combination.
- 9. device according to claim 8, it is characterised in that the association mining unit includes：Computation subunit, as candidate, calculated for " history emotion word-history feature word " frequent two item collection by described in Frequent binomial concentrates the support each combined；Subelement is filtered, for according to support descending sort, will be greater than being equal to minimum each combination that frequent binomial is concentrated The combination of support threshold forms frequent binomial table；Tree structure subelement,, successively will be each according to the frequently binomial table for creating FP-Growth root vertex null " history emotion word-history feature word " combination extracted out as an affairs from table, combined with the root node, composition one from Root node null->Feature Words->The path of emotion word, passage path encode to the affairs, and by all sections on path The frequency count of point is 1, generates some paths, forms frequent mode FP-Growth trees；Association mining subelement, for according to the frequent mode FP-Growth trees, generating correlation rule by frequent episode table, obtaining " the history feature word-history emotion word " collectively formed to the history emotion word associated by the history feature word it is most frequent same Now relation.
- 10. device according to claim 9, it is characterised in that the association mining subelement includes：Confidence calculations block, putting for each " history emotion word-history feature word " combination is concentrated for calculating the frequently binomial Reliability；Condition tree structure block, for the support and confidence calculations according to each " history emotion word-history feature word " combination As a result, structure conditional pattern base and condition FP-Growth trees are combined to each " history emotion word-history feature word ", traversal is whole Individual frequent two item collection, until FP-Growth trees only include a single path for sky, or FP-Growth trees；Mode combinations block, for using each combination in the combination for generating all subpaths in the path as a frequent mould The most frequent cooccurrence relation of formula, i.e. " history feature word-history emotion word ".
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710473766.5A CN107526721B (en) | 2017-06-21 | 2017-06-21 | Ambiguity elimination method and device for comment vocabularies of e-commerce products |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710473766.5A CN107526721B (en) | 2017-06-21 | 2017-06-21 | Ambiguity elimination method and device for comment vocabularies of e-commerce products |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107526721A true CN107526721A (en) | 2017-12-29 |
| CN107526721B CN107526721B (en) | 2020-07-10 |
Family
ID=60748736
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710473766.5A Active CN107526721B (en) | 2017-06-21 | 2017-06-21 | Ambiguity elimination method and device for comment vocabularies of e-commerce products |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107526721B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108319586A (en) * | 2018-01-31 | 2018-07-24 | 天闻数媒科技（北京）有限公司 | A kind of generation of information extraction rule and semantic analysis method and device |
| CN108596637A (en) * | 2018-04-24 | 2018-09-28 | 北京航空航天大学 | A kind of electric business service problem discovery system |
| CN109344150A (en) * | 2018-08-03 | 2019-02-15 | 昆明理工大学 | A kind of spatiotemporal data structure analysis method based on FP- tree |
| CN110362670A (en) * | 2019-07-19 | 2019-10-22 | 中国联合网络通信集团有限公司 | Item property abstracting method and system |
| CN112818682A (en) * | 2021-01-22 | 2021-05-18 | 深圳大学 | E-commerce data analysis method, equipment, device and computer-readable storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110202334A1 (en) * | 2001-03-16 | 2011-08-18 | Meaningful Machines, LLC | Knowledge System Method and Apparatus |
| CN102609424A (en) * | 2011-01-21 | 2012-07-25 | 日电（中国）有限公司 | Method and equipment for extracting assessment information |
| CN103631961A (en) * | 2013-12-17 | 2014-03-12 | 苏州大学张家港工业技术研究院 | Method for identifying relationship between sentiment words and evaluation objects |
| CN104699766A (en) * | 2015-02-15 | 2015-06-10 | 浙江理工大学 | Implicit attribute mining method integrating word correlation and context deduction |
| CN106650225A (en) * | 2016-10-25 | 2017-05-10 | 康美药业股份有限公司 | FP growth algorithm model-based traditional Chinese medicine formula data mining method and system |
-
2017
- 2017-06-21 CN CN201710473766.5A patent/CN107526721B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110202334A1 (en) * | 2001-03-16 | 2011-08-18 | Meaningful Machines, LLC | Knowledge System Method and Apparatus |
| CN102609424A (en) * | 2011-01-21 | 2012-07-25 | 日电（中国）有限公司 | Method and equipment for extracting assessment information |
| CN103631961A (en) * | 2013-12-17 | 2014-03-12 | 苏州大学张家港工业技术研究院 | Method for identifying relationship between sentiment words and evaluation objects |
