CN106570164A - Integrated foodstuff safety text classification method based on deep learning - Google Patents
Integrated foodstuff safety text classification method based on deep learning Download PDFInfo
- Publication number
- CN106570164A CN106570164A CN201610976304.0A CN201610976304A CN106570164A CN 106570164 A CN106570164 A CN 106570164A CN 201610976304 A CN201610976304 A CN 201610976304A CN 106570164 A CN106570164 A CN 106570164A
- Authority
- CN
- China
- Prior art keywords
- language material
- corpus
- food safety
- label
- deep learning
- 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
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
- G06F16/35—Clustering; Classification
-
- 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
-
- 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
- G06N3/08—Learning methods
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)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides an integrated foodstuff safety text classification method based on deep learning. The invention relates to a Chinese-character foodstuff safety text classification method which comprises the following steps of S1, performing news corpus acquisition, and obtaining three kinds of files, namely foodstuff safety files, foodstuff-safety-unrelated files and no-label files; S2, processing the corpus for dividing into a training corpus and a testing corpus; S3, in a training period, firstly dividing negative samples in the training corpus into N sets, combining the negative samples and the positive samples in each set for obtaining a training corpus subset, secondly, performing training according to a deep learning method by means of the subset for generating a base classifier, and finally, forming a combined classifier based on deep learning by the N base classifiers; and S4, in a testing period, performing classification on the testing corpus by means of the combined classifier, thereby obtaining N classification labels, and performing ballot on the labels according to a majority voting system, thereby obtaining a final classification label. The integrated foodstuff safety text classification method can settle a single-kind classification problem in Chinese-character foodstuff safety corpus and accurately screens foodstuff safety related reports from network news.
Description
Technical field
The present invention relates to natural language processing field, more particularly to the integrated form food safety text based on deep learning
Sorting technique.
Background technology
After people's living standard is improved, food safety increasingly becomes popular focus of interest, according to " 2015
Chinese comprehensive well-to-do level index " its " ten large focal spot problem of greatest concern " investigation shows that food safety is with 44.8% attention rate
Occupy first place.The varieties of food items security incidents such as pesticide residues exceeded, fowl poultry kind " short-term training ", the illegal additive of food of fruit and vegerable
Again and again occur so that food-safety problem is more severe, exacerbate popular worry safe to food.Simultaneously as the Internet
Convenience and ageing, increasing food safety affair selects to be exposed on the network media.So, supervise network food
Food Safety Analysis are reported and carried out to security news, and this has become a kind of important method of food-safety problem research.But,
The first step of this method seeks to obtain food safety related news (i.e. Chinese food automatically from the internet news of magnanimity
The safe text classification of product), this but not a duck soup.
At present most of text classification problems are many classification (or two classification) problems, in order to reach higher classifying quality,
Current Algorithm of documents categorization adopts full supervised classification method mostly.But, this dividing based on full supervised classification method
Class effect is highly dependent on the quality of artificial mark language material, and the transplantability of disaggregated model is not high.For Chinese food safety
For text classification problem, at present the food safety news corpus of artificial mark only include food safety class document, not including non-
Food safety class document.This classification problem comprising a class example is commonly known as single class classification problem, its corresponding skill
Art research is also less.At present list class classification problem often adopts unsupervised segmentation technology, but its sorting technique effect is bad.Cause
This, needs exploitation for the Automatic document classification method of Chinese food safety.
The content of the invention
(1) technical problem to be solved
The technical problem to be solved in the present invention is to provide a kind of Chinese food safety file classification method, can be from network
Food safety related news are automatically identified, the Chinese food safety news corpus for solving artificial mark at present are single class classification languages
Material, and for bad the two problems of sorting technique effect of single class classification problem.
