KR20190084169A - The system and method for the extensive production data, transaction, ads and survey in the crowd-work of the terminal users and how it is used in AI learning system and method - Google Patents

Abstract

The present invention relates to a mass data production, processing, advertising and survey system, and a method through a group work of terminal users, and an artificial intelligence (AI) learning system and a method thereof using the same. The system and method cover the cost of producing meaningful data with an advertising revenue, and learn, for example, AI related to text recognition by using the meaningful data, and further advance the learning of AI through feedback by the method in which terminal users respond to the parts difficult for a computer to answer, such as character recognition, object identification in image data, questionnaire, etc. by using the idle time of the terminal users (e.g., loading time, advertising time of the program in the terminal, etc.), and providing a device which processes and accumulates data in the form of ″question of a user-answer of a user″ and allows users to view advertisements.

Description

[0001] The present invention relates to a system and method for mass data production, processing, advertisement, questionnaire system, and method using group work of terminal users, and an artificial intelligence learning system and method using the same of terminal users and how it is used in AI learning system and method}

The present invention relates to mass data production, processing, advertisement, a questionnaire system and method, and an artificial intelligence learning system and method using the same, and more particularly, to an idle time of terminal users (for example, , Advertisement time, etc.), the terminal users respond to the parts of the computer that are difficult to answer, for example, character recognition, object identification in image data, questionnaire, etc. For example, &Quot;, and by providing a device by which users can view the advertisement in the process, the cost of producing the "meaningful data " is covered by the advertisement cost and the meaningful data is used Learning of artificial intelligence related to character recognition and learning of artificial intelligence again through feedback And more particularly,

Classic advertising methods were in the form of signs such as signs, flyers, and banners. Such a classical advertising method can not specify the target of the advertisement and requires a place and manpower to be able to expose the advertisement to the target of the advertisement.

However, as the spread of personal computers, smart devices, and communication networks such as the Internet are developed, online advertising methods or software based advertising methods are recognized as more efficient advertisement methods than conventional advertisement methods.

Specifically, in the case of the online advertisement method, the advertisement is exposed when the user searches for specific information, or the advertisement is exposed to the user using the software in the case of the advertisement method using the software, . As a result, a large advertisement effect can be obtained as compared with a classical advertisement method in which an advertisement is delivered to an unspecified number of persons. In addition, since the advertisement method using the online advertisement method and the software does not require a separate advertisement, the place and manpower are not required.

2. Description of the Related Art Wireless terminals such as smart phones, notebooks, and smart pads have undergone tremendous development due to the development of the Internet and wireless communication technologies. Among them, wireless terminals such as smart phones have been widely used as portable telephones performing communication functions, This is a multimedia device that is made up of an information terminal and makes it possible to access desired information anytime, anywhere.

In particular, in response to such a trend, wireless terminal manufacturers and distributors have developed and distributed an application program, which is a variety of application programs that can be executed in accordance with an operating program of a wireless terminal. The companies that provide services also develop their own applications that can be executed on wireless terminals and distribute them to users.

As the Apple iPhone developed popularity, manufacturers / distributors / users / companies are developing applications that are applications that can run on such smartphones, and these applications are available in millions of ways Is coming. Smartphones that provide similar functions to Apple's iPhone are Android based on Google's mobile operating system, and a variety of applications are also being developed and provided for handsets using the Android operating system.

Among these various applications, there have been provided applications that can easily make users' schedule, memo, scrap, and the like. However, in addition to a terminal such as a smart phone, a digital pen or a paper with a code printed thereon can be used in addition to the terminal. In addition, there was a contradiction that had to be transmitted on-line through a digital device in order to share data that was produced, that is, data converted from analog data to digital data. Again, analog data (for example, notes with simple content, pictures, pictures, etc.) can be easily transferred without using a digital device, but digital data is not. Anyone who has used a digital terminal should have experienced the situation where they can not share data because there is no battery right away.

In the Internet industry, there have been a number of successful business items that make life easier for people. Amazon's "Mechanical Turk" is one of them. The name comes from the 18th century chess play organization 'The Turk'. It looks as though the machine is holding the chess, but in fact it is a mechanism in which the person enters and adjusts. What did Amazon's services have in common with Turk? The outline of this service is simple. For example, when a questionnaire response of about 600 people is needed for writing a paper in economics or psychology, it is solved at a low cost in just four hours. It translates a large amount of text into a desired language in a short time and changes it into a voice. It seems that powerful software works to alleviate the burden on the customer, but in fact, a number of 'tuckers' in the 'Minnican Turk' are doing it. So this is called artificial artificial intelligence, and it is called crowd-work or digital piecework.

One of the most notable areas of artificial intelligence (AI) is deep-learning. Deep running is not much different from the old Artificial Neural Network (ANN). This is a structure that is repeatedly constructed by linear fitting and nonlinear transformation or activation. In other words, an artificial neural network is a structure that draws lines that can distinguish data well and repeats the addition and subtraction of these spaces well.

For example, when a computer is supposed to detect a cat in a photograph, the abstract image of a computer, called 'cat', is a result of a combination of various factors such as line, surface, shape, color, and size. This is not a problem that can be identified as a simple linear distinction, such as 'a circle, a square,' or 'a yellow cat, or a red cat'. Deep learning aims at creating an optimal divider in a complex image space by repeating 'line drawing, distortion and summing' in this process.

As shown in Fig.

* The above description is a mixture of deep running and classification problems, and in fact deep running can be applied to the problem of regression that deals with real-value.

It is the data that is based on the process of drawing the above-mentioned line and distorting the space. It is optimized based on the data so that it draws a rough line and better results. Deep learning is data learning through a lot of data and a very long time optimization.

Fig. 2 shows that complex neural functions (linear fit + nonlinear transform) can be estimated when a complex function is combined (Pascal Vincent Professeur Aggregate / Associate Professor http://www.iro.umontreal.ca/) ~ vincentp /)

This artificial neural network was actually developed in the 1940s, and in the 1980s, an optimization method called back propagation method was introduced, and artificial neural network research reached its peak. Since then, artificial neural networks have been applied to a variety of fields such as image processing, artificial intelligence, and control. In the 1990s, the research reached saturation, and within the limits, it soon began to meet the dark. In the early 2000's thesis examination, the word 'artificial neural network' was rejected as an old subject. Thus, artificial neural networks have disappeared. In the 2000s, various kernel methods using nonlinear functions (eg, support vector machines, Gaussian processes) are becoming a mainstay of machine learning.

