Intelligent annotation method and system

Abstract

The invention relates to the technical field of man-machine interaction, and discloses an intelligent annotation method and system, and the method comprises the following steps: collecting multi-modal behavior data of a user; taking the user multi-modal behavior data as input of a user behavior analysis model, and enabling the user behavior analysis model to analyze the user multi-modal behavior data so as to predict an original user operation expectation; performing goodness of fit analysis on the original user operation expectation and the user operation mode report to screen the original user operation expectation meeting the requirement, and further obtaining a final user operation expectation; and dynamically adjusting the annotation tool and / or the annotation interface based on the final user operation expectation, and responding to the annotation operation of the user. According to the method, personalized dynamic adjustment of the annotation interface and the annotation tool is realized, the user operation steps are simplified, the user can obtain good personalized experience, the technical problems of inconvenient operation and lack of personalized experience in the prior art are solved, and the user experience and the operation efficiency are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of human-computer interaction, and in particular to an intelligent annotation method and system thereof. Background Art

[0002] At present, reviewing and annotating documents through electronic systems is a basic requirement for designers to carry out their work and collaborative design. Establishing a communication platform between designers and between designers and auditors through review and annotation plays an important role in improving the speed and quality of modern product design.

[0003] The existing annotation system adjusts the annotation interface and annotation tools with the version update. The annotation interface layout is fixed and there are many types of annotation tools. The annotation interface and annotation tools cannot be adjusted in real time according to user habits. In addition, the wide variety of annotation tools makes it difficult to simplify the calling steps, which brings about the inconvenience of annotation operation and lack of personalized experience, resulting in poor user experience and low operation efficiency. Therefore, how to design an effective annotation method to improve user experience and operation efficiency is an urgent problem to be solved. Summary of the invention

[0004] The present invention aims to provide an intelligent annotation method to solve the above-mentioned technical problems, so that the annotation interface and annotation tools can be dynamically adjusted according to user needs without having to make fixed changes with version updates, so as to simplify user operations and improve user experience and operating efficiency.

[0005] In order to solve the above technical problems, the present invention provides an intelligent annotation method, comprising the following steps:

[0006] Collect user multimodal behavior data;

[0007] Using the user multimodal behavior data as input to a preset user behavior analysis model, so that the user behavior analysis model analyzes the user multimodal behavior data and further predicts the original user operation expectation;

[0008] Perform a consistency analysis between the original user operation expectations and the preset user operation mode report to screen the original user operation expectations that meet the preset requirements, and then obtain the final user operation expectations;

[0009] Dynamically adjust the annotation tool and / or annotation interface based on the end-user operation expectations and respond to the user's annotation operation.

[0010] The above scheme makes full use of the obtained user multimodal behavior data to predict user operation expectations, and then realizes personalized dynamic adjustment of the annotation interface and annotation tools, so that users can obtain the required annotation tools and adjust the annotation interface in a timely manner, which simplifies the user operation steps and allows users to get a good personalized experience. It solves the technical problems of inconvenient operation and lack of personalized experience in the existing technology, and significantly improves the user experience and operation efficiency.

[0011] Furthermore, the preset requirement may be that the matching rate between the original user operation expectation and the preset user operation mode report reaches a relatively high percentage.

[0012] Furthermore, the method for obtaining the user operation mode report may be:

[0013] Collect user multimodal behavior data; use the user multimodal behavior data as input to a preset user behavior analysis model, so that the user behavior analysis model analyzes the user multimodal behavior data to obtain a user operation mode report.

[0014] Preferably, the above scheme further includes screening and / or generating matching annotation tools and annotation layouts according to the user operation mode report data, and then recommending the annotation tools and annotation layouts to the user.

[0015] Preferably, the user multimodal behavior data described in the above scheme includes at least one of the following: stylus position, pressure, speed, voice command, gesture; the device used to collect the user multimodal behavior data includes at least one of the following: touch sensor, microphone, camera.

[0016] Preferably, the above scheme also includes preprocessing the user multimodal behavior data, and the preprocessing method includes denoising and normalization processing to improve the accuracy of user behavior prediction.

