CN106294662A - Inquiry based on context-aware theme represents and mixed index method for establishing model - Google Patents

Inquiry based on context-aware theme represents and mixed index method for establishing model Download PDF

Info

Publication number
CN106294662A
CN106294662A CN201610634174.2A CN201610634174A CN106294662A CN 106294662 A CN106294662 A CN 106294662A CN 201610634174 A CN201610634174 A CN 201610634174A CN 106294662 A CN106294662 A CN 106294662A
Authority
CN
China
Prior art keywords
context
inquiry
theme
aware
model
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.)
Pending
Application number
CN201610634174.2A
Other languages
Chinese (zh)
Inventor
贺樑
陈琴
胡琴敏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
East China Normal University
Original Assignee
East China Normal University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by East China Normal University filed Critical East China Normal University
Priority to CN201610634174.2A priority Critical patent/CN106294662A/en
Publication of CN106294662A publication Critical patent/CN106294662A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3344—Query execution using natural language 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
    • G06F16/33—Querying
    • G06F16/3331—Query processing

Abstract

The invention discloses a kind of inquiry based on context-aware theme to represent and mixed index method for establishing model, comprise the steps: step one: based on the keyword set inquired about, obtain the pseudo-linear filter document of inquiry, from pseudo-linear filter document, choose context associated with the query；Step 2: introduce context-aware topic model, context is incorporated in context-aware topic model, the subject information implied based on corpus Topics Crawling contextual window, obtain its corresponding theme vector；Step 3: inquiry is combined with keyword set with theme vector and represents, based on theme vector and keyword set, set up mixed index model, obtain final retrieval score.

