CN109858037A - A kind of pair of OCR recognition result carries out the method and system of structuring output - Google Patents
A kind of pair of OCR recognition result carries out the method and system of structuring output Download PDFInfo
- Publication number
- CN109858037A CN109858037A CN201910145824.0A CN201910145824A CN109858037A CN 109858037 A CN109858037 A CN 109858037A CN 201910145824 A CN201910145824 A CN 201910145824A CN 109858037 A CN109858037 A CN 109858037A
- Authority
- CN
- China
- Prior art keywords
- entity
- name
- text
- recognition result
- label
- 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
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000012549 training Methods 0.000 claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 6
- 239000007787 solid Substances 0.000 claims abstract description 6
- 238000002372 labelling Methods 0.000 claims description 17
- 230000001568 sexual effect Effects 0.000 claims description 17
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 230000000306 recurrent effect Effects 0.000 claims description 6
- 230000006403 short-term memory Effects 0.000 claims description 6
- 238000000605 extraction Methods 0.000 abstract description 2
- 238000012015 optical character recognition Methods 0.000 description 30
- 238000013461 design Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000015654 memory Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Landscapes
- Character Discrimination (AREA)
Abstract
The invention discloses the method and system that a kind of pair of OCR recognition result carries out structuring output, and method includes: to be identified using ID Card Image of the OCR to acquisition, carry out processing to recognition result and obtain text sequence;Entity recognition is named to the text sequence by trained name physical model, the name solid data of tape label is obtained, exports corresponding structured text.A kind of pair of OCR recognition result of the present invention carries out the method and system of structuring output, pass through the text of a large amount of tape label of generation, training obtains Named Entity Extraction Model, each entity in OCR recognition result can fast and efficiently be extracted, the output of structuring is obtained, very big help is brought to the typing of identity information.
Description
Technical field
The present invention relates to pictographs to identify field, and in particular to a kind of pair of OCR recognition result carries out structuring output
Method and system.
Background technique
OCR (Optical Character Recognition, optical character identification) technology is mainly by the text in image
Word is identified as editable character string.What the OCR technique of early stage mainly identified is some simple file and pictures, due to depth
The development of habit, current OCR technique have been widely used for the Text region of image under various complex scenes.
In recent years, perfecting with system, more and more occasions need us to carry out system of real name, based on OCR technique
Development, usually we only need to shoot or image of the upload containing identity card, system carry out image using OCR technique
Identification, the result that then will identify that carry out typing.However, the result that OCR technique identifies is only a string of editable words
Symbol string, does not include any structured message.Generally require to establish series of rules screening items for result to typing, or
Direct labor's typing.The former robustness is very poor, and can not establish a set of complete rule to carry out the screening of every terms of information.Afterwards
Person's low efficiency causes the waste of great human cost.In addition, the information for being in same level direction for text on picture can
The case where recognition result dislocation can occur.
Summary of the invention
It is a primary object of the present invention to propose that a kind of pair of OCR recognition result carries out the method and system of structuring output,
Each name entity in OCR recognition result can be fast and efficiently extracted, the output of structuring is obtained, to identity information
Typing bring very big help.
The present invention adopts the following technical scheme:
On the one hand, the method that a kind of pair of OCR recognition result of the present invention carries out structuring output, comprising:
It is identified using ID Card Image of the OCR to acquisition, processing is carried out to recognition result and obtains text sequence;
Entity recognition is named to the text sequence by trained name physical model, obtains the life of tape label
Name solid data, exports corresponding structured text.
Preferably, the training method of physical model is named, comprising:
A), the sample text sequence of several tape labels is generated;
Each sample text sequence includes the name entity of name, gender, nationality, birth, address and citizenship number,
Respectively name entity, sexual entities, national entity, birth entity, address entity and citizenship number entity;Each sample
Text sequence further include ' name ', ' gender ', ' nationality ', ' birth ', ' address ' and ' citizenship number ' text items;Its
In, ' name ' is corresponding with name entity, and ' gender ' is corresponding with sexual entities, and ' nationality ' is corresponding with national entity, ' out
It is raw ' it is corresponding with birth entity, ' address ' is corresponding with address entity, ' citizenship number ' and citizenship number entity phase
It is corresponding;
Each text items and it is each name entity start-up portion label for labelling be B-entityName, each text items and
The label for labelling of each name entity other parts is I-entityName;Wherein entityName is customized character
String;
B), using the sample text sequence training name physical model of the tape label.