[0017] The data preprocessing step improves the quality and reliability of user multimodal behavior data, thereby improving the performance and accuracy of the user behavior prediction model.

[0018] Preferably, the above scheme also includes obtaining user annotation content based on the user annotation operation; using the user annotation content as input of a preset natural language processing model, so that the natural language processing model performs semantic analysis on the user annotation content, and then identifies the annotation subject and emotion; based on the annotation subject and emotion, provides preset annotation suggestions and classifies the user annotation content.

[0019] The annotation suggestions may include: highlighting keywords in the text, generating a brief summary of the document, adding sentiment tags to the text, and providing relevant supplementary information and / or references based on the context. Providing annotation suggestions to users through semantic analysis can help users better understand and process text content. At the same time, classifying user annotation content can improve users' personal experience and annotation efficiency in the process of document review and annotation, and collaborative annotation.

[0020] Preferably, the above scheme also includes collecting user operation instructions; inputting the user operation instructions into a preset virtual assistant so that the preset virtual assistant automatically performs annotation operations and / or provides preset annotation suggestions; wherein the user operation instructions include voice instructions and text instructions.

[0021] The introduction of the virtual assistant uses command recognition methods such as voice command recognition and text command recognition to automatically perform complex annotation operations on behalf of the user, thereby eliminating the steps of manual annotation operations required by traditional annotation methods, thereby improving the user experience.

[0022] Preferably, the above scheme further includes obtaining user annotation content based on the user annotation operation; and automatically saving the user annotation content as an annotation history version, so that the user can view and trace back the annotation history version.

[0023] The automatic saving of the annotation content can help prevent the loss of annotation data and ensure that the user's work progress is saved and backed up in a timely manner, thereby eliminating the need for the user to frequently manually save the annotation content, thereby improving user experience and operating efficiency.

[0024] Preferably, the above scheme also includes obtaining user annotation content based on the user annotation operation; using the user annotation content as input of a preset language model, and the language model automatically detects grammatical and / or logical errors in the user annotation content according to a preset language rule database, so as to issue corresponding prompts and modification suggestions to the user.

[0025] The language model detects grammatical and / or logical errors in the annotation content, which can significantly improve the quality and accuracy of the user's annotated documents, thereby improving the user's annotation efficiency.

[0026] Preferably, the above scheme also includes obtaining user annotation content based on user annotation operations; uploading user multimodal behavior data and user annotation content to the cloud database in real time; when users log in to different devices to make annotations, data synchronization of user multimodal behavior data and user annotation content is achieved through data transmission between the cloud database and the device local database; wherein the cloud database can be used for multiple devices to download its stored cloud data.

[0027] The cross-device data synchronization function of the above-mentioned different devices enables users to process and annotate documents on different devices at any time without having to worry about the preservation and transfer of files, which simplifies user operations and thus improves user annotation efficiency and annotation experience.

[0028] Furthermore, after uploading the user multimodal behavior data and user annotation content to the cloud database in real time, when multiple users log in to multiple devices to make annotations, the annotation content collected by each device is uploaded to the cloud database via the Internet, so that the annotation content can be sorted by the cloud server, and the sorted annotation content can be synchronously transmitted to each user device via the Internet; wherein, the cloud server is connected to the cloud database.

[0029] Furthermore, when multiple users log in to multiple devices to annotate, the network connection status of the devices is judged; when the network connection is stable, the annotation content collected by each device is uploaded to the cloud database via the Internet, so that the annotation content is sorted by the cloud server, and the sorted annotation content is synchronously transmitted to each user device via the Internet; when the network connection is unstable, the annotation content collected by each device is saved locally, waiting for the network connection to become stable again.

[0030] The multi-user multi-device data synchronization function enables multiple users to annotate the same document at the same time, realizing multi-user collaborative annotation and annotation content sharing. At the same time, data synchronization through the Internet ignores the physical distance between users and improves the efficiency of user teamwork. In addition, users can also invite other users to join the annotation work of the same document and limit the scope of annotation content sharing.

