A method for automatically annotating unstructured data based on neural networks

Abstract

The present invention discloses a method for automatically annotating unstructured data based on a neural network, which relates to the technical field of unstructured data, including: establishing a preprocessing annotation method based on NLP technology and CV technology; obtaining an annotation processing interval using the preprocessing annotation method; establishing a neural network and establishing an in-depth annotation method; training the neural network to obtain an optimal neural network; and automatically annotating unannotated data based on the optimal neural network. The present invention is used to solve the problem that the existing method for annotating unstructured data cannot solve the problem that when there are many formats and data sources in the unstructured data, each data source needs to be analyzed independently, which makes it difficult to establish a unified standard across data types, thereby reducing the annotation efficiency and the user information receiving efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of unstructured data, and in particular to a method for automatically annotating unstructured data based on a neural network. Background Art

[0002] Unstructured data is data with an irregular or incomplete structure, lacking a predefined data model, and difficult to represent using a two-dimensional database logical table. This includes office documents of all formats, text, images, HTML, various reports, graphics, audio, and video information. Unstructured data has a wide variety of formats and standards, and is technically more difficult to standardize and understand than structured information. Therefore, its storage, retrieval, publication, and utilization require more intelligent IT technologies. Annotation of unstructured data is primarily achieved through methods such as data cleaning, data standardization, and image processing.

[0003] The existing method for annotating unstructured data is usually to summarize the unstructured data through analysis tools after receiving the unstructured data, and display the summary information to the user, and the user modifies the summary information and interacts with the unstructured data analysis system through the annotation layer to achieve annotation of the unstructured data. Although this improved method can help users fully understand the unstructured data, when there are many formats of data in the unstructured data, summarizing the unstructured data only through analysis tools will make it difficult to establish a unified standard for cross-type data due to the large number of data types. As a result, when there are many data sources, each data source needs to be analyzed independently, which leads to reduced annotation efficiency and reduced user information reception efficiency. For example, in the publication number CN1073685 The patent application of 06A discloses an unstructured data analysis system and method. This solution is to apply one or more analysis tools to unstructured data and display summary information to one or more users, allowing one or more users to modify the granularity of the summary information and interact with the unstructured data analysis system through the annotation layer at the same time. Other improvements for annotating unstructured data are usually based on improvements in data recognition efficiency of transmission, which still cannot solve the problem that when there are many data formats and data sources in unstructured data, each data source needs to be analyzed independently, resulting in difficulty in establishing a unified standard across types of data, thereby reducing annotation efficiency and user information reception efficiency. In view of this, it is necessary to improve the existing method of annotating unstructured data. Summary of the Invention

[0004] The present invention aims to solve, at least to a certain extent, one of the technical problems in the prior art. By proposing a method for automatically annotating unstructured data based on a neural network, the present invention is used to solve the problem that the existing methods for annotating unstructured data cannot solve the problem that when there are many formats of data in the unstructured data and many data sources, each data source needs to be analyzed independently, which makes it difficult to establish a unified standard across types of data, thereby reducing the annotation efficiency and the efficiency of user information reception.

[0005] To achieve the above objectives, the present application provides a method for automatically annotating unstructured data based on a neural network, comprising the following steps:

[0006] Obtain unannotated unstructured data, recorded as unannotated data; establish a preprocessing annotation method based on NLP technology and CV technology; use the preprocessing annotation method to preprocess the unannotated data, and obtain the annotation processing interval based on the preprocessing results;

[0007] Obtain annotated unstructured data, recorded as annotated data; establish a neural network, introduce a pre-processing annotation method and an annotation processing interval into the neural network, and establish a deep annotation method in the neural network; train the neural network based on unannotated data and annotated data, and obtain an optimal neural network based on the training results;

[0008] Step S3: Automatically annotate the unannotated data based on the optimal neural network.

[0009] Furthermore, the preprocessing annotation method includes:

[0010] Obtain unannotated data, extract text from the unannotated data based on NLP technology, and record the obtained independent text segments as data text SW1 to data text SW based on the extraction order of independent text segments in the extraction results. n , where n is the number of independent text segments;

[0011] Based on CV technology, images, audio and video in unannotated data are extracted, and the extracted images, audio and video are recorded as data images, data audio and data video respectively.