Professor Geoffrey Hinton, a scholar who has walked through artificial neural networks when everyone turned away from artificial neural networks, pointed out that "optimization" is not easy because artificial neural networks are one of the limitations of the outside world. In other words, in order to use artificial neural networks properly, tens of thousands of virtual neurons must be assigned appropriate values to these lines, connected by tens of thousands of wires. This is called parameter training.

Parameter training should be done through optimization algorithms, and it is even impossible to figure out where to go if the optimization algorithm reaches the wrong optimal value, if not the real optimal value.

Hinton, however, has shown that these traps (local minima) can be largely solved through pre-training of the data. This study is a 2006 paper entitled "A fast learning algorithm for deep belief nets (2006)". In this paper, Prof. Hinthon has been able to fine-tune each layer of the artificial neural network first through unsupervised learning method, and by performing artificial neural network optimization by stacking such pre-processed data, And that it can produce great results.

The artificial neural network research opens new electricity from this paper. In particular, artificial neural networks emit more light through big data. If many studies prior to 2006 had made efforts to understand the specific shape of the data, after 2006, supply unlimited data to artificial neural networks of structures hundreds or thousands of times larger than before. And by using the optimization techniques developed since 2006, we have learned a lot of time so that we can produce the best results using this data.

Deep learning techniques are currently being used in a variety of machine learning applications, and are currently being studied for image processing and speech recognition.

The four factors of success of the deep learning method are as follows

1. Preprocessing process using non-bipod learning method

This is what Professor Hinton proposed in 2006. A simple way to learn how to learn Bijido is to have a large number of images of apples, cats, and people entered without distinction like "This is an apple", "This is a cat" or "This person". At this point, the computer finds a way of clustering similar things. At this time, the unique cases are boldly ignored, resulting in the effect of noise reduction. This way of learning non-geographic data, you can fine-tune the data evenly (preprocessing), and putting it into a deep learning network greatly reduces the chances of getting into the wrong optimization.

2. Evolution of Convolutional Neural Networks

Machine learning can proceed directly from "data → knowledge", but it is common to learn from the stage of "data → feature → knowledge" through an intermediate step, feature extraction. For example, in order to recognize an object in a photograph, first a characteristic line or a characteristic color distribution is first extracted from a pixel value, and then, based on this, it is judged as "This is an apple" or "This is a banana."

This intermediate representation step is called a feature map. The performance of machine learning depends on how well the features are extracted (this applies to most machine learning, such as speech recognition and natural language analysis, as well as image processing).

In Fig. 3, the original image (right side) and the characteristic map extracted by the convolutional network are shown.

One of the success factors of Deep Learning is that even these features, which were once picked by human predictions, are extracted using machine learning. In the past, people created algorithms for features such as 'circle', 'line', and 'color'. Now, both 'feature extraction' and 'learning' are included in the deep learning algorithm.

Convolutional Neural Network, which extracts features from multiple stages, is a major trend of deep learning at present and has made great progress especially in image recognition.

3. Recurrent neural network for time series data

Deep learning algorithms can be roughly divided into three categories.

- Deep Belief Network (Deep Auto-encoder) based on non-bipy learning method

- Various variants of convolutive neural networks

- Recurrent Neural Network (RNN) and gate units (eg Long-Short Term Memory (LSTM)) for time series data.

Time-series data refers to data that changes with the passage of time. For example, stock prices also change with time, people's movements change with time, and video changes with time. The deep-running method that shows excellent performance in this time-series data is the recurrent neural network (RNN).

RNN builds an artificial neural network structure every unit time. For example, if 50 hours are spent, 50 artificial neural networks are stacked. It is also called deepest structure during deep run.

The old RNN was so difficult to learn because of the vanishing gradient problem that the artificial neural network was so deep that it was long ago that the long-short term memory gate unit of Professor Jurgen Schmidhuber was placed per node, Overcome the problem and is now positioned to be one of the strongest competitors in convolutive networks.

4. Advances in the emergence and learning of GPU parallel computing

In fact, if you used a lot of data and thought that using a particular algorithm would perform well in the depth run, it was not easy to implement. It might not work well, but it could not be running for hundreds or thousands of computers for months. However, after the concept of General-Purpose Computing on Graphics Processing Units (GPGPU) has emerged, GPU products have been released that enable CPUs and parallel processing at a low cost. Language structures (eg, CuDA) The computing time has been reduced to "tens of minutes." The data pool you can use for your research is so large that you can pull in 10 million high-resolution photos (such as ImageNet) and, if necessary, data from Google or YouTube.

The algorithmic problem of artificial neural network is that the development of Rectified Linear Unit (ReLU) used for nonlinear transformation and the discovery of drop-out that selectively learns a large network greatly improves the performance of deep learning.

Deep learning is not much different from artificial neural networks that have existed for a long time, but due to the development of algorithms, the development of hardware, and the power of big data, it has been evaluated as a machine learning method with the highest performance and is emerging as a hope of future artificial intelligence .

1. Method for exposing an advertisement using an application program and method and apparatus for exposing an advertisement using the same Patent Application No. 10-2007-0051192 2. A reward advertisement apparatus based on a mobile ad-phi phi eye and a method thereof, and a method thereof. [0002] 3. Method for constructing a parsing model using deep learning and a device for performing the same (Patent Application No. 10-2014-0160143

https://deepmind.com/ http://yann.lecun.com/ https://www.linkedin.com/in/andrewyng http://colah.github.io/posts/2014-03-NN-Manifolds-Topology/ http://www.iro.umontreal.ca/~vincentp/ http://boundless.utoronto.ca/?s=our-supporters/geoffrey-hinton/ https://www.cs.toronto.edu/~hinton/absps/fastnc.pdf http://www.shivonzilis.com/machineintelligence http://slownews.kr/41461

4 is a diagram showing companies involved in machine learning and deep learning. Among them, Facebook, Google, Baidu, and Amazon all are deeply committed to deep running. They are spending astronomical costs to build / secure data related to deep running, and to stand out from the competition.