Preferably, the entityName of the entityName of each text items and its corresponding name entity includes part phase
Same character string.
Preferably, the corresponding structured text of the output, specifically:
Every a line exports the text items and its label or a name entity and its label;The text items
Corresponding name entity is adjacent, and text items output is in the previous row of its corresponding name entity.
Preferably, the name physical model is two-way long short-term memory Recognition with Recurrent Neural Network Bi-LSTM and condition random
The model that field CRF is combined.
Second aspect, the system that a kind of pair of OCR recognition result of the present invention carries out structuring output, comprising:
OCR output obtains module, for being identified using ID Card Image of the OCR to acquisition, carries out to recognition result
Processing obtains text sequence;
Structuring output module, for being named entity to the text sequence by trained name physical model
Identification, obtains the name solid data of tape label, exports corresponding structured text.
Preferably, the training method of physical model is named, comprising:
A), the sample text sequence of several tape labels is generated;
Each sample text sequence includes the name entity of name, gender, nationality, birth, address and citizenship number,
Respectively name entity, sexual entities, national entity, birth entity, address entity and citizenship number entity;Each sample
Text sequence further include ' name ', ' gender ', ' nationality ', ' birth ', ' address ' and ' citizenship number ' text items;Its
In, ' name ' is corresponding with name entity, and ' gender ' is corresponding with sexual entities, and ' nationality ' is corresponding with national entity, ' out
It is raw ' it is corresponding with birth entity, ' address ' is corresponding with address entity, ' citizenship number ' and citizenship number entity phase
It is corresponding;
Each text items and it is each name entity start-up portion label for labelling be B-entityName, each text items and
The label for labelling of each name entity other parts is I-entityName;Wherein entityName is customized character
String;
B), using the sample text sequence training name physical model of the tape label.
Preferably, the entityName of the entityName of each text items and its corresponding name entity includes part phase
Same character string.
Preferably, the corresponding structured text of the output, specifically:
Every a line exports the text items and its label or a name entity and its label;The text items
Corresponding name entity is adjacent, and text items output is in the previous row of its corresponding name entity.
Preferably, the name physical model is two-way long short-term memory Recognition with Recurrent Neural Network Bi-LSTM and condition random
The model that field CRF is combined.
Compared with prior art, beneficial effects of the present invention are as follows:
(1) a kind of pair of OCR recognition result of the present invention carries out the method and system of structuring output, by ID card information
Analysis, devise the text data generating algorithm of a set of tape label, the text of a large amount of tape label can be generated;
(2) a kind of pair of OCR recognition result of the present invention carries out the method and system of structuring output, passes through a large amount of of generation
The text of tape label, training obtain Named Entity Extraction Model, can fast and efficiently extract each in OCR recognition result
A entity obtains the output of structuring, brings very big help to the typing of identity information.
The above description is only an overview of the technical scheme of the present invention, in order to more clearly understand technology hand of the invention
Section, so as to be implemented in accordance with the contents of the specification, and in order to allow above and other objects, features and advantages of the invention
It can be more clearly understood, be exemplified below a specific embodiment of the invention.
According to the following detailed description of specific embodiments of the present invention in conjunction with the accompanying drawings, those skilled in the art will be brighter
Above-mentioned and other purposes of the invention, advantages and features.
Detailed description of the invention
Fig. 1 is the flow chart of the method that structuring output is carried out to OCR recognition result of the embodiment of the present invention;
Fig. 2 is the recognition result with dislocation that the OCR of the embodiment of the present invention is exported;
Fig. 3 is the output of the name physical model of the embodiment of the present invention;
Fig. 4 is that the structuring of the embodiment of the present invention exports;
Fig. 5 is the structural block diagram of the system that structuring output is carried out to OCR recognition result of the embodiment of the present invention.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached drawing to embodiment party of the present invention
Formula is described in further detail.
Method shown in Figure 1, that a kind of pair of OCR recognition result of the present invention carries out structuring output, comprising:
It is identified using ID Card Image of the OCR to acquisition, processing is carried out to recognition result and obtains text sequence;
Entity recognition is named to the text sequence by trained name physical model, obtains the life of tape label
Name solid data, exports corresponding structured text.