[0031] Preferably, the above scheme also includes collecting user suggestions and feedback to optimize the above algorithm and model according to the user suggestions and feedback, so as to further improve any of the above intelligent annotation methods.

[0032] The present invention also provides an intelligent annotation system, comprising:

[0033] User data collection module, used to collect user multimodal behavior data;

[0034] A user behavior analysis module, used to use the user multimodal behavior data as an input of a preset user behavior analysis model, so that the user behavior analysis model analyzes the user multimodal behavior data and further predicts the original user operation expectation;

[0035] The user operation expectation prediction module is used to analyze the consistency between the original user operation expectation and the preset user operation mode report to screen the original user operation expectation that meets the preset requirements, and then obtain the final user operation expectation;

[0036] The dynamic adjustment module is used to dynamically adjust the annotation tool and / or the annotation interface based on the end user's operation expectations and respond to the user's annotation operation.

[0037] The above system architecture is simple and easy to implement. It can realize an intelligent annotation method, realize personalized dynamic adjustment of the annotation interface, simplify the user operation steps, and provide users with a personalized annotation experience.

[0038] Preferably, the system also includes an intelligent semantic analysis module with a preset natural language processing model, which is used to obtain user annotation content and use the user annotation content as input to the natural language processing model so that the natural language processing model performs semantic analysis on the user annotation content and further identifies the annotation subject and sentiment; based on the annotation subject and sentiment, provides preset annotation suggestions and classifies the user annotation content.

[0039] Preferably, the system also includes a virtual assistant module with a preset virtual assistant, which is used to collect user operation instructions and input the user operation instructions into the virtual assistant so that the preset virtual assistant automatically performs annotation operations and / or provides preset annotation suggestions.

[0040] Preferably, the system further comprises an annotation backtracking control module for acquiring user annotation content and automatically saving the user annotation content as an annotation history version, so that the user can view and backtrack the annotation history version.

[0041] Preferably, the system also includes an intelligent feedback module for obtaining user annotation content, taking the user annotation content as input of a preset language model, the language model automatically detecting grammatical and / or logical errors in the user annotation content according to a preset language rule database; and issuing corresponding prompts and modification suggestions to the user based on the grammatical and / or logical errors.

[0042] Preferably, the system also includes a cross-platform synchronization module for obtaining user annotation content and uploading user multimodal behavior data and user annotation content to the cloud database in real time; when users log in to different devices to make annotations, data synchronization of user multimodal behavior data and user annotation content is achieved through data transmission between the cloud database and the device local database; wherein the cloud database can be used for multiple devices to download its stored cloud data.

[0043] Preferably, the system also includes a real-time collaboration module, which is used for uploading the annotation content collected by each device to the cloud database via the Internet when multiple users log in to multiple devices to make annotations, so as to organize the annotation content using the cloud server, and synchronously transmit the organized annotation content to each user device via the Internet; wherein the cloud server is connected to the cloud database.

[0044] An intelligent annotation method and system provided by the present invention has at least the following advantages compared with the prior art:

[0045] It can realize personalized dynamic adjustment of annotation interface and annotation tools, so that users can obtain the required annotation tools and adjust the annotation interface in time, simplifying the user operation steps and giving users a good personalized experience. Furthermore, by introducing intelligent semantic analysis, virtual assistant assistance, annotation history version backtracking, annotation intelligent error correction and suggestions, cross-platform data synchronization, and multi-user collaborative sharing of annotation functions, the user annotation experience is significantly improved and the user annotation efficiency is effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flowchart of an intelligent annotation method provided by an embodiment of the present invention;

[0047] Figure 2 It is a schematic diagram of an intelligent annotation system architecture provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0048] Below in conjunction with the drawings and Examples, the specific embodiments of the present invention are described in further detail. The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those of ordinary skill in the art to which the present invention belongs. The terms used in the present invention are only to describe specific embodiments, and are not intended to limit exemplary embodiments according to the present invention.

