KR20220057268A - Method, device and computer program for classifying patent documents using neural network - Google Patents

Abstract

According to one embodiment of the present disclosure, disclosed is a computer program stored in a computer-readable storage medium. The computer program performs operations for extracting a technology field by using a neural network when executed on at least one processor. The operations may comprise: an operation of inputting a document to a first model for processing a natural language; an operation of acquiring target data to be classified from output of the first model; an operation of inputting the target data to be classified to a second model for classification; and a step of acquiring classification for the document from output of the second model.

Description

Method, device and computer program for classifying patent documents using neural networks

The present disclosure relates to data processing using a computing device, and more particularly, to a technology for classifying patent documents using a neural network.

BACKGROUND With the development of computer and Internet technologies, a large amount of data is being accumulated. The importance of technology for efficiently storing large amounts of data and finding data easily and quickly with desired search conditions is increasing. Current storage media allow for storing and retrieving a large amount of information.

When the storage medium contains a large amount of data, it may be difficult for a user to obtain data of interest. Also, it may take a relatively long time for the computing device to search for data of interest to the user.

In artificial intelligence technology, especially in the field of deep learning, remarkable progress has been made in data automation processing. With the recent development of deep learning technology, there is a trend to use deep learning to classify data.

Republic of Korea Patent Publication No. 2019-0061984 discloses a classification code recommendation system using deep learning.

The present disclosure has been made in response to the above-described background technology, and is intended to provide a computing device for data processing.

According to an embodiment of the present disclosure for realizing the above-described problems, a computer program including instructions stored in a computer-readable storage medium to cause the computer to perform the following operations is disclosed. inputting the document into a first model for natural language processing; obtaining classification target data from the output of the first model; inputting the classification target data into a second model for classification; and obtaining a classification for the document from the output of the second model.

In an alternative embodiment, the document may include at least one of a first part document for claiming the right or a second part document for explaining the content of the claim.

In an alternative embodiment, before the operation of inputting the document into the first model for natural language processing, the operation of analyzing the document may include dividing the document into the first partial document and the second partial document.

In an alternative embodiment, the classification target data may include at least one of a first element including at least a part of the first partial document or a second element including at least a part of the second partial document.

In an alternative embodiment, the second element may include at least one of text or an image having a correlation with the first element of a certain level or higher.

In an alternative embodiment, the classification target data may include a portion selected in advance according to a predetermined criterion of the second partial document.

In an alternative embodiment, the first model may include a recurrent neural network for natural language processing.

In an alternative embodiment, the first model comprises: a first submodel for extracting features of a first partial document; a second submodel for extracting features of the second partial document; a third submodel for deriving a correlation between a first partial document and a second partial document based on outputs of the first submodel and the second submodel; and a fourth submodel for extracting a second element having the highest correlation with the first element based on the correlation.

In an alternative embodiment, before the operation of inputting the classification target data into a second model for classification, the operation may include performing pre-processing on the classification target data.

In an alternative embodiment, the pre-processing of the classification target data may include a process of converting the classification target data into image data.

In an alternative embodiment, the second model may include a model trained using a training data set using the classification target data as training input data and a classification corresponding to the classification target data as a label.

In an alternative embodiment, the second model may include at least one of a plurality of convolutional blocks, a plurality of pooling layers, or a plurality of fully connected layers. .

In an alternative embodiment, the obtaining of the classification for the document from the output of the second model may include: inputting an output of a convolutional block to a pooling layer; inputting an output of the pooling layer to a fully connected layer; and obtaining a classification for the document from the output of the fully connected layer.

In an alternative embodiment, the convolutional block includes a plurality of convolutional layers, the plurality of convolutional layers have the same size, and the size is a size for forming a deep structure of the convolutional layer. may include

According to another embodiment, a data processing method using a neural network is disclosed. A method of processing data using a neural network, the method comprising: inputting a document into a first model for natural language processing; obtaining classification target data from the output of the first model; inputting the classification target data into a second model for classification; and obtaining a classification for the document from the output of the second model.

A computing device is disclosed according to another embodiment of the present disclosure. one or more processors; and a memory storing instructions executable by the one or more processors; including, wherein the one or more processors input a document to a first model for natural language processing, obtain classification target data from an output of the first model, and input the classification target data to a second model for classification and a classification for the document may be obtained from the output of the second model.

According to an embodiment of the present disclosure, a computing device for processing data using a neural network may be provided.

BRIEF DESCRIPTION OF THE DRAWINGS So that the above-mentioned features of the present disclosure may be understood in detail, with a more specific description, with reference to the following embodiments, some of the embodiments are shown in the accompanying drawings. Also, like reference numerals in the drawings are intended to refer to the same or similar functions throughout the various aspects. However, it should be noted that the appended drawings show only certain typical embodiments of the present disclosure and are not to be considered as limiting the scope of the present disclosure, and other embodiments having the same effect may be sufficiently recognized. Be careful.
1 is a block diagram of a computing device for data processing according to an embodiment of the present disclosure.
2 is a schematic diagram illustrating a network function for data processing according to an embodiment of the present disclosure;
3 is an exemplary diagram for explaining an operation of processing data using a neural network according to an embodiment of the present disclosure.
4 is an exemplary diagram for describing a document according to an embodiment of the present disclosure.
5 is an exemplary diagram for describing classification target data according to an embodiment of the present disclosure.
6 is a flowchart of a method of processing data using a neural network according to an embodiment of the present disclosure.
7 is a block diagram illustrating a module for processing data using a neural network according to an embodiment of the present disclosure.
8 is a simplified, general schematic diagram of an exemplary computing environment in which embodiments of the present disclosure may be implemented.

Various embodiments are now described with reference to the drawings. In this specification, various descriptions are presented to provide an understanding of the present disclosure. However, it is apparent that these embodiments may be practiced without these specific descriptions.

The terms “component,” “module,” “system,” and the like, as used herein, refer to a computer-related entity, hardware, firmware, software, a combination of software and hardware, or execution of software. For example, a component can be, but is not limited to being, a process running on a processor, a processor, an object, a thread of execution, a program, and/or a computer. For example, both an application running on a computing device and the computing device may be a component. One or more components may reside within a processor and/or thread of execution. A component may be localized within one computer. A component may be distributed between two or more computers. In addition, these components can execute from various computer readable media having various data structures stored therein. Components may communicate via a network such as the Internet with another system, for example via a signal having one or more data packets (eg, data and/or signals from one component interacting with another component in a local system, distributed system, etc.) may communicate via local and/or remote processes depending on the data being transmitted).

In addition, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or.” That is, unless otherwise specified or clear from context, "X employs A or B" is intended to mean one of the natural implicit substitutions. That is, X employs A; X employs B; or when X employs both A and B, "X employs A or B" may apply to either of these cases. It should also be understood that the term “and/or” as used herein refers to and includes all possible combinations of one or more of the listed related items.

Also, the terms "comprises" and/or "comprising" should be understood to mean that the feature and/or element in question is present. However, it should be understood that the terms "comprises" and/or "comprising" do not exclude the presence or addition of one or more other features, elements and/or groups thereof. Also, unless otherwise specified or unless it is clear from context to refer to a singular form, the singular in the specification and claims should generally be construed to mean “one or more”.

And, the term "at least one of A or B" should be interpreted as meaning "when including only A", "when including only B", and "when combined with the configuration of A and B".

Those skilled in the art will further appreciate that the various illustrative logical blocks, configurations, modules, circuits, means, logics, and algorithm steps described in connection with the embodiments disclosed herein may be implemented in electronic hardware, computer software, or combinations of both. It should be recognized that they can be implemented with To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, configurations, means, logics, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application. However, such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

Descriptions of the presented embodiments are provided to enable those skilled in the art to use or practice the present invention. Various modifications to these embodiments will be apparent to those skilled in the art of the present disclosure. The generic principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Therefore, the present invention is not limited to the embodiments presented herein. This invention is to be interpreted in its widest scope consistent with the principles and novel features presented herein.

1 is a block diagram of a computing device for data processing according to an embodiment of the present disclosure.

1 shows the configuration of a computing device 100 according to an embodiment of the present invention. The computing device 100 illustrated in FIG. 1 is only a simplified example of the computing device 100 according to an embodiment of the present invention. In an embodiment of the present disclosure, the computing device 100 may include other components for performing the computing environment of the computing device 100 , and only some of the disclosed components may configure the computing device 100 .

The computing device 100 may include a processor 110 , a memory 130 , and a network unit 150 .

The processor 110 may include one or more cores, and a central processing unit (CPU) of a computing device, a general purpose graphics processing unit (GPGPU), and a tensor processing unit (TPU). unit) and the like, and may include a processor for data analysis and deep learning. The processor 110 may read a computer program stored in the memory 130 to perform data processing for machine learning according to an embodiment of the present disclosure. According to an embodiment of the present disclosure, the processor 110 may perform an operation for learning the neural network. The processor 110 performs learning of the neural network such as processing of input data for learning in deep learning (DL), extracting features from input data, calculating an error, and updating the weight of the neural network using backpropagation. calculations can be performed for At least one of a CPU, a GPGPU, and a TPU of the processor 110 may process learning of a network function. For example, the CPU and the GPGPU can process learning of a network function and data classification using the network function. Also, in an embodiment of the present disclosure, learning of a network function and data classification using the network function may be processed by using the processors of a plurality of computing devices together. In addition, the computer program executed in the computing device according to an embodiment of the present disclosure may be a CPU, GPGPU or TPU executable program.