(2) technical scheme
In order to solve above-mentioned technical problem, the invention provides a kind of integrated form food safety text based on deep learning
Sorting technique, the method comprising the steps of：
Step S1：From《Viewpoint of Chinese food safety event set (2001-2011)》With acquisition Chinese on some food news websites
News corpus, carry out pretreatment, obtain food safety class document language material and without label class document language material；
Step S2：It is combined by the food safety class document language material and without label class document language material, obtains training language
Material and testing material；
Step S3：The corpus are divided into training subset, base grader is trained using training subset, according to integrated
Learning method trains assembled classifier；
Step S4：The testing material is classified according to the assembled classifier, is instructed according to majority voting system
Practice the final classification label of example, so as to obtain food safety class document.
Further, step S1 is specifically included：
It is right《Viewpoint of Chinese food safety event set (2001-2011)》In news documents be formatted process, extract wherein
Title, date, source, summary, textual data, be saved in data base using unified form, obtain food safety class text
Shelves language material；
News is collected from some food news websites, the information such as title therein, date, source, summary, text are extracted,
It is saved in data base using unified form, is obtained without label class document language material；
Further, step S2 is specifically included：
The food safety class document language material is divided into into the positive example language material of training and the positive example language material of test；
By described training is divided into without label language material and test without label language material without label class document language material；
The training is born into example language material without label class document language material as training；
Using the training without label language material in without label example as negative example (i.e. pseudo- negative example), form training pseudo- negative
Example language material；
Testing material should include not bearing example language in positive example language material and negative example language material, but the data for collecting
Material, therefore go out a part without random choose in label language material from the test and manually marked, obtain testing negative example language material；
It is using the pseudo- negative example language material of the positive example language material of the training and the training as corpus, the test is positive real
Illustrative phrase material and the negative example language material of the test are used as testing material；
Further, step S3 is specifically included：
For the corpus, all words in the title and text in each document are extracted as the spy of the document
Reference ceases；
Because critical noisy is asked present in the disequilibrium and pseudo- negative example of the positive example and pseudo- negative example quantity
Topic, therefore according to the characteristic information, using the multiple base graders of LSTM depth sorting Algorithm for Training, by the classification of the plurality of base
Device is combined, and constitutes the assembled classifier based on deep learning, concretely comprises the following steps：
The pseudo- negative example language material is N times of the positive example language material, so the pseudo- negative example language material is divided into into N groups,
The negative example language material of per group of puppet and the positive example language material are a training subset, obtain N number of training subset；
For described each training subset is learnt by LSTM depth sorting algorithms, N number of base grader is obtained；
N number of base classifiers combination is obtained into assembled classifier；
Further, step S4 is specifically included：
For each test case in the testing material, one can be obtained using the base grader classification described in
Label, so N number of base grader can obtain N number of tag along sort, according to majority voting system (if poll 1:1 label is 0) to carry out
Ballot, used as the final classification label of the test case, final classification label is food peace for the test case of " 1 " to voting results
Universal class document, final classification label is non-food safety class document for the test case of " 0 ".
Description of the drawings
Fig. 1 is based on the integrated form food safety file classification method flow chart of deep learning；
Fig. 2 is based on the document classification model support composition of integrated learning approach.
Specific embodiment
To make present disclosure clearer, embodiment of the present invention is carried out specifically below in conjunction with accompanying drawing
It is bright.
The integrated form food safety file classification method based on deep learning that the present invention is provided, can be automatically new from network
Food safety report is recognized in news, and recognition result accuracy increases compared to existing textual classification model.Its work
Make flow chart as shown in Figure 1.
