CN111159396A - Method for establishing text data classification hierarchical model facing data sharing exchange - Google Patents
Method for establishing text data classification hierarchical model facing data sharing exchange Download PDFInfo
- Publication number
- CN111159396A CN111159396A CN201911224374.0A CN201911224374A CN111159396A CN 111159396 A CN111159396 A CN 111159396A CN 201911224374 A CN201911224374 A CN 201911224374A CN 111159396 A CN111159396 A CN 111159396A
- Authority
- CN
- China
- Prior art keywords
- data
- text
- classification
- establishing
- bigru
- 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
Images
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
- G06F16/35—Clustering; Classification
-
- 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
- G06F16/35—Clustering; Classification
- G06F16/355—Class or cluster creation or modification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a method for establishing a text data classification and classification model for data sharing exchange, which comprises the following steps: vectorizing text data to form a structured description of the data; classifying the data after opposite quantization by using a BiGRU-CNN serial mixed classification model; thirdly, carrying out security grade division on the classified data; and step four, establishing a data grading classification model. Compared with the prior art, the invention has the following positive effects: the data classification is carried out by combining artificial intelligence technology with artificial subjective judgment, data classification is carried out from the safety angle, a more appropriate and accurate safety strategy is provided when fine-grained protection is carried out on different types or level data in the sharing exchange process, the data protection efficiency is improved while the data safety protection is improved, the automation degree of the processing process is high, and the classification result accuracy is high.
Description
Technical Field
The invention relates to a method for establishing a text data classification and classification model for data sharing exchange.
Background
The information security level protection system is implemented in China at present, and the proposed protection idea of dividing regions and giving emphasis is an effective means for solving the current information security problem. Data management is more chaotic in the current data sharing exchange process, the same or similar protective measures are taken for different data, the protection granularity is thicker, great hidden danger is brought to data sharing exchange safety, and once some sensitive data fall into the hands of the people centering on the measurement, the interest of an enterprise or even the national security is seriously influenced. Therefore, fine-grained protection of data is an important content of information security, research on the aspect at home and abroad is few, classification of data proposed in the guidelines of the examination committee of the federal financial institution in the united states is part of risk assessment, but classification and protection methods are not provided, the requirements of relevant standards of national grade protection on data protection are few, and no mature mode can be used for reference at present for classifying and classifying data from the security perspective.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention provides a method for establishing a text data classification hierarchical model facing data sharing exchange, aiming at the requirements of fine-grained protection in the existing data sharing exchange process, and mainly solving the following technical problems:
(1) the text data quantization identification method does not have the capability of forming a structural description by carrying out quantization identification on the text data;
(2) there is no clear quantitative evaluation method for the grading of data;
(3) the text data classification method does not have the capability of automatically classifying and grading the text data, and cannot effectively perform difference protection on the data.
The technical scheme adopted by the invention for solving the technical problems is as follows: a method for establishing a text data classification and classification model facing data sharing exchange comprises the following steps:
vectorizing text data to form a structured description of the data;
classifying the data after opposite quantization by using a BiGRU-CNN serial mixed classification model;
thirdly, carrying out security grade division on the classified data;
and step four, establishing a data grading classification model.
Compared with the prior art, the invention has the following positive effects:
(1) the data classification is carried out by combining artificial intelligence technology with artificial subjective judgment, the data classification is carried out from the safety perspective, a more appropriate and accurate safety strategy is provided when fine-grained protection is carried out on data of different classes or levels in the sharing exchange process, the data protection efficiency is improved while the data safety protection is improved, the automation degree of the processing process is high, and the classification result is high in accuracy;
(2) the method can be widely applied to data sharing and exchanging products, different safety prevention and control strategies can be automatically and accurately formulated according to the classification and classification model, and important protection can be performed on the heavy data.
Drawings
The invention will now be described, by way of example, with reference to the accompanying drawings, in which:
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a flow chart of a BiGRU-CNN serial mixed classification model.
Detailed Description
A method for establishing a text data classification hierarchical model facing data sharing exchange is shown in figure 1.