The present invention relates to mass data production, processing, advertisement, a questionnaire system and method, and an artificial intelligence learning system and method using the same, and more particularly, to an idle time of terminal users (for example, , Advertisement time, etc.) to provide a system and method by which terminal users respond to parts that are difficult for a computer to answer, for example, character recognition, number in image data, character, object discrimination,

Also, by processing the data in the form of "user's query-user's answer" through the above process, the user can view the advertisement in this process, And to provide a method of covering them.

In addition, by using the data, it is possible to learn artificial intelligence related to printed characters, artificial intelligence related to numerical recognition, handwritten characters, numbers, artificial intelligence related to pictures, and object recognition, The present invention provides a system and method for enhancing learning of artificial intelligence through feedback.

The present invention also provides a system and method for digitizing various analog data of users using the hardware of a smart phone and easily transmitting the digitized data to an analog method before using the digital device, .

Further, the present invention provides motivation for the users to voluntarily digitize various analog data, thereby continuously accumulating "analog data" and "analog data-digital data" necessary for machine learning, deep learning, And to provide a system and method that can be used.

The present invention also provides a system and a method for allowing terminal users to obtain a profit through idle time by making a request of a user requesting a survey to the survey using the idle time of terminal users, have.

In addition, the present invention provides qualitative and quantitative analysis of tendencies, desires, hobbies, and imports of terminal users through a survey, and provides contents (for example, advertisements, questionnaires, etc.) The present invention provides a system and a method for providing a necessary information to a user of a surveying unit in a short time by providing a means for executing a proper survey and a response to a desired surveying group in a short period of time.

In order to solve the above problems, the present invention identifies the type of users and configures a system capable of performing the following.

USER Type A - A type that uses images (text, photographs) to produce and use data. The server transmits the data to the server, receives the data as ASCII character data, and transmits the service as the user or another user pre-sets in the application according to the present invention (for example, playing a specific video, Down, etc.) can be used.

USER Type B - A type that ASCIIizes the image (text, picture) and evaluates whether the ASCII data of the Type B user matches with the image (text, picture). USER Type A and USER Type B are only roles given at that moment, and USER Type according to the present invention is basically A and B. The USER Type A is a means for using a specific service to produce an image, and the produced data is transmitted to a server. The data is transmitted to USER Type B, and USER Type B is a character converted into an ASCII code Data is input and transmitted to the server. The server then passes the ASCII coded character to USER Type A, and the terminal of USER Type A performs the specific action as predefined.

In this description, USER Type A and USER Type B may basically mean a terminal of USER Type A and a terminal of USER Type B. This is preferably contextually interpreted.

In this process, USER Type B generates revenue.

*

USER Type C - You can get what you want by Type (USER Type A) and USER Type (B), which promotes companies and organizations through images (texts, pictures) and systems according to the present invention.

USER Type D - A type that advertises USER Type A and USER Type B using the advertisement system according to the present invention

USER Type E - A type that uses the system according to the present invention because it requires ASCII characterization for the image (character, photograph, etc.) source that it owns. USER Type B allows you to get what you want. These may be companies, organizations, individuals who are making the following services. The image sources that they primarily request are business cards, receipts, etc.,

USER Type F - A type of company or organization that requires a survey. These surveys are performed by USER Type B users.

And provides a system according to the present invention so that the users can achieve the purpose through the client terminal and the server.

By learning the AI using the above data, the system according to the present invention gradually reduces the dependence on the human being and solves it through AI.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Reference will now be made in detail to the preferred embodiments of the present invention, examples of which are illustrated in the accompanying drawings.

The present invention provides a system in which terminal users respond to parts that are difficult for a computer to answer, for example, character recognition, numbers in image data, characters, object discrimination, and the like.

Also, through the above system and method, terminal users are provided with a system for generating revenue using idle time (e.g., loading time of a program in a terminal, advertisement time, etc.).

Through the above process, highly reliable image data required for artificial intelligence learning and ASCII code data paired therewith can be accumulated in the form of "user's query-user's response ".

Also, through the present invention, the costs incurred in the above process can be covered by advertisement revenue.

In addition, by using the data, it is possible to learn artificial intelligence related to printed characters, artificial intelligence related to numerical recognition, handwritten characters, numbers, artificial intelligence related to pictures, and object recognition, Feedback can be used to enhance the learning of artificial intelligence.

In addition, through the present invention, it is possible to digitize various analog data of users by using the hardware of a smart phone, and to easily transmit the digitized data and general digital data to an analog method before using the digital device Method.

In addition, the present invention allows a request of a user requesting a survey to be utilized in a survey by utilizing the idle time of the terminal users, thereby making it possible for the terminal users to earn profit through the idle time.