Step S1：The collection of language material is carried out, from《Viewpoint of Chinese food safety event set (2001-2011)》And food partner's net
Data are collected with some related web sites, the data to collecting carry out pretreatment, including：
It is right《Viewpoint of Chinese food safety event set (2001-2011)》In news documents be formatted process, extract wherein
Title, date, source, summary, textual data, be saved in data base using unified form, obtain food safety class text
Shelves language material, totally 2398 documents；
News is collected from some food news websites, the information such as title therein, date, source, summary, text are extracted,
It is saved in data base using unified form, is obtained without label class document language material, totally 11388 documents；
Step S2：Experimental data is divided into corpus and testing material；Concretely comprise the following steps：
2708 are randomly selected from the class document language material without label manually to be marked, until obtaining 500 non-foods
Product security classes document, as the negative example sample language material of test；
Pacify from 500 food safety class documents of random choose in the food safety class document, and 500 non-food stuffs
Universal class document is used as test set (totally 1000 document datas)；
From the food safety class document in remaining 1898 food safety class documents and the class document without label
Remaining 11388 are without label class document as training set (totally 13286 document datas)；
Step S3：Because data are present in corpus disequilibrium problem and noise problem, using integrated study
Method assembled classification model is trained, concretely comprise the following steps：
Negative example total sample number is approximately N (=6) times of positive example total sample number, in the training stage, bears example sample random
It is divided into N groups；
Per group of negative example sample and positive example sample are combined into a corpus subset, in each corpus subset
Sample distribution it is balanced, obtain N number of training subset；
One base classification is generated using deep learning method by LSTM models using a described corpus subset
Device, carries out N number of training subset after n times and obtains N number of base grader, N number of base grader is combined and obtain combination point
Class device；
Step S4：The testing material is classified according to the grader, obtains food safety class document；Concrete step
Suddenly it is：
Classification is carried out to testing material using N number of base grader and obtains N number of classification results；
According to majority voting system (if poll 1:1 label is set to " 0 ") N number of classification results are voted, vote
As a result as the final classification result of the test document：If " 1 ", the test document is a food safety class document；
If " 0 ", the test document is a non-food safety class document.
In order to detect the effectiveness of the integrated form food safety file classification method based on deep learning disclosed by the invention,
This patent in the textual classification model and the present invention based on LSTM to improving based on the Chinese food safety of LSTM integrated studies
Textual classification model is tested, and test result is respectively as shown in Table 1 and Table 2.+ ,-positive and negative example sample experiment knot is represented respectively
Really, Precision, Recall, F1 represent respectively precision ratio, recall ratio and F values；Consider two in Average F1 synthesis F
Class words recognition effect, is averaged by F1 and F0 and is obtained.Accuracy represents accuracy.
Table 1
Table 2
Relatively learnt by Tables 1 and 2, in general, compare baseline state (based on LSTM textual classification models), be based on
The textual classification model of LSTM integrated studies all increases in every evaluation index, and wherein accuracy improves 0.9%,
Average F1 improve 1.7%.This shows that integrated learning approach can be the imbalance problem in effectively solving data distribution and to make an uproar
Mail is inscribed.Specifically, for aligning example sample, baseline state is compared, the textual classification model based on LSTM integrated studies
Accuracy rate improves 3.7%, and recall rate improves 0.7%, F1 and improves 2.4%.For negative example sample, baseline shape is compared
State, based on the textual classification model accuracy rate of LSTM integrated studies 0.7% is improve, and recall rate improves 1.9%, F1 and improves
0.9%.So learning by comparing, traditional textual classification model based on LSTM, the text based on LSTM integrated studies are compared
This disaggregated model effect in every respect is all significantly improved.
In addition by textual classification model of the present invention based on LSTM integrated studies and the text classification based on SVM integrated studies
Model is compared, and comparative result is as shown in table 3.
Table 3
From table 3 it can be seen that in general, the textual classification model based on SVM integrated studies is compared, it is integrated based on LSTM
The textual classification model accuracy of study improves 1.6%, average F1 and improves 1.6%.This shows in terms of text classification
LSTM is better than SVM.Specifically, for aligning example sample, the textual classification model based on SVM integrated studies is compared, is based on
The textual classification model accuracy rate of LSTM integrated studies improves 3.5%, F1 and improves 1.5%.For negative example sample, phase
Than the textual classification model based on SVM integrated studies, improve based on the textual classification model accuracy rate of LSTM integrated studies
0.3%, recall rate improves 3.1%, F1 and improves 1.6%.So can be seen that from comparative result compare it is integrated based on SVM
The textual classification model of study, the effect in every respect of the textual classification model based on LSTM integrated studies is all significantly improved.