According to an embodiment of the present disclosure, the processor 110 may input a document into the first model for natural language processing.

According to an embodiment of the present disclosure, the document may include physically integrated information designed for communication. The document may include a textual representation of a person's thoughts by providing information in symbolic symbols. A document may include a digital file in a specific format according to the development of computers. A specific type of digital file may include, for example, a file such as a music file, an image file, a text file, and the like. The foregoing document is merely an example, and the present disclosure is not limited thereto.

A document according to an embodiment of the present invention may include two or more arbitrary partial documents. Any union of two or more may be combined to include all documents according to an embodiment of the present invention. In another embodiment, the union of two or more arbitrary partial documents may include only a part of the document according to an embodiment of the present invention. Two or more arbitrary partial documents may share a common part. The foregoing partial document is merely an example, and the present disclosure is not limited thereto.

In an embodiment of the present invention, two or more arbitrary partial documents included in a document may be defined as a first partial document and a second partial document. The first partial document and the second partial document may have a correlation based on content of the document. For example, at least a portion of the first partial document may be identical to or substantially corresponding to any portion of the second partial document. In an embodiment of the present invention, the first partial document may be created based on all or a part of the document or the second partial document. For example, the first partial document may be a claim of a patent specification, a summary of a thesis, a table of contents of a book, a review of a book, a summary of a plot of a book, and the like.

In an embodiment of the present invention, the document may include a first partial document including core information of the document and a second partial document having a substantially corresponding relationship with the first partial document including core information of the document. The core information of the document may include information that compressively expresses the content that the document intends to convey. The core information of the document may include information that enables a person or a computing device to know what the document intends to convey even if the person or computing device does not read the entire document. The core information of the document is, for example, a partial document containing the content of abstract or introduction in the thesis, a judgment matter in a judgment, a partial document containing the summary of the judgment, a partial document containing the content of claims in the patent specification, and design registration Drawings in the application, trademarks and designated products in the application for trademark registration, a table of contents or summary of a plot in a book (eg, a magazine, a poem, a novel, a workbook, a textbook) may be included.

In an embodiment of the present disclosure, the second partial document may substantially correspond to the first partial document. The second partial document may have a correlation with the first partial document. The second partial document may have a correlation of at least a certain level with the first partial document. The second part document may include content supporting or specifically explaining the first part document. The second part document may include, for example, an experimental method in a thesis, an experimental result, a reason for a judgment in a judgment, and an invention in a patent specification It may include a specific description for implementing the design, a partial document describing the design in the design registration application, a partial document describing the trademark in the trademark registration application, the contents corresponding to the table of contents in the book, or all of the book.

The document will be specifically described below with reference to FIG. 4 .

According to an embodiment of the present disclosure, the document may include at least one of a first partial document for claiming the right and a second partial document for explaining the content of the claim.

According to an embodiment of the present disclosure, the first part document for claiming rights may include at least a part of the document describing the part to be legally protected according to the laws and regulations of each country. Rights may include patent rights, utility model rights, design rights, trademark rights, and copyrights. For example, if the right is a patent right, the first part of the document for claiming the right may include matters described in the claims. The above-described first partial document is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the second part document for explaining the content of the claim may include at least a part of the document for specifically explaining the part to be legally protected according to the laws and regulations of each country. For example, if the right is a patent right, the second part document for explaining the content of the claim is the title of the invention, the technical field, the technology behind the invention, the content of the invention, a brief description of the drawings, It may include specific content, explanations of symbols, drawings, summaries, and representative views. The title of the invention may include a name that allows categories of the invention to be divided in consideration of the content of the invention. The technical field may include a clear and concise description of the technical field of the invention for which a patent is sought. The technology that is the background of the invention may include a description of the prior art that is thought to be useful for understanding and research of the invention. The content of the invention may include the problem to be solved, the means for solving the problem, and the effect of the invention. A brief description of the drawings may include descriptions of the state of the drawings, the types and parts of the drawings, so that it is easy to understand what the drawings are specifically intended to represent. Specific content for carrying out the invention may include at least one specific description for carrying out the invention so that those skilled in the art can easily understand how the invention is practiced. Description of reference numerals may include descriptions of reference numerals indicating main parts of the drawing. The drawings may include images for explaining specific contents for carrying out the invention. The summary may include a simple description of the invention content within a certain number of characters (eg, within 300 characters). Representative drawings may include reference numbers representing the content of the invention in at least one or more drawings. The above-described second partial document is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the processor 110 may analyze a document and divide it into a first partial document and a second partial document. The processor 110 may divide the first partial document and the second partial document based on a predetermined criterion. Certain criteria may be included based on sub-headings included in the document. For example, when the document is a patent document, the processor 110 recognizes the sub-title [claims], and the contents included in the [claims] are the first partial document, the sub-title [contents of the invention], [simple in the drawings] Description] and [Specific contents for carrying out the invention] can be divided into the second part of the document. The foregoing is merely an example, and the present disclosure is not limited thereto.

An operation of analyzing a document and dividing the document into a first partial document and a second partial document will be described later with reference to FIG. 4 .

According to an embodiment of the present disclosure, the classification target data may include at least one of a first element including at least a part of the first partial document and a second element including at least a part of the second partial document. The first element may be a part of the first partial document selected according to a predetermined criterion. The predetermined criterion may include, for example, a criterion determined based on whether the document is a patent document and is a patent document disclosed to non-confidential persons, or whether the document is a registered patent document. If the document is a patent document and is a patent document disclosed to persons without obligation of confidentiality, the processor 110 may include the independent claim 1 and the dependent claims of the independent claim 1 in the first element. For example, when the independent claim 1 and the dependent claims of the independent claim 1 are claims 2 to 4 in the first partial document, the processor 110 may include the claims 1 to 4 in the first element. When the document is a patent document and a registered patent document, the processor 110 may include only the independent claim in the first element. For example, when independent claims are claims 1, 5, and 7 in the first partial document, the processor 110 may include only claims 1, 5, and 7 in the first element. If the document is a patent document, it is unlikely that the subject matter of the invention will be concentrated in the independent claim (i.e., the subject matter of the invention may be set forth in the dependent claim) in a patent document disclosed to non-disclosure parties. high), the registered patent document has a high probability that the core content of the invention has been collected and registered in the independent claim, so that the processor 110 may select the content included in the first element based on the predetermined criteria as described above. The foregoing is merely an example, and the present disclosure is not limited thereto.

The classification target data will be described later with reference to FIG. 5 .

According to an embodiment of the present disclosure, the second element may include at least one of text or an image having a correlation with the first element of a predetermined level or higher. The correlation between the second element and the first element may include a correlation determined based on a result obtained by the processor 110 using the first model. (A specific process in which the processor 110 derives the correlation between the first element and the second element using the first model will be described later.) The result obtained by the processor 110 using the first model is 0 It can be expressed as a probability value between and 1. For example, the processor 110 may obtain a larger probability value as the correlation between the first element and the second element increases. Accordingly, when the correlation between the first element and the second element 1 is 0.9, the correlation between the first element and the second element 2 is 0.8, and the correlation between the first element and the second element 3 is 0.5, the processor 110 is Only the second element greater than or equal to 0.7 may be selected, and finally, classification target data may include the second element 1 and the second element 2 having high correlation with the first element. The text may include specific descriptions for practicing the invention, for example if the document is a patent document. The text may include, for example, a formula for describing the invention when the document is a patent document. Images may include drawings, for example, if the document is a patent document. The foregoing is merely an example, and the present disclosure is not limited thereto.

The second element is described below with reference to FIG. 5 .