[0049] See also Figure 1 The embodiment of the present invention preferably provides an intelligent annotation method, comprising the following steps:

[0050] S1: Collect user multimodal behavior data;

[0051] S2: using the user multimodal behavior data as input to a preset user behavior analysis model, so that the user behavior analysis model analyzes the user multimodal behavior data and further predicts the original user operation expectation;

[0052] S3: Performing a consistency analysis between the original user operation expectations and the preset user operation mode report to screen the original user operation expectations that meet the preset requirements, and then obtain the final user operation expectations;

[0053] S4: Dynamically adjust the annotation tool and / or annotation interface based on the end-user's operation expectations and respond to the user's annotation operation.

[0054] This embodiment uses the obtained user multimodal behavior data to predict the user's operation expectations, and then realizes personalized dynamic adjustment of the annotation interface and annotation tools, so that the user can obtain the required annotation tools and adjust the annotation interface in a timely manner, which simplifies the user operation steps and allows the user to obtain a good personalized experience. It solves the technical problems of the existing technology of inconvenient operation and lack of personalized experience, and significantly improves the user experience and operation efficiency.

[0055] Preferably, the preset requirement may be that the matching rate between the original user operation expectation and the preset user operation mode report reaches a relatively high percentage.

[0056] Specifically, when the original user operation expectation and the preset user operation mode report have a high degree of consistency, it usually means that the original user operation expectation can well explain and predict the user's actual operation expectation, so it can be screened out as the final user operation expectation. This consistency evaluation can be performed in a variety of ways, including:

[0057] Data comparison method: By comparing the original user operation expectation data and the data recorded in the user operation mode report, the deviation between the two is intuitively displayed.

[0058] Statistical analysis method: Use statistical methods to analyze the differences between the original user operation expectations and the user operation mode reports, such as calculating the root mean square error (RMSE), mean value, standard deviation, etc.

[0059] The consistency evaluation method provided in this embodiment can effectively evaluate the accuracy of the original user operation expectation based on the user operation mode report, thereby improving the accuracy of user behavior prediction.

[0060] Preferably, the method for obtaining the user operation mode report may be:

[0061] Collect user multimodal behavior data; use the user multimodal behavior data as input to a preset user behavior analysis model, so that the user behavior analysis model analyzes the user multimodal behavior data to obtain a user operation mode report.

[0062] Preferably, this embodiment further includes filtering and / or generating matching annotation tools and annotation layouts according to the user operation mode report data, and then recommending the annotation tools and annotation layouts to the user.

[0063] Preferably, the user multimodal behavior data described in this embodiment includes at least one of the following: stylus position, pressure, speed, voice command, gesture; the device used to collect the user multimodal behavior data includes at least one of the following: touch sensor, microphone, camera.

[0064] Preferably, the present embodiment further includes preprocessing the user multimodal behavior data, wherein the preprocessing method includes denoising and normalization processing to improve the accuracy of user behavior prediction.

[0065] Specifically, the method for preprocessing the data of the stylus position may be: standardizing the coordinate values ​​to keep them within the same range, thereby eliminating the differences between styluses of different models.

[0066] Specifically, the method for preprocessing the data of the stylus pen pressure may be: normalizing the pressure value to ensure that the data is on a uniform scale.

[0067] Specifically, the method for preprocessing the data of the stylus pen speed may be: calculating the speed between the touch points, and performing smoothing processing to reduce noise.

[0068] Specifically, the method for data preprocessing of voice commands may be denoising, which includes using a filter to remove background noise; feature extraction, which includes extracting features such as Mel-frequency cepstral coefficients (MFCC) for subsequent voice recognition; and normalization, which includes standardizing the amplitude of an audio signal.

[0069] Specifically, the method of preprocessing the gesture data may be smoothing, which includes using a filter to smooth the gesture trajectory and reduce jitter; feature extraction, which includes extracting key points and motion features, such as speed and acceleration.

[0070] The data preprocessing method provided in this embodiment can improve the quality and reliability of user multimodal behavior data, reduce the abnormal impact of data collection on subsequent user behavior analysis and user behavior prediction, thereby improving the performance and accuracy of the user behavior prediction model.