Description

Inquiry based on context-aware theme represents and mixed index method for establishing model
Technical field
The present invention relates to internet information retrieval technique field, particularly relate to a kind of based on context-aware topic model Inquiry represents and mixed index method for establishing model.
Background technology
Inquiry represents the core of always information retrieval field, and most common of which problem is that user inquires about the shortest (only bag Containing several key words), easily cause relevant documentation in retrieving and do not mate with inquiry.Such as " hydropenia " this user is looked into Ask, if containing words associated with the query such as " arid " in document, although dependency is the highest, but owing to closing without original inquiry Keyword " hydropenia ", final matching degree will be the lowest, and then the accuracy rate of impact inquiry.
Common solution is query expansion based on pseudo-linear filter.The method sets up the base in preliminary search result On plinth, it is assumed that come K document (referred to as " pseudo-linear filter document ") above relevant to former inquiry, key therein Word can use related algorithm to extract to represent for query expansion.But the method is unsupervised, easy band is served The word unrelated with inquiry.Although the sorting technique of supervision can be used in theory, consider the various features of expansion word, choose Select real word associated with the query.But, this method depends on Feature Engineering and mark training set, the cost of actual application Higher.
Some researchs recently begin to focus on the independent extensions how utilizing various contextual information to alleviate during inquiry represents Word introduces problem.Contextual information source mainly include high-quality external data source (such as encyclopedia, domain body etc.) and Pseudo-linear filter document based on data set itself.The former is owing to being only suitable for partial query, and external data source is in most cases Update slow, obtain difficulty, so actual application is the most extensive.And the latter's pseudo-linear filter based on data set self document is actual On also provide to inquiry context describe, there is bigger Research Prospects.Such as, for " hydropenia " this inquiry, Pseudo-linear filter document 1 describes: " Britain will face water shortage problem the coming years, so asking using water wisely, repair your fire hose Head.”；Pseudo-linear filter document 2 describes: " drought-resistant agriculture: a kind of method alleviating arid and water shortage problem ".These two is all to close In the counter-measure of water shortage problem, these contextual informations may serve to nonproductive poll and represent.But existing extension selected ci poem Access method typically only considered expansion word and former query word co-occurrence degree in the contextual window of pseudo-linear filter, however it remains Problems with: (1) needs explicitly to select which word to be used as final query expansion, still can introduce one in the case of unsupervised A little unrelated words, even " harmful word ".Such as: in the article relating to various environmental resource, there is relatively frequency in key word " hydropenia " Numerous, but its context also there will be similar " hydroelectric generation ", " natural gas ", original query can be deviateed, reduce inquiry Accuracy；(2) final inquiry represents still based on dictionary space, have ignored the semantic information that inquiry is implicit, such as potential theme； (3) retrieval model represented based on this inquiry mainly considers Keywords matching, and have ignored document with inquiry at semantic hierarchies On coupling.
Summary of the invention
It is an object of the invention to propose for the deficiencies in the prior art is a kind of based on context-aware topic model Inquiry represents and mixed index design methods, incorporates context theme based on pseudo-linear filter letter in inquiry represents Breath, thus on the basis of original retrieval model based on Keywords matching, increase theme coupling, promote the accuracy of retrieval result.
The present invention proposes a kind of inquiry based on context-aware theme and represents and mixed index method for establishing model, bag Include following steps:
Step one: keyword set based on inquiry, obtains the pseudo-linear filter document of described inquiry, from described spurious correlation The context relevant to described inquiry chosen by feedback document；
Step 2: introduce context-aware topic model, described context is incorporated described context-aware topic model In, the subject information implied based on contextual window described in corpus Topics Crawling, obtain its corresponding theme vector；
Step 3: described inquiry is combined with described keyword set with described theme vector and represents；Based on described theme Keyword set described in vector sum, sets up mixed index model, obtains final retrieval score.
Described based on context-aware theme the inquiry that the present invention proposes represents and in mixed index method for establishing model, Described pseudo-linear filter document is divided into multiple sliding window by step one, and calculates the phase of each window and described inquiry Guan Xing, takes the dependency window higher than threshold value as the contextual window relevant to described inquiry.
Described based on context-aware theme the inquiry that the present invention proposes represents and in mixed index method for establishing model, Described context selected threshold associated with the query is the meansigma methods of all window dependencys under this inquiry.
Described based on context-aware theme the inquiry that the present invention proposes represents and in mixed index method for establishing model, Described context-aware topic model is according to designed by inquiry related context and whole corpus, utilizes described context sense Know that topic model assumes the master that contextual window is same with the pseudo-linear filter document sharing at its place in theme modeling process Topic distribution, obtains the theme vector of context.
Described based on context-aware theme the inquiry that the present invention proposes represents and in mixed index method for establishing model, Described pseudo-linear filter document uses retrieval model Keywords matching score to calculate and obtains.
Described based on context-aware theme the inquiry that the present invention proposes represents and in mixed index method for establishing model, Described retrieval score represents with equation below:
S = ( 1 - λ ) Σ q i ∈ Q s ( q i , d ) + λ · s ′ ( Q ′ , d )
Wherein, s represents score based on Keywords matching in conventional retrieval model, and s ' represents based on new inquiry expression Q ' Theme matching score, λ is the weight parameter between both scores, is also the balance coefficient of two kinds of matching ways.
The beneficial effects of the present invention is: the present invention takes full advantage of corpus and is themselves based on the context of pseudo-linear filter Information, solves the problem that high-quality external data source is difficult to obtain.And by pseudo-linear filter document is divided into one by one Contextual window, and therefrom select and inquire about more relevant context fragment for inquiring about expression, decrease " noise " and draw Enter and inquire about drift, being a kind of novelty behave inquired about and represent quality control.The context-aware theme proposed in the present invention Model, has fully excavated the subject information that context associated with the query is corresponding, breaches tradition and is based only upon key word aspect Understand, contribute to more comprehensively, be more fully understood from user's inquiry.Traditional retrieval model is based primarily upon Keywords matching, and neglects Omit the semantic dependency of profound level.The mixed index model of present invention design has considered Keywords matching and theme Joining, this diversified matching way helps lend some impetus to the lifting of retrieval effectiveness.Inquiry method for expressing that the present invention proposes and mixed Close retrieval model on the data set of Microblog Track 2011-2014, be all proved to be effective, incorporate in queries Context subject information, its MAP value finally retrieved has exceeded some up-to-date inquiry method for expressing.