In the present embodiment, the training method of physical model is named, comprising:
A), the sample text sequence of several tape labels is generated;
Each sample text sequence includes the name entity of name, gender, nationality, birth, address and citizenship number,
Respectively name entity, sexual entities, national entity, birth entity, address entity and citizenship number entity;Each sample
Text sequence further include ' name ', ' gender ', ' nationality ', ' birth ', ' address ' and ' citizenship number ' text items;Its
In, ' name ' is corresponding with name entity, and ' gender ' is corresponding with sexual entities, and ' nationality ' is corresponding with national entity, ' out
It is raw ' it is corresponding with birth entity, ' address ' is corresponding with address entity, ' citizenship number ' and citizenship number entity phase
It is corresponding;
Each text items and it is each name entity start-up portion label for labelling be B-entityName, each text items and
The label for labelling of each name entity other parts is I-entityName, and the unit marked is word, i.e., each word has
One label;Wherein entityName is customized character string;
B), using the sample text sequence training name physical model of the tape label.
The entityName of the entityName of each text items and its corresponding name entity includes the identical word in part
Symbol string.
If the label for labelling of the word ' surname ' of ' name ' start-up portion is B-name, the label for labelling of ' name ' is I-name,
The label for labelling of the word ' Lee ' of corresponding name entity ' Li Ming ' start-up portion is B-e_name, and the label for labelling of ' bright ' is I-e_
name。
Specifically, name entity, sexual entities, national entity, birth entity, address are real when name physical model training
The create-rule of body and citizenship number entity is as follows.
Name entity: it is generated in such a way that surname adds text random combine, label B-e_name/I-e_name;This
Outside, certain limitation is arranged in the length of name, and certain ratio is arranged in name length, if length of name is no more than 4, name length
Ratio for 2/3/4 is 7:2:1 etc..
Sexual entities: the ratio of male/female is 1:1, label B-e_gender/I-e_gender;
National entity: range is 56 nationalitys.Label is B-e_nation/I-e_nation;Han nationality sets with other nationalitys
Set certain ratio, such as 8:2.
Birth entity: the range of year of birth is to push ahead the several years from the time grown up recently, such as be instantly
2018, label B-e_birth/I-e_birth;
Address entity: borough information is saved according to the whole nation and carries out random combine, label B-e_address/I-e_
address。
Citizenship number entity: first 14 of identification number are determined that latter four are number by address and birthdate
Word random combine, label B-e_idnum/I-e_idnum.
Remaining ' name ', ' gender ', ' nationality ', ' birth ', ' address ', ' citizenship number ' this several mark respectively
For ' B-name/I-name ', ' B-gender/I-gender ', ' B-nation/I-nation ', ' B-birth/I-birth ',
‘B-address/I-address’、‘B-idnum/I-idnum’。
In addition, also needing the sequence for exchanging each name entity by certain probability when naming physical model training, increasing
Add the diversity for generating sample data.
It is shown in Figure 2, such as when identity card identification, the information for being in same level direction for text can go out
The case where now misplacing, i.e., the sequence of text that we identify is directly obtained according to the position of text box in fact, in this way
It determines each entity just very not robust according to this text box, therefore increases this name entity (text) sequence of exchanging
Sample trains name physical model, enhances the robustness of model, to solve the problems, such as this rule-based insurmountable.
The text data ginseng of the tape label generated after being adjusted according to above-mentioned name entity create-rule and name entity sequence
As shown in Figure 3.
Further, several pieces sample is generated after adjusting according to above-mentioned name entity create-rule and/or name entity sequence
Then the data of generation are sent into the name physical model and are trained, wherein the name physical model by notebook data
Specially two-way long short-term memory Recognition with Recurrent Neural Network (Bi-directional Long short-time Memory, Bi-
LSTM) the model combined with condition random field (condition random field) CRF.
After the name physical model trains, ID Card Image is identified by OCR first, then recognition result is carried out
Integration obtains a text sequence, and the text sequence after integration is as shown in Figure 2;The text training is directly made a gift to someone again and is trained
Name physical model be named Entity recognition, the result of name physical model output is as shown in Figure 3.