1 is a schematic diagram illustrating a data learning structure of an artificial neural network;
Figure 2 shows that complex neural functions (linear fit + nonlinear transform) can be estimated when a combination of neurons
3 is a diagram showing an original image (right side) and a feature map (left side) extracted by a convolutional network
Fig. 4 is a diagram showing companies related to machine learning
5 is a schematic diagram of an overall system configuration according to the present invention;
Figure 6 is a schematic diagram of a server according to the present invention;
7 is a schematic diagram of an AI server according to the present invention.
8 is a schematic diagram of a user terminal according to the present invention.
9 is a view showing an example of a use screen of USER Type A in the present invention
10 is a view showing an example of a use screen of USER Type B in the present invention
11 is a view showing an example of a use screen of USER Type C in the present invention
12 is a view showing an example of a use screen of USER Type D in the present invention
13 is a diagram showing an example of offline data to be provided to a system according to the present invention for converting USER Type E into digital data in the present invention
14 is a view showing an example of a use screen of USER Type F in the present invention
15 is a diagram showing an example of the operation sequence of USER Type A in the present invention
16 is a diagram showing an example of the operation sequence of USER Type B in the present invention
17 is a diagram showing an example of the operation sequence of the server in the present invention
18 is a diagram showing an example of a server system in the present invention.
19 is a diagram showing an example of the relationship structure of USER Type B in the present invention
20 is a diagram illustrating an example of a method in which USER Type B generates profit in the present invention.
FIG. 21 is a flowchart illustrating an example of a method of generating profit by USER Type B in the present invention.
22 is a diagram showing an example of a server system in the present invention.
FIG. 23 is a diagram showing an example in which, in the present invention, the USER Type A is analog data 1
FIG. 24 is a flow chart showing the process of generating analog image data 2
FIG. 25 is a diagram showing an example in which, in the present invention, the analog image data 3
26 is a diagram showing an example of a method of calculating the length of all characters in analog image data produced by USER Type A in the present invention
27 is a diagram showing an example of a method of calculating the length of each character constituting a sentence in analog image data produced by USER Type A in the present invention
28 shows an example of a method of calculating length and height of each character constituting a sentence in analog image data produced by USER Type A in the present invention,
29 is a diagram illustrating an example of a method of calculating the length and height of each character constituting a sentence in analog image data produced by USER Type A in the present invention
30 is a diagram showing an example of the relationship structure of USER Type A in the present invention
FIG. 31 is a diagram illustrating an example in which USER Type A receives GPS data when photographing / scanning analog image data and analog image data that USER Type A intends to scan according to the received GPS data can be recognized in the system according to the present invention 1 showing an example of displaying an area where
FIG. 32 is a diagram illustrating an example in which USER Type A receives GPS data when photographing / scanning analog image data, and analog image data that USER Type A desires to scan according to received GPS data can be recognized in the system according to the present invention 2 showing an example of displaying an area where
FIG. 33 is a diagram illustrating an example in which USER Type A receives GPS data when photographing / scanning analog image data and analog image data that USER Type A desires to scan according to the received GPS data can be recognized in the system according to the present invention 3 showing an example of displaying an area where
FIG. 34 is a flowchart illustrating an example of a work performed in a terminal of USER Type A in which USER Type A photographs / scans analog image data and is in communication with a server. FIG.
FIG. 35 is a flowchart illustrating an example of a process in which data produced by the USER Type A is verified several times by the USER Type B and stored in the DB, and the AI server learns the data again. .
FIG. 36 is a flowchart showing an example of the operation of a process in which the data produced by the USER Type A is stored in the DB, verified by the USER Type B, and then re-learned by the AI Server.
FIG. 37 is a diagram illustrating an example of a process in which AI server performs reinforcement learning through data subjected to secondary verification in the present invention; FIG.
38 is a diagram showing an example in which USER Type A records handwriting characters + handwriting numbers in the analog data created in the memo.
FIG. 39 is a diagram showing a USER Type A showing a handwriting character + handwriting number on a note
40 is a view for photographing / scanning the note of FIG. 38 with a terminal of USER Type A
FIG. 41 is a view for photographing / scanning a note of FIG. 39 with a terminal of USER Type A
42 is a diagram showing that USER Type A records the handwriting character + primary handwriting code in the analog data created in the memo.
FIG. 43 is a diagram showing a case where USER Type A indicates a handwriting character + primary handwriting code on a note
FIG. 44 is a view for photographing / scanning the note of FIG. 42 with a terminal of USER Type A
45 is a view for photographing / scanning a note of FIG. 43 with a terminal of USER Type A
46 is a diagram showing that USER Type A records the handwriting character + secondary handwriting code in the analog data created in the memo.
47 is a diagram showing a USER Type A showing a handwriting character + secondary handwriting code on a note
Fig. 48 is a view for photographing / scanning the note of Fig. 46 with a terminal of USER Type A
Fig. 49 is a diagram showing a photograph taken / scanned by the terminal of USER Type A in the note of Fig. 47
50 is a flowchart showing an example of an operation for assigning image character data from USER Type B's in a program according to the present invention and calculating a fee according to the inputted order
51 is a diagram illustrating a process in which a USER Type B verifies image character + ASCII character data through a terminal in a program according to the present invention and stores the data in the DB.
52 is a diagram showing that a user having an ID of sonsory in the USER Type B in the program according to the present invention can purchase a desired item through points and benefits received by participating in the program according to the present invention
FIG. 53 is a flowchart illustrating a method of using the application according to the present invention. A drawing showing an example of a lock screen of USER Type A
54 is a flowchart illustrating a method of using an application according to the present invention. 1 showing an example of a lock screen of USER Type B;
55 is a flowchart illustrating a method of using an application according to the present invention. 2 showing an example of a lock screen of USER Type B;
FIG. 56 is a flowchart illustrating a method of using the application according to the present invention. 3 showing the lock screen of USER Type B;
57 is a flowchart illustrating a method of using an application according to the present invention. 4 shows an example of a lock screen of USER Type B;
58 shows a method of dividing scan data before delivering offline data taken / scanned by a server or a client application according to the present invention to USER Type B, and transferring the scan data to several USER Type B according to the number of segments One
59 shows a method of dividing offline data shot / scanned by a server or a client application according to the present invention before delivering the data to USER Type B and transferring the offline data to several USER Type B according to the number of divisions
FIG. 60 is a flowchart illustrating a method of using the application according to the present invention. Figure 1 shows a screen where USER Type B participates in the survey.
FIG. 61 is a flowchart illustrating a method of using an application according to the present invention. USER Type B participates in the survey.
62 is a flowchart illustrating a method of using an application according to the present invention. USER Type B shows a screen participating in the survey.
63 is a flowchart illustrating a method of using an application according to the present invention. A drawing showing an example of a lock screen of USER Type A, B
FIG. 64 is a diagram illustrating the relationship between a questionnaire and an advertisement in the present invention. FIG.
65 is a diagram showing the relationship between the questionnaire and advertisement in the present invention;
Figure 66 illustrates an example service-translation service that may be implemented through the present invention.
67 shows an example 1 of USER Type B screen when an application according to the present invention is applied to a PC.
68 shows an example 2 of USER Type B screen when an application according to the present invention is applied to a PC.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a detailed description of preferred embodiments of the present invention will be given with reference to the accompanying drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

As used herein, the term " one "is to be interpreted as" one "or" more than one ".

As used herein, the term "plurality" is to be interpreted as "two" or "two or more ".

As used herein, the term "different" is preferably interpreted from at least "two or more comparison objects ".

It is to be understood that the words "comprising" or "having" as used herein are to be interpreted as "including (openly)".

As used herein, the word "connection" may not necessarily be "direct ", and does not necessarily imply" mechanical "connection.

In the present specification, when an element 'transmits' data or a signal to another element, the element can transmit the data or signal directly to another element, and data Or signals can be transmitted to other components.

