CN1936892A - Image content semanteme marking method - Google Patents
Image content semanteme marking method Download PDFInfo
- Publication number
- CN1936892A CN1936892A CN 200610053867 CN200610053867A CN1936892A CN 1936892 A CN1936892 A CN 1936892A CN 200610053867 CN200610053867 CN 200610053867 CN 200610053867 A CN200610053867 A CN 200610053867A CN 1936892 A CN1936892 A CN 1936892A
- Authority
- CN
- China
- Prior art keywords
- image
- semantic
- mark
- mapping ruler
- 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.)
- Granted
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
Using techniques of image processing, machine learning, and semantic processing natural language etc, the method combines semantic labeling for visual character of image with semantic labeling text character of image to carry out semantic labeling content of image. Moreover, based on labeling characteristic of specific user, the method also supports to correct mapping rule base of label at bottom layer so that the labeled result is more accorded with labeling requirement of specific user. The method is widely applicable to each applications of need to carry out image searches. The method raises labeling precision, and expands range of application.
Description
Technical field
The present invention relates to a kind of image content semanteme marking method of image labeling, be particularly related to and use image processing techniques, natural language processing technique and machine learning techniques, utilize attribute informations such as the visual signature of picture material and related text that picture material is carried out semantic tagger.
Background technology
Along with the simple and easy availability of improved day by day digital image technology and the Internet, the popularization of digital picture increases rapidly in recent years, and there have every day increasing digital image to become to be available.Design one rapidly and accurately the method for the image that needs of retrieval user huge realistic meaning is arranged.Mainly contain two kinds of image retrieval modes at present.A kind of retrieval that is based on image keyword, another kind are based on the image retrieval CBIR of content, and the difference between them is the mark mode difference of picture material.
The image keyword mark that is applied to the image keyword retrieval mainly contains two kinds of methods at present and generates: the manual key word mark of selecting of one or more people, image keyword mark automatically generating device generate automatically.
1) manually generating image keyword mark is main method in the early stage image search method, marks automatic generation method with image keyword and compares the characteristics that have degree of accuracy high.But manually generate the image keyword mask method and mainly contain two shortcomings: the one, need manual each image of checking and carefully mark, these steps need a large amount of work and cost very high, especially along with the Internet popularize and situation that the digital picture scale is increasing under; The 2nd, different users is because separately world outlook and professional domain knowledge, understanding separately arranged thereby image made different semantic taggers for the content of identical image, and this has caused the inconsistency of image content semanteme marking.
2) image keyword mark automatically generating device mainly is to utilize other attribute information except that picture material to generate the image keyword mark.At present the image keyword automatic marking method is primarily aimed at the image of relevant informations such as some subsidiary rich texts such as the Internet.This method is compared maximum advantage with manual method be not need artificial interference, and shortcoming is that the degree of accuracy of image is lower with respect to manual mode.
Some uses in recent years begin to form based on the image management system of the image retrieval of picture material.Usually, based on the visual signatures such as color, texture and shape of the image indexing system abstract image of picture material mark, find out in the time of image retrieval and the approximate one or more images of Image Visual Feature that are retrieved are used as result for retrieval and are returned as picture material.Need calculate on a large scale for the visual signature of abstract image and by visual signature similarity between the computed image, and can not distinguish for the people based on the visual signature that the image indexing system of picture material extracts, do not possess visually and differentiability semantically, therefore be difficult to the image retrieval condition is described.
Thus, need a kind of usable range of invention extensively, be easy to calculate and the method for the semantic tagger of the picture material that degree of accuracy is high.
Summary of the invention
The object of the present invention is to provide a kind of usable range extensively, be easy to calculate and the semanteme marking method of the picture material that degree of accuracy is high.Image content semanteme marking method among the present invention is used natural language semantic processes technology and Image Visual Feature semanteme marking method and image text feature semanteme marking method is combined picture material is carried out semantic tagger.Different users is for the understanding property of there are differences of identical picture material, the also property of there are differences of Dui Ying image content semanteme marking therewith along with user's difference, but by image content semanteme marking study interface, image content semanteme marking method among the present invention utilizes machine learning and natural language processing correlation technique, can be for the specific user set up the image content semanteme marking preference pattern, make the image content semanteme marking result more near the mark preference of particular user.