According to an embodiment of the present disclosure, the classification target data may include a portion selected in advance according to a predetermined criterion of the second partial document. The predetermined criterion of the second partial document may include a criterion determined based on whether the processor 110 can obtain an accurate classification using the second model. For example, when the document is a patent document, the technology that is the background of the technical field and the invention may include content that can help the processor 110 to obtain an accurate classification result of the patent document. Since the technical field may include contents or technical classification in which the technical field of the invention for which a patent is sought is clearly and concisely described, the processor 110 may obtain an accurate classification result based on the contents described in the technical field. Accordingly, the part selected in advance according to the predetermined criteria of the second part document may include the technical field and the background technology of the invention in order to obtain an accurate classification result. The processor 110 provides a classification result with higher accuracy by including a pre-selected part (eg, a technical field, a technology that is a background of the invention) in the classification target data regardless of the correlation with the first element. can be obtained The above-described classification target data is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the first model may include a recurrent neural network for natural language processing. The recurrent neural network may include a neural network using time series data as input data. The time series data may include a sequence of data arranged at regular time intervals. The recurrent neural network may also include a neural network with natural language as input data. The processor 110 may divide the natural language at regular intervals (eg, a word may be determined at one regular interval) and input it to the recurrent neural network. The above-described recurrent neural network is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the recurrent neural network may include a Long Short Term Memory Network. The short-term memory network may include a neural network including an input gate that outputs the influence of data input at the current time, and a forget gate that outputs the influence of data input before the current time. . Through this, information of data received in the past more distant than the present time can be remembered without forgetting as time passes, and it is possible to reflect the forgetfulness of information in the data as time goes by. The processor 110 may output a predicted value in consideration of data prior to the current time using a long-term storage network. The above-described recurrent neural network is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the first model includes a first submodel for extracting features of a first partial document, a second submodel for extracting features of a second partial document, a first submodel and a second Based on the output of the second submodel, a third submodel for deriving a correlation between the first partial document and the second partial document and extracting a second element having the highest correlation with the first element based on the correlation It may include a fourth submodel for The above-described first model is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the first submodel may include a recurrent neural network. The recurrent neural network may include a short-term memory network. The processor 110 may extract the features of the first partial document by inputting the first partial document into the long and short-term storage network. The features of the first partial document may include a first submodel encoding matrix. The first submodel encoding matrix may include a vector obtained by inputting a document into a long-term storage network. The obtained vector may include a vector having a probability value of each word as an element of the vector. The processor 110 may obtain a first submodel encoding matrix by using a vector obtained from the first submodel. For example, the set of words included in the first partial document is (x1, x2, x3, x4, .... , xn) and a vector corresponding to each word (the processor enters the word into the first submodel to If the obtained vector) is (v1, v2, v3, v4, .... ,vn), the first submodel encoding matrix can be expressed as [v1, v2, v3, v4, ... ,vn] there is. The above-described first sub-model is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the second submodel may include a recurrent neural network. The recurrent neural network may include a short-term memory network. The long and short-term memory network of the second submodel may include the same network as the long-term and short-term memory network of the first submodel. That is, the parameters of the long and short-term memory network of the second sub-model may be the same as the parameters of the long- and short-term memory network of the first sub model. As the first submodel and the second submodel share the parameters of the network, the processor 110 inputs the same word sequence (eg, sentence) to the first submodel and the second submodel to represent the same or similar vector. can be obtained. The above-described second submodel is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the processor 110 may extract the features of the second partial document by inputting the second partial document to the long-term storage network. The features of the second partial document may include a second submodel encoding matrix. The second submodel encoding matrix may include a vector obtained by inputting the document into a long-term storage network. The obtained vector may include a vector having a probability value of each word as a component of the vector. The processor 110 may obtain a second submodel encoding matrix by using a vector obtained from the second submodel. For example, the set of words included in the second partial document is (y1, y2, y3, y4, ... , yn) and a vector corresponding to each word (acquired by the processor entering the word into the second submodel) one vector) is (v'1, v'2, v'3, v'4, ... ,v'n), the second submodel encoding matrix is [v'1, v'2, v' 3, v'4, ... ,v'n]. The above-described first sub-model is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the processor 110 derives a correlation between the first partial document and the second partial document based on the outputs of the first submodel and the second submodel using the third submodel can do. The processor 110 may derive a correlation between the first partial document and the second partial document based on the first submodel encoding matrix and the second submodel encoding matrix. The processor 110 inputs the result obtained by calculating the transpose of the second submodel encoding matrix and the first submodel encoding matrix to the softmax function, and the words included in the first partial document and the second partial document It is possible to obtain a correlation of words included in . The above-described third submodel is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the processor 110 may extract a second element having the highest correlation with the first element by using the fourth submodel. The processor 110 may obtain a correlation between the first element and at least one or more second elements by using the third submodel. The processor 110 may select one second element having the highest correlation with the first element and add it to the classification target data. The processor 110 may select the top 3% second elements having the highest correlation with the first element and add them to the classification target data. The processor 110 may obtain location information of the selected second element by using the fourth submodel. When the second element is in the tenth paragraph of the document, the processor 110 may obtain position information of the tenth paragraph. The processor 110 may obtain a start position and an end position of the second element. For example, when the start position of the second element is number 10 and the end position is number 20, (start position=10, end position=20) information may be acquired. The processor 110 may efficiently search for the second element in the document using the location information and add it to the classification target data. The above-described fourth submodel is merely an example, and the present disclosure is not limited thereto.

For a detailed description of the method for obtaining classification target data using the above-described first model, see the paper Dynamic Coattention Networks for using Question Answering (published on March 6, 2018, author: Caiming Xiong, Victoor Zhong, Richard Socher).

According to an embodiment of the present disclosure, the processor 110 may perform pre-processing on the classification target data. The preprocessing of the classification target data may include a process for inputting the classification target data into the second model. For example, when the second model is a Convolutional Neural Network (CNN), since the size of the input data is limited, the preprocessing may include a process of making the size of the input data to a size for input to the CNN. In addition, when the second model is implemented based on Tensorflow, data input to the second model may be input in the form of TFRecord. The TFRecord data type may be a binary data type for storing training data of Tensor Flow. Accordingly, the processor 110 may convert the classification target data into a TFRecord form in order to input the classification target data to the second model. Through this, the processor 110 may improve the prediction accuracy of the second model through preprocessing and reduce the learning time. The above-described pre-processing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the pre-processing of the classification target data may include a process of converting the classification target data into image data. The data in the form of an image may include a form of data expressed in the form of a table. Data in the form of an image may include information on at least one or more pixels included in the image. The above-described pre-processing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, data in the form of an image may include data in the form of a table expressed in the form of a table. When the classification target data is text, the processor 110 may make the text into a table form to input the text into the second model. In a table format, a row may include words included in text, and a column may include each component of a word vector corresponding to each word. When the size of input data input to the second model is 500 words, the processor 110 may generate classification target data within 500 words. For example, when classification target data is 600 words and there are three second elements having a relatively high correlation with the first element, by excluding two second elements having a relatively low correlation among them, the classification target data can make the number of words less than 500 words. When the classification target data is 460 words or less, the processor 110 may create a table form by expressing a vector component of 500-460=40 words as 0. The processor 110 may obtain a classification result by inputting text as well as an image to the CNN through pre-processing of making the text into a table form. The above-described pre-processing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the second model may include a model trained using a training data set in which classification target data is used as training input data and a classification corresponding to the classification target data is used as a label. The training data set may include a data set generated by the processor 110 labeling a classification matching the classification target data. The processor 110 may generate a training data set by labeling the classification obtained by inputting the classification target data into the second model. The processor 110 may exclude classification target data incorrectly classified using the second model from the training data set. (For example, erroneously labeled classification target data may be excluded from the training data set by using a data dropout method.) According to another embodiment of the present disclosure, the training data set is configured by an operator to label the classification target data. Data sets obtained through the process may be included. The above-described training data set is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the second model may include at least one of a plurality of convolutional blocks, a plurality of pooling layers, or a plurality of fully connected layers. can The pooling layer may include a layer that extracts only strong features of an image by reducing the dimension of input data. The fully connected layer may include a layer for obtaining classification. The above-described second model is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, a convolutional block may include a plurality of convolutional layers. The convolutional layer may include a layer that extracts features of input data through a convolution operation. A convolutional block may include, for example, two convolutional layers. The processor 110 may batch normalize the result obtained from the convolutional layer. The processor 110 may input the result obtained by inputting the batch normalized result to an activation function to the next convolutional layer. The activation function may include a ReLU activation function. The above-described convolutional block is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the plurality of convolutional layers may have the same size, and the size may include a size for forming a deep structure of the convolutional layer. The convolutional layer may include a kernel. The kernel size may be 3. When the kernel of the convolutional layer included in the second model has the same kernel size 3, the processor 110 uses a neural network having a deep structure (eg, a structure having 26 convolutional layers) with high accuracy. classification can be obtained. According to another embodiment of the present disclosure, when the kernel size of the convolutional layer is large (eg, the size is 7), the processor 110 uses the convolutional layer to have a shallow layer structure (eg, , the classification can be obtained using a neural network having a structure with five convolutional layers). The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the processor 110 may input the preprocessed classification target data to the convolutional block. The processor 110 may obtain a feature corresponding to the classification target data by inputting the preprocessed classification target data to the convolutional block. The convolutional block may include at least one or more convolutional layers. The processor 110 may acquire features by inputting the preprocessed classification target data to the convolutional layer. After the features obtained using the convolutional layer are subjected to a batch normalization process, the output of the convolutional block may be finally obtained using an activation function (eg, ReLU activation function). The processor 110 may extract only strong features by inputting the output obtained from the convolutional block to the pooling layer. The pooling layer may include a max pooling layer. The processor 110 may obtain a classification by inputting the output of the max pooling layer to the fully connected layer. The foregoing is merely an example, and the present disclosure is not limited thereto.

For a detailed description of the method for obtaining a classification using the second model described above, see the paper Very Deep Convolutional Neural Networks for Text Classification (published on January 27, 2017, author: Alexis), which is incorporated herein by reference in its entirety. Conneau, Holger Schwenk, Yann Le Cun, Loic Barrault).

According to an embodiment of the present disclosure, the processor 110 may obtain a classification for a document from the output of the fully connected layer. The processor 110 may obtain a classification labeled with a feature of the classification target data by using the fully connected layer. For example, when the document is a patent document and the patent document includes an electrical/electronic invention, the processor 110 may obtain an electrical/electronic classification using a fully connected layer. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the classification may be divided into a large classification, a medium classification, and a small classification based on the industrial technology classification. The major categories may include machinery, materials, electrical and electronic equipment, information and communication technology, chemistry, biotechnology, medical care, energy resources, knowledge services, ceramics, and the like. The middle classification of large categories of information and communication may include, for example, mobile communication, digital broadcasting, satellite radio, home network, broadband integrated network, software, U-computing, digital content, knowledge information security, information communication modules and parts, and RFID. there is. Categorized software may include, for example, embedded software, software solutions, system integration, and Internet software. The above classification is merely an example, and the present disclosure is not limited thereto.