In this patent, analog image data refers to handwritten or drawn texts, pictures, or printed prints. If the digital image data is captured / scanned by a digital device and stored, transferred, or shared in a digital device, .

In some workflow diagrams, the items in the border may be about the item or process that covers the possibility of always running at all times when the process is going on. The distinction is that each item is a work item, , And should be understood in the overall picture according to each situation.

A server may refer to all servers, including AI servers and network management servers, or to individual servers and their groups that are separated separately.

1 is a schematic diagram illustrating a data learning structure of an artificial neural network;

Figure 2 shows that complex neural functions (linear fit + nonlinear transform) can be estimated when a combination of neurons

3 is a diagram showing an original image (right side) and a feature map (left side) extracted by a convolutional network

Fig. 4 is a diagram showing companies related to machine learning

5 is a schematic diagram of an overall system configuration according to the present invention;

Figure 6 is a schematic diagram of a server according to the present invention;

7 is a schematic diagram of an AI server according to the present invention.

8 is a schematic diagram of a user terminal according to the present invention.

9 is a view showing an example of a use screen of USER Type A in the present invention

10 is a view showing an example of a use screen of USER Type B in the present invention

11 is a view showing an example of a use screen of USER Type C in the present invention

12 is a view showing an example of a use screen of USER Type D in the present invention

13 is a diagram showing an example of offline data to be provided to a system according to the present invention for converting USER Type E into digital data in the present invention

14 is a view showing an example of a use screen of USER Type F in the present invention

15 is a diagram showing an example of the operation sequence of USER Type A in the present invention

16 is a diagram showing an example of the operation sequence of USER Type B in the present invention

17 is a diagram showing an example of the operation sequence of the server in the present invention

18 is a diagram showing an example of a server system in the present invention.

19 is a diagram showing an example of the relationship structure of USER Type B in the present invention

20 is a diagram illustrating an example of a method in which USER Type B generates profit in the present invention.

FIG. 21 is a flowchart illustrating an example of a method of generating profit by USER Type B in the present invention.

22 is a diagram showing an example of a server system in the present invention.

FIG. 23 is a diagram showing an example in which, in the present invention, the USER Type A is analog data 1

FIG. 24 is a flow chart showing the process of generating analog image data 2

FIG. 25 is a diagram showing an example in which, in the present invention, the analog image data 3

26 is a diagram showing an example of a method of calculating the length of all characters in analog image data produced by USER Type A in the present invention

27 is a diagram showing an example of a method of calculating the length of each character constituting a sentence in analog image data produced by USER Type A in the present invention

28 shows an example of a method of calculating length and height of each character constituting a sentence in analog image data produced by USER Type A in the present invention,

29 is a diagram illustrating an example of a method of calculating the length and height of each character constituting a sentence in analog image data produced by USER Type A in the present invention

30 is a diagram showing an example of the relationship structure of USER Type A in the present invention

FIG. 31 is a diagram illustrating an example in which USER Type A receives GPS data when photographing / scanning analog image data and analog image data that USER Type A intends to scan according to the received GPS data can be recognized in the system according to the present invention 1 showing an example of displaying an area where

FIG. 32 is a diagram illustrating an example in which USER Type A receives GPS data when photographing / scanning analog image data, and analog image data that USER Type A desires to scan according to received GPS data can be recognized in the system according to the present invention 2 showing an example of displaying an area where

FIG. 33 is a diagram illustrating an example in which USER Type A receives GPS data when photographing / scanning analog image data and analog image data that USER Type A desires to scan according to the received GPS data can be recognized in the system according to the present invention 3 showing an example of displaying an area where

FIG. 34 is a flowchart illustrating an example of a work performed in a terminal of USER Type A in which USER Type A photographs / scans analog image data and is in communication with a server. FIG.

FIG. 35 is a flowchart illustrating an example of a process in which data produced by the USER Type A is verified several times by the USER Type B and stored in the DB, and the AI server learns the data again. .

FIG. 36 is a flowchart showing an example of the operation of a process in which the data produced by the USER Type A is stored in the DB, verified by the USER Type B, and then re-learned by the AI Server.

FIG. 37 is a diagram illustrating a process in which an AI server performs reinforcement learning (to be deleted thereafter) (a process of recognizing a randomly changed image character -> validation to USER Type B) through data subjected to secondary verification in the present invention.

38 is a diagram showing that USER Type A records handwriting characters + handwriting numbers in the analog data created in the memo. (To be deleted thereafter) and shooting / scanning it, so that USER Type A is set to the server according to the present invention or the client preliminarily set in the client application according to the present invention (the specific cloud, for example, Edit, etc.) automatically. In this process, the handwriting + number image is transmitted to the USER Type B through the server, and character (or character + number) data is received as an ASCII code value recognizable by the terminal of USER Type A, It will run according to the options you set in advance.

FIG. 39 is a diagram showing a USER Type A showing a handwriting character + handwriting number on a note

40 is a view for photographing / scanning the note of FIG. 38 with a terminal of USER Type A

Usually upload data.

FIG. 41 is a view for photographing / scanning a note of FIG. 39 with a terminal of USER Type A

It usually downloads data.

42 is a diagram showing that USER Type A records the handwriting character + primary handwriting code in the analog data created in the memo.

FIG. 43 is a diagram showing a case where USER Type A indicates a handwriting character + primary handwriting code on a note

FIG. 44 is a view for photographing / scanning the note of FIG. 42 with a terminal of USER Type A

45 is a view for photographing / scanning a note of FIG. 43 with a terminal of USER Type A

46 is a diagram showing that USER Type A records the handwriting character + secondary handwriting code in the analog data created in the memo.

47 is a diagram showing a USER Type A showing a handwriting character + secondary handwriting code on a note

Fig. 48 is a view for photographing / scanning the note of Fig. 46 with a terminal of USER Type A

Fig. 49 is a diagram showing a photograph taken / scanned by the terminal of USER Type A in the note of Fig. 47

50 is a flowchart showing an example of an operation for assigning image character data from USER Type B's in a program according to the present invention and calculating a fee according to the inputted order

51 is a diagram illustrating a process in which a USER Type B verifies image character + ASCII character data through a terminal in a program according to the present invention and stores the data in the DB.