The technical scheme that the present invention solves its technical matters employing is as follows:
1. the step of image content semanteme marking method is among the present invention,
1) at first, with original image data input image data processing layer, extracts Image Visual Feature data and image text characteristic by image, semantic mark interface;
2) secondly, the Image Visual Feature data input picture visual signature that step 1) is extracted marks layer, this module can be finished following function: at first the Image Visual Feature of visit foundation in advance marks the mapping ruler storehouse, extract the mapping ruler that conforms to Image Visual Feature, access images contents semantic mark shines upon accumulation layer then, and therefrom the mark of taking-up and mapping ruler correspondence is as the semantic tagger of Image Visual Feature;
3) once more, with the image text characteristic input picture text feature mark layer that step 1) extracts, this module utilizes the natural language semantic processes technology in the natural language semantic processes layer to extract the semantic tagger of image text characteristic;
4) last, at the image content semanteme marking layer, use natural language semantic processes technology, the semantic tagger of the semantic tagger of Image Visual Feature and image text feature merged be used as image content semanteme marking output, whether through one is the condition judgment of mode of learning: if be under the mode of learning, the semantic tagger result is fed back to mark mapping ruler learning layer; If not being under the mode of learning, the result exports to the user with semantic tagger.
2. the mark of the user images contents semantic among the present invention preference pattern learning procedure is:
1) at first, whether the semantic tagger of the picture material that is calculated by the step 4) of claim 1 is the condition judgment of mode of learning through one, is imported under mode of learning in the mark mapping ruler learning layer;
2) secondly, the user imports user-defined image content semanteme marking data by mapping ruler study interface;
3) once more, in mark mapping ruler learning layer, calculation procedure 1) annotation results and the step 2 that obtain in) in otherness between the self-defined annotation results of user input, if otherness is bigger, use the correlation machine learning art to create corresponding mark mapping ruler in corresponding mark mapping ruler or the correction mark mapping ruler storehouse;
4) repeated execution of steps 1) to step 3), the otherness in step 3) is very little, or reach the predetermined iteration upper limit, the semantic tagger mapping that will have user preference at last is stored in the middle of the image content semanteme marking mapping accumulation layer.
The present invention compares with traditional picture material mask method, and the beneficial effect that has is:
In the high advantage of the accuracy that the present invention possesses when keeping directly using Image Visual Feature as mark, when having solved the high defective of its computation complexity and Image Visual Feature as mark effectively by directly visual signature being mapped to key word visually and semantically can not the property distinguished.
The present invention carries out roughing to view data and has obtained Image Visual Feature data and image text characteristic, utilize machine learning techniques and image processing techniques from Image Visual Feature, to obtain the Image Visual Feature semantic tagger then, utilize natural language processing technique from image text characteristic and Image Visual Feature semantic tagger, to obtain image content semanteme marking.Owing to made full use of image data information, improved the degree of accuracy of existing picture material mark.
The inventive method possesses effective learning functionality, can set up the particular user of the corresponding to mark mapping ruler of preference when carrying out semantic tagger with to(for) image satisfying the mark demand of different user, this makes image content semanteme marking method among the present invention have better robustness and extensive applicability more.
Description of drawings
Accompanying drawing is the diagrammatic representation of the general frame of image content semanteme marking method.
Embodiment
Image content semanteme marking method of the present invention provides the user two functions: image content semanteme marking function and user images contents semantic mark preferential learning function.User images contents semantic mark preferential learning function be to the image content semanteme marking function replenish and perfect.
1. the implementation step of picture material semanteme
Image content semanteme marking method of the present invention as shown in drawings has four steps when carries out image marks: original image data processing, Image Visual Feature data semantic mark, image text characteristic semantic tagger and image content semanteme marking.
