A text annotation recommendation method, device, equipment and storage medium

Abstract

The present invention is applicable to the field of computer technology, and provides a text annotation recommendation method, device, equipment and storage medium. The method includes: obtaining a text to be annotated, preprocessing the text to be annotated to obtain a word set, obtaining a word feature set representing the text to be annotated through a pre-trained model based on the word set, obtaining sentence features representing each sentence of the text to be annotated according to the word feature set, clustering the sentence features using a preset clustering algorithm to generate pseudo labels corresponding to each category, performing keyword mining on each category, determining an actual category label and a corresponding keyword set for each category, and generating annotation suggestions for the text to be annotated according to the actual category label of each sentence and the corresponding keyword set, thereby improving the accuracy of text annotation.

Description

Technical Field

[0001] The present invention belongs to the field of computer technology, and in particular relates to a text annotation recommendation method, device, equipment and storage medium. Background Art

[0002] The Internet has developed into the world's largest information base and the main channel for disseminating information worldwide. With the large-scale popularization of the Internet and the improvement of the degree of enterprise informatization, various data resources have grown explosively. In order to mine useful data from massive data files, these data texts must be analyzed first. Data analysis is usually based on machine learning or deep learning models. First, the collected text data needs to be subject-labeled, and then the labeled data is used to train the machine learning model or deep learning model, and then the trained model is applied to the classification and prediction of unlabeled data.

[0003] Text annotation is usually done manually with the help of annotation tools. In addition, there are also open source annotation tools and commercial annotation platforms based on crowdsourcing strategies, such as DeepDive, LabelBox annotation tools, Amazon Mechanical Turk, CrowdFlower, Baidu Crowd Testing, Alibaba Crowdsourcing, etc. Crowdsourcing can annotate a huge amount of text, but its limitation is that its annotators come from a wide range of sources and the quality of annotation varies. In highly professional fields, since topic annotation requires the intervention of professional knowledge, especially the need for field experts to conduct in-depth analysis of the text. Therefore, traditional crowdsourcing strategies are not suitable for highly professional data annotation. When using crowdsourcing strategies, if intelligent annotation suggestions can be provided to annotators to assist them in learning related field keywords and field characteristics, it will help improve the accuracy and efficiency of annotation. Summary of the invention

[0004] The purpose of the present invention is to provide a text annotation recommendation method, device, equipment and storage medium, aiming to solve the problem that the existing text annotation method has low accuracy.

[0005] In one aspect, the present invention provides a text annotation recommendation method, the method comprising:

[0006] Acquire a text to be annotated, and preprocess the text to be annotated to obtain a word set;

[0007] Based on the word set, a word feature set representing the text to be annotated is obtained through a pre-trained model, and sentence features representing each sentence of the text to be annotated are obtained according to the word feature set;

[0008] Clustering the sentence features using a preset clustering algorithm to generate pseudo labels corresponding to each category;

[0009] Performing keyword mining on each of the categories to determine an actual category label and a corresponding keyword set for each category;

[0010] Generate annotation suggestions for the text to be annotated according to the actual category label of each sentence and the corresponding theme word set.

[0011] Preferably, after the step of performing keyword mining on each category and determining the actual category label and the corresponding keyword set for each category, and before the step of generating annotation suggestions for the text to be annotated according to the actual category label and the corresponding keyword set of each sentence, the method includes:

[0012] Receive modification and / or merging requests for the actual category labels from domain experts, and based on the modification and / or merging requests, modify the actual category labels of the corresponding categories, and / or merge the actual category labels of the corresponding categories and the corresponding subject word sets to obtain updated actual category labels and corresponding subject word sets.

[0013] Preferably, after the step of modifying the actual category label of the corresponding category and / or merging the actual category label of the corresponding category and the corresponding subject word set, the method further includes:

[0014] The updated actual category label and the corresponding theme word set are input into the pre-trained model so that the pre-trained model is trained according to the updated actual category label and the corresponding theme word set to update the pre-trained model.

[0015] Preferably, after the step of performing keyword mining on each category and determining the actual category label and the corresponding keyword set for each category, and before the step of generating annotation suggestions for the text to be annotated according to the actual category label and the corresponding keyword set of each sentence, the method further includes:

[0016] The actual category labels determined for each category and the corresponding subject word set are stored in a preset category-subject word database.

[0017] Preferably, after the step of generating annotation suggestions for the text to be annotated according to the category label of each sentence and the corresponding keyword set, the following steps are included:

[0018] updating the corresponding category labels and subject word sets in the category-subject word database according to the annotation results input by the annotator;

[0019] The updated category labels and theme word sets in the category-theme word database are input into the pre-trained model so that the pre-trained model is trained according to the updated category labels and theme word sets to update the pre-trained model.

[0020] In another aspect, the present invention provides a text annotation recommendation device, the device comprising:

[0021] A preprocessing unit, used for obtaining a text to be annotated, and preprocessing the text to be annotated to obtain a word set;

[0022] A feature conversion unit, configured to obtain a word feature set representing the text to be annotated based on the word set through a pre-trained model, and obtain sentence features representing each sentence of the text to be annotated according to the word feature set;

[0023] A clustering unit, used to cluster the sentence features using a preset clustering algorithm to generate pseudo labels corresponding to each category;

[0024] A subject word mining unit, used to perform subject word mining on each of the categories, and determine an actual category label and a corresponding subject word set for each category; and

[0025] The annotation suggestion generating unit is used to generate annotation suggestions for the text to be annotated according to the actual category label of each sentence and the corresponding theme word set.

[0026] Preferably, the text annotation recommendation device further includes:

[0027] The optimization unit is used to receive the modification and / or merging request of the actual category label from the domain expert, and based on the modification and / or merging request, modify the actual category label of the corresponding category, and / or merge the actual category label of the corresponding category and the corresponding subject word set to obtain an updated actual category label and a corresponding subject word set.

[0028] On the other hand, the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the functions of the system as described above when executing the computer program.

[0029] On the other hand, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the functions of the system as described above are realized.

[0030] After obtaining the text to be annotated, the present invention preprocesses the text to be annotated to obtain a word set, based on the word set, obtains a word feature set representing the text to be annotated through a pre-trained model, obtains sentence features representing each sentence of the text to be annotated according to the word feature set, clusters the sentence features using a preset clustering algorithm to generate pseudo labels corresponding to each category, performs keyword mining on each category, determines the actual category label and the corresponding keyword set for each category, generates annotation suggestions for the text to be annotated according to the actual category label of each sentence and the corresponding keyword set, thereby improving the accuracy of text annotation. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is a flowchart of the implementation of the text annotation recommendation method provided in the first embodiment of the present invention;

[0032] Figure 2 is a flowchart of the implementation of the text annotation recommendation method provided in the second embodiment of the present invention;

[0033] Figure 3 is a flowchart of the implementation of the text annotation recommendation method provided in the third embodiment of the present invention;

[0034] Figure 4 is a structural diagram of a text annotation recommendation device provided by a fourth embodiment of the present invention; and

[0035] Figure 5 It is a schematic diagram of the structure of the computing device provided in Embodiment 5 of the present invention. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0037] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments:

[0038] Embodiment 1:

[0039] Figure 1 The implementation process of the text annotation recommendation method provided in the first embodiment of the present invention is shown. For the convenience of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0040] In step S101, a text to be annotated is obtained, and the text to be annotated is preprocessed to obtain a word set;

[0041