52 is a diagram showing that a user having an ID of sonsory in the USER Type B in the program according to the present invention can purchase a desired item through points and benefits received by participating in the program according to the present invention

53 is a flowchart illustrating a method of using an application according to the present invention. A drawing showing an example of a lock screen of USER Type A

54 is a flowchart illustrating a method of using an application according to the present invention. 1 shows an example of a lock screen of USER Type B (character input for a character image transferred from USER Type A).

55 is a flowchart illustrating a method of using an application according to the present invention. 2 showing an example of a lock screen of USER Type B (evaluating whether or not the character inputted by another USER Type B user is correct).

FIG. 56 is a flowchart illustrating a method of using the application according to the present invention. 3 shows an example of a lock screen of USER Type B (direct advertisement input);

57 is a flowchart illustrating a method of using an application according to the present invention. Example of lock screen of USER Type B (event advertisement -> click of advertisement, event of free download of paid app, etc.) 4.

58 illustrates a method of dividing offline data shot / scanned by a server or a client application according to the present invention before delivering it to USER Type B and delivering it to various USER Types B in accordance with the number of divisions

59 shows a method of dividing offline data shot / scanned by a server or a client application according to the present invention before delivering the data to USER Type B and transferring the offline data to several USER Type B according to the number of divisions

FIG. 60 is a flowchart illustrating a method of using the application according to the present invention. Figure 1 (also shows a direct input lock screen) showing a screen in which USER Type B participates in the survey.

FIG. 61 is a flowchart illustrating a method of using an application according to the present invention. USER Type B shows a screen for participating in the survey. (NOPE and YEP can be displayed on the left and right, and a lock screen may be displayed.)

62 is a flowchart illustrating a method of using an application according to the present invention. USER Type B Figure 3 (Multiple choice, locked screen) showing a screen participating in the survey.

63 is a flowchart illustrating a method of using an application according to the present invention. A drawing showing an example of a lock screen of USER Type A, B

FIG. 64 is a diagram illustrating the relationship between a questionnaire and an advertisement in the present invention. FIG. (The service operator according to the present invention conducts a direct survey or uses the data of the survey companies with the consent,

65 is a diagram showing the relationship between the questionnaire and advertisement in the present invention;

Figure 66 illustrates an example service-translation service that may be implemented through the present invention.

67 shows an example 1 of a USER Type B screen when an application according to the present invention is applied and used in a PC. For example, when you run an explorer or run an MS word or the like, the data to be input manually appears along with the advertisement on the screen to be executed. It is possible to input image characters with the same ID in a portable terminal such as a smart phone, a notebook computer, and a computer, and all of the USER Type B operations can be performed in the system according to the present invention described above. That is, from the secondary verification of image data, it is possible to input the value of the scan data such as business card, receipt, and participate in the survey. This is self-evident.

68 shows an example 2 of USER Type B screen when an application according to the present invention is applied to a PC.

The terminal

- Data production, management, processing

User A type is primarily for producing character images for the pic- ture, but it can often be photographed and uploaded for mathematical problems, or a sentence written in one language may be translated into another language. This means that if a certain symbol +, -, x, /, etc. exist in the dictionary, it is accepted as a mathematical problem solving request. In the case of a translation for a language, a translation target language (source language, original language) , Or a translation language), and the terminal may automatically select a translation target language through an image produced by the user A type.

If a user of the user A type is set in Korean to English in case of translation in the presetting, the user can shoot the terminal, whether it is printed English or handwritten by his hand, If it is judged that the language is different from the native language at the time of setting, it is automatically transmitted to the user B type capable of translation. User B type is also a terminal user. User B type You can also pre-select the language you write or use when you sign up.

* The user A type and the user B type can be considered to be divided according to how the application according to the present invention is used.

The application is based on a generic application, but the ultimate goal of this item can be to take advantage of it in the terminal's lock screen (the initial screen).

- Distribution method

 The same image data (this image data may be regarded as a small unit of image data produced through a special process on the server) can be sent to one person only, sent to three or more persons, and sent to one person on the system Within a certain time, you can only send it to one person (in this case, it is proper to systematically process only one person to enter data), and if you send it to three people, you can input only the above three people within a certain time. This could be a solution to problems that might occur when user Btypes did not fit into the above one or three order when they entered data. It would seem to be a problem not to provide any points or caches because you did not get into the rankings even though you entered a fairly fast time. Such

A valid system (sent to three if you need to input to three, omission if there is no answer within a reasonable time, finish the case)

- Advertising

- Fixed ad

 When user B type is text input for request of user A type

- Interactive ads

- survey

Open source code is available in API form, allowing more people to create a myriad of programs that voluntarily become user A type or user B type.

How earnings are calculated

  - How to check through the survey

- Advertising

- It can be used offline.

If you want to use offline even if you want to use it, download the data that can be worked offline in the user B type (for example, a piece of data requiring translation of one book), and if the online connection is available, You can also upload data as you choose.

The user may be user A type, user B type, user C type, user D type, user E type, or user F type, depending on the purpose of using the system according to the present invention .

In the present invention, the types of users are specified as follows and a system capable of performing the types is configured.

USER Type A - A type that uses images (text, photographs) to produce and use data. The server transmits the data to the server, receives the data as ASCII character data, and transmits the service as the user or another user pre-sets in the application according to the present invention (for example, playing a specific video, Down, etc.) can be used.

USER Type B - A type that ASCIIizes the image (text, picture) and evaluates whether the ASCII data of the Type B user matches with the image (text, picture). USER Type A and USER Type B are only roles given at that moment, and USER Type according to the present invention is basically A and B. The USER Type A is a means for using a specific service to produce an image, and the produced data is transmitted to a server. The data is transmitted to USER Type B, and USER Type B is a character converted into an ASCII code Data is input and transmitted to the server. The server then passes the ASCII coded character to USER Type A, and the terminal of USER Type A performs the specific action as predefined.

In this description, USER Type A and USER Type B may basically mean a terminal of USER Type A and a terminal of USER Type B. This is preferably contextually interpreted.

In this process, USER Type B generates revenue.

USER Type C - You can get what you want by Type (USER Type A) and USER Type (B), which promotes companies and organizations through images (texts, pictures) and systems according to the present invention.