[0012] Furthermore, the preprocessing annotation method also includes:

[0013] For any data image: the data image with data text SW is recorded as a text image, and the data image without data text SW is recorded as a normal image;

[0014] For any audio data: convert the audio data into text based on NLP technology, and record the obtained text as audio text; use AI to identify the background noise in the audio data, and record the audio data with background noise higher than the standard recognition decibel as noise audio;

[0015] The audio text in the audio text that is in the noise audio is recorded as the interference audio text, and the audio text in the audio text that is not in the noise audio is recorded as the normal audio text.

[0016] Furthermore, the preprocessing annotation method also includes:

[0017] For any data video: extract all images from the data video based on CV technology, and record the obtained images as video images;

[0018] Based on NLP technology, the audio of the data video is converted into text and the text of the video image is extracted, and the obtained text is recorded as video audio text and video image text respectively;

[0019] Record video images with video, audio, and text as video text images, and record video images without video, audio, and text as video standard images; obtain background noise in data videos based on AI recognition, and record data videos with background noise greater than the standard recognition decibel as noise videos;

[0020] The video and audio text in the video and audio text that is in the noise video is recorded as interference video text, and the video and audio text that is not in the noise audio is recorded as regular video text.

[0021] Furthermore, the unannotated data is preprocessed using a preprocessing annotation method, and the annotation processing interval is obtained based on the preprocessing result.

[0022] A set of unannotated data is recorded as single test data, and the single test data is preprocessed using the preprocessing annotation method, and the obtained data text SW, regular audio text and regular video text are recorded as clear processed text;

[0023] Record the interfering audio text and the interfering video text as fuzzy text;

[0024] The text image and the video text image are recorded as text processing images; the regular image and the video standard image are recorded as regular processing images.

[0025] Furthermore, the unannotated data is preprocessed using a preprocessing annotation method, and the annotation processing interval is obtained based on the preprocessing result. The processing interval also includes:

[0026] Multiple groups of unannotated data are recorded as parameter test data; for any parameter test data: based on the preprocessing annotation method, clear processed text, fuzzy processed text, text processed image and regular processed image of the parameter test data are obtained respectively, and the acquisition time is recorded as clear text time, fuzzy text time, text image time and regular image time respectively, and the memory space occupied by the parameter test data is recorded as test memory;

[0027] Obtain the clear text time, fuzzy text time, text image time, regular image time and test memory corresponding to all parameter test data, and record the values corresponding to the minimum test memory and the maximum test memory as XX1 and XX2 respectively.

[0028] Furthermore, the unannotated data is preprocessed using a preprocessing annotation method, and the annotation processing interval is obtained based on the preprocessing result. The processing interval also includes:

[0029] Establish a plane rectangular coordinate system, recorded as the parameter analysis coordinate system, where the unit of the X-axis of the parameter analysis coordinate system is MB, and the unit of the Y-axis is ms; for any parameter test data: in the parameter analysis coordinate system, use the test memory of the parameter test data as the horizontal coordinate, and use the clear text time, fuzzy text time, text image time, and regular image time of the parameter test data as the vertical coordinates to punctuate, and record them as clear text points, fuzzy text points, text image points, and regular image points respectively;

[0030] The curve obtained by fitting all clear text points is recorded as curve QW; the curve obtained by fitting all fuzzy text points is recorded as curve MW; the curve obtained by fitting all text image points is recorded as curve WT; the curve obtained by fitting all regular image points is recorded as curve CT;

[0031] For any horizontal coordinate X0 from XX1 to XX2 in the X-axis, the point with the horizontal coordinate X0 and the smallest vertical coordinate in the curve QW, curve MW, curve WT and curve CT is recorded as the low-parameter point; the curve composed of all low-parameter points in the parameter analysis coordinate system is recorded as the parameter analysis curve.

[0032] Furthermore, the unannotated data is preprocessed using a preprocessing annotation method, and the annotation processing interval is obtained based on the preprocessing result. The processing interval also includes:

[0033] The interval formed by the horizontal coordinates of the area in the parameter analysis curve that coincides with the curve QW is recorded as the QW preferred interval; the interval formed by the horizontal coordinates of the area in the parameter analysis curve that coincides with the curve MW is recorded as the MW preferred interval;

[0034] The interval formed by the abscissa of the area in the parameter analysis curve that coincides with the curve WT is recorded as the WT preferred interval;

[0035] The interval formed by the abscissa of the area in the parameter analysis curve that coincides with the curve CT is recorded as the CT optimal interval;

[0036] The QW preferred interval, the MW preferred interval, the WT preferred interval, and the CT preferred interval are recorded as annotation processing intervals.