[0071] Preferably, this embodiment also includes obtaining user annotation content based on the user annotation operation; using the user annotation content as input to a preset natural language processing model so that the natural language processing model performs semantic analysis on the user annotation content, and then identifies the annotation subject and emotion; based on the annotation subject and emotion, provides preset annotation suggestions and classifies the user annotation content.

[0072] Specifically, the annotation suggestions may include: highlighting keywords in the text, generating a brief summary of the document, adding sentiment tags to the text, and providing relevant supplementary information and / or reference materials according to the context.

[0073] Specifically, the natural language processing model includes a word embedding vector model (Word Embedding), such as Word2Vec, GloVe and FastText, etc.; wherein, the training method used by Word2Vec can be Bag-of-Words (CBOW) and Skip-Gram. These models learn word vectors by training corpora, so that semantically similar words are closer in the vector space. Applying the word vector model in the annotation method is conducive to calculating the semantic similarity of the annotation content, thereby identifying the annotation theme and sentiment and classifying the annotation text content.

[0074] This embodiment performs semantic analysis through the natural language processing model to provide annotation suggestions to users, which can help users better understand and process text content. At the same time, classifying user annotation content can improve users' personal experience and annotation efficiency in the process of document review and annotation and collaborative annotation.

[0075] Preferably, this embodiment also includes collecting user operation instructions; inputting the user operation instructions into a preset virtual assistant so that the preset virtual assistant automatically performs annotation operations and / or provides preset annotation suggestions; wherein the user operation instructions include voice instructions and text instructions.

[0076] Specifically, the algorithms and models used by the virtual assistant include:

[0077] Word embedding models, such as Word2Vec, GloVe, and FastText. The training methods used by Word2Vec can be Bag-of-Words (CBOW) and Skip-Gram.

[0078] Recurrent Neural Network (RNN), the basic structure of which includes an input layer, a hidden layer, and an output layer. The hidden layer is processed using a recurrent neural unit (RU). The recurrent neural unit can memorize the previous hidden state and realize the processing of sequence data;

[0079] Self-Attention Mechanism: The self-attention mechanism can help the model better focus on the key information in the input sequence. The calculation formula of self-attention is shown in the following formula 1:

[0080]

[0081] Where Q is the query vector, K is the key vector, V is the value vector, and d k is the dimension of the key vector;

[0082] Transformer, the core structure of Transformer is Multi-Head Self-Attention and Position-wise Feed-Forward Networks. Multi-Head Self-Attention can process different relationships in the input sequence in parallel, improving the computational efficiency of the model. Position-wise Feed-Forward Networks is a one-layer fully connected neural network that can process the position information in the sequence.

[0083] The virtual assistant provided in this embodiment achieves accurate identification of user instructions through technical means such as word embedding vector models, recurrent neural networks, self-attention mechanisms, and Transformers, thereby automatically performing complex annotation operations on behalf of users, thereby eliminating the need for users to manually perform annotation operations in traditional annotation methods and improving user experience.

[0084] Preferably, the present embodiment further comprises obtaining user annotation content based on the user annotation operation; and automatically saving the user annotation content as an annotation history version, so that the user can view and trace back the annotation history version.

[0085] Preferably, this embodiment also includes obtaining user annotation content based on the user annotation operation; using the user annotation content as input to a preset language model, and the language model automatically detects grammatical and / or logical errors in the user annotation content according to a preset language rule database to issue corresponding prompts and modification suggestions to the user.

[0086] Specifically, the language model is the natural language processing model mentioned above; the language rule database can be an SQL language rule database, a natural language processing (NLP) rule database, a programming language rule database, etc.

[0087] The language model provided in this embodiment detects grammatical and / or logical errors in the annotation content, which can significantly improve the quality and accuracy of the user's annotated documents, thereby improving the user's annotation efficiency.

[0088] Preferably, this embodiment also includes obtaining user annotation content based on user annotation operations; uploading user multimodal behavior data and user annotation content to a cloud database in real time; when users log in to different devices to make annotations, data synchronization of user multimodal behavior data and user annotation content is achieved through data transmission between the cloud database and the device local database; wherein the cloud database can be used for multiple devices to download its stored cloud data.