According to another embodiment of the present disclosure, the classification may include an artificial intelligence classification. The AI classification may include, for example, an AI learning method, an AI research field, an AI domain, and an AI application field. The artificial intelligence learning method may include a case where there is no limitation in teacher learning, comparative teacher learning, reinforcement learning, and learning methods. The field of artificial intelligence research may include preprocessing, training methods, network structures, and inventions using artificial intelligence (eg, inventions that achieve special purposes using artificial intelligence). The artificial intelligence domain may include the type of data to be processed using artificial intelligence. For example, it may include images, natural language, time series data (eg, sensor data), and cases without domain limitation. Artificial intelligence application fields may include classifications based on tasks that artificial intelligence intends to perform using artificial intelligence technology in industries. For example, it may include medical care, security, management, gaming, autonomous driving, translation, smart factory, and general (eg, when the use of artificial intelligence is not disclosed). The above classification is merely an example, and the present disclosure is not limited thereto.

Recently, as the industry develops, technologies in various fields are developed, and the number of cases of patent applications for legal protection of new technologies is increasing. Among these numerous patents, it may consume a lot of money, manpower, and time to find a patent corresponding to a technology field of interest.

According to the present disclosure, by obtaining a classification of a document using a neural network, it is possible to obtain a document corresponding to a technical field of interest from among numerous documents at a low cost and time. Documents related to patent trend research, documents related to technology development direction research of companies engaged in the technology field of interest, documents containing prior art content similar to the research direction, etc. can be created within a short period of time, allowing individuals and/or companies to use a vast amount of data usefully in the direction they want to utilize.

2 is a schematic diagram illustrating a network function for data processing according to an embodiment of the present disclosure;

Throughout this specification, computational model, neural network, network function, and neural network may be used interchangeably. A neural network may be composed of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network is configured to include at least one or more nodes. Nodes (or neurons) constituting the neural networks may be interconnected by one or more links.

In the neural network, one or more nodes connected through a link may relatively form a relationship between an input node and an output node. The concepts of an input node and an output node are relative, and any node in an output node relationship with respect to one node may be in an input node relationship in a relationship with another node, and vice versa. As described above, an input node-to-output node relationship may be created around a link. One or more output nodes may be connected to one input node through a link, and vice versa.

In the relationship between the input node and the output node connected through one link, the value of the output node may be determined based on data input to the input node. Here, a node interconnecting the input node and the output node may have a weight. The weight may be variable, and may be changed by a user or an algorithm in order for the neural network to perform a desired function. For example, when one or more input nodes are interconnected to one output node by respective links, the output node sets values input to input nodes connected to the output node and links corresponding to the respective input nodes. An output node value may be determined based on the weight.

As described above, in a neural network, one or more nodes are interconnected through one or more links to form an input node and an output node relationship in the neural network. The characteristics of the neural network may be determined according to the number of nodes and links in the neural network, the correlation between the nodes and the links, and the value of a weight assigned to each of the links. For example, when the same number of nodes and links exist and there are two neural networks having different weight values between the links, the two neural networks may be recognized as different from each other.

A neural network may include one or more nodes. Some of the nodes constituting the neural network may configure one layer based on distances from the initial input node. For example, a set of nodes having a distance of n from the initial input node may constitute n layers. The distance from the initial input node may be defined by the minimum number of links that must be passed to reach the corresponding node from the initial input node. However, the definition of such a layer is arbitrary for description, and the order of the layer in the neural network may be defined in a different way from the above. For example, a layer of nodes may be defined by a distance from the final output node.

The initial input node may mean one or more nodes to which data is directly input without going through a link in a relationship with other nodes among nodes in the neural network. Alternatively, in a relationship between nodes based on a link in a neural network, it may mean nodes that do not have other input nodes connected by a link. Similarly, the final output node may refer to one or more nodes that do not have an output node in relation to other nodes among nodes in the neural network. In addition, the hidden node may mean nodes constituting the neural network other than the first input node and the last output node. The neural network according to an embodiment of the present disclosure may be a neural network in which the number of nodes in the input layer may be the same as the number of nodes in the output layer, and the number of nodes decreases and then increases again as progresses from the input layer to the hidden layer. can Also, in the neural network according to another embodiment of the present disclosure, the number of nodes in the input layer may be less than the number of nodes in the output layer, and the number of nodes may be reduced as the number of nodes progresses from the input layer to the hidden layer. there is. In addition, the neural network according to another embodiment of the present disclosure may be a neural network in which the number of nodes in the input layer may be greater than the number of nodes in the output layer, and the number of nodes increases as the number of nodes progresses from the input layer to the hidden layer. can The neural network according to another embodiment of the present disclosure may be a neural network in a combined form of the aforementioned neural networks.

A deep neural network (DNN) may refer to a neural network including a plurality of hidden layers in addition to an input layer and an output layer. Deep neural networks can be used to identify the latent structures of data. In other words, it can identify the potential structure of photos, texts, videos, voices, and music (e.g., what objects are in the photos, what the text and emotions are, what the texts and emotions are, etc.) . Deep neural networks include convolutional neural networks (CNNs), recurrent neural networks (RNNs), auto encoders, generative adversarial networks (GANs), and restricted Boltzmann machines (RBMs). boltzmann machine), a deep belief network (DBN), a Q network, a U network, a Siamese network, and the like. The description of the deep neural network described above is only an example, and the present disclosure is not limited thereto.

In an embodiment of the present disclosure, the network function may include an auto-encoder. The auto-encoder may be a kind of artificial neural network for outputting output data similar to input data. The auto encoder may include at least one hidden layer, and an odd number of hidden layers may be disposed between the input/output layers. The number of nodes in each layer may be reduced from the number of nodes of the input layer to an intermediate layer called the bottleneck layer (encoding), and then expanded symmetrically with reduction from the bottleneck layer to the output layer (symmetrically with the input layer). In this case, although the dimensionality reduction layer and the dimension reconstruction layer are illustrated as being symmetrical in the example of FIG. 2 , the present disclosure is not limited thereto, and the nodes of the dimension reduction layer and the dimension reconstruction layer may or may not be symmetrical. The auto-encoder can perform non-linear dimensionality reduction. The number of input layers and output layers may correspond to the number of sensors remaining after pre-processing of input data. In the auto-encoder structure, the number of nodes of the hidden layer included in the encoder may have a structure that decreases as the distance from the input layer increases. If the number of nodes in the bottleneck layer (the layer with the fewest nodes located between the encoder and decoder) is too small, a sufficient amount of information may not be conveyed, so a certain number or more (e.g., more than half of the input layer, etc.) ) may be maintained.

The neural network may be trained by at least one of supervised learning, unsupervised learning, and semi-supervised learning. The training of the neural network is to minimize the error in the output. In the training of a neural network, iteratively input the training data into the neural network, calculate the output of the neural network and the target error for the training data, and calculate the error of the neural network from the output layer of the neural network to the input layer in the direction to reduce the error. It is a process of updating the weight of each node in the neural network by backpropagation in the direction. In the case of teacher learning, learning data in which the correct answer is labeled in each learning data is used (ie, labeled learning data), and in the case of comparative learning, the correct answer may not be labeled in each learning data. That is, for example, learning data in the case of teacher learning related to data classification may be data in which categories are labeled in each of the learning data. The labeled training data is input to the neural network, and an error can be calculated by comparing the output (category) of the neural network with the label of the training data. As another example, in the case of comparison learning related to data classification, an error may be calculated by comparing the input training data with the neural network output. The calculated error is back propagated in the reverse direction (ie, from the output layer to the input layer) in the neural network, and the connection weight of each node of each layer of the neural network may be updated according to the back propagation. The change amount of the connection weight of each node to be updated may be determined according to a learning rate. The computation of the neural network on the input data and the backpropagation of errors can constitute a learning cycle (epoch). The learning rate may be applied differently according to the number of repetitions of the learning cycle of the neural network. For example, in the early stage of learning of a neural network, a high learning rate can be used to enable the neural network to quickly obtain a certain level of performance, thereby increasing efficiency, and using a low learning rate at a later stage of learning can increase accuracy.

In the training of neural networks, in general, the training data may be a subset of real data (that is, data to be processed using the trained neural network), and thus, the error on the training data is reduced, but the error on the real data is reduced. There may be increasing learning cycles. Overfitting is a phenomenon in which errors on actual data increase by over-learning on training data as described above. For example, a phenomenon in which a neural network that has learned a cat by seeing a yellow cat does not recognize that it is a cat when it sees a cat other than yellow may be a type of overfitting. Overfitting can act as a cause of increasing errors in machine learning algorithms. In order to prevent such overfitting, various optimization methods can be used. In order to prevent overfitting, methods such as increasing training data, regularization, or dropout in which a part of nodes in the network are omitted in the process of learning, may be applied.

A computer-readable medium storing a data structure is disclosed according to an embodiment of the present disclosure.

The data structure may refer to the organization, management, and storage of data that enables efficient access and modification of data. A data structure may refer to an organization of data to solve a specific problem (eg, data retrieval, data storage, and data modification in the shortest time). A data structure may be defined as a physical or logical relationship between data elements designed to support a particular data processing function. The logical relationship between data elements may include a connection relationship between data elements that a user thinks. Physical relationships between data elements may include actual relationships between data elements physically stored on a computer-readable storage medium (eg, a hard disk). A data structure may specifically include a set of data, relationships between data, and functions or instructions applicable to data. Through an effectively designed data structure, the computing device can perform an operation while using the resource of the computing device to a minimum. Specifically, the computing device may increase the efficiency of operations, reads, insertions, deletions, comparisons, exchanges, and retrievals through effectively designed data structures.