The method comprises the following steps:
1) performing quantitative identification on text data by taking key words as main parts through a text vectorization technology to form a structural description of the data;
1.1) input text set P ═ { P ═ P1,P2,...,Pt,...,PMIn which P istFor the t-th text, M is the total number of texts. Setting the number of extracted keywords of each text as MkeyDamping factor d, sliding window width omega, iteration stop threshold sigma, maximum iteration numberGmax;
1.2) performing word segmentation on each text in the text set P by using a word segmentation algorithm and stopping words;
1.3) calculating the TF-IDF value of each corresponding word in the text set P by using a TF-IDF algorithm;
1.4) to text PtVectorization is performed, t is 1, 2.., M;
1.4.1) based on the width omega, sliding a window to sweep through a text phrase, and constructing an edge between any two points by adopting a co-occurrence relation, thereby constructing a PtGraph G formed by Chinese wordst;
1.4.2) iteratively calculating graph G according totScoring each node sigma until the difference of the two scoring is less than sigma;
wherein, W (V)i)TF-IDFIs the word ViTF-IDF value of, ViAnd VjRespectively show diagram GtMiddle ith and j node, S (V)i) Is the node weight score of the ith node, I (V)i) Is PtAll of which are connected to ViA set of nodes of (a); o (V)j) Is PtMiddle node ViA collection of connected nodes.
1.4.3) pairs of GtThe weight scores of the nodes are arranged in descending order, and M is taken as the topkeyThe words corresponding to the nodes form a keyword set K;
1.4.4) calculating the text P according totThe word vector of (2);
wherein VkAnd W (V)k)TF-IDFAnd respectively representing the word vector of the kth keyword in the K and the TF-IDF value corresponding to the keyword.
2) And classifying the vectorized data according to a BiGRU-CNN serial mixed classification model. The process of the BiGRU-CNN serial mixed classification model is shown in FIG. 2. The method comprises the following specific steps:
2.1.1) embedding layer: the word vectors of the text are spliced to generate a sentence-level word vector matrix, and the generation formula is as follows:
Sij=V1⊕V2⊕…⊕Vi
wherein SijA word vector matrix representing the jth sentence spliced by the i word vectors, and ⊕ representing the word vector splicing operation.
2.1.2) BiGRU: sentence matrix S with Embedding layerijAnd inputting the input sequence from the positive direction and the negative direction of the BiGRU layer, extracting the upper information characteristic of the text and the lower information characteristic of the text through the hidden layer, and finally splicing the hidden layer outputs in the two directions through the following formula.
hij=BiGRU(Sijt)
2.1.3) Conv: local semantic features were extracted for each of 128 CNN networks using filters 3 x 100, 4 x 100, 5 x 100 and calculated by the following equation.
Oijt=CNN(hijt)
Wherein h isijtRepresents the output of the BiGRU at time t; o isijtIndicating the output at time CNN.
2.1.4) reducing the dimension of the output matrix by utilizing global average pooling, fusing local semantic feature semantics extracted by different convolution kernels in the CNN, and finally connecting by using a full connection layer.
2.1.5) output: text classification is performed by a Softmax function, which is shown as follows:
yi=Softmax(widijt+bi)
wherein: w is aiWeight coefficient matrix, w, representing the Dense layer to the output layeriRepresenting the corresponding offset, dijtIndicating the input of the Dense layer at time tOutputting a vector;
2.2) further training a BiGRU-CNN text classification model through classification actual combat simulation, increasing decision data and improving the capability of the classification model to cope with abnormity.
2.2.1) classifying the actual text service data by using a BiGRU-CNN model, and recording the classified result;
2.2.2) adding the classified text data serving as new training data into a training sample set;
and 2.3) carrying out auxiliary judgment on the text data with the abnormal classification result manually, and inputting the data into a classification model after manual decision making to be used as a training sample for relearning.
3) Referring to the relevant hierarchical mode in the 'equal security', the security level of the classified data is divided into 4 levels, namely a first level, a second level, a third level and a fourth level. Each grade is assigned a numerical value: first-grade 3.5 ═ t ═ 4.0, second-grade 3 ═ t <3.5, third-grade 1.5 ═ t <3 and fourth-grade t < 1.5.
4) And establishing a data classification model, which is equivalent to providing a data protection reference standard for making a data security strategy when fine-grained protection is carried out on data in the data sharing exchange process. The method comprises the following specific steps:
4.1) grading the classified data from three dimensions, namely data type, influence range and object and loss influence, wherein the grading descriptions of the three dimensions are shown in the table 1.
TABLE 1 three-dimensional Scoring description of evaluation data levels
Because the data grade is 4 grades, the score of each dimension scoring data type is the highest and can only be 4 scores, the lowest score is 1 score, the score of each dimension data type is determined by the importance of the data, for example, the cadre investigation material data can be assigned 4 scores, the internal audit report data can be assigned 3 scores, the important customer data can be assigned 3 scores, the market research report data can be assigned 2 scores, and the retail price and other publicly available data can be assigned 1 score.