[0037] Further, in-depth annotation methods include:

[0038] The neurons in the input layer of the neural network receive data, and the memory space occupied by the received data is recorded as the annotation memory; the interval where the annotation memory is located in the annotation processing interval is recorded as the preferred analysis interval, and the data type corresponding to the preferred analysis interval is recorded as the preferred screening type, where the data type includes clear processing text, fuzzy processing text, text processing image and conventional processing image;

[0039] Based on the preprocessing annotation method, the preferred screened-out types of data are preferentially obtained, and the data types other than the preferred screened-out types are obtained;

[0040] The data types obtained by the preprocessing annotation method are annotated based on target detection boxes, named entity recognition, and speaker separation technologies, and the annotated data are output by the neurons in the output layer of the neural network.

[0041] Furthermore, training a neural network based on the unannotated data and the annotated data, and obtaining a preferred neural network based on the training results includes:

[0042] For any value X0 between XX1 and XX2, the unannotated data and the annotated data whose annotation is X0 are respectively input into the neural network and annotation processing is performed, wherein the method of annotating the annotated data is recorded as the annotated method. When the annotated data is input into the neural network, the neural network re-annotates the annotated data based on the annotated method.

[0043] When the time that the unannotated data is annotated by the neural network is less than or equal to the time that the annotated data is annotated by the neural network, X0 is recorded as optimized memory and the annotated method corresponding to X0 is removed from the neural network;

[0044] When the time for the unannotated data to be annotated by the neural network is greater than the time for the annotated data to be annotated by the neural network, X0 is recorded as the memory to be learned and the annotated method corresponding to X0 is recorded as the preferred annotation method of X0;

[0045] When all values in XX1 to XX2 are recorded as optimized memory, or recorded as memory to be learned and the corresponding preferred annotation method is entered into the neural network, the neural network is recorded as the preferred neural network;

[0046] For unannotated data annotated by a neural network: when the annotation memory is the optimized memory, the unannotated data is annotated based on the preprocessing annotation method and the annotation processing interval; when the annotation memory is the memory to be learned, the unannotated data is annotated based on the preferred annotation method of the memory to be learned.

[0047] Beneficial effects of the present invention: The present application first obtains unannotated unstructured data, recorded as unannotated data; establishes a preprocessing annotation method based on NLP technology and CV technology; uses the preprocessing annotation method to preprocess the unannotated data, and obtains the annotation processing interval based on the preprocessing result. The advantage of this is that by establishing the preprocessing annotation method based on NLP technology and CV technology, when there are many formats and data sources in the unstructured data, the text, pictures, audio and video in the unstructured data can be effectively preprocessed by the preprocessing annotation method to make them pure text or pictures, thereby establishing a unified standard for cross-type data, thereby improving the subsequent annotation efficiency; and by obtaining the annotation processing interval, the type of data extraction with the highest efficiency when using the preprocessing annotation method can be obtained based on the size of the memory occupied by the unstructured data, which helps to preprocess the unstructured data with the fastest preprocessing efficiency when facing unstructured data of different sizes;

[0048] The present application also obtains annotated unstructured data and records it as annotated data; establishes a neural network, introduces the preprocessing annotation method and the annotation processing interval into the neural network, and establishes an in-depth annotation method in the neural network; trains the neural network based on unannotated data and annotated data, and obtains the preferred neural network based on the training results; finally, automatically annotates the unannotated data based on the preferred neural network. The advantage of this is that after establishing and training the neural network, the preferred neural network is used for automatic annotation, which can effectively improve the annotation efficiency of unstructured data in the case of more data formats and more data sources in the unstructured data, thereby reducing the user's information processing time while improving the user's efficiency in receiving data. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a flow chart of the steps of the method of the present invention;

[0050] Figure 2 A schematic diagram of the parameter analysis coordinate system of the present invention;

[0051] Figure 3 Schematic diagram of the structure of the electronic device of the present invention. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0053] Example 1, please refer to Figure 1 As shown, the present application provides a method for automatically annotating unstructured data based on a neural network, comprising the following steps:

[0054] Step S1, obtaining unannotated unstructured data, recorded as unannotated data; establishing a preprocessing annotation method based on NLP technology and CV technology; using the preprocessing annotation method to preprocess the unannotated data, and obtaining an annotation processing interval based on the preprocessing result;

[0055] Step S1 includes: Step S101, the pre-processing annotation method includes: Step S1011, obtaining unannotated data, extracting text from the unannotated data based on NLP technology, and recording the obtained independent text segments in the extraction result as data text SW1 to data text SW n , where n is the number of independent text segments;

[0056] In a specific implementation, the data analysis SW may include text in the unannotated data that is within the image and text outside the image in the unannotated data;

[0057] Step S1012: extracting images, audio, and video from the unannotated data based on CV technology, and recording the extracted images, audio, and video as data images, data audio, and data video, respectively;

[0058] Step S1013: for any data image: the data image with the data text SW is recorded as a text image, and the data image without the data text SW is recorded as a regular image;

[0059] Step S1014: For any audio data, the audio data is converted into text based on NLP technology, and the resulting text is recorded as audio text; background noise in the audio data is obtained based on AI recognition, and audio data with background noise exceeding the standard recognition decibel level is recorded as noise audio;

[0060] In the specific implementation process, in this embodiment, the standard recognition decibel is set to 40 decibels. The standard recognition decibel can be adjusted according to the actual level of background noise in the audio of the unstructured data during actual analysis and the actual environment in which the audio is located. For example, if the actual environment in which the audio is located is a relatively quiet place such as an auditorium or classroom, the standard recognition decibel can be lowered. By obtaining background noise and marking the noisy audio, it is possible to distinguish between audio in the data where text can be clearly recognized and audio in which text cannot be clearly recognized, thereby facilitating the distinction between clear text and fuzzy text during subsequent analysis and annotation.

[0061] Step S1015: Record the audio text in the noise audio as interference audio text, and record the audio text in the non-noise audio as regular audio text;

[0062] Step S1016: for any data video: extract all pictures in the data video based on CV technology, and record the obtained pictures as video pictures;

[0063] Step S1017: Based on NLP technology, the audio of the data video is converted into text and the text of the video image is extracted, and the obtained texts are recorded as video audio text and video image text respectively;

[0064] Step S1018: Record the video image with video, audio, and text as a video text image, and record the video image without video, audio, and text as a video standard image; obtain the background noise in the data video based on AI recognition, and record the data video with background noise greater than the standard recognition decibel as a noise video;

[0065] Step S1019: Record the video and audio text in the noise video as interference video text, and record the video and audio text not in the noise audio as regular video text;

[0066] In the specific implementation process, by processing the text, images, audio and video in the unannotated data, and obtaining data text SW, regular audio text, interference audio text, regular video text, interference video text, text image, video text image, regular image and video standard image, and in subsequent analysis, unifying them into clear processed text, fuzzy processed text, text processed image and regular processed image, it is possible to achieve the goal of converting the text, images, audio and video in the unstructured data into pure text or images when there are many formats and data sources in the unstructured data, thereby establishing a unified standard for cross-type data and improving the subsequent annotation efficiency;

[0067] Step S102: record a set of unannotated data as single test data, preprocess the single test data using a preprocessing annotation method, and record the obtained data text SW, conventional audio text, and conventional video text as clear processed text;

[0068] Step S1021: Record the interfering audio text and the interfering video text as fuzzy text;

[0069] Step S1022: record the text image and the video text image as text-processed images; and record the regular image and the video standard image as regular-processed images.

[0070] Step S1 further includes: step S103, recording multiple groups of unannotated data as parameter test data; for any parameter test data: obtaining clear processed text, fuzzy processed text, text processed image, and regular processed image of the parameter test data based on the preprocessing annotation method, and recording the acquisition time as clear text time, fuzzy text time, text image time, and regular image time, respectively, and recording the memory space occupied by the parameter test data as test memory;

[0071] Step S104, obtaining the clear text time, fuzzy text time, text image time, normal image time and test memory corresponding to all parameter test data, and recording the values corresponding to the minimum test memory and the maximum test memory as XX1 and XX2 respectively;

[0072] In a specific implementation process, for example, during a data analysis, the test memory occupied by all parameter test data is 2MB, 231MB, 322MB, 500MB, and 12MB respectively. Then, through analysis, it can be obtained that XX1 and XX2 are 2MB and 500MB respectively;

[0073] Step S105: Establish a plane rectangular coordinate system, recorded as the parameter analysis coordinate system, wherein the unit of the X-axis of the parameter analysis coordinate system is MB, and the unit of the Y-axis is ms. For any parameter test data: in the parameter analysis coordinate system, use the test memory of the parameter test data as the horizontal axis, and use the clear text time, fuzzy text time, text image time, and regular image time of the parameter test data as the vertical axis to punctuate points, and record them as clear text points, fuzzy text points, text image points, and regular image points, respectively.