A data structure may be classified into a linear data structure and a non-linear data structure according to the type of the data structure. The linear data structure may be a structure in which only one piece of data is connected after one piece of data. The linear data structure may include a list, a stack, a queue, and a deck. A list may mean a set of data in which an order exists internally. The list may include a linked list. The linked list may be a data structure in which data is linked in such a way that each data is linked in a line with a pointer. In a linked list, a pointer may contain information about a link with the next or previous data. A linked list may be expressed as a single linked list, a doubly linked list, or a circularly linked list according to a shape. A stack can be a data enumeration structure with limited access to data. A stack can be a linear data structure in which data can be manipulated (eg, inserted or deleted) at only one end of the data structure. The data stored in the stack may be a data structure (LIFO-Last in First Out). A queue is a data listing structure that allows limited access to data. Unlike a stack, the queue may be a data structure that comes out later (FIFO-First in First Out) as data stored later. A deck can be a data structure that can process data at either end of the data structure.

The non-linear data structure may be a structure in which a plurality of data is connected after one data. The nonlinear data structure may include a graph data structure. A graph data structure may be defined as a vertex and an edge, and the edge may include a line connecting two different vertices. A graph data structure may include a tree data structure. The tree data structure may be a data structure in which one path connects two different vertices among a plurality of vertices included in the tree. That is, it may be a data structure that does not form a loop in the graph data structure.

Throughout this specification, computational model, neural network, network function, and neural network may be used interchangeably. (Hereinafter, the neural network is unified and described.) The data structure may include a neural network. And the data structure including the neural network may be stored in a computer-readable medium. Data structures, including neural networks, may also include data input to the neural network, weights of the neural network, hyperparameters of the neural network, data obtained from the neural network, activation functions associated with each node or layer of the neural network, and loss functions for learning the neural network. there is. A data structure comprising a neural network may include any of the components disclosed above. That is, the data structure including the neural network includes all or all of the data input to the neural network, the weights of the neural network, hyperparameters of the neural network, data obtained from the neural network, the activation function associated with each node or layer of the neural network, and the loss function for training the neural network. may be configured including any combination of In addition to the above-described configurations, a data structure including a neural network may include any other information that determines a characteristic of a neural network. In addition, the data structure may include all types of data used or generated in the operation process of the neural network, and is not limited to the above. Computer-readable media may include computer-readable recording media and/or computer-readable transmission media. A neural network may be composed of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network is configured by including at least one or more nodes.

The data structure may include data input to the neural network. A data structure including data input to the neural network may be stored in a computer-readable medium. The data input to the neural network may include learning data input in a neural network learning process and/or input data input to the neural network in which learning is completed. Data input to the neural network may include pre-processing data and/or pre-processing target data. The preprocessing may include a data processing process for inputting data into the neural network. Accordingly, the data structure may include data to be pre-processed and data generated by pre-processing. The above-described data structure is merely an example, and the present disclosure is not limited thereto.

The data structure may include the weights of the neural network. (In this specification, weight and parameter may be used interchangeably.) And the data structure including the weight of the neural network may be stored in a computer-readable medium. The neural network may include a plurality of weights. The weight may be variable, and may be changed by a user or an algorithm in order for the neural network to perform a desired function. For example, when one or more input nodes are interconnected to one output node by respective links, the output node sets values input to input nodes connected to the output node and links corresponding to the respective input nodes. An output node value may be determined based on the parameter. The above-described data structure is merely an example, and the present disclosure is not limited thereto.

By way of example and not limitation, the weight may include a weight variable in a neural network learning process and/or a weight in which neural network learning is completed. The variable weight in the neural network learning process may include a weight at a time point at which a learning cycle starts and/or a weight variable during the learning cycle. The weight for which neural network learning is completed may include a weight for which a learning cycle is completed. Accordingly, the data structure including the weights of the neural network may include a data structure including the weights that vary in the process of learning the neural network and/or the weights on which the learning of the neural network is completed. Therefore, it is assumed that the above-described weights and/or combinations of weights are included in the data structure including the weights of the neural network. The above-described data structure is merely an example, and the present disclosure is not limited thereto.

The data structure including the weights of the neural network may be stored in a computer-readable storage medium (eg, memory, hard disk) after being serialized. Serialization can be the process of converting a data structure into a form that can be reconstructed and used later by storing it on the same or a different computing device. The computing device may serialize the data structure to send and receive data over the network. A data structure including weights of the serialized neural network may be reconstructed in the same computing device or in another computing device through deserialization. The data structure including the weight of the neural network is not limited to serialization. Furthermore, the data structure including the weights of the neural network is a data structure to increase computational efficiency while using the resources of the computing device to a minimum (e.g., B-Tree, Trie, m-way search tree, AVL tree, Red-Black Tree). The foregoing is merely an example, and the present disclosure is not limited thereto.

The data structure may include hyper-parameters of the neural network. In addition, the data structure including the hyperparameters of the neural network may be stored in a computer-readable medium. The hyper parameter may be a variable variable by a user. Hyperparameters are, for example, learning rate, cost function, number of iterations of the learning cycle, weight initialization (e.g., setting the range of weight values to be initialized for weights), Hidden Unit The number (eg, the number of hidden layers, the number of nodes of the hidden layer) may be included. The above-described data structure is merely an example, and the present disclosure is not limited thereto.

3 is an exemplary diagram for explaining an operation of processing data using a neural network according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the computing device 100 may input the document 310 into the first model 330 . According to an embodiment of the present disclosure, the document 310 may include at least one of a first partial document for claiming the right and a second partial document for describing the content of the claim. The first part of the document for claiming rights may include at least a part of the document describing the part to be legally protected according to the laws and regulations of each country. The second part of the document to explain the content of the claim may include at least a part of the document to specifically explain the part to be protected legally according to the laws of each country.

According to an embodiment of the present disclosure, the computing device 100 may acquire the classification target data 350 using the first model 330 . The first model 330 includes a first submodel for extracting features of a first partial document, a second submodel for extracting features of a second partial document, and outputs of the first submodel and the second submodel. Based on the third submodel for deriving the correlation between the first partial document and the second partial document, and a fourth submodel for extracting the second element having the highest correlation with the first element based on the correlation, may include The above-described first model is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may extract a feature of the first partial document by inputting the first partial document into the first submodel. The features of the first partial document may include a first submodel encoding matrix. The first submodel encoding matrix may include a vector obtained by inputting a document into a long-term storage network. The obtained vector may include a vector having a probability value of each word as a component of the vector. The computing device 100 may obtain a first submodel encoding matrix by using a vector obtained from the first submodel. For example, the set of words included in the first partial document is (x1, x2, x3, x4, ... , xn) and a vector corresponding to each word (acquired by the processor entering the word into the first submodel) When a vector) is (v1, v2, v3, v4, ... ,vn), the first submodel encoding matrix may be expressed as [v1, v2, v3, v4, ... ,vn]. The above-described first sub-model is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may extract a feature of the second partial document by inputting the second partial document into the second submodel. The features of the second partial document may include a second submodel encoding matrix. The second submodel encoding matrix may include a vector obtained by inputting the document into a long-term storage network. The obtained vector may include a vector having a probability value of each word as an element of the vector. The computing device 100 may obtain a second submodel encoding matrix by using a vector obtained from the second submodel. For example, the set of words included in the second partial document is (y1, y2, y3, y4, ... , yn) and a vector corresponding to each word (acquired by the processor entering the word into the second submodel) one vector) is (v'1, v'2, v'3, v'4, ... ,v'n), the second submodel encoding matrix is [v'1, v'2, v' 3, v'4, ... ,v'n]. The above-described first sub-model is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 calculates the correlation between the first partial document and the second partial document based on the outputs of the first submodel and the second submodel using the third submodel. can be derived The computing device 100 may derive a correlation between the first partial document and the second partial document based on the first submodel encoding matrix and the second submodel encoding matrix. The computing device 100 may calculate a transpose matrix of the second submodel encoding matrix and the first submodel encoding matrix. The computing device 100 inputs the result of calculating the transpose matrix of the second submodel encoding matrix and the first submodel encoding matrix into the softmax function, and includes the word included in the first partial document and the second partial document It is possible to obtain the correlation of the words. The above-described third submodel is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may extract a second element having the highest correlation with the first element by using the fourth submodel. The computing device 100 may obtain a correlation between the first element and at least one or more second elements by using the third submodel. The computing device 100 may select one second element having the highest correlation with the first element and add it to the classification target data. The computing device 100 may select the top 3% second elements having the highest correlation with the first element and add them to the classification target data. The computing device 100 may obtain location information of the selected second element by using the fourth submodel. When the second element is in the tenth paragraph of the document, the computing device 100 may obtain location information of the tenth paragraph. The computing device 100 may obtain a start position and an end position of the second element. For example, when the start position of the second element is number 10 and the end position is number 20, (start position=10, end position=20) information may be acquired. The computing device 100 may efficiently search for the second element in the document using the location information and add it to the classification target data. The above-described fourth submodel is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may pre-process the classification target data 350 ( 360 ). The preprocessing 360 of the classification target data 350 may include a process for inputting the data into the second model 370 . The pre-processing 360 of the classification target data 350 may include a process of converting the classification target data 350 into image data. The data in the form of an image may include a form of data expressed in the form of a table. In a table format, a row may include words included in text, and a column may include a vector component corresponding to each word. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may input the preprocessed classification target data 350 to the second model 370 . The second model 370 may include at least one of a plurality of convolutional blocks, a plurality of pooling layers, and a plurality of fully connected layers. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may input preprocessed classification target data into the convolutional block. The computing device 100 may obtain a feature corresponding to the classification target data by inputting the preprocessed classification target data to the convolutional block. The convolutional block may include at least one or more convolutional layers. The computing device 100 may acquire features by inputting the preprocessed classification target data to the convolutional layer. After the features obtained using the convolutional layer are subjected to a batch normalization process, the output of the convolutional block may be finally obtained using an activation function (eg, ReLU activation function). The computing device 100 may extract only strong features by inputting the output obtained from the convolutional block to the pooling layer. The pooling layer may include a max pooling layer. The computing device 100 may obtain a classification by inputting the output of the max pooling layer to the fully connected layer. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may obtain a classification 390 . For example, when the document is a patent document and the content described in the patent document is a semiconductor, the computing device 100 may output a classification of “electrical and electronic”. When the document is a patent document and the content described in the patent document is an internal combustion engine, the computing device 100 may output a classification of “machine”. When the document is a patent document and the content described in the patent document is a fertilizer manufacturing method, the computing device 100 may output a classification of “chemistry”. When the document is a patent document and the content described in the patent document is a method of generating solar energy, the computing device 100 may output a classification of “energy resource”. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may obtain a classification 390 . For example, when the document is a patent document, the first element contains the word "computer program" or "computer program", and the content described in the first element is a "software" related invention, the computing device 100 may " software" can be obtained. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, when the document is a patent document and the prior art related to software (eg, software technology development history, past software technology) is described in the technology related to the technical field and the background of the invention, the computing device (100) may obtain a classification of "software". In order to improve classification accuracy, the computing device 100 may finally perform a process of verifying whether the classification of “software” is accurate in consideration of the contents described in the first partial document. Furthermore, in order to improve classification accuracy, the computing device 100 may additionally perform a process of finally verifying whether the classification of “software” is accurate in consideration of the contents described in the second element having a high correlation with the first element. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to another embodiment of the present disclosure, the computing device 100 may obtain an artificial intelligence classification. The AI classification may include, for example, an AI learning method, an AI research field, an AI domain, and an AI application field. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may obtain an AI learning method classification. When the document is a patent document and the first element includes “learning content using data including labels”, the computing device 100 may obtain an AI learning method classification of “teacher learning”. If the computing device 100 does not find the content for classifying the AI learning method in the first element (eg, when there is no basis for determining teacher learning, comparative tutor learning, and reinforcement learning in the first element), In the second element having a high correlation with the first element, content for classifying the AI learning method may be searched. Therefore, when "learning content using data including labels" is included in the second element having a high correlation with the first element, the computing device 100 may obtain an AI learning method classification called "teacher learning". . The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may output an artificial intelligence research field. When the document is a patent document and the first element includes "specific method content related to network learning" (for example, when a process of updating parameters of a network function in the learning process is specifically described), the computing device ( 100) may acquire a classification of "training method". The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may output an artificial intelligence domain. When the document is a patent document and the first element includes "input data is an image", the computing device 100 may obtain a classification that the artificial intelligence domain is "image". The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may output an artificial intelligence application field. When the document is a patent document and the first element includes “content related to autonomous vehicle driving, drone autonomous flight, and unmanned moving object”, the computing device 100 may acquire a classification of “autonomous driving”. The foregoing is merely an example, and the present disclosure is not limited thereto.