In Fig. 2, recognition result is the sequence that text is determined according to the position of text box in OCR identification, is identified in OCR
As a result in, sexual entities and national entity since there is dislocation in the detection of text box, i.e., ' male ' this text box ratio it is same
The text box of ' gender ' of a line and ' the national Chinese ' wants low.The character order identified is caused also to misplace.
And in the output result of Fig. 3, due to outputing label for labelling, ' Chinese ' can be identified as B-e_nation
(national entity), ' male ' can be identified as ' B-e_gender ' (sexual entities).
It is shown in Figure 4, it further, can be according to the text of corresponding tag identifier export structure.
It is described to export corresponding structured text, specifically:
Every a line exports the text items and its label or a name entity and its label;The text items
Corresponding name entity is adjacent, and text items output is in the previous row of its corresponding name entity.
In Fig. 4, to be text importing convenient for showing the tag replacement of entity.That is (name is real by name (name), e_name
Body), gender (gender), e_gender (sexual entities), nation (nationality), e_nation (national entity), birth (out
It is raw), e_birth (birth entity), address (address), e_address (address entity), idnum (citizen ID certificate number)
With e_idnum (ID card No. entity).
It is shown in Figure 5, second aspect, the system that a kind of pair of OCR recognition result of the present invention carries out structuring output, packet
It includes:
OCR output obtain module 501, for being identified using ID Card Image of the OCR to acquisition, to recognition result into
Row processing obtains text sequence;
Structuring output module 502, for being named by trained name physical model to the text sequence
Entity recognition obtains the name solid data of tape label, exports corresponding structured text.
In the present embodiment, the training method of physical model is named, comprising:
A), the sample text sequence of several tape labels is generated;
Each sample text sequence includes the name entity of name, gender, nationality, birth, address and citizenship number,
Respectively name entity, sexual entities, national entity, birth entity, address entity and citizenship number entity;Each sample
Text sequence further include ' name ', ' gender ', ' nationality ', ' birth ', ' address ' and ' citizenship number ' text items;Its
In, ' name ' is corresponding with name entity, and ' gender ' is corresponding with sexual entities, and ' nationality ' is corresponding with national entity, ' out
It is raw ' it is corresponding with birth entity, ' address ' is corresponding with address entity, ' citizenship number ' and citizenship number entity phase
It is corresponding;
Each text items and it is each name entity start-up portion label for labelling be B-entityName, each text items and
The label for labelling of each name entity other parts is I-entityName;Wherein entityName is customized character
String;
B), using the sample text sequence training name physical model of the tape label.
The entityName of the entityName of each text items and its corresponding name entity includes the identical word in part
Symbol string.
It is described to export corresponding structured text, specifically:
Every a line exports the text items and its label or a name entity and its label;The text items
Corresponding name entity is adjacent, and text items output is in the previous row of its corresponding name entity.
Preferably, the name physical model is two-way long short-term memory Recognition with Recurrent Neural Network Bi-LSTM and condition random
The model that field CRF is combined.
The above is only a specific embodiment of the present invention, but the design concept of the present invention is not limited to this, all to utilize this
Design makes a non-material change to the present invention, and should all belong to behavior that violates the scope of protection of the present invention.
Claims (10)
1. the method that a kind of pair of OCR recognition result carries out structuring output characterized by comprising
It is identified using ID Card Image of the OCR to acquisition, processing is carried out to recognition result and obtains text sequence;
Entity recognition is named to the text sequence by trained name physical model, the name for obtaining tape label is real
Volume data exports corresponding structured text.
2. the method according to claim 1 for carrying out structuring output to OCR recognition result, which is characterized in that name is real
The training method of body Model, comprising:
A), the sample text sequence of several tape labels is generated;
Each sample text sequence includes the name entity of name, gender, nationality, birth, address and citizenship number, respectively
For name entity, sexual entities, national entity, birth entity, address entity and citizenship number entity;Each sample text
Sequence further include ' name ', ' gender ', ' nationality ', ' birth ', ' address ' and ' citizenship number ' text items;Wherein, ' surname
Name ' it is corresponding with name entity, ' gender ' is corresponding with sexual entities, and ' nationality ' is corresponding with national entity, ' birth ' and goes out
Raw entity is corresponding, and ' address ' is corresponding with address entity, and ' citizenship number ' is corresponding with citizenship number entity;
The label for labelling of each text items and each name entity start-up portion is B-entityName, each text items and each
The label for labelling for naming entity other parts is I-entityName;Wherein entityName is customized character string;
B), using the sample text sequence training name physical model of the tape label.