Accompanying drawing explanation
Fig. 1 is that present invention inquiry based on context-aware theme represents and the flow process of mixed index method for establishing model Figure.
Fig. 2 is that flow chart chosen in context based on pseudo-linear filter.
Fig. 3 is that the graph model of context-aware topic model represents.
Detailed description of the invention
In conjunction with specific examples below and accompanying drawing, the present invention is described in further detail.Implement the present invention process, Condition, experimental technique etc., outside the lower content mentioned specially, be universal knowledege and the common knowledge of this area, this Bright content is not particularly limited.
As it is shown in figure 1, present invention inquiry based on context-aware theme represents and mixed index method for establishing model bag Include following steps:
Step one: keyword set based on inquiry, obtains the pseudo-linear filter document of inquiry, from pseudo-linear filter document In choose context associated with the query；
Step 2: introduce context-aware topic model, context is incorporated in context-aware topic model, based on language The subject information that material storehouse Topics Crawling contextual window is implied, obtains its corresponding theme vector；
Step 3: inquiry is combined with keyword set with theme vector and represents；Based on theme vector and keyword set, Set up mixed index model, obtain final retrieval score.
(1), related context based on pseudo-linear filter is chosen
Being easily obtained due to pseudo-linear filter document and comprise a lot of content associated with the query, the present invention will therefrom choose Going out to the more relevant context of inquiry for inquiring about expression, its idiographic flow is shown in accompanying drawing 2.
First, pseudo-linear filter document is carried out cutting, obtain the contextual window that multiple size is n.Definition Q={q1, q2..., q|Q|It is an inquiry, wherein qiRepresenting a searching keyword, | Q | represents the number of key word in this inquiry.It is pseudo-linear filter collection of document corresponding for inquiry Q, during retrieval, comes the literary composition of top k i.e. for the first time Shelves.For a pseudo-linear filter documentBy the form with sliding window, it be divided into as shown in Figure 2 some Individual size is the contextual window (comprising n word) of n, i.e. Qc1, Qc2..., Qcl, I represents the number of contextual window.
Secondly, the dependency of contextual window and former inquiry is calculated.For an inquiry with contextual window to (Q, Qc), The present invention comprehensively uses multiple method to the dependency R (Q, the Q that calculate between themc), such as equalization point mutual trust based on Term co-occurrence Cease (Pointwise Mutual Information), Jaccard similarity based on set of words, based on term vector word2vec Semantic similarity etc., finally take its meansigma methods.
Then, context associated with the query is filtered out.First dependency derived above is normalized.Connect , arranging threshold value is the meansigma methods of all window dependencys under this inquiry, filters out the dependency context window less than this threshold value Mouthful, remaining context more relevant to inquiry will be further used as the modeling of context-aware theme.
(2), the modeling of context theme-aware and inquiry represent
The context associated with the query obtained in given (one) and whole corpus, the present invention designs a context sense Know topic model, in order to be dissolved in topic model by contextual information associated with the query, generate new inquiry and represent.
Inspired by correlational study, due to the contextual window chosen in (one) and its place pseudo-linear filter document all It is closely-related, thus, it is supposed that they share same theme distribution with inquiry.Under this assumption, traditional LDA master is improved Topic model, thus obtain context-aware topic model CAT, its graph model represents such as accompanying drawing 3.The related symbol related in model Illustrate such as table 1.This model is one and generates model, and concrete modeling process is shown in algorithm 1.
Related symbol explanation in table 1 context-aware topic model CAT
For the parameter in solving model, the present invention uses widely used gibbs sampler (Gibbs sampling) algorithm.
First, according to gibbs sampler algorithm, in document, the word is assigned to the probability of theme and represents with equation below (1):
Wherein,Represent the theme allocation vector of other all words not including current i-th word,Represent document D is assigned to the word number (not including current word) of theme k,Represent word wiIn whole language material, it is assigned to theme k's Number of times (does not include current word).During symbol is represented disappearance subscript or subscript (asWith) represent and this disappearance is tieed up Degree summation, 1 is the vector that an element is all 1.
Similarly, in document d, jth contextual window associated with the query is assigned to the probability of theme k and can use down The formula (2) in face represents:
Wherein,Represent the master of other all windows not including current jth contextual window associated with the query Topic allocation vector,Represent the number (not including current window) of all contextual windows relevant to inquiry Q in theme k, θD, kRepresent the probability of theme k in document d, can calculate by equation below further:
θ d , k = n k d + α k n ( · ) d + α T 1 - - - ( 3 )
Wherein,Represent the total word number being assigned to theme k in document d.
When model convergence or when reaching default iterations, following distribution will be obtained: " document-theme " distribution θ, " theme-word " distribution Φ and " theme-inquiry context " distribution η.Each list of η shows that certain all related context inquired about exists Distribution situation on theme, this be also obtain newly inquire about expression.Visible, this expression the most simultaneously by contextual information and Subject information merges, and will be better than method for expressing to each self-modeling respectively in theory.
(3), mixed index modelling
The present invention newly inquires about expression based on obtain, designs a kind of mixing simultaneously considering Keywords matching and theme coupling Retrieval model, its retrieval score computing formula is as follows:
S = ( 1 - λ ) Σ q i ∈ Q s ( q i , d ) + λ · s ′ ( Q ′ , d ) - - - ( 4 )
Wherein s score based on Keywords matching in representing conventional retrieval model, as language model retrieves score Or BM25 retrieves score, s ' represents the theme matching score representing Q ' based on new inquiry, and λ is the weight ginseng between both scores Number, is also the balance coefficient of two kinds of matching ways.
About theme matching score, multiple computational methods can be used.Specifically, given new inquiry represents and the master of document Topic distribution vector, can be obtained, such as Jensen-Shannon by the theme distribution similarity calculated between the two Divergence (JSD) and cosine similarity (Cosine similarity).
The protection content of the present invention is not limited to above example.Under the spirit and scope without departing substantially from inventive concept, this Skilled person it is conceivable that change and advantage be all included in the present invention, and with appending claims for protect Protect scope.