When the document is a patent document, a process of obtaining a classification of the patent document may consume a lot of manpower, time, and cost. In order to obtain a classification of patent documents, it may be necessary to read patent documents containing vast amounts of text and images. However, when the classification of patent documents is obtained using artificial intelligence according to the present disclosure, time and cost are reduced, and manpower is also reduced, so that the work can be performed within a short time. By obtaining the classification of patent documents in a short time, it is possible to obtain reliable data analysis data by creating patent maps, patent trend research reports, and technological development flow charts based on more patent documents. In addition, by using artificial intelligence, classification is performed according to a uniform standard, so that the accuracy of classification and the reliability of the classification result can be improved. As a result, by classifying many documents with high accuracy in a short time, it is possible to efficiently use a large amount of documents to obtain a desired data analysis result.

4 is an exemplary diagram for describing a document according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the document 310 may include at least one of the first partial document 410 and the second partial document 430 . If the document 310 is a patent document, the first partial document 410 may include claims. A claim may include at least one or more claims. Claim 1 (411) may be an independent claim since it does not refer to another claim. Claim 2(413) cites Claim 1 and thus may be a dependent Claim. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, when the document 310 is a patent document, the second partial document 430 is the title of the invention, the technical field 431 , the background technology 433 of the invention, and the content of the invention. , a brief description of the drawings (435), specific content for carrying out the invention (437), a description of symbols, drawings, summary, and representative views may be included. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the computing device 100 may analyze a document and divide it into the first partial document and the second partial document. The computing device 100 may divide the first partial document and the second partial document based on the sub-title of the document. When the document sub-title is a claim, the computing device may recognize claims that are the sub-title and determine claims 1 (411) and 2 (413) included in the claims as content included in the first partial document. When the document sub-title is the technical field 431, the technology behind the invention 433, the problem to be solved, a brief description of the drawing 435, and the specific content 437 for carrying out the invention, the computing device 100 is the sub-title The content included in each of the technical field 431, the background technology 433 of the invention, the problem to be solved, the brief description of the drawings 435, and the specific content 437 for carrying out the invention is included in the second part document content can be determined. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to the present disclosure, by dividing the first part document 410 for claiming the right and the second part document 430 for explaining the content of the claim, it is possible to determine which part of the content included in the document to be given importance. there is. When the document is a patent document, the first part document contains content for claiming rights, so it may be a key part (eg, a part where an invention is differentiated from prior art) in determining the scope of rights. Accordingly, by dividing the patent document into a first partial document and a second partial document, core content may be included in the classification target data while the computing device 100 generates the classification target data using the first model. By including the core content in the classification target data, the computing device 100 may obtain a classification result with higher accuracy by using the second model.

According to an embodiment of the present disclosure, there is a description in the technical field 431 that "this disclosure relates to data processing using a computing device, and more specifically, to a technology for processing data using a neural network" If there is, the computing device 100 may obtain a classification of "invention applied with artificial intelligence" in the field of artificial intelligence research by using the first model and the second model. An invention to which artificial intelligence is applied may include, for example, an invention that achieves a special purpose using artificial intelligence. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, in the technology 433 that is the background of the invention, "In artificial intelligence technology, especially in the field of deep learning, remarkable progress has been made in data automation processing. Recently, as deep learning technology develops, the classification of data is also When there is a description of "the trend of using deep learning", the computing device 100 may obtain a classification of "artificial intelligence" for the document 310 using the first model and the second model. The computing device 100 may obtain a detailed artificial intelligence classification based on the contents described in the first partial document 310 or the second partial document 310 to obtain a detailed artificial intelligence classification. The foregoing is merely an example, and the present disclosure is not limited thereto.

5 is an exemplary diagram for describing classification target data according to an embodiment of the present disclosure.

5 exemplarily shows detailed items of the classification target data 350 .

According to an embodiment of the present disclosure, the classification target data may include a portion selected in advance according to a predetermined criterion of the second partial document 430 . The pre-selected portion may include the technology 550 that is the background of the technical field and the invention. The technical field may include a clear and concise description of the technical field of the invention for which a patent is sought. The technology that is the background of the invention may include a description of the prior art that is thought to be useful for understanding and research of the invention. Therefore, regardless of the correlation with the first element, the technology 550 that is the background of the technical field and invention may help to obtain a classification with less information. Accordingly, it may be helpful for the computing device 100 to perform classification with high accuracy using the second model. In addition, when the operator labels the classification target data to generate the training data set of the second model, the classification target data can be labeled within a short time by using the technology 550 that is the background of the technical field and invention. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, reference numeral 510 may include a first element 1 511 and a second element 1 513 having a correlation with the first element 1 511 at a certain level or higher. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the classification target data may include the first element 1 511 and the second element 1 513 having a correlation with the first element 1 511 or higher. The first element 1 (511) may include claim 1 (411). Claim 1 (411) includes "the operation of inputting a document into a first model for natural language processing; the operation of obtaining classification target data from an output of the first model; inputting the classification target data into a second model for classification operation; and an operation of obtaining a classification for the document from the output of the second model; ". Specific content 515 for carrying out the invention may include a description, "According to an embodiment of the present disclosure, the processor 110 may acquire classification target data using the first model." The computing device 100 may obtain a correlation between the first element 1 511 and the content included in the detailed content 515 for carrying out the invention by using the first model. According to another embodiment of the present disclosure, the second element 1 513 may include a drawing 517 . Since the computing device 100 discloses the content included in claim 1 (411) that is included in the specific content 515 for carrying out the invention, the content included in the specific content 515 for carrying out the invention is It can be determined that the content included in claim 1 (411) has a high correlation. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to another embodiment of the present disclosure, a brief description 435 of the drawing is an exemplary diagram for explaining an operation of processing data using a neural network according to an embodiment of the present disclosure in relation to FIG. 3 . It may include the description "is. The computing device 100 calculates the correlation between the contents described in the brief description 435 of the drawing and the first element 1 511 by using the first model, thereby calculating the correlation between the drawing and the first element 1 511 . can be obtained Accordingly, when the correlation between the content described in the brief description 435 of the drawing and the first element 1 511 is high, the computing device 100 indicates that the claim 1 411 and FIG. 3 included in the drawing 517 have a high correlation. relationship can be considered. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, reference numeral 530 may include a first element 2 531 and a second element 2 533 having a correlation with the first element 2 531 or higher. The foregoing is merely an example, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the classification target data may include the first element 2 531 and the second element 2 533 having a correlation with the first element 2 531 or higher. The first element 2 (531) may include claim 2 (413). 2(413) is "The computer of claim 1, wherein the document is stored in a computer-readable storage medium comprising at least one of a first part document for claiming a right or a second part document for describing content of a claim. program." The second element 2 533 may include specific content 151 for implementing the invention. Specific content 535 for practicing the invention is "According to an embodiment of the present disclosure, a document may include at least one of a first partial document for claiming a right or a second partial document for explaining the content of the claim. The first part for claiming the right The document may include at least a part of the document that describes the part to be legally protected according to the laws of each country. For example, if the right is a patent right, the document for claiming the right A one-part document may include items described in the claims." Since the second element 2 533 includes the sentence "a document is a first part document for claiming rights or a second part document for explaining content of claims" described in the first element 2 531 as it is, the computing device (100) may determine that the second element 2 533 and the first element 2 531 have a high correlation using the first model 330 . The foregoing is merely an example, and the present disclosure is not limited thereto.