4.2) the final score of the data is a three-dimensional composite value, and the composite score will obtain a data grade score t ═ data type score × 25% + impact range and object × 25% + loss impact estimation score × 50% according to the following formula. Each dimension weight division is obtained according to the average value of three dimension weight scores given by professionals of a sampling survey institution, and the method has objectivity, scientificity and usability.
4.3) rating the data according to the data comprehensive scores.
Claims (8)
1. A method for establishing a text data classification and classification model facing data sharing exchange is characterized in that: the method comprises the following steps:
vectorizing text data to form a structured description of the data;
classifying the data after opposite quantization by using a BiGRU-CNN serial mixed classification model;
thirdly, carrying out security grade division on the classified data;
and step four, establishing a data grading classification model.
2. The method for establishing the text data classification hierarchical model facing the data sharing exchange according to the claim 1, is characterized in that: step one, the method for vectorizing the text data comprises the following steps:
1.1) input text set P ═ { P ═ P1,P2,...,Pt,...,PMIn which P istSetting the number of extracted keywords of each text as M for the t-th text and M for the total number of the textskeyDamping factor d, sliding window width ω, iteration stop threshold σ, maximum iteration number Gmax;
1.2) performing word segmentation on each text in the text set P by using a word segmentation algorithm and removing stop words;
1.3) calculating the TF-IDF value of each corresponding word in the text set P by using a TF-IDF algorithm;
1.4) vectorizing each text in the text set P to obtain a word vector of each text;
1.5) outputting a text vector set of the text set P.
3. The method for establishing the text data classification grading model facing the data sharing exchange according to the claim 2, characterized in that: the method for vectorizing each text comprises the following steps:
1.4.1) based on the width omega, sliding a window to sweep through a text phrase, and constructing an edge between any two points by adopting a co-occurrence relation, thereby constructing a PtGraph G composed of words in (1)t;
1.4.2) iteratively calculating graph G according totThe score of each node sigma until the difference of the two scores is less than sigma:
wherein, W (V)i)TF-IDFIs the word ViTF-IDF value of, ViAnd VjRespectively show diagram GtMiddle ith and j node, S (V)i) Is the node weight score of the ith node, I (V)i) Is PtAll of which are connected to ViA set of nodes of (a); o (V)j) Is PtMiddle node ViA set of connected nodes;
1.4.3) pairs of GtThe weight scores of the nodes are arranged in descending order, and M is taken as the topkeyThe words corresponding to the nodes form a keyword set K;
1.4.4) calculating the text P according totThe word vector of (2);
wherein VkAnd W (V)k)TF-IDFAnd respectively representing the word vector of the kth keyword in the set K and the TF-IDF value corresponding to the keyword.
4. The method for establishing the text data classification hierarchical model facing the data sharing exchange according to the claim 1, is characterized in that: step two, the method for classifying the data after the opposite quantization comprises the following steps:
2.1) establishing a BiGRU-CNN serial mixed classification model;
2.2) further training a BiGRU-CNN serial mixed classification model through classification actual combat simulation;
and 2.3) carrying out auxiliary judgment on the text data with the abnormal classification result manually, and inputting the data into a classification model after manual decision making to be used as a training sample for relearning.
5. The method for establishing the text data classification grading model facing the data sharing exchange according to the claim 4, characterized in that: the BiGRU-CNN serial mixed classification model comprises the following components:
2.1.1) embedding layer: splicing word vectors of the text to generate a sentence-level word vector matrix:
wherein SijA word vector matrix representing the jth sentence spliced by the i word vectors;splicing operation of the expression word vectors;
2.1.2) BiGRU: with SijInputting an input sequence from the positive direction and the negative direction of the input sequence for the input of the BiGRU layer, extracting the upper information characteristic of a text and the lower information characteristic of the text through a hidden layer, and finally splicing the hidden layer outputs in the two directions through the following formula:
hij=BiGRU(Sijt)
2.1.3) Conv: local semantic features were extracted for each of 128 CNN networks 3 × 100, 4 × 100, 5 × 100 using filters and calculated by:
Oijt=CNN(hijt)
wherein h isijtRepresents the output of the BiGRU at time t; o isijtRepresents the output of CNN at time t;
2.1.4) reducing the dimension of the output matrix by utilizing global average pooling, fusing local semantic feature semantics extracted by different convolution kernels in the CNN, and finally connecting by using a full connection layer;
2.1.5) output: text classification is performed by a Softmax function, which is shown as follows:
yi=Softmax(widijt+bi)
wherein: w is aiWeight coefficient matrix, w, representing the Dense layer to the output layeriRepresenting the corresponding offset, dijtThe output vector at the sense layer at time t is shown.