Can be seen that from both the above comparative result and be based in the present invention textual classification model of LSTM integrated studies each
Aspect has very big advantage, and automatically accurately document can be classified.
Embodiment of above is merely to illustrate the present invention, and not limitation of the present invention, about the technology of technical field
Personnel, in the case of without departing from the inventive method and scope, can also make a variety of changes, therefore the technical side of all equivalents
Case falls within scope of the invention, and the scope of patent protection proper right of the present invention requires to limit.
Claims (6)
1. a kind of integrated form food safety file classification method based on deep learning, comprises the steps：
Language material is obtained, and pretreatment is carried out to Chinese news corpus, obtain food safety class document language material and without label class document
Language material；
Using the food safety class document language material as positive example language material, example is born as pseudo- without label class document language material using described
Language material, the positive example language material and pseudo- example language material are combined, and obtain corpus and testing material；
N groups are randomly divided into by example language material is born in the corpus, per group of negative example sample and positive example sample are combined into
One corpus subset.Using N number of corpus subset, an assembled classifier based on deep learning is trained；
Chinese food safety document is obtained from the testing material using the assembled classifier based on deep learning.
2. the integrated form food safety file classification method based on deep learning according to claim 1, it is characterised in that
Language material is obtained, and pretreatment is carried out to Chinese news corpus, obtain food safety class document language material and without label class document language material,
Specifically include：
It is right《Viewpoint of Chinese food safety event set (2001-2011)》In news documents be formatted process, extract mark therein
Topic, date, source, summary, textual data, are saved in data base using unified form, obtain food safety class document language
Material；
News is collected from the main medium websites such as food partner net, title therein, date, source, summary, text etc. is extracted
Information, is saved in data base using unified form, is obtained without label class document language material.
3. the integrated form food safety file classification method based on deep learning according to claim 1, it is characterised in that
It is combined by the food safety class document language material and without label class document language material, obtains corpus and testing material, has
Body includes：
The food safety class document language material is divided into into the positive example language material of training and the positive example language material of test；
By described training is divided into without label language material and test without label language material without label class document language material；
Using the training without label language material in without label example as negative example (i.e. pseudo- negative example), form the pseudo- negative example of training
Language material；
Choose the test manually to be marked without label example without the part in label language material, form the negative example language material of test；
The pseudo- negative example language material of the positive example language material of the training and the training is combined into into corpus, by the positive example of the test
Language material and the negative example language material of the test are combined into testing material.
4. the integrated form food safety file classification method based on deep learning according to claim 1, it is characterised in that
An assembled classifier based on deep learning is trained using the corpus, is specifically included：
For the corpus, title and text in each document are extracted as the characteristic information of the document；
According to the characteristic information, using the multiple base graders of LSTM depth sorting Algorithm for Training, by the plurality of base grader
It is combined, constitutes the assembled classifier based on deep learning.
5. the integrated form food safety file classification method based on deep learning according to claim 4, it is characterised in that
According to the characteristic information, using the multiple base graders of LSTM depth sorting Algorithm for Training, the plurality of base grader is carried out
Combination, constitutes the assembled classifier based on deep learning, specifically includes：
The negative example language material of puppet in the corpus is randomly divided into into N groups, per group of the negative example language material of puppet and positive example language material
It is combined into a corpus subset；
For a corpus subset, learnt by LSTM depth sorting algorithms, obtained a base grader；In repetition
Face step n times, then obtain N number of base grader；
N number of base grader is combined, the assembled classifier based on deep learning is obtained.