1) at first, the image labeling interface that utilizes image content semanteme marking method of the present invention with associated picture raw data input image data processing layer to extract Image Visual Feature data and image text characteristic:
A) color characteristic of abstract image raw data, textural characteristics and shape facility are as the Image Visual Feature data.The color characteristic of image has multiple expression mode, for example adopts color histogram, color matrix or color dependency graph representation.Similarly the texture of image can adopt Tamura texture, autoregression texture or co-occurrence matrix textural characteristics to represent, and shape facility can adopt Fourier descriptors method or the irrelevant matrix method of shape to describe.These eigenwerts but do not pay close attention to which kind of mode of concrete employing in the method for the invention do not generate these eigenwerts or adopt which kind of form to show these eigenwerts, as long as can be described image content corresponding and the image labeling method that is effectively applied among the present invention accurately;
B) use various semantic dictionaries to have the feature extraction of the semantic text of semantic nature to come out to form the image text characteristic with possessing, this step can be fallen many interference text filterings, improves the degree of accuracy and the efficient of follow-up mark work.WordNet can be adopted at the English Semantic dictionary, and HowNet can be adopted at Chinese semantic dictionary;
2) the Image Visual Feature data input picture visual signature mark layer that secondly, step 1) is extracted.This module can be visited the mark mapping ruler storehouse of an Image Visual Feature of setting up in advance, take out the mark mapping ruler consistent with the present image visual signature, at this moment the Image Visual Feature data have been converted to the mark mapping ruler of some correspondences, afterwards according to these mark mapping rulers, access images contents semantic mark mapping accumulation layer obtains mark with the mapping ruler correspondence as the Image Visual Feature semantic tagger.Machine learning and the foundation of natural language semantic processes correlation technique such as decision tree, neural network, support vector machine and statistical language probability model can be used in above-mentioned Image Visual Feature semantic tagger mapping ruler storehouse, in addition now also exist multiple other to set up the method for this rule base, but the present invention does not pay close attention to the foundation of adopting which kind of concrete grammar to realize this rule base, and only the mapping ruler that need be created can satisfy and exactly Image Visual Feature is mapped to that this functional requirement gets final product on the semantic tagger;
3) the image text characteristic input picture text feature mark layer that once more, step 1) is extracted.This module extracts the semantic tagger of text feature according to the various attributes of image text characteristic.For example for from the also image text characteristic of webpage, the source of the text data that these text feature data attribute information comprise, form, with the relative position of image, whether be web page title, whether adopt italics and boldface letter etc.The attribute information of these text feature data can be applied to calculating the weights of respective image text feature data, the text feature data that weights are big more are good more to the semantic description of picture material, therefrom take out the semantic tagger of some text feature data of weights maximum as the image text feature;
4) last, the image content semanteme marking module is used natural language semantic processes technology, similarity between the semantic tagger of use semantic dictionary computed image visual signature and the semantic tagger of image text feature, then similar high semantic tagger has been merged and be used as the image content semanteme marking result, whether through one
After the condition judgment of under mode of learning, moving,, annotation results is exported to the user, otherwise annotation results is input to mark mapping ruler learning layer if condition judgment is false.The concrete fusion steps of the semantic tagger of the semantic tagger of Image Visual Feature and image text feature is: extract from the semantic tagger of image text feature with the high mark of the semantic tagger similarity of Image Visual Feature and form mark collection X, extract and mark the high mark of collection X similarity then and form mark collection Y from the semantic tagger of Image Visual Feature, last X+Y is exactly a picture material mark of exporting to the user.
2. image content semanteme marking mapping ruler learning functionality
The user is by the raw data of image, semantic mark interface input picture, and the mark by mapping ruler study interface input picture, can set up the Image Visual Feature data to the mapping ruler between the Image Visual Feature semantic tagger.The user is by input and have the image content semanteme marking data that oneself mark preference, can utilize this mapping ruler learning functionality foundation of the inventive method to have the mark mapping ruler that the user marks preference.On the basis of image content semanteme marking implementation step, the study of image content semanteme marking mapping ruler is divided into three steps as shown in drawings: image content semanteme marking, image labeling result input, the study of image content semanteme marking mapping ruler.
1) at first, original image data is input to image, semantic mark interface, by calculating the contents semantic annotation results of image, by one whether be the condition judgment of mode of learning for after true, the image content semanteme marking result is imported in the mark mapping ruler learning layer;
2) secondly, the image content semanteme marking result with user's input is input in the mark mapping ruler learning layer;
3) last, to have the otherness that the user marks between the semantic tagger of preference bigger if image content semanteme marking result who calculates in the step 1) and user provide, then use machine learning correlation techniques such as neural network or decision tree, learn repeatedly and produce new mapping to advise this process, the otherness between the semantic tagger that the image content semanteme marking result who obtains according to new mapping ruler and user provide is less.Finish after the learning process, new mapping ruler is compared with mapping ruler before, and annotation results is more near the mark preference of particular user.At last new mark mapping ruler is stored in the middle of the visual signature mark mapping ruler storehouse, finishes the correction in visual signature mark mapping ruler storehouse.Except technology such as neural network above-mentioned, also exist several different methods can realize the learning functionality of mapping ruler, the inventive method is not paid close attention to concrete mapping ruler learning method, only needs it can satisfy the function that the user marks preference mapping ruler study effectively and gets final product.