Claims (6)

1. an inquiry based on context-aware theme represents and mixed index method for establishing model, it is characterised in that include Following steps:
Step one: keyword set based on inquiry, obtains the pseudo-linear filter document of described inquiry, from described pseudo-linear filter Document is chosen the context relevant to described inquiry；
Step 2: introduce context-aware topic model, described context is incorporated in described context-aware topic model, base In the subject information that contextual window described in corpus Topics Crawling is implied, obtain its corresponding theme vector；
Step 3: described inquiry is combined with described keyword set with described theme vector and represents, based on described theme vector With described keyword set, set up mixed index model, obtain final retrieval score.
2. inquiry based on context-aware theme as claimed in claim 1 represents and mixed index method for establishing model, its It is characterised by, described pseudo-linear filter document is divided into multiple sliding window by step one, and calculates each window and institute State the dependency of inquiry, take the dependency window higher than threshold value as the contextual window relevant to described inquiry.
3. inquiry based on context-aware theme as claimed in claim 2 represents and mixed index method for establishing model, its Being characterised by, described context selected threshold associated with the query is the meansigma methods of all window dependencys under this inquiry.
4. inquiry based on context-aware theme as claimed in claim 1 represents and mixed index method for establishing model, its Being characterised by, described context-aware topic model is according to designed by inquiry related context and whole corpus, utilizes institute State context-aware topic model and in theme modeling process, assume the pseudo-linear filter document at contextual window and its place altogether Enjoy same theme distribution, obtain the theme vector of context.
5. inquiry based on context-aware theme as claimed in claim 1 represents and mixed index method for establishing model, its Being characterised by, described pseudo-linear filter document uses retrieval model Keywords matching score to calculate and obtains.
6. inquiry based on context-aware theme as claimed in claim 1 represents and mixed index method for establishing model, its Being characterised by, described retrieval score represents with equation below:
S = ( 1 - λ ) Σ q i ∈ Q s ( q i , d ) + λ · s ′ ( Q ′ , d )
Wherein, s represents score based on Keywords matching in conventional retrieval model, and s ' represents the theme representing Q ' based on new inquiry Matching score, λ is the weight parameter between both scores, is also the balance coefficient of two kinds of matching ways.
CN201610634174.2A 2016-08-05 2016-08-05 Inquiry based on context-aware theme represents and mixed index method for establishing model Pending CN106294662A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610634174.2A CN106294662A (en) 2016-08-05 2016-08-05 Inquiry based on context-aware theme represents and mixed index method for establishing model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610634174.2A CN106294662A (en) 2016-08-05 2016-08-05 Inquiry based on context-aware theme represents and mixed index method for establishing model