6. the integrated form food safety file classification method based on deep learning according to claim 1, it is characterised in that
Chinese food safety document is obtained using the assembled classifier based on deep learning, is specifically included：
For each test case in the corpus, obtained using the assembled classifier classification based on deep learning
N number of tag along sort；
Voted according to majority voting system, voting results as the test case final classification label (if number of votes obtained is 1:
1, then final classification label be set to " 0 ").If final classification label is " 1 ", the test case is a food safety class text
Shelves；If final classification label is " 0 ", the test case is a non-food safety class document.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610976304.0A CN106570164A (en) | 2016-11-07 | 2016-11-07 | Integrated foodstuff safety text classification method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610976304.0A CN106570164A (en) | 2016-11-07 | 2016-11-07 | Integrated foodstuff safety text classification method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106570164A true CN106570164A (en) | 2017-04-19 |
Family
ID=58540223
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610976304.0A Pending CN106570164A (en) | 2016-11-07 | 2016-11-07 | Integrated foodstuff safety text classification method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106570164A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107330009A (en) * | 2017-06-14 | 2017-11-07 | 腾讯科技（深圳）有限公司 | Descriptor disaggregated model creation method, creating device and storage medium |
| CN107491433A (en) * | 2017-07-24 | 2017-12-19 | 成都知数科技有限公司 | Electric business exception financial products recognition methods based on deep learning |
| CN108509596A (en) * | 2018-04-02 | 2018-09-07 | 广州市申迪计算机系统有限公司 | File classification method, device, computer equipment and storage medium |
| CN108846018A (en) * | 2018-05-07 | 2018-11-20 | 中国农业大学 | A kind of Chinese food safety media event Information Automatic Extraction method towards news |
| CN109284383A (en) * | 2018-10-09 | 2019-01-29 | 北京来也网络科技有限公司 | Text handling method and device |
| WO2019019860A1 (en) * | 2017-07-24 | 2019-01-31 | 华为技术有限公司 | Method and apparatus for training classification model |
| CN109471941A (en) * | 2018-11-07 | 2019-03-15 | 中国电子科技集团公司第二十八研究所 | A kind of charge classification method for coping with class imbalance |
| WO2019184217A1 (en) * | 2018-03-26 | 2019-10-03 | 平安科技（深圳）有限公司 | Hotspot event classification method and apparatus, and storage medium |
| CN110795564A (en) * | 2019-11-01 | 2020-02-14 | 南京稷图数据科技有限公司 | Text classification method lacking negative cases |
| CN112199503A (en) * | 2020-10-28 | 2021-01-08 | 南京信息工程大学 | Feature enhancement based unbalanced Bi-LSTM Chinese text classification method |
| WO2021212753A1 (en) * | 2020-04-23 | 2021-10-28 | 平安科技（深圳）有限公司 | Computer performance data determining method and apparatus, computer device, and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521656A (en) * | 2011-12-29 | 2012-06-27 | 北京工商大学 | Integrated transfer learning method for classification of unbalance samples |
| JP2012155394A (en) * | 2011-01-24 | 2012-08-16 | Kddi Corp | Document classification learning control device, document classification device and computer program |
-
2016
- 2016-11-07 CN CN201610976304.0A patent/CN106570164A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2012155394A (en) * | 2011-01-24 | 2012-08-16 | Kddi Corp | Document classification learning control device, document classification device and computer program |
| CN102521656A (en) * | 2011-12-29 | 2012-06-27 | 北京工商大学 | Integrated transfer learning method for classification of unbalance samples |
Non-Patent Citations (3)
| Title |
|---|
| 张冲: "《基于Attention-Based LSTM模型的文本分类技术的研究》", 《中国优秀硕士学位论文全文数据库信息科技》 * |
| 王璐林: "《面向不平衡样本的Boosting分类算法研究》", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 祁仲昂: "《有限且非完美的标签学习》", 《中国博士学位论文全文数据库信息科技辑》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107330009A (en) * | 2017-06-14 | 2017-11-07 | 腾讯科技（深圳）有限公司 | Descriptor disaggregated model creation method, creating device and storage medium |
| CN107491433A (en) * | 2017-07-24 | 2017-12-19 | 成都知数科技有限公司 | Electric business exception financial products recognition methods based on deep learning |
| WO2019019860A1 (en) * | 2017-07-24 | 2019-01-31 | 华为技术有限公司 | Method and apparatus for training classification model |