3. the method according to claim 2 for carrying out structuring output to OCR recognition result, which is characterized in that Mei Gewen
The entityName of the entityName of this item and its corresponding name entity includes the identical character string in part.
4. the method according to claim 2 for carrying out structuring output to OCR recognition result, which is characterized in that described defeated
Corresponding structured text out, specifically:
Every a line exports the text items and its label or a name entity and its label;The text items and its
Corresponding name entity adjacent rows each other, and text items output is in the previous row of its corresponding name entity.
5. the method according to claim 1 for carrying out structuring output to OCR recognition result, which is characterized in that the life
Name physical model is the model that two-way long short-term memory Recognition with Recurrent Neural Network Bi-LSTM is combined with condition random field CRF.
6. the system that a kind of pair of OCR recognition result carries out structuring output characterized by comprising
OCR output is obtained module and handled for being identified using ID Card Image of the OCR to acquisition recognition result
Obtain text sequence;
Structuring output module is known for being named entity to the text sequence by trained name physical model
Not, the name solid data for obtaining tape label, exports corresponding structured text.
7. the system according to claim 6 for carrying out structuring output to OCR recognition result, which is characterized in that name is real
The training method of body Model, comprising:
A), the sample text sequence of several tape labels is generated;
Each sample text sequence includes the name entity of name, gender, nationality, birth, address and citizenship number, respectively
For name entity, sexual entities, national entity, birth entity, address entity and citizenship number entity;Each sample text
Sequence further include ' name ', ' gender ', ' nationality ', ' birth ', ' address ' and ' citizenship number ' text items;Wherein, ' surname
Name ' it is corresponding with name entity, ' gender ' is corresponding with sexual entities, and ' nationality ' is corresponding with national entity, ' birth ' and goes out
Raw entity is corresponding, and ' address ' is corresponding with address entity, and ' citizenship number ' is corresponding with citizenship number entity;
The label for labelling of each text items and each name entity start-up portion is B-entityName, each text items and each
The label for labelling for naming entity other parts is I-entityName;Wherein entityName is customized character string;
B), using the sample text sequence training name physical model of the tape label.
8. the system according to claim 7 for carrying out structuring output to OCR recognition result, which is characterized in that Mei Gewen
The entityName of the entityName of this item and its corresponding name entity includes the identical character string in part.
9. the system according to claim 7 for carrying out structuring output to OCR recognition result, which is characterized in that described defeated
Corresponding structured text out, specifically:
Every a line exports the text items and its label or a name entity and its label;The text items and its
Corresponding name entity is adjacent, and text items output is in the previous row of its corresponding name entity.
10. the system according to claim 6 for carrying out structuring output to OCR recognition result, which is characterized in that the life
Name physical model is the model that two-way long short-term memory Recognition with Recurrent Neural Network Bi-LSTM is combined with condition random field CRF.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910145824.0A CN109858037A (en) | 2019-02-27 | 2019-02-27 | A kind of pair of OCR recognition result carries out the method and system of structuring output |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910145824.0A CN109858037A (en) | 2019-02-27 | 2019-02-27 | A kind of pair of OCR recognition result carries out the method and system of structuring output |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109858037A true CN109858037A (en) | 2019-06-07 |
Family
ID=66899180
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910145824.0A Pending CN109858037A (en) | 2019-02-27 | 2019-02-27 | A kind of pair of OCR recognition result carries out the method and system of structuring output |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109858037A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110348021A (en) * | 2019-07-17 | 2019-10-18 | 湖北亿咖通科技有限公司 | Character string identification method, electronic equipment, storage medium based on name physical model |
CN112185520A (en) * | 2020-09-27 | 2021-01-05 | 志诺维思(北京)基因科技有限公司 | Text structured processing system and method for medical pathology report picture |