[0074] Step S106: The curve obtained by fitting all clear text points is recorded as curve QW; the curve obtained by fitting all fuzzy text points is recorded as curve MW; the curve obtained by fitting all text image points is recorded as curve WT; and the curve obtained by fitting all regular image points is recorded as curve CT.

[0075] Step S107: For any abscissa X0 between XX1 and XX2 in the X-axis, record the point with the abscissa X0 and the smallest ordinate in the curves QW, MW, WT, and CT as a low-parameter point; and record the curve formed by all the low-parameter points in the parameter analysis coordinate system as a parameter analysis curve.

[0076] In the specific implementation process, for example, during a data analysis, the obtained parameter analysis coordinate system is as follows Figure 2 As shown in PP1 in the figure, through analysis, it can be obtained that the curve CF in PP2 is a parameter analysis curve. By obtaining the parameter analysis curve, the type of data extraction with the highest efficiency when using the preprocessing annotation method can be obtained based on the size of the memory occupied by the unstructured data, which helps to preprocess the unstructured data with the highest preprocessing efficiency when facing unstructured data of different sizes. For example, for point CO1 in the curve CF, through analysis, it can be obtained that point CO1 is in the curve WT, indicating that when the memory occupied by the unstructured data is NC, the text image time is the smallest, that is, the processing efficiency of the text processing image is the fastest. Therefore, when the memory occupied by the unstructured data is CO1, the text processing image can be extracted first to achieve the purpose of improving the annotation efficiency.

[0077] Step S108: record the interval formed by the horizontal coordinates of the area in the parameter analysis curve that overlaps with the curve QW as the QW preferred interval; record the interval formed by the horizontal coordinates of the area in the parameter analysis curve that overlaps with the curve MW as the MW preferred interval;

[0078] Step S109, recording the interval formed by the horizontal coordinate of the area in the parameter analysis curve that overlaps with the curve WT as the WT preferred interval;

[0079] The interval formed by the abscissa of the area in the parameter analysis curve that coincides with the curve CT is recorded as the CT optimal interval;

[0080] The QW preferred interval, MW preferred interval, WT preferred interval, and CT preferred interval are recorded as annotation processing intervals;

[0081] In the specific implementation process, through Figure 2 It can be obtained that the QW preferred interval, MW preferred interval, WT preferred interval and CT preferred interval are [C1, C2], [C2, C3], [C3, C4] and [C4, C5] respectively; for memory that is in two intervals at the same time, such as C2 that is in both the QW preferred interval and the MW preferred interval, in subsequent analysis, when the space occupied by unstructured data is C2, either the QW preferred interval or the MW preferred interval can be selected for analysis.

[0082] Step S2, obtaining annotated unstructured data, recorded as annotated data; establishing a neural network, introducing the pre-processing annotation method and the annotation processing interval into the neural network, and establishing a deep annotation method in the neural network; training the neural network based on the unannotated data and the annotated data, and obtaining an optimal neural network based on the training results;

[0083] Step S2 has the following sub-steps: Step S201, the in-depth annotation method includes: Step S2011, receiving data by neurons in the input layer of the neural network, and recording the memory space occupied by the received data as annotation memory; recording the interval where the annotation memory is located in the annotation processing interval as the preferred analysis interval, and recording the data type corresponding to the preferred analysis interval as the preferred screening type, wherein the data type includes clear processing text, fuzzy processing text, text processing image and conventional processing image;

[0084] Step S2012: obtaining the preferred screened-out types of data based on the preprocessing annotation method, and obtaining data types other than the preferred screened-out types;

[0085] Step S2013, annotating the data type obtained by the pre-processing annotation method based on the object detection box, named entity recognition, and speaker separation technology, and outputting the annotated data through the neurons of the output layer of the neural network;

[0086] In the specific implementation process, the data types obtained by the preprocessing annotation method can be annotated according to the annotation tools that can be applied in the actual analysis, so as to ensure that the annotation is more comprehensive and efficient.