6 is a flowchart of a method of processing data using a neural network according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the computing device 100 may input 610 a document into a first model for natural language processing.

In an embodiment of the present disclosure, the document may include at least one of a first partial document for claiming the right and a second partial document for explaining the content of the claim.

According to an embodiment of the present disclosure, the computing device 100 may obtain ( 620 ) classification target data from the output of the first model.

In an embodiment of the present disclosure, the classification target data may include at least one of a first element including at least a part of the first partial document and a second element including at least a part of the second partial document. there is.

In an embodiment of the present disclosure, the second element may include at least one of a text or an image having a correlation with the first element at a certain level or higher.

In an embodiment of the present disclosure, the classification target data may include a portion selected in advance according to a predetermined criterion of the second partial document.

In an embodiment of the present disclosure, the first model may include a recurrent neural network for natural language processing.

In an embodiment of the present disclosure, the first model includes: a first submodel for extracting features of a first partial document; a second submodel for extracting features of the second partial document; a third submodel for deriving a correlation between a first partial document and a second partial document based on outputs of the first submodel and the second submodel; and a fourth submodel for extracting a second element having the highest correlation with the first element based on the correlation; may include

The computing device 100 may input 630 the classification target data to a second model for classification according to an embodiment of the present disclosure.

In an embodiment of the present disclosure, the second model may include a model trained using a training data set using the classification target data as learning input data and a classification corresponding to the classification target data as a label. there is.

In an embodiment of the present disclosure, the second model includes at least one of a plurality of convolutional blocks, a plurality of pooling layers, or a plurality of fully connected layers. can

In an embodiment of the present disclosure, the convolutional block includes a plurality of convolutional layers, the plurality of convolutional layers have the same size, and the size is to form a deep structure of the convolutional layer size may be included.

According to an embodiment of the present disclosure, the computing device 100 may obtain ( 640 ) a classification for the document from the output of the second model.

According to the present disclosure, by obtaining the classification of a document using artificial intelligence, the computing device 100 obtains the classification according to the uniform standard of the artificial intelligence, so that the reliability and consistency of the classification result may be high. (When multiple operators perform classification, there may be cases in which it is difficult to classify clearly, even when classifying using the same criteria. Therefore, when it is difficult to classify clearly, several people may perform classification according to different criteria. Therefore, classification consistency may be lacking.) In addition, excessive manpower waste can be reduced by using artificial intelligence, and the effect of saving money and time in terms of cost and time can be expected.

7 is a block diagram illustrating a module for processing data using a neural network according to an embodiment of the present disclosure.

A data processing method using a neural network in an embodiment of the present disclosure may be implemented by the following modules. a module 710 for inputting a document into a first model for natural language processing; a module 720 for obtaining classification target data from the output of the first model; a module 730 for inputting the classification target data into a second model for classification; and a module 740 for obtaining a classification for the document from the output of the second model.

In an alternative embodiment for processing data using a neural network, it may include a module for analyzing the document and dividing it into the first partial document and the second partial document.

In an alternative embodiment for processing data using a neural network, a module for performing pre-processing on the classification target data may be included.

In an alternative embodiment for processing data using a neural network, the pre-processing of the classification target data may include processing of converting the classification target data into image data.

In an alternative embodiment for processing data using a neural network, the module 740 for obtaining a classification for the document from the output of the second model is a module for inputting the output of the convolutional block to a pooling layer ; a module for inputting an output of the pooling layer to a fully connected layer; and a module for obtaining a classification for a document from the output of the fully connected layer.

According to an embodiment of the present disclosure, a module for processing data using a neural network may be implemented by means, circuits or logic for implementing a computing device. Those skilled in the art will further appreciate that the various illustrative logical blocks, configurations, modules, circuits, means, logics, and algorithm steps described in connection with the embodiments disclosed herein may be combined with electronic hardware, computer software, or combinations of both. It should be recognized that it can be implemented as To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, configurations, means, logics, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

8 is a simplified, general schematic diagram of an exemplary computing environment in which embodiments of the present disclosure may be implemented.

Although the present disclosure has been described above as being generally capable of being implemented by computing device 100 , those skilled in the art will appreciate that this disclosure may be implemented in hardware and/or in combination with computer-executable instructions and/or other program modules that may be executed on one or more computers. It will be well understood that it can be implemented as a combination of and software.

Generally, program modules include routines, programs, components, data structures, etc. that perform particular tasks or implement particular abstract data types. In addition, those skilled in the art will appreciate that the methods of the present disclosure can be applied to single-processor or multiprocessor computer systems, minicomputers, mainframe computers as well as personal computers, handheld computing devices, microprocessor-based or programmable consumer electronics, etc. (each of which is It will be appreciated that other computer system configurations may be implemented, including those that may operate in connection with one or more associated devices.

The described embodiments of the present disclosure may also be practiced in distributed computing environments where certain tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote memory storage devices.

Computers typically include a variety of computer-readable media. Any medium accessible by a computer can be a computer-readable medium, and such computer-readable media includes volatile and nonvolatile media, transitory and non-transitory media, removable and non-transitory media. including removable media. By way of example, and not limitation, computer-readable media may include computer-readable storage media and computer-readable transmission media. Computer readable storage media includes volatile and nonvolatile media, temporary and non-transitory media, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. includes media. A computer-readable storage medium may be RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital video disk (DVD) or other optical disk storage device, magnetic cassette, magnetic tape, magnetic disk storage device, or other magnetic storage device. device, or any other medium that can be accessed by a computer and used to store the desired information.

Computer readable transmission media typically embodies computer readable instructions, data structures, program modules or other data, etc. in a modulated data signal such as a carrier wave or other transport mechanism, and Includes all information delivery media. The term modulated data signal means a signal in which one or more of the characteristics of the signal is set or changed so as to encode information in the signal. By way of example, and not limitation, computer-readable transmission media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared, and other wireless media. Combinations of any of the above are also intended to be included within the scope of computer-readable transmission media.

An example environment 1100 implementing various aspects of the disclosure is shown including a computer 1102 , the computer 1102 including a processing unit 1104 , a system memory 1106 , and a system bus 1108 . do. A system bus 1108 couples system components, including but not limited to system memory 1106 , to the processing device 1104 . The processing device 1104 may be any of a variety of commercially available processors. Dual processor and other multiprocessor architectures may also be used as processing unit 1104 .

The system bus 1108 may be any of several types of bus structures that may further interconnect a memory bus, a peripheral bus, and a local bus using any of a variety of commercial bus architectures. System memory 1106 includes read only memory (ROM) 1110 and random access memory (RAM) 1112 . A basic input/output system (BIOS) is stored in non-volatile memory 1110, such as ROM, EPROM, EEPROM, etc., the BIOS is the basic input/output system (BIOS) that helps transfer information between components within computer 1102, such as during startup. contains routines. RAM 1112 may also include high-speed RAM, such as static RAM, for caching data.

The computer 1102 may also include an internal hard disk drive (HDD) 1114 (eg, EIDE, SATA) - this internal hard disk drive 1114 may also be configured for external use within a suitable chassis (not shown). Yes—a magnetic floppy disk drive (FDD) 1116 (eg, for reading from or writing to removable diskette 1118), and an optical disk drive 1120 (eg, a CD-ROM) for reading from, or writing to, disk 1122, or other high capacity optical media, such as DVD. The hard disk drive 1114 , the magnetic disk drive 1116 , and the optical disk drive 1120 are connected to the system bus 1108 by the hard disk drive interface 1124 , the magnetic disk drive interface 1126 , and the optical drive interface 1128 , respectively. ) can be connected to The interface 1124 for implementing an external drive includes at least one or both of Universal Serial Bus (USB) and IEEE 1394 interface technologies.

These drives and their associated computer-readable media provide non-volatile storage of data, data structures, computer-executable instructions, and the like. In the case of computer 1102, drives and media correspond to storing any data in a suitable digital format. Although the description of computer readable media above refers to HDDs, removable magnetic disks, and removable optical media such as CDs or DVDs, those skilled in the art will use zip drives, magnetic cassettes, flash memory cards, cartridges, etc. It will be appreciated that other tangible computer-readable media such as etc. may also be used in the exemplary operating environment and any such media may include computer-executable instructions for performing the methods of the present disclosure.