Claims (2)
1. image content semanteme marking method is characterized in that:
1) at first, with original image data input image data processing layer, extracts Image Visual Feature data and image text characteristic by image, semantic mark interface;
2) secondly, the Image Visual Feature data input picture visual signature that step 1) is extracted marks layer, this module can be finished following function: at first the visual signature of visit foundation in advance marks the mapping ruler storehouse, extract the mapping ruler that conforms to Image Visual Feature, access images contents semantic mark shines upon accumulation layer then, and therefrom the mark of taking-up and mapping ruler correspondence is as the semantic tagger of Image Visual Feature;
3) once more, with the image text characteristic input picture text feature mark layer that step 1) extracts, this module utilizes the natural language semantic processes technology in the natural language semantic processes layer to extract the semantic tagger of image text characteristic;
4) last, at the image content semanteme marking layer, use natural language semantic processes technology, the semantic tagger of the semantic tagger of Image Visual Feature and image text feature merged be used as image content semanteme marking output, whether through one is the condition judgment of mode of learning: if be under the mode of learning, the semantic tagger result is fed back to mark mapping ruler learning layer; If not being under the mode of learning, the result exports to the user with semantic tagger.
2. image content semanteme marking method according to claim 1 is characterized in that:
1) at first, whether the semantic tagger of the picture material that is calculated by the step 4) of claim 1 is the condition judgment of mode of learning through one, is imported under mode of learning in the mark mapping ruler learning layer;
2) secondly, the user imports user-defined image content semanteme marking data by mapping ruler study interface;
3) once more, in mark mapping ruler learning layer, calculation procedure 1) annotation results and the step 2 that obtain in) in otherness between the self-defined labeled data of user input, if otherness is bigger, use the correlation machine learning art to create corresponding mark mapping ruler in corresponding mark mapping ruler or the correction mark mapping ruler storehouse;
4) repeated execution of steps 1) to step 3), the otherness in step 3) is very little, or reach the predetermined iteration upper limit, the semantic tagger mapping that will have user preference at last is stored in the middle of the image content semanteme marking mapping accumulation layer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB2006100538679A CN100437582C (en) | 2006-10-17 | 2006-10-17 | Image content semanteme marking method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB2006100538679A CN100437582C (en) | 2006-10-17 | 2006-10-17 | Image content semanteme marking method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN1936892A true CN1936892A (en) | 2007-03-28 |
CN100437582C CN100437582C (en) | 2008-11-26 |
Family
ID=37954396
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNB2006100538679A Expired - Fee Related CN100437582C (en) | 2006-10-17 | 2006-10-17 | Image content semanteme marking method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN100437582C (en) |
Cited By (24)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101853295A (en) * | 2010-05-28 | 2010-10-06 | 天津大学 | Image search method |
CN102142089A (en) * | 2011-01-07 | 2011-08-03 | 哈尔滨工程大学 | Semantic binary tree-based image annotation method |
CN102254043A (en) * | 2011-08-17 | 2011-11-23 | 电子科技大学 | Semantic mapping-based clothing image retrieving method |
CN102422286A (en) * | 2009-03-11 | 2012-04-18 | 香港浸会大学 | Automatic and semi-automatic image classification, annotation and tagging through the use of image acquisition parameters and metadata |
CN102662953A (en) * | 2012-03-01 | 2012-09-12 | 倪旻 | Semantic annotation system and method integrated with input method |
CN102880612A (en) * | 2011-07-14 | 2013-01-16 | 富士通株式会社 | Image annotation method and device thereof |
CN103246688A (en) * | 2012-12-03 | 2013-08-14 | 苏州大学 | Method for systematically managing images by aid of semantic hierarchical model on basis of sparse representation for salient regions |