[0089] Furthermore, after uploading the user multimodal behavior data and user annotation content to the cloud database in real time, when multiple users log in to multiple devices to make annotations, the annotation content collected by each device is uploaded to the cloud database via the Internet, so that the annotation content can be sorted by the cloud server, and the sorted annotation content can be synchronously transmitted to each user device via the Internet; wherein, the cloud server is connected to the cloud database.

[0090] Furthermore, when multiple users log in to multiple devices to annotate, the network connection status of the devices is judged; when the network connection is stable, the annotation content collected by each device is uploaded to the cloud database via the Internet, so that the annotation content is sorted by the cloud server, and the sorted annotation content is synchronously transmitted to each user device via the Internet; when the network connection is unstable, the annotation content collected by each device is saved locally, waiting for the network connection to become stable again.

[0091] Preferably, this embodiment also includes collecting user suggestions and feedback to optimize the above algorithms and models according to the user suggestions and feedback, so as to further improve any of the above intelligent annotation methods.

[0092] See also Figure 2 This embodiment also proposes an intelligent annotation system, including:

[0093] User data collection module, used to collect user multimodal behavior data;

[0094] A user behavior analysis module, used to use the user multimodal behavior data as an input of a preset user behavior analysis model, so that the user behavior analysis model analyzes the user multimodal behavior data and further predicts the original user operation expectation;

[0095] The user operation expectation prediction module is used to analyze the consistency between the original user operation expectation and the preset user operation mode report to screen the original user operation expectation that meets the preset requirements, and then obtain the final user operation expectation;

[0096] A dynamic adjustment module, used to dynamically adjust the annotation tool and / or the annotation interface based on the end user's operation expectations and respond to the user's annotation operation;

[0097] Preferably, the system also includes an intelligent semantic analysis module with a preset natural language processing model, which is used to obtain user annotation content and use the user annotation content as input to the natural language processing model so that the natural language processing model performs semantic analysis on the user annotation content and further identifies the annotation subject and sentiment; based on the annotation subject and sentiment, provides preset annotation suggestions and classifies the user annotation content.

[0098] Preferably, the system also includes a virtual assistant module with a preset virtual assistant, which is used to collect user operation instructions and input the user operation instructions into the virtual assistant so that the preset virtual assistant automatically performs annotation operations and / or provides preset annotation suggestions.

[0099] Preferably, the system further comprises an annotation backtracking control module for acquiring user annotation content and automatically saving the user annotation content as an annotation history version, so that the user can view and backtrack the annotation history version.

[0100] Preferably, the system also includes an intelligent feedback module for obtaining user annotation content, taking the user annotation content as input of a preset language model, the language model automatically detecting grammatical and / or logical errors in the user annotation content according to a preset language rule database; and issuing corresponding prompts and modification suggestions to the user based on the grammatical and / or logical errors.

[0101] Preferably, the system also includes a cross-platform synchronization module for obtaining user annotation content and uploading user multimodal behavior data and user annotation content to the cloud database in real time; when users log in to different devices to make annotations, data synchronization of user multimodal behavior data and user annotation content is achieved through data transmission between the cloud database and the device local database; wherein the cloud database can be used for multiple devices to download its stored cloud data.

[0102] Preferably, the system also includes a real-time collaboration module, which is used for uploading the annotation content collected by each device to the cloud database via the Internet when multiple users log in to multiple devices to make annotations, so as to organize the annotation content using the cloud server, and synchronously transmit the organized annotation content to each user device via the Internet; wherein the cloud server is connected to the cloud database.

[0103] The system architecture provided by this embodiment is simple and easy to implement. It can realize an intelligent annotation method, realize personalized dynamic adjustment of the annotation interface, and simplify the user operation steps. The system can provide users with annotation suggestions and classify annotation content through intelligent semantic analysis; help users perform annotation operations through virtual assistants; save annotation content for users to view and trace back; issue corresponding prompts and modification suggestions to users based on grammatical and / or logical errors in the annotation content; provide cross-platform data synchronization and multi-user collaborative sharing of annotation functions, which significantly improves the user annotation experience and effectively improves the user annotation efficiency.