A number of program modules may be stored in the drive and RAM 1112 , including an operating system 1130 , one or more application programs 1132 , other program modules 1134 , and program data 1136 . All or portions of the operating system, applications, modules, and/or data may also be cached in RAM 1112 . It will be appreciated that the present disclosure may be implemented in various commercially available operating systems or combinations of operating systems.

A user may enter commands and information into the computer 1102 via one or more wired/wireless input devices, for example, a pointing device such as a keyboard 1138 and a mouse 1140 . Other input devices (not shown) may include a microphone, IR remote control, joystick, game pad, stylus pen, touch screen, and the like. Although these and other input devices are connected to the processing unit 1104 through an input device interface 1142 that is often connected to the system bus 1108, parallel ports, IEEE 1394 serial ports, game ports, USB ports, IR interfaces, It may be connected by other interfaces, etc.

A monitor 1144 or other type of display device is also coupled to the system bus 1108 via an interface, such as a video adapter 1146 . In addition to the monitor 1144, the computer typically includes other peripheral output devices (not shown), such as speakers, printers, and the like.

Computer 1102 may operate in a networked environment using logical connections to one or more remote computers, such as remote computer(s) 1148 via wired and/or wireless communications. Remote computer(s) 1148 may be workstations, computing device computers, routers, personal computers, portable computers, microprocessor-based entertainment devices, peer devices, or other common network nodes, and are typically connected to computer 1102 . Although it includes many or all of the components described for it, only memory storage device 1150 is shown for simplicity. The logical connections shown include wired/wireless connections to a local area network (LAN) 1152 and/or a larger network, eg, a wide area network (WAN) 1154 . Such LAN and WAN networking environments are common in offices and companies, and facilitate enterprise-wide computer networks, such as intranets, all of which can be connected to a worldwide computer network, for example, the Internet.

When used in a LAN networking environment, the computer 1102 is coupled to the local network 1152 through a wired and/or wireless communication network interface or adapter 1156 . Adapter 1156 may facilitate wired or wireless communication to LAN 1152 , which LAN 1152 also includes a wireless access point installed therein for communicating with wireless adapter 1156 . When used in a WAN networking environment, the computer 1102 may include a modem 1158, be connected to a communication computing device on the WAN 1154, or establish communications over the WAN 1154, such as over the Internet. have other means. A modem 1158 , which may be internal or external and a wired or wireless device, is coupled to the system bus 1108 via a serial port interface 1142 . In a networked environment, program modules described for computer 1102 , or portions thereof, may be stored in remote memory/storage device 1150 . It will be appreciated that the network connections shown are exemplary and other means of establishing a communication link between the computers may be used.

The computer 1102 may be associated with any wireless device or object that is deployed and operates in wireless communication, for example, a printer, scanner, desktop and/or portable computer, portable data assistant (PDA), communication satellite, wireless detectable tag. It operates to communicate with any device or place, and phone. This includes at least Wi-Fi and Bluetooth wireless technologies. Accordingly, the communication may be a predefined structure as in a conventional network or may simply be an ad hoc communication between at least two devices.

Wi-Fi (Wireless Fidelity) makes it possible to connect to the Internet, etc. without a wired connection. Wi-Fi is a wireless technology such as cell phones that allows these devices, eg, computers, to transmit and receive data indoors and outdoors, ie anywhere within range of a base station. Wi-Fi networks use a radio technology called IEEE 802.11 (a, b, g, etc.) to provide secure, reliable, and high-speed wireless connections. Wi-Fi can be used to connect computers to each other, to the Internet, and to wired networks (using IEEE 802.3 or Ethernet). Wi-Fi networks may operate in unlicensed 2.4 and 5 GHz radio bands, for example at 11 Mbps (802.11a) or 54 Mbps (802.11b) data rates, or in products that include both bands (dual band). .

One of ordinary skill in the art of this disclosure will understand that information and signals may be represented using any of a variety of different technologies and techniques. For example, data, instructions, instructions, information, signals, bits, symbols, and chips that may be referenced in the above description are voltages, currents, electromagnetic waves, magnetic fields or particles, optical field particles or particles, or any combination thereof.

Those of ordinary skill in the art of the present disclosure will recognize that the various illustrative logical blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein include electronic hardware, (convenience For this purpose, it will be understood that it may be implemented by various forms of program or design code (referred to herein as software) or a combination of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. A person skilled in the art of the present disclosure may implement the described functionality in various ways for each specific application, but such implementation decisions should not be interpreted as a departure from the scope of the present disclosure.

The various embodiments presented herein may be implemented as methods, apparatus, or articles of manufacture using standard programming and/or engineering techniques. The term article of manufacture includes a computer program, carrier, or media accessible from any computer-readable storage device. For example, computer-readable storage media include magnetic storage devices (eg, hard disks, floppy disks, magnetic strips, etc.), optical disks (eg, CDs, DVDs, etc.), smart cards, and flash drives. memory devices (eg, EEPROMs, cards, sticks, key drives, etc.). Also, various storage media presented herein include one or more devices and/or other machine-readable media for storing information.

It is to be understood that the specific order or hierarchy of steps in the presented processes is an example of exemplary approaches. Based on design priorities, it is to be understood that the specific order or hierarchy of steps in the processes may be rearranged within the scope of the present disclosure. The appended method claims present elements of the various steps in a sample order, but are not meant to be limited to the specific order or hierarchy presented.

The description of the presented embodiments is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Thus, the present disclosure is not intended to be limited to the embodiments presented herein, but is to be construed in the widest scope consistent with the principles and novel features presented herein.

Claims (17)

A computer program stored in a computer-readable storage medium, when the computer program is executed on one or more processors to perform the following operations for extracting a technical field using a neural network, the operation comprising:
inputting the document into a first model for natural language processing;
obtaining classification target data from the output of the first model;
inputting the classification target data into a second model for classification; and
obtaining a classification for the document from the output of the second model;
containing,
A computer program stored on a computer-readable storage medium.
The method of claim 1,
wherein the document includes at least one of a first partial document including key information of the document and a second partial document having a substantial correspondence with the first partial document;
A computer program stored on a computer-readable storage medium.
The method of claim 1,
The document includes at least one of a first part document for claiming the right or a second part document for explaining the content of the claim,
A computer program stored on a computer-readable storage medium.
4. The method of claim 3,
Prior to inputting the document into the first model for natural language processing,
dividing the document into the first partial document and the second partial document by analyzing the document;
containing,
A computer program stored on a computer-readable storage medium.
4. The method of claim 3,
The classification target data is
at least one of a first element including at least a portion of the first partial document or a second element including at least a portion of the second partial document;
A computer program stored on a computer-readable storage medium.
6. The method of claim 5,
The second element is
Containing at least one of text or an image having a correlation with the first element at a certain level or higher,
A computer program stored on a computer-readable storage medium.
The method of claim 1,
The classification target data is
a second part comprising a predetermined part according to a predetermined criterion of the document;
A computer program stored on a computer-readable storage medium.
The method of claim 1,
The first model is
Including a recurrent neural network for natural language processing,
A computer program stored on a computer-readable storage medium.
The method of claim 1,
The first model is
a first submodel for extracting features of the first partial document;
a second submodel for extracting features of the second partial document;
a third submodel for deriving a correlation between a first partial document and a second partial document based on outputs of the first submodel and the second submodel; and
a fourth submodel for extracting a second element having the highest correlation with the first element based on the correlation;
containing,
A computer program stored on a computer-readable storage medium.
The method of claim 1,
Before the operation of inputting the classification target data into the second model for classification,
performing pre-processing on the classification target data;
containing,
A computer program stored on a computer-readable storage medium.
11. The method of claim 10,
The pre-processing of the classification target data is
Including processing of converting the classification target data into image data,
A computer program stored on a computer-readable storage medium.
The method of claim 1,
The second model is
Using the classification target data as learning input data, and including a model trained using a training data set using a classification corresponding to the classification target data as a label,
A computer program stored on a computer-readable storage medium.
The method of claim 1,
The second model is
Containing at least one of a plurality of convolutional blocks, a plurality of pooling layers, or a plurality of fully connected layers,
A computer program stored on a computer-readable storage medium.
The method of claim 1,
The operation of obtaining the classification for the document from the output of the second model comprises:
inputting the output of the convolutional block to the pooling layer;
inputting an output of the pooling layer to a fully connected layer; and
obtaining a classification for a document from the output of the fully connected layer;
containing,
A computer program stored on a computer-readable storage medium.
15. The method of claim 14,
The convolutional block is
including a plurality of convolutional layers,
The plurality of convolutional layers have the same size,
The size is a size for forming the deep structure of the convolutional layer,
A computer program stored on a computer-readable storage medium.
A method of extracting a technical field using a neural network is,
inputting the document into a first model for natural language processing;
obtaining classification target data from the output of the first model;
inputting the classification target data into a second model for classification; and
obtaining a classification for the document from the output of the second model;
containing,
A method of extracting technical fields using neural networks.
A computing device for extracting technical fields using a neural network, comprising:
one or more processors; and
a memory storing instructions executable by the one or more processors;
including,
The one or more processors,
input the document into a first model for natural language processing;
obtaining classification target data from the output of the first model;
inputting the data to be classified into a second model for classification, and
obtaining a classification for the document from the output of the second model;
A computing device for extracting technical fields using neural networks.

Images