USER Type D - A type that advertises USER Type A and USER Type B using the advertisement system according to the present invention

USER Type E - A type that uses the system according to the present invention because it requires ASCII characterization for the image (character, photograph, etc.) source that it owns. USER Type B allows you to get what you want. These may be companies, organizations, individuals who are making the following services. The image sources that they primarily request are business cards, receipts, etc.,

USER Type F - A type of company or organization that requires a survey. These surveys are performed by USER Type B users.

Server - The server according to the present invention does not end at the level of simply helping to handle requests among users but it is a system that continuously validates the algorithm (that is, self learning) through the data of users and repeats the deep learning, Developed to a better environment.

Capture system - For non-public data such as business cards, the program can not capture the screen itself, and the data is divided and sent to B type users. Infringement of personal information is minimized.

For example, partitioning of data and its operation may mean:

By default, business cards have a layout. Due to the layout, business cards can be divided into several areas, which can be reasonably divided by the following criteria.

"Use as a baseline to set and divide a hypothetical line without characters, but divide into n regions with the average width of the divided regions being most similar"

According to the above criterion, a business card can be divided into n equal parts (n is a natural number, and 4 is suitable). Each individual area of the divided card image is transmitted to the terminal of the B type users, and the data in the image is transferred from the "image data to the digital character data having the ASCII code value", and then it is retransmitted to the first requestor F type user.

- Similarly, protection measures may be necessary because personal information about user A may exist in data that user A has photographed / scanned. For example, it can be divided into several kinds of user B type by dividing it according to syllable (Korean) consonant, vowel (roman language series, Japanese, etc.), and a word and it may be possible that the user himself inserts a separator such as a special character The promised system among service providers, and can be distinguished by a letter)

- Benefits can be used for billing, etc. that can use the Internet according to specific countries, and can be used for specific groups, campaigns, donations to support people.

- When the user A directly inputs the number written with the handwriting character in the application according to the present invention by using the preset service according to the present invention (by inputting the special character including the special character, (You can connect to a specific app, go to a specific web screen, or download data). It is not necessary to go through User B in this process. That is, if it is possible to cope with the data stored in the terminal A or the server of the user A as much as possible, the user B may not go through the user B, and if the AI server receives the image character data - a set of ASCII character data The user A can store all the image character data captured / scanned by the user A in the server and convert the ASCII character data to the AI server through the user B.

The image character data refers to image data including characters among the images taken / scanned by the user A.

The ASCII character data means data in which characters existing in the image data photographed / scanned by the user A are input and stored in the ASCII code readable by the user B by the computer.

An advertisement technique that shows an advertisement having a keyword corresponding to the ASCII character or similar to the user A when the user B has made an ASCII character for a character existing in the image character data shot / scanned by the specific user A. The advertisement is applied when the user A plays the role of the user B.

- Ads can be provided differently according to the situations when specific APP is installed, when they are moved to the actual URL, when they are exposed, when they are clicked.

For the secondary verification, Figure 51. Even when user B is in offline environment, he downloads a specific amount of "image character + ASCII character DB" that requires a large amount of secondary verification (Eg, 7 days, 2 weeks, etc.) for the "image character + ASCII character DB" for a certain period of time - for example, to prevent other user B from verifying it. You can also receive rewards like points by submitting to the server.

Drawings (listed in sequence, regulated again)

- GPS coverage related.

 B general coverage, modified coverage .. continued correction, some formula ..

For c modified form, modified to circle, modified to cam form. When other codes and regions overlap?

* Drawing a sentence.

Drawing of dividing a business card

- Character input and advertisement screen, survey screen,

Acquiring information to acquire characters

Code information - can be non-code or numeric

GPS information [see document 160702-3]

Character length information

Frequency of use of b friends

If the information related to the reliability of the author (ie, all of the above conditions are the same, then this information can be a big differentiator.

유사 Similar age, gender, interest ... Etc. - Membership information - Acquisition of information by acquisition of subscription or survey.

ㅡ # translation version learning tool for Korean and # English. If you apply this a bit more, you can become a learning tool to understand mathematical problems with mathematical symbols and solve mathematical problems.

It can be used as a short translation tool.

Ads that are not linked to other sites? There may be things that are needed. For example, there are lock apps in English, so it would be nice to load specific pages, profit share. This way?

- The first page of the site seems to be able to introduce this.

The first screen of execution of the application

Intermediate middle loading screen

Similarly, the video may be executed separately from the input request, the input during execution, and the input before executing the video.

*

Benefit is to change to a cache when points and points accumulate to some extent, and it seems to have to join with various cyber caches.

B Benefits can be various. It can be music, a movie, or something.

For example, you can use the Internet.

It may be linked to a bank account - It can be linked with FinTech providers.

 - B type user, advertisement ranking by input ranking

Even if you are looking at b different ads, you can pay advertising costs in the same format as ad rate x according to your order

The system can accept up to three people, but you do not have to wait until all three are recruited. If you enter one, you can pass the value to the A type user through the server, and let the A type user decide directly. (Systematically, the B type user can implement the P2P that directly transfers to the A type user. It is not this way of conveying what is important.) If a user is false or false, I have to enter the same again, which is usually not a concern. Also, through the revalidation process of B type users, it should not be called "B type user". But, after all, B type users constitute a huge relationship between them. I think it would be better to do it.

ㅡ B type user verifies the first verified image character by other B type users

It is targeted to B type users who use the corresponding language.

B) verify the same image character sent by the user in the Choi type A type

B Dating App - Tinder's UI will be used.

It would be worth 1 won for one verification.

ㄴ (-2 won if wrong choice)

Relationship between B type users - Pyramid type relationship

To strengthen responsibility, maintain reliability

For example, an upper-level user can only verify image characters that are verified first by the lower-level users. Therefore, securing a large number of lower-level users is linked to their own profits. However, the general pyramid type of business method is given that income is unearned, in the first place we lined with such a structure.

 To create a strong relationship with the parent user and the child user, the parent user can process the secondary verification image that the parent user has not processed in the pyramid to which the user belongs. - In this structure, the more you acquire each other, and the more diligent you get, the more profit you can get.

This results in mutual feedback (evaluating the users in my pyramid, that is, the ones in my pyramid), reliability, voluntary verification.

The more you have, the better the number of images you can have. However, the number of images you can process is limited because of the number of images you can process (even if you do, you will cherish your account for as long as you use it, I think it would be nice to give a little more benefit in the second verification. -> In fact, long-term people have a high level of credibility.