6. The method for establishing the text data classification grading model facing the data sharing exchange according to the claim 4, characterized in that: the method for further training the BiGRU-CNN serial mixed classification model comprises the following steps:
2.2.1) classifying the actual text service data by using a BiGRU-CNN model, and recording the classified result;
2.2.2) adding the classified text data serving as new training data into a training sample set.
7. The method for establishing the text data classification hierarchical model facing the data sharing exchange according to the claim 1, is characterized in that: thirdly, the method for classifying the safety grade of the classified data comprises the following steps:
3.1) dividing the safety level of the classified data into 4 levels according to the grading mode of grade protection, wherein the 4 levels are respectively a first level, a second level, a third level and a fourth level;
3.2) assign the following to each class:
first-grade 3.5 ═ t ═ 4.0, second-grade 3 ═ t <3.5, third-grade 1.5 ═ t <3 and fourth-grade t < 1.5.
8. The method for establishing the text data classification hierarchical model facing the data sharing exchange according to the claim 1, is characterized in that: step four, the method for establishing the data classification model comprises the following steps:
4.1) grading the classified data from three dimensions of data type, influence range and object and loss influence estimation respectively;
4.2) calculating the comprehensive score of the classified data:
a composite score t ═ data type score × 25% + impact range and an object score × 25% + loss impact estimate score × 50%;
4.3) rating the data according to the data comprehensive scores.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911224374.0A CN111159396B (en) | 2019-12-04 | 2019-12-04 | Method for establishing text data classification hierarchical model facing data sharing exchange |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911224374.0A CN111159396B (en) | 2019-12-04 | 2019-12-04 | Method for establishing text data classification hierarchical model facing data sharing exchange |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111159396A true CN111159396A (en) | 2020-05-15 |
CN111159396B CN111159396B (en) | 2022-04-22 |
Family
ID=70556419
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911224374.0A Active CN111159396B (en) | 2019-12-04 | 2019-12-04 | Method for establishing text data classification hierarchical model facing data sharing exchange |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111159396B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113240024A (en) * | 2021-05-20 | 2021-08-10 | 贾晓丰 | Data classification method and device based on deep learning clustering algorithm |
CN113792202A (en) * | 2021-08-31 | 2021-12-14 | 中国电子科技集团公司第三十研究所 | Screening method for user classification |
CN114172636A (en) * | 2020-09-11 | 2022-03-11 | 军事科学院系统工程研究院网络信息研究所 | Hybrid secure communication method for encrypting critical data quanta |
CN117278252A (en) * | 2023-08-17 | 2023-12-22 | 武汉易通商融信息技术有限公司 | Big data based secure information sharing processing method, system and storage medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106408140A (en) * | 2015-07-27 | 2017-02-15 | 广州西麦信息科技有限公司 | Grading and classifying model method based on power grid enterprise data |
AU2018101513A4 (en) * | 2018-10-11 | 2018-11-15 | Hui, Bo Mr | Comprehensive Stock Prediction GRU Model: Emotional Index and Volatility Based |
CN109189925A (en) * | 2018-08-16 | 2019-01-11 | 华南师范大学 | Term vector model based on mutual information and based on the file classification method of CNN |
CN109902175A (en) * | 2019-02-20 | 2019-06-18 | 上海方立数码科技有限公司 | A kind of file classification method and categorizing system based on neural network structure model |
CN110110080A (en) * | 2019-03-29 | 2019-08-09 | 平安科技(深圳)有限公司 | Textual classification model training method, device, computer equipment and storage medium |
CN110502749A (en) * | 2019-08-02 | 2019-11-26 | 中国电子科技集团公司第二十八研究所 | A kind of text Relation extraction method based on the double-deck attention mechanism Yu two-way GRU |
-
2019
- 2019-12-04 CN CN201911224374.0A patent/CN111159396B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106408140A (en) * | 2015-07-27 | 2017-02-15 | 广州西麦信息科技有限公司 | Grading and classifying model method based on power grid enterprise data |
CN109189925A (en) * | 2018-08-16 | 2019-01-11 | 华南师范大学 | Term vector model based on mutual information and based on the file classification method of CNN |