Publications (1)

Publication Number Publication Date
CN106294662A true CN106294662A (en) 2017-01-04

Family

ID=57664982

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610634174.2A Pending CN106294662A (en) 2016-08-05 2016-08-05 Inquiry based on context-aware theme represents and mixed index method for establishing model

Country Status (1)

Country Link
CN (1) CN106294662A (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108121699A (en) * 2017-12-21 2018-06-05 北京百度网讯科技有限公司 For the method and apparatus of output information
CN108520033A (en) * 2018-03-28 2018-09-11 华中师范大学 Enhancing pseudo-linear filter model information search method based on superspace simulation language
CN108710611A (en) * 2018-05-17 2018-10-26 南京大学 A kind of short text topic model generation method of word-based network and term vector
CN108804443A (en) * 2017-04-27 2018-11-13 安徽富驰信息技术有限公司 A kind of judicial class case searching method based on multi-feature fusion
CN110333700A (en) * 2019-05-24 2019-10-15 蓝炬兴业(赤壁)科技有限公司 Industrial computer server remote management platform system and method
CN110427400A (en) * 2019-06-21 2019-11-08 贵州电网有限责任公司 Search method is excavated based on operation of power networks information interactive information user's demand depth
CN111897928A (en) * 2020-08-04 2020-11-06 广西财经学院 Chinese query expansion method for embedding expansion words into query words and counting expansion word union
CN112685440A (en) * 2020-12-31 2021-04-20 王程 Structural query information expression method for marking search semantic role
WO2021250488A1 (en) * 2020-06-08 2021-12-16 International Business Machines Corporation Refining a search request to a content provider

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102750315A (en) * 2012-04-25 2012-10-24 北京航空航天大学 Rapid discovering method of conceptual relations based on sovereignty iterative search
CN103678412A (en) * 2012-09-21 2014-03-26 北京大学 Document retrieval method and device
CN103927177A (en) * 2014-04-18 2014-07-16 扬州大学 Characteristic-interface digraph establishment method based on LDA model and PageRank algorithm
CN104050235A (en) * 2014-03-27 2014-09-17 浙江大学 Distributed information retrieval method based on set selection
CN104391942A (en) * 2014-11-25 2015-03-04 中国科学院自动化研究所 Short text characteristic expanding method based on semantic atlas

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102750315A (en) * 2012-04-25 2012-10-24 北京航空航天大学 Rapid discovering method of conceptual relations based on sovereignty iterative search
CN103678412A (en) * 2012-09-21 2014-03-26 北京大学 Document retrieval method and device
CN104050235A (en) * 2014-03-27 2014-09-17 浙江大学 Distributed information retrieval method based on set selection
CN103927177A (en) * 2014-04-18 2014-07-16 扬州大学 Characteristic-interface digraph establishment method based on LDA model and PageRank algorithm
CN104391942A (en) * 2014-11-25 2015-03-04 中国科学院自动化研究所 Short text characteristic expanding method based on semantic atlas