CN116229493A (en) * | 2022-12-14 | 2023-06-06 | 国家能源集团物资有限公司 | Cross-modal picture text named entity recognition method and system and electronic equipment |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105844274A (en) * | 2016-03-25 | 2016-08-10 | 浪潮通用软件有限公司 | Method for automatically identifying picture |
US20170330048A1 (en) * | 2016-05-13 | 2017-11-16 | Abbyy Development Llc | Optical character recognition of series of images |
CN107644014A (en) * | 2017-09-25 | 2018-01-30 | 南京安链数据科技有限公司 | A kind of name entity recognition method based on two-way LSTM and CRF |
CN108536679A (en) * | 2018-04-13 | 2018-09-14 | 腾讯科技(成都)有限公司 | Name entity recognition method, device, equipment and computer readable storage medium |
-
2019
- 2019-02-27 CN CN201910145824.0A patent/CN109858037A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105844274A (en) * | 2016-03-25 | 2016-08-10 | 浪潮通用软件有限公司 | Method for automatically identifying picture |
US20170330048A1 (en) * | 2016-05-13 | 2017-11-16 | Abbyy Development Llc | Optical character recognition of series of images |
CN107644014A (en) * | 2017-09-25 | 2018-01-30 | 南京安链数据科技有限公司 | A kind of name entity recognition method based on two-way LSTM and CRF |
CN108536679A (en) * | 2018-04-13 | 2018-09-14 | 腾讯科技(成都)有限公司 | Name entity recognition method, device, equipment and computer readable storage medium |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110348021A (en) * | 2019-07-17 | 2019-10-18 | 湖北亿咖通科技有限公司 | Character string identification method, electronic equipment, storage medium based on name physical model |
CN112185520A (en) * | 2020-09-27 | 2021-01-05 | 志诺维思(北京)基因科技有限公司 | Text structured processing system and method for medical pathology report picture |
CN112185520B (en) * | 2020-09-27 | 2024-06-07 | 志诺维思(北京)基因科技有限公司 | Text structuring processing system and method for medical pathology report picture |
CN116229493A (en) * | 2022-12-14 | 2023-06-06 | 国家能源集团物资有限公司 | Cross-modal picture text named entity recognition method and system and electronic equipment |
CN116229493B (en) * | 2022-12-14 | 2024-02-09 | 国家能源集团物资有限公司 | Cross-modal picture text named entity recognition method and system and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109034159B (en) | Image information extraction method and device | |
Wu et al. | Harvesting discriminative meta objects with deep CNN features for scene classification | |
CN105631468B (en) | A kind of picture based on RNN describes automatic generation method | |
CN109858037A (en) | A kind of pair of OCR recognition result carries out the method and system of structuring output | |
CN108664474B (en) | Resume analysis method based on deep learning | |
Xu et al. | Learning unified binary codes for cross-modal retrieval via latent semantic hashing | |
CN114092964A (en) | Cross-domain pedestrian re-identification method based on attention guidance and multi-scale label generation | |
CN112818951B (en) | Ticket identification method | |
CN112380435A (en) | Literature recommendation method and recommendation system based on heterogeneous graph neural network | |
CN109710907A (en) | A kind of generation method and equipment of electronic document | |
Bai et al. | Explain me the painting: Multi-topic knowledgeable art description generation | |
Huang et al. | Isolated Handwritten Pashto Character Recognition Using a K‐NN Classification Tool based on Zoning and HOG Feature Extraction Techniques | |
CN108345901A (en) | A kind of graphical diagram node-classification method based on own coding neural network | |
Huang et al. | Learning social image embedding with deep multimodal attention networks | |
CN109117891B (en) | Cross-social media account matching method fusing social relations and naming features | |
CN113762050A (en) | Image data processing method, apparatus, device and medium | |
CN103455479A (en) | Method and terminal for creating contacts | |
Chang et al. | Efficient graph-based word sense induction by distributional inclusion vector embeddings | |
Liu et al. | Graph convolution network with node feature optimization using cross attention for few-shot learning | |
Jia et al. | A hybrid attention mechanism for blind automatic modulation classification | |
CN116681128A (en) | Neural network model training method and device with noisy multi-label data | |
CN116245106A (en) | Cross-domain named entity identification method based on autoregressive model | |
Satwashil et al. | English text localization and recognition from natural scene image | |
CN112381149B (en) | Reasonable countermeasure analysis method for source camera identification based on deep learning | |
CN114155387A (en) | Similarity Logo discovery method by utilizing Logo mark graphic and text information |
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 |
Application publication date: 20190607 |
|
RJ01 | Rejection of invention patent application after publication |