CN103377381A (en) * | 2012-04-26 | 2013-10-30 | 富士通株式会社 | Method and device for identifying content attribute of image |
CN103632388A (en) * | 2013-12-19 | 2014-03-12 | 百度在线网络技术(北京)有限公司 | Semantic annotation method, device and client for image |
CN104008177A (en) * | 2014-06-09 | 2014-08-27 | 华中师范大学 | Method and system for rule base structure optimization and generation facing image semantic annotation |
CN104156433A (en) * | 2014-08-11 | 2014-11-19 | 合肥工业大学 | Image retrieval method based on semantic mapping space construction |
CN102112987B (en) * | 2008-05-30 | 2015-03-04 | 微软公司 | Statistical approach to large-scale image annotation |
WO2016150328A1 (en) * | 2015-03-25 | 2016-09-29 | 阿里巴巴集团控股有限公司 | Data annotation management method and apparatus |
CN106650775A (en) * | 2016-10-12 | 2017-05-10 | 南京理工大学 | Image labeling method for mining visual and semantic similarity at the same time |
CN106649610A (en) * | 2016-11-29 | 2017-05-10 | 北京智能管家科技有限公司 | Image labeling method and apparatus |
GB2544379A (en) * | 2015-11-11 | 2017-05-17 | Adobe Systems Inc | Structured knowledge modeling, extraction and localization from images |
CN103793498B (en) * | 2014-01-22 | 2017-08-25 | 百度在线网络技术(北京)有限公司 | Image searching method, device and search engine |
CN109271539A (en) * | 2018-08-31 | 2019-01-25 | 华中科技大学 | A kind of image automatic annotation method and device based on deep learning |
CN109543690A (en) * | 2018-11-27 | 2019-03-29 | 北京百度网讯科技有限公司 | Method and apparatus for extracting information |
US10460033B2 (en) | 2015-11-11 | 2019-10-29 | Adobe Inc. | Structured knowledge modeling, extraction and localization from images |
CN111222500A (en) * | 2020-04-24 | 2020-06-02 | 腾讯科技(深圳)有限公司 | Label extraction method and device |
CN113128509A (en) * | 2019-12-31 | 2021-07-16 | 广东爱因智能数字营销有限公司 | Image semantic element extraction method |
CN113255665A (en) * | 2021-06-04 | 2021-08-13 | 明品云(北京)数据科技有限公司 | Target text extraction method and system |
US11514244B2 (en) | 2015-11-11 | 2022-11-29 | Adobe Inc. | Structured knowledge modeling and extraction from images |
-
2006
- 2006-10-17 CN CNB2006100538679A patent/CN100437582C/en not_active Expired - Fee Related
Cited By (33)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102112987B (en) * | 2008-05-30 | 2015-03-04 | 微软公司 | Statistical approach to large-scale image annotation |
CN102422286A (en) * | 2009-03-11 | 2012-04-18 | 香港浸会大学 | Automatic and semi-automatic image classification, annotation and tagging through the use of image acquisition parameters and metadata |
CN101853295A (en) * | 2010-05-28 | 2010-10-06 | 天津大学 | Image search method |
CN102142089A (en) * | 2011-01-07 | 2011-08-03 | 哈尔滨工程大学 | Semantic binary tree-based image annotation method |
CN102880612A (en) * | 2011-07-14 | 2013-01-16 | 富士通株式会社 | Image annotation method and device thereof |
CN102880612B (en) * | 2011-07-14 | 2015-05-06 | 富士通株式会社 | Image annotation method and device thereof |
CN102254043A (en) * | 2011-08-17 | 2011-11-23 | 电子科技大学 | Semantic mapping-based clothing image retrieving method |
CN102254043B (en) * | 2011-08-17 | 2013-04-03 | 电子科技大学 | Semantic mapping-based clothing image retrieving method |
CN102662953A (en) * | 2012-03-01 | 2012-09-12 | 倪旻 | Semantic annotation system and method integrated with input method |
CN102662953B (en) * | 2012-03-01 | 2016-04-06 | 倪旻 | With the semantic tagger system and method that input method is integrated |
CN103377381A (en) * | 2012-04-26 | 2013-10-30 | 富士通株式会社 | Method and device for identifying content attribute of image |
CN103377381B (en) * | 2012-04-26 | 2016-09-28 | 富士通株式会社 | The method and apparatus identifying the contents attribute of image |
CN103246688A (en) * | 2012-12-03 | 2013-08-14 | 苏州大学 | Method for systematically managing images by aid of semantic hierarchical model on basis of sparse representation for salient regions |
CN103632388A (en) * | 2013-12-19 | 2014-03-12 | 百度在线网络技术(北京)有限公司 | Semantic annotation method, device and client for image |
CN103793498B (en) * | 2014-01-22 | 2017-08-25 | 百度在线网络技术(北京)有限公司 | Image searching method, device and search engine |
CN104008177A (en) * | 2014-06-09 | 2014-08-27 | 华中师范大学 | Method and system for rule base structure optimization and generation facing image semantic annotation |
CN104008177B (en) * | 2014-06-09 | 2017-06-13 | 华中师范大学 | Rule base structure optimization and generation method and system towards linguistic indexing of pictures |