AU2018101513A4 (en) * | 2018-10-11 | 2018-11-15 | Hui, Bo Mr | Comprehensive Stock Prediction GRU Model: Emotional Index and Volatility Based |
CN109902175A (en) * | 2019-02-20 | 2019-06-18 | 上海方立数码科技有限公司 | A kind of file classification method and categorizing system based on neural network structure model |
CN110110080A (en) * | 2019-03-29 | 2019-08-09 | 平安科技(深圳)有限公司 | Textual classification model training method, device, computer equipment and storage medium |
CN110502749A (en) * | 2019-08-02 | 2019-11-26 | 中国电子科技集团公司第二十八研究所 | A kind of text Relation extraction method based on the double-deck attention mechanism Yu two-way GRU |
Non-Patent Citations (3)
Title |
---|
ZHANG JINGREN等: "feature fusion text classification model combining CNN and Bigru with multi-attention mechanism", 《FUTURE INTERNET》 * |
苟先太等: "基于GRU-Attention神经网络的空中群组态势识别方法", 《计算机与现代化》 * |
陈洁等: "基于并行混合神经网络模型的短文本情感分析", 《计算机应用》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114172636A (en) * | 2020-09-11 | 2022-03-11 | 军事科学院系统工程研究院网络信息研究所 | Hybrid secure communication method for encrypting critical data quanta |
CN114172636B (en) * | 2020-09-11 | 2024-02-20 | 军事科学院系统工程研究院网络信息研究所 | Hybrid safety communication method for key data quantum encryption |
CN113240024A (en) * | 2021-05-20 | 2021-08-10 | 贾晓丰 | Data classification method and device based on deep learning clustering algorithm |
CN113792202A (en) * | 2021-08-31 | 2021-12-14 | 中国电子科技集团公司第三十研究所 | Screening method for user classification |
CN113792202B (en) * | 2021-08-31 | 2023-05-05 | 中国电子科技集团公司第三十研究所 | User classification screening method |
CN117278252A (en) * | 2023-08-17 | 2023-12-22 | 武汉易通商融信息技术有限公司 | Big data based secure information sharing processing method, system and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN111159396B (en) | 2022-04-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111159396B (en) | Method for establishing text data classification hierarchical model facing data sharing exchange | |
CN109684440B (en) | Address similarity measurement method based on hierarchical annotation | |
CN109657947B (en) | Enterprise industry classification-oriented anomaly detection method | |
CN110852856B (en) | Invoice false invoice identification method based on dynamic network representation | |
CN110929034A (en) | Commodity comment fine-grained emotion classification method based on improved LSTM | |
CN110135459B (en) | Zero sample classification method based on double-triple depth measurement learning network | |
CN109960727B (en) | Personal privacy information automatic detection method and system for unstructured text | |
CN111831824A (en) | Public opinion positive and negative face classification method | |
CN105069072A (en) | Emotional analysis based mixed user scoring information recommendation method and apparatus | |
CN109240258A (en) | Vehicle failure intelligent auxiliary diagnosis method and system based on term vector | |
CN104021300B (en) | Comprehensive assessment method based on effect of distribution type electrical connection on power distribution network | |
CN109102157A (en) | A kind of bank's work order worksheet processing method and system based on deep learning | |
CN111597331A (en) | Judgment document classification method based on Bayesian network | |
CN111177402A (en) | Evaluation method and device based on word segmentation processing, computer equipment and storage medium | |
CN107392129A (en) | Face retrieval method and system based on Softmax | |
TWI477987B (en) | Methods for sentimental analysis of news text | |
CN114757557A (en) | On-site operation risk assessment prediction method and device based on electric work ticket | |
CN112417893A (en) | Software function demand classification method and system based on semantic hierarchical clustering | |
CN113157918A (en) | Commodity name short text classification method and system based on attention mechanism | |
CN110826315B (en) | Method for identifying timeliness of short text by using neural network system | |
CN115238685A (en) | Combined extraction method for building engineering change events based on position perception | |
CN111079582A (en) | Image recognition English composition running question judgment method | |
CN107577738A (en) | A kind of FMECA method by SVM text mining processing datas | |
CN107480126B (en) | Intelligent identification method for engineering material category | |
CN111191029B (en) | AC construction method based on supervised learning and text classification |
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 |