[0087] Step S2 further includes: step S202, for any value X0 from XX1 to XX2, inputting the unannotated data and the annotated data whose annotation is X0 into the neural network respectively, and performing annotation processing, wherein the method for annotating the annotated data is recorded as the annotated method, and when the annotated data is input into the neural network, the neural network re-annotates the annotated data based on the annotated method;

[0088] Step S203, when the time for which the unannotated data is annotated by the neural network is less than or equal to the time for which the annotated data is annotated by the neural network, X0 is recorded as optimized and the annotated method corresponding to X0 is removed from the neural network;

[0089] Step S204, when the time for which the unannotated data is annotated by the neural network is greater than the time for which the annotated data is annotated by the neural network, X0 is recorded as a memory to be learned and the annotated method corresponding to X0 is recorded as the preferred annotation method of X0;

[0090] In the specific implementation process, for example, during a data analysis, the time it takes for the unannotated data to be annotated by the neural network is 1 minute, and the time it takes for the annotated data to be annotated by the neural network is 0.8 minutes. This means that the annotation efficiency of the annotated method for the annotated data is better than that of the neural network. Therefore, the neural network can be used to learn the annotated method, thereby improving the neural network. In subsequent annotations, the unannotated data can be annotated quickly and accurately based on the preferred annotation method learned by the neural network, the pre-processed annotation method stored by the neural network, and the annotation processing interval.

[0091] Step S205 , when all values in XX1 to XX2 are recorded as optimized memory, or recorded as memory to be learned and the corresponding preferred annotation method is entered into the neural network, the neural network is recorded as the preferred neural network;

[0092] Step S206, for the unannotated data annotated by the neural network: when the annotation occupancy is the optimized occupancy, the unannotated data is annotated based on the preprocessing annotation method and the annotation processing interval; when the annotation occupancy is the to-be-learned occupancy, the unannotated data is annotated based on the preferred annotation method of the to-be-learned occupancy.

[0093] Step S3: Automatically annotate the unannotated data based on the optimal neural network.

[0094] Example 2, please refer to Figure 3 As shown, Figure 3 The present invention provides a schematic diagram of the structure of an electronic device, which may include: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus. The memory stores computer-readable instructions, and the processor can call the instructions in the memory. When the computer-readable instructions are executed by the processor, the steps in the method for automatically annotating unstructured data based on a neural network are executed to achieve the following functions: first, unannotated unstructured data is obtained, recorded as unannotated data; a preprocessing annotation method is established based on NLP technology and CV technology; the unannotated data is preprocessed using the preprocessing annotation method, and an annotation processing interval is obtained based on the preprocessing result; then, annotated unstructured data is obtained, recorded as annotated data; a neural network is established, the preprocessing annotation method and the annotation processing interval are introduced into the neural network, and a deep annotation method is established in the neural network; the neural network is trained based on the unannotated data and the annotated data, and a preferred neural network is obtained based on the training result; and finally, the unannotated data is automatically annotated based on the preferred neural network.

[0095] In addition, the logical instructions in the above-mentioned memory can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application, or the part that contributes to the existing technology, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0096] Example 3. The present application also provides a computer program product, which includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the method for automatically annotating unstructured data based on a neural network provided by the above methods, the method including: first obtaining unannotated unstructured data, recorded as unannotated data; establishing a preprocessing annotation method based on NLP technology and CV technology; using the preprocessing annotation method to preprocess the unannotated data, and obtaining an annotation processing interval based on the preprocessing result; then obtaining annotated unstructured data, recorded as annotated data; establishing a neural network, introducing the preprocessing annotation method and the annotation processing interval into the neural network, and establishing an in-depth annotation method in the neural network; training the neural network based on the unannotated data and the annotated data, and obtaining a preferred neural network based on the training results; finally, automatically annotating the unannotated data based on the preferred neural network.