CN104156433A (en) * | 2014-08-11 | 2014-11-19 | 合肥工业大学 | Image retrieval method based on semantic mapping space construction |
CN104156433B (en) * | 2014-08-11 | 2017-05-17 | 合肥工业大学 | Image retrieval method based on semantic mapping space construction |
WO2016150328A1 (en) * | 2015-03-25 | 2016-09-29 | 阿里巴巴集团控股有限公司 | Data annotation management method and apparatus |
GB2544379A (en) * | 2015-11-11 | 2017-05-17 | Adobe Systems Inc | Structured knowledge modeling, extraction and localization from images |
US11514244B2 (en) | 2015-11-11 | 2022-11-29 | Adobe Inc. | Structured knowledge modeling and extraction from images |
GB2544379B (en) * | 2015-11-11 | 2019-09-11 | Adobe Inc | Structured knowledge modeling, extraction and localization from images |
US10460033B2 (en) | 2015-11-11 | 2019-10-29 | Adobe Inc. | Structured knowledge modeling, extraction and localization from images |
CN106650775B (en) * | 2016-10-12 | 2020-04-10 | 南京理工大学 | Image annotation method capable of mining visual and semantic similarity simultaneously |
CN106650775A (en) * | 2016-10-12 | 2017-05-10 | 南京理工大学 | Image labeling method for mining visual and semantic similarity at the same time |
CN106649610A (en) * | 2016-11-29 | 2017-05-10 | 北京智能管家科技有限公司 | Image labeling method and apparatus |
CN109271539A (en) * | 2018-08-31 | 2019-01-25 | 华中科技大学 | A kind of image automatic annotation method and device based on deep learning |
CN109543690A (en) * | 2018-11-27 | 2019-03-29 | 北京百度网讯科技有限公司 | Method and apparatus for extracting information |
CN113128509A (en) * | 2019-12-31 | 2021-07-16 | 广东爱因智能数字营销有限公司 | Image semantic element extraction method |
CN111222500A (en) * | 2020-04-24 | 2020-06-02 | 腾讯科技(深圳)有限公司 | Label extraction method and device |
CN113255665A (en) * | 2021-06-04 | 2021-08-13 | 明品云(北京)数据科技有限公司 | Target text extraction method and system |
CN113255665B (en) * | 2021-06-04 | 2021-12-21 | 明品云(北京)数据科技有限公司 | Target text extraction method and system |
Also Published As
Publication number | Publication date |
---|---|
CN100437582C (en) | 2008-11-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN100437582C (en) | Image content semanteme marking method | |
CN105718586B (en) | The method and device of participle | |
CN102955848B (en) | A kind of three-dimensional model searching system based on semanteme and method | |
CN107291783B (en) | Semantic matching method and intelligent equipment | |
CN103049446B (en) | A kind of image search method and device | |
CN102262634B (en) | Automatic questioning and answering method and system | |
Sousa et al. | Sketch-based retrieval of drawings using spatial proximity | |
Wilkinson et al. | Neural Ctrl-F: segmentation-free query-by-string word spotting in handwritten manuscript collections | |
CN105956053B (en) | A kind of searching method and device based on the network information | |
CN104199972A (en) | Named entity relation extraction and construction method based on deep learning | |
CN107392143A (en) | A kind of resume accurate Analysis method based on SVM text classifications | |
CN104933039A (en) | Entity link system for language lacking resources | |
CN106407180A (en) | Entity disambiguation method and apparatus | |
CN101620615A (en) | Automatic image annotation and translation method based on decision tree learning | |
WO2008008213A2 (en) | Interactively crawling data records on web pages | |
CN112800239B (en) | Training method of intention recognition model, and intention recognition method and device | |
CN102193946A (en) | Method and system for adding tags into media file | |
CN109857912A (en) | A kind of font recognition methods, electronic equipment and storage medium | |
CN106980620A (en) | A kind of method and device matched to Chinese character string | |
CN110347857A (en) | The semanteme marking method of remote sensing image based on intensified learning | |
CN103678288A (en) | Automatic proper noun translation method | |
CN106469188A (en) | A kind of entity disambiguation method and device | |
CN112131341A (en) | Text similarity calculation method and device, electronic equipment and storage medium | |
CN112989811B (en) | History book reading auxiliary system based on BiLSTM-CRF and control method thereof | |
CN112148735B (en) | Construction method for structured form data knowledge graph |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20081126 Termination date: 20151017 |
|
EXPY | Termination of patent right or utility model |