[0104] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention. These improvements and substitutions should also be regarded as the scope of protection of the present invention.

Claims

1. An intelligent annotation method, characterized in that: The following steps are involved: Collect user multimodal behavior data; Using the user multimodal behavior data as input to a preset user behavior analysis model, so that the user behavior analysis model analyzes the user multimodal behavior data and further predicts the original user operation expectation; Perform a consistency analysis between the original user operation expectations and the preset user operation mode report to screen the original user operation expectations that meet the preset requirements, and then obtain the final user operation expectations; Dynamically adjust the annotation tool and / or annotation interface based on the end-user operation expectations and respond to the user's annotation operation.

2. The intelligent annotation method according to claim 1, characterized in that: Also includes: Obtain user annotation content based on user annotation operation; Using the user's annotation content as an input to a preset natural language processing model, so that the natural language processing model performs semantic analysis on the user's annotation content, and then identifies the annotation theme and sentiment; Provide preset annotation suggestions and classify user annotation content based on annotation topics and sentiments.

3. The intelligent annotation method according to claim 1, characterized in that: Also includes: Collect user operation instructions; The user operation instruction is input into a preset virtual assistant so that the preset virtual assistant automatically performs the annotation operation and / or provides preset annotation suggestions.

4. The intelligent annotation method according to claim 1, characterized in that: Also includes: Obtain user annotation content based on user annotation operation; The user's annotation content is automatically saved as an annotation history version, so that the user can view and trace back the annotation history version.

5. The intelligent annotation method according to claim 1, characterized in that: Also includes: Obtain user annotation content based on user annotation operation; The user's annotation content is used as an input to a preset language model, and the language model automatically detects grammatical and / or logical errors in the user's annotation content according to a preset language rule database, so as to issue corresponding prompts and modification suggestions to the user.

6. An intelligent annotation system, characterized in that: include: User data collection module, used to collect user multimodal behavior data; A user behavior analysis module, which is preset with a user behavior analysis model, and is used to use the user multimodal behavior data as an input of the user behavior analysis model, so that the user behavior analysis model analyzes the user multimodal behavior data and further predicts the original user operation expectation; A user operation expectation prediction module, which is preset with a user operation mode report and is used to perform a consistency analysis between the original user operation expectations and the user operation mode report to screen the original user operation expectations that meet the preset requirements, and then obtain the final user operation expectations; The dynamic adjustment module is used to dynamically adjust the annotation tool and / or annotation interface based on the end user's operation requirements and operation intentions, and respond to the user's annotation operations.

7. The intelligent annotation system according to claim 6, characterized in that: It also includes an intelligent semantic analysis module with a preset natural language processing model, which is used to obtain user annotation content and use the user annotation content as input to the natural language processing model so that the natural language processing model can perform semantic analysis on the user annotation content and further identify the annotation subject and emotion; based on the annotation subject and emotion, provide preset annotation suggestions and classify the user annotation content.

8. The intelligent annotation system according to claim 6, characterized in that: It also includes a virtual assistant module with a preset virtual assistant, which is used to collect user operation instructions and input the user operation instructions into the preset virtual assistant so that the virtual assistant automatically performs annotation operations and / or provides preset annotation suggestions.

9. The intelligent annotation system according to claim 6, characterized in that: It also includes an annotation backtracking control module for obtaining user annotation content; automatically saving the user annotation content as an annotation history version, so that the user can view and backtrack the annotation history version.

10. The intelligent annotation system according to claim 6, characterized in that: It also includes an intelligent feedback module for obtaining user annotation content; using the user annotation content as input to a preset language model, the language model automatically detecting grammatical and / or logical errors in the user annotation content based on a preset language rule database; and issuing corresponding prompts and modification suggestions to the user based on the grammatical and / or logical errors.

Images