[0097] Example 4. The present application also provides a computer-readable storage medium. The present application provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps in the method for automatically annotating unstructured data based on a neural network are executed to achieve the following functions: first, unannotated unstructured data is obtained, recorded as unannotated data; a preprocessing annotation method is established based on NLP technology and CV technology; the unannotated data is preprocessed using the preprocessing annotation method, and an annotation processing interval is obtained based on the preprocessing result; then, annotated unstructured data is obtained, recorded as annotated data; a neural network is established, the preprocessing annotation method and the annotation processing interval are introduced into the neural network, and an in-depth annotation method is established in the neural network; the neural network is trained based on the unannotated data and the annotated data, and a preferred neural network is obtained based on the training results; finally, the unannotated data is automatically annotated based on the preferred neural network.

[0098] Through the description of the above embodiments, the embodiments of the present invention can be provided as methods, systems, or computer program products. Based on this understanding, the essence of the above technical solutions or the portion that contributes to the prior art can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for enabling a computer device (such as a personal computer, server, or network device) to execute the methods described in various embodiments or certain portions of the embodiments.

[0099] In the embodiments provided in this application, it should be understood that the disclosed system or method can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of modules or units is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or units can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of systems, modules and units can be electrical, mechanical or other forms.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for automatically annotating unstructured data based on a neural network, characterized in that: The steps include: Obtain unannotated unstructured data, recorded as unannotated data; establish a preprocessing annotation method based on NLP technology and CV technology; Use the preprocessing annotation method to preprocess the unannotated data, and obtain the annotation processing interval based on the preprocessing results; Obtain annotated unstructured data, record it as annotated data; establish a neural network, introduce the pre-processing annotation method and the annotation processing interval into the neural network, and establish a deep annotation method in the neural network; Training a neural network based on unannotated data and annotated data, and obtaining a preferred neural network based on the training results; Automatic annotation of unannotated data based on optimized neural networks; Preprocessing annotation methods include: Obtain unannotated data, extract text from the unannotated data based on NLP technology, and record the obtained independent text segments as data text SW1 to data text SW based on the extraction order of independent text segments in the extraction results. n , where n is the number of independent text segments; Extract images, audio and video from unannotated data based on CV technology, and record the extracted images, audio and video as data images, data audio and data video respectively; Preprocessing annotation methods also include: For any data image: the data image with data text SW is recorded as a text image, and the data image without data text SW is recorded as a normal image; For any audio data: convert the audio data into text based on NLP technology, and record the obtained text as audio text; use AI to identify the background noise in the audio data, and record the audio data with background noise higher than the standard recognition decibel as noise audio; The audio text in the audio text that is in the noise audio is recorded as the interference audio text, and the audio text in the audio text that is not in the noise audio is recorded as the normal audio text.

2. The method for automatically annotating unstructured data based on a neural network according to claim 1, characterized in that: Preprocessing annotation methods also include: For any data video: extract all images from the data video based on CV technology, and record the obtained images as video images; Based on NLP technology, the audio of the data video is converted into text and the text of the video image is extracted, and the obtained text is recorded as video audio text and video image text respectively; Record video images with video, audio, and text as video text images, and record video images without video, audio, and text as video standard images; obtain background noise in data videos based on AI recognition, and record data videos with background noise greater than the standard recognition decibel as noise videos; The video and audio text in the video and audio text that is in the noise video is recorded as interference video text, and the video and audio text that is not in the noise audio is recorded as regular video text.

3. The method for automatically annotating unstructured data based on a neural network according to claim 2, characterized in that: Use the preprocessing annotation method to preprocess the unannotated data and obtain the annotation processing interval based on the preprocessing results. A set of unannotated data is recorded as single test data, and the single test data is preprocessed using the preprocessing annotation method, and the obtained data text SW, regular audio text and regular video text are recorded as clear processed text; Record the interfering audio text and the interfering video text as fuzzy text; The text image and the video text image are recorded as text processing images; the regular image and the video standard image are recorded as regular processing images.

4. The method for automatically annotating unstructured data based on a neural network according to claim 3, characterized in that: Use the preprocessing annotation method to preprocess the unannotated data and obtain annotations based on the preprocessing results. The processing interval also includes: Multiple groups of unannotated data are recorded as parameter test data; for any parameter test data: based on the preprocessing annotation method, clear processed text, fuzzy processed text, text processed image and regular processed image of the parameter test data are obtained respectively, and the acquisition time is recorded as clear text time, fuzzy text time, text image time and regular image time respectively, and the memory space occupied by the parameter test data is recorded as test memory; Obtain the clear text time, fuzzy text time, text image time, regular image time and test memory corresponding to all parameter test data, and record the values corresponding to the minimum test memory and the maximum test memory as XX1 and XX2 respectively.