| US11151182B2 (en) | 2017-07-24 | 2021-10-19 | Huawei Technologies Co., Ltd. | Classification model training method and apparatus |
| WO2019184217A1 (en) * | 2018-03-26 | 2019-10-03 | 平安科技（深圳）有限公司 | Hotspot event classification method and apparatus, and storage medium |
| CN108509596B (en) * | 2018-04-02 | 2021-06-04 | 广州市申迪计算机系统有限公司 | Text classification method and device, computer equipment and storage medium |
| CN108509596A (en) * | 2018-04-02 | 2018-09-07 | 广州市申迪计算机系统有限公司 | File classification method, device, computer equipment and storage medium |
| CN108846018A (en) * | 2018-05-07 | 2018-11-20 | 中国农业大学 | A kind of Chinese food safety media event Information Automatic Extraction method towards news |
| CN109284383A (en) * | 2018-10-09 | 2019-01-29 | 北京来也网络科技有限公司 | Text handling method and device |
| CN109471941A (en) * | 2018-11-07 | 2019-03-15 | 中国电子科技集团公司第二十八研究所 | A kind of charge classification method for coping with class imbalance |
| CN110795564A (en) * | 2019-11-01 | 2020-02-14 | 南京稷图数据科技有限公司 | Text classification method lacking negative cases |
| CN110795564B (en) * | 2019-11-01 | 2022-02-22 | 南京稷图数据科技有限公司 | Text classification method lacking negative cases |
| WO2021212753A1 (en) * | 2020-04-23 | 2021-10-28 | 平安科技（深圳）有限公司 | Computer performance data determining method and apparatus, computer device, and storage medium |
| CN112199503A (en) * | 2020-10-28 | 2021-01-08 | 南京信息工程大学 | Feature enhancement based unbalanced Bi-LSTM Chinese text classification method |
| CN112199503B (en) * | 2020-10-28 | 2023-04-28 | 南京信息工程大学 | Feature-enhanced unbalanced Bi-LSTM-based Chinese text classification method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106570164A (en) | Integrated foodstuff safety text classification method based on deep learning | |
| US10445357B2 (en) | Document classification system, document classification method, and document classification program | |
| CN108030494B (en) | Electrocardiosignal error flag training sample recognition methods based on cross validation | |
| CN104239485B (en) | A kind of dark chain detection method in internet based on statistical machine learning | |
| CN104239436B (en) | It is a kind of that method is found based on the network hotspot event of text classification and cluster analysis | |
| CN102693219B (en) | Method and system for extracting Chinese event | |
| CN104268134B (en) | Subjective and objective classifier building method and system | |
| CN102298646B (en) | Method and device for classifying subjective text and objective text | |
| CN103064971A (en) | Scoring and Chinese sentiment analysis based review spam detection method | |
| CN103886108B (en) | The feature selecting and weighing computation method of a kind of unbalanced text set | |
| CN111104526A (en) | Financial label extraction method and system based on keyword semantics | |
| Chatzakou et al. | Detecting aggressors and bullies on Twitter | |
| CN109840532A (en) | A kind of law court's class case recommended method based on k-means | |
| CN109471932A (en) | Rumour detection method, system and storage medium based on learning model | |
| CN110472257A (en) | A kind of MT engine assessment preferred method and system based on sentence pair | |
| CN105975518A (en) | Information entropy-based expected cross entropy feature selection text classification system and method | |
| Malik et al. | Man vs. machine: A comparative analysis for signature verification | |
| CN103106346A (en) | Character prediction system based on off-line writing picture division and identification | |
| CN102663435A (en) | Junk image filtering method based on semi-supervision | |
| CN107992550A (en) | A kind of network comment analysis method and system | |
| Abercrombie et al. | Identifying opinion-topics and polarity of parliamentary debate motions | |
| CN106650696A (en) | Handwritten electrical element identification method based on singular value decomposition | |
| CN106649264A (en) | Text information-based Chinese fruit variety information extracting method and device | |
| CN102902984B (en) | Remote-sensing image semi-supervised projection dimension reducing method based on local consistency | |
| CN110489753A (en) | Improve the corresponding cross-cutting sensibility classification method of study of neuromechanism of feature selecting |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170419 |