It is possible to evaluate the first verification according to the second verification of b.

If the second verification is 1 won per one time, it will be appropriate. You can earn 100 won per minute. 6,000 won per hour! Can be set within the limit not violating the minimum threshold.

- B type Provides the programmer with random changes to the first verified image character to give the user a variety of examples to re-verify and learn AI (I think this would be useful only at the beginning of the system)

It is targeted to B type users who use the corresponding language.

This is not the same as verifying the same image character.

A combination of letters and handwritten binary codes

I am going to insert the b non-code patent claiming priority, but mention it here once again.

A combination of letters and numbers

It is said that the number is already reading at the level of the person. If you use it, it will be a very convenient level of on-line integrated management system. And the letter recognition before the numbers is pretty easy, it seems to be almost non-code level.

- After the first verification, the next verification will display the advertisement with the similarity to the previous verification character. Compared to users who did not like this, it would be a meaningful experiment. -> This can be linked to the survey to create meaningful data. Later, it will be used as useful data in advertisement execution.

ㅡ B type An event where a user is working for a certain period of time and then the event disappears according to the advertisement of the advertiser. - An event where a specific app can be downloaded.

* PPI (Pay per Install) or CPI (Cost per Install)

B Type user can be anyone. B type user when installing app by default. When loading the first time, ads will appear and image characters and fields will appear soon. When a character suited to an image character is inserted, the corresponding ASCII code value is stored in the server as data that is paired with the image character. It can be delivered to the A type user after it is delivered to the server, or it can be delivered to the A type user as P2P separately from the server. Likewise, the image data of the A type user may be transmitted to the B type user without going through the server. In this case, the server may only receive a pair of data consisting of the image character-ASCII code value from the B type user. Of course, the advertisement is executed !!

*

- How to check domain length [

A type When a user writes a non-code on a terminal, the corresponding domain is immediately searched for and completed, and vice versa. Then, according to the search condition of the previously used non-code or domain, Very few to be derived.

A type user, B type user need not be a person. It could be a program-based bot.

 These may be people, but they may be programs that perform a series of processes in a terminal or a server. In other words, if A type user is a program and can not read the value of a specific part, the program might ask for help from a person, ie, B type user.

The order may be in the form A -> server -> B. Then A can solve the specific problem by entering the correct value (ie, from B) from the server. [Refer to document 160701-2]

Non-code-domain pairs, once used, can be used repeatedly, there is no reason not to find.

[Refer to document 160703-1]

- Ability to pre-download data that requires a few hours of secondary verification when you are offline, when you are on an airplane, and to access and submit online during a specific period - During this period, Does not allocate the data to other users until the period of the corresponding user expires (or excludes the number of users who use the offline service in the number of persons, if the second verification is not one but two).

- It is also an advertisement that allows you to directly input important phrases of an advertisement.

- It would be convenient to expand the description with the lock screen app while explaining it as a general app form.

About the questionnaire, [Refer to document 160705]

==========

The overall emphasis is on semi-automation, that is, semi-automation, that is, firstly, a variety of information to be input by a person, followed by various kinds of text, and various information of the user to read the text, When you shoot / scan the text ... The present invention explains in detail the part that compares the input information with the first inputted information to find the text. In other words, it will have the effect of autonomously appearing to read a lot of characters. The advantage of this is that it does not recognize characters directly when recognizing text. That is, it does not distinguish whether the character is English, Korean, Japanese, Arabic or Chinese. This will be the most optimal device, system for obtaining language data for language learning ... .

This is not likely to lead to a mere result. Automatically generated image character - The value of ASCII code is absolutely necessary data to enhance learning of AI, and it is strengthened once again through verification activities of B type users. In addition, in this process, it is possible to make the AI judgment by sending to the B type user what the AI judges, or to make the intersection judgment that the AI judges the image character-and-ASCII code character judged by the B type user, Continuously shot / scanned image character data in the authenticated area will teach AI how various characters, such as diverse weather, various devices, various angles, various environments, can be varied in various time zones for each user. ...

By enabling the tool to be used at the start and end of almost any device or program, users can make money on idle time, advertisers can do the best marketing (optimized targeting due to various surveys), AI Deep Learning You can create a learning model.

** Deep learning tools. Regardless of the language of any country in the world ..

** Learning the language should be given data to learn new patterns every day. 10 years ago, the average font of Koreans and the average font of Koreans today are different. And the writing language is different. You should focus on this point. In other words, the language continues to change, but the data at a certain point can not recognize the text and understand the context. Continued learning ... .

In the present invention, a system in which terminal users respond to parts that are difficult for a computer to answer, for example, character recognition, numbers in images data, characters, object discrimination, and the like, Can be used.

In addition, mass data production, processing, advertisement, questionnaire system and method, and artificial intelligence learning system and method using them can be used industrially.

In addition, by using the data, it is possible to learn artificial intelligence related to printed characters, artificial intelligence related to numerical recognition, handwritten characters, numbers, artificial intelligence related to pictures, and object recognition, By enhancing the learning of artificial intelligence through feedback, for example, unmanned small vehicles, unmanned robots, drones, etc. can deliver the function of correctly reading text directly besides GPS information when delivering, Can be used industrially.

In addition, the present invention can easily access specific data on-line through photographing / scanning texts, pictures, and symbols written on paper, floor, and walls using the system provided by the present invention and hardware devices of a smartphone This function allows easy access to specific data (eg introduction to the construction company, homepage, construction description, etc.) on the construction site or on the road after the construction using letters, pictures and symbols. It is possible to provide a means (or customers can easily access) which can be very useful industrially. This can be applied in various ways to industry.

100: Internet / Network
400: terminal, server

Claims (1)

The present invention relates to mass data production, processing, advertisement, questionnaire system and method, and artificial intelligence learning system and method using the same. The present invention relates to an idle time of terminal users (for example, Etc.), for example, through a method in which terminal users respond to parts that are difficult for the computer to answer, for example, character recognition, object identification in image data, questionnaires, etc., By providing a device by which users can view advertisements in the course of processing and accumulating data and by using the meaningful data as the cost of producing the "meaningful data " To learn the artificial intelligence related to recognition and to improve learning of artificial intelligence again through feedback. Stem and method

Images