5. The method for automatically annotating unstructured data based on a neural network according to claim 4, characterized in that: Use the preprocessing annotation method to preprocess the unannotated data and obtain annotations based on the preprocessing results. The processing interval also includes: Establish a plane rectangular coordinate system, recorded as the parameter analysis coordinate system, where the unit of the X-axis of the parameter analysis coordinate system is MB, and the unit of the Y-axis is ms; for any parameter test data: in the parameter analysis coordinate system, use the test memory of the parameter test data as the horizontal coordinate, and use the clear text time, fuzzy text time, text image time, and regular image time of the parameter test data as the vertical coordinates to punctuate, and record them as clear text points, fuzzy text points, text image points, and regular image points respectively; The curve obtained by fitting all clear text points is recorded as curve QW; the curve obtained by fitting all fuzzy text points is recorded as curve MW; the curve obtained by fitting all text image points is recorded as curve WT; the curve obtained by fitting all regular image points is recorded as curve CT; For any horizontal coordinate X0 from XX1 to XX2 in the X-axis, the point with the horizontal coordinate X0 and the smallest vertical coordinate in the curve QW, curve MW, curve WT and curve CT is recorded as the low-parameter point; the curve composed of all low-parameter points in the parameter analysis coordinate system is recorded as the parameter analysis curve.

6. The method for automatically annotating unstructured data based on a neural network according to claim 5, characterized in that: Use the preprocessing annotation method to preprocess the unannotated data and obtain annotations based on the preprocessing results. The processing interval also includes: The interval formed by the horizontal coordinates of the area in the parameter analysis curve that coincides with the curve QW is recorded as the QW preferred interval; the interval formed by the horizontal coordinates of the area in the parameter analysis curve that coincides with the curve MW is recorded as the MW preferred interval; The interval formed by the abscissa of the area in the parameter analysis curve that coincides with the curve WT is recorded as the WT preferred interval; The interval formed by the abscissa of the area in the parameter analysis curve that coincides with the curve CT is recorded as the CT optimal interval; The QW preferred interval, the MW preferred interval, the WT preferred interval, and the CT preferred interval are recorded as annotation processing intervals.

7. The method for automatically annotating unstructured data based on a neural network according to claim 6, characterized in that: In-depth annotation methods include: The neurons in the input layer of the neural network receive data, and the memory space occupied by the received data is recorded as the annotation memory; the interval where the annotation memory is located in the annotation processing interval is recorded as the preferred analysis interval, and the data type corresponding to the preferred analysis interval is recorded as the preferred screening type, where the data type includes clear processing text, fuzzy processing text, text processing image and conventional processing image; Based on the preprocessing annotation method, the preferred screened-out types of data are preferentially obtained, and the data types other than the preferred screened-out types are obtained; The data types obtained by the preprocessing annotation method are annotated based on target detection boxes, named entity recognition, and speaker separation technologies, and the annotated data are output by the neurons in the output layer of the neural network.

8. The method for automatically annotating unstructured data based on a neural network according to claim 7, characterized in that: Training a neural network based on unannotated data and annotated data, and obtaining a preferred neural network based on the training results includes: For any value X0 between XX1 and XX2, the unannotated data and the annotated data whose annotation is X0 are respectively input into the neural network and annotation processing is performed, wherein the method of annotating the annotated data is recorded as the annotated method. When the annotated data is input into the neural network, the neural network re-annotates the annotated data based on the annotated method. When the time that the unannotated data is annotated by the neural network is less than or equal to the time that the annotated data is annotated by the neural network, X0 is recorded as optimized memory and the annotated method corresponding to X0 is removed from the neural network; When the time for the unannotated data to be annotated by the neural network is greater than the time for the annotated data to be annotated by the neural network, X0 is recorded as the memory to be learned and the annotated method corresponding to X0 is recorded as the preferred annotation method of X0; When all values in XX1 to XX2 are recorded as optimized memory, or recorded as memory to be learned and the corresponding preferred annotation method is entered into the neural network, the neural network is recorded as the preferred neural network; For unannotated data annotated by a neural network: when the annotation memory is the optimized memory, the unannotated data is annotated based on the preprocessing annotation method and the annotation processing interval; when the annotation memory is the memory to be learned, the unannotated data is annotated based on the preferred annotation method of the memory to be learned.

Images