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108804443A (en) * 2017-04-27 2018-11-13 安徽富驰信息技术有限公司 A kind of judicial class case searching method based on multi-feature fusion
CN108121699A (en) * 2017-12-21 2018-06-05 北京百度网讯科技有限公司 For the method and apparatus of output information
CN108520033A (en) * 2018-03-28 2018-09-11 华中师范大学 Enhancing pseudo-linear filter model information search method based on superspace simulation language
CN108710611B (en) * 2018-05-17 2021-08-03 南京大学 Short text topic model generation method based on word network and word vector
CN108710611A (en) * 2018-05-17 2018-10-26 南京大学 A kind of short text topic model generation method of word-based network and term vector
CN110333700A (en) * 2019-05-24 2019-10-15 蓝炬兴业(赤壁)科技有限公司 Industrial computer server remote management platform system and method
CN110427400A (en) * 2019-06-21 2019-11-08 贵州电网有限责任公司 Search method is excavated based on operation of power networks information interactive information user's demand depth
WO2021250488A1 (en) * 2020-06-08 2021-12-16 International Business Machines Corporation Refining a search request to a content provider
US11238052B2 (en) 2020-06-08 2022-02-01 International Business Machines Corporation Refining a search request to a content provider
GB2611237A (en) * 2020-06-08 2023-03-29 Ibm Refining a search request to a content provider
AU2021289542B2 (en) * 2020-06-08 2023-06-01 International Business Machines Corporation Refining a search request to a content provider
CN111897928A (en) * 2020-08-04 2020-11-06 广西财经学院 Chinese query expansion method for embedding expansion words into query words and counting expansion word union
CN112685440A (en) * 2020-12-31 2021-04-20 王程 Structural query information expression method for marking search semantic role
CN112685440B (en) * 2020-12-31 2022-03-22 上海欣兆阳信息科技有限公司 Structural query information expression method for marking search semantic role

Similar Documents

Publication Publication Date Title
CN106294662A (en) Inquiry based on context-aware theme represents and mixed index method for establishing model
CN104239513A (en) Semantic retrieval method oriented to field data
CN112434169A (en) Knowledge graph construction method and system and computer equipment
CN105653840A (en) Similar case recommendation system based on word and phrase distributed representation, and corresponding method
CN105095433A (en) Recommendation method and device for entities
CN102708100A (en) Method and device for digging relation keyword of relevant entity word and application thereof
CN105045933B (en) Mapping method between apparel maintenance support information relationship database schema and body
WO2007143898A1 (en) Method for information retrieval and processing based on ternary model
CN108038133A (en) Personalized recommendation method
Cowan et al. Named entity recognition in travel-related search queries
Koopman et al. Graph-based concept weighting for medical information retrieval
Zhang Review Essay: How to Merge Western Theories and Chinese Indigenous Theories to Study Chinese Politics?
CN105956158A (en) Automatic extraction method of network neologism on the basis of mass microblog texts and use information
MacAvaney et al. Characterizing question facets for complex answer retrieval
Zadrozny et al. Remarks on various aspects of bipolarity in database querying
An et al. A heuristic approach on metadata recommendation for search engine optimization
CN112069306B (en) Paper partner recommendation method based on author writing tree and graph neural network
CN103886049B (en) Method for mining heterogeneous related data set in data space
Liang et al. Finding knowledgeable groups in enterprise corpora
Zhang et al. Improving entity linking in Chinese domain by sense embedding based on graph clustering
Pang et al. Query expansion and query fuzzy with large-scale click-through data for microblog retrieval
Han et al. HLJIT at TREC 2017 Real-Time Summarization.
Zhao et al. Entity-Based Language Model Smoothing Approach for Smart Search
Wang et al. A thesaurus and online encyclopedia merging method for large scale domain-ontology automatic construction
Gao et al. Mining frequent sets using fuzzy multiple-level association rules

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20170104