| CN104699766A (en) * | 2015-02-15 | 2015-06-10 | 浙江理工大学 | Implicit attribute mining method integrating word correlation and context deduction |
| CN106650225A (en) * | 2016-10-25 | 2017-05-10 | 康美药业股份有限公司 | FP growth algorithm model-based traditional Chinese medicine formula data mining method and system |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108319586A (en) * | 2018-01-31 | 2018-07-24 | 天闻数媒科技（北京）有限公司 | A kind of generation of information extraction rule and semantic analysis method and device |
| CN108319586B (en) * | 2018-01-31 | 2021-09-24 | 天闻数媒科技（北京）有限公司 | Information extraction rule generation and semantic analysis method and device |
| CN108596637A (en) * | 2018-04-24 | 2018-09-28 | 北京航空航天大学 | A kind of electric business service problem discovery system |
| CN108596637B (en) * | 2018-04-24 | 2022-05-06 | 北京航空航天大学 | Automatic E-commerce service problem discovery system |
| CN109344150A (en) * | 2018-08-03 | 2019-02-15 | 昆明理工大学 | A kind of spatiotemporal data structure analysis method based on FP- tree |
| CN110362670A (en) * | 2019-07-19 | 2019-10-22 | 中国联合网络通信集团有限公司 | Item property abstracting method and system |
| CN112818682A (en) * | 2021-01-22 | 2021-05-18 | 深圳大学 | E-commerce data analysis method, equipment, device and computer-readable storage medium |
| CN112818682B (en) * | 2021-01-22 | 2023-01-03 | 深圳大学 | E-commerce data analysis method, equipment, device and computer-readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107526721B (en) | 2020-07-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110096570B (en) | Intention identification method and device applied to intelligent customer service robot | |
| CN107704892B (en) | A kind of commodity code classification method and system based on Bayesian model | |
| CN107515877B (en) | Sensitive subject word set generation method and device | |
| WO2019080863A1 (en) | Text sentiment classification method, storage medium and computer | |
| CN107526721A (en) | A kind of disambiguation method and device to electric business product review vocabulary | |
| CN111353306B (en) | Entity relationship and dependency Tree-LSTM-based combined event extraction method | |
| CN109614620B (en) | HowNet-based graph model word sense disambiguation method and system | |
| CN111414763A (en) | Semantic disambiguation method, device, equipment and storage device for sign language calculation | |
| CN106570180A (en) | Artificial intelligence based voice searching method and device | |
| CN113095080B (en) | Theme-based semantic recognition method and device, electronic equipment and storage medium | |
| CN110188359B (en) | Text entity extraction method | |
| CN109522396B (en) | Knowledge processing method and system for national defense science and technology field | |
| CN114997288A (en) | Design resource association method | |
| CN113821605A (en) | Event extraction method | |
| CN112860896A (en) | Corpus generalization method and man-machine conversation emotion analysis method for industrial field | |
| CN112347339A (en) | Search result processing method and device | |
| CN111611404A (en) | Method for predicting mineral products in target area based on geological text corpus | |
| CN114491062A (en) | Short text classification method fusing knowledge graph and topic model | |
| CN112632272B (en) | Microblog emotion classification method and system based on syntactic analysis | |
| CN110688540B (en) | Cheating account screening method, device, equipment and medium | |
| CN111091009B (en) | Document association auditing method based on semantic analysis | |
| CN110377753B (en) | Relation extraction method and device based on relation trigger word and GRU model | |
| CN107562774A (en) | Generation method, system and the answering method and system of rare foreign languages word incorporation model | |
| CN107818078B (en) | Semantic association and matching method for Chinese natural language dialogue | |
| CN111813927A (en) | Sentence similarity calculation method based on topic model and LSTM |
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 | ||
| CP03 | Change of name, title or address |
Address after: 528200 north of the third floor of plant 5, Midea Global Innovation Center, Industrial Avenue, Beijiao community neighborhood committee, Beijiao Town, Shunde District, Foshan City, Guangdong Province Patentee after: Foshan Meiyun Zhishu Technology Co.,Ltd. Address before: 528311 7th floor, Zone D, Midea headquarters building, No. 6, Midea Avenue, Beijiao Town, Shunde District, Foshan City, Guangdong Province Patentee before: SHENZHEN MEICLOUD TECHNOLOGY CO.,LTD. |
|
| CP03 | Change of name, title or address | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20220422 Address after: 528200 third floor of plant 5, Midea Global Innovation Center, Industrial Avenue, Beijiao Town, Shunde District, Foshan City, Guangdong Province Patentee after: Guangdong Meiyun Zhishu Technology Co.,Ltd. Address before: 528200 north of the third floor of plant 5, Midea Global Innovation Center, Industrial Avenue, Beijiao community neighborhood committee, Beijiao Town, Shunde District, Foshan City, Guangdong Province Patentee before: Foshan Meiyun Zhishu Technology Co.,Ltd. |
|
| TR01 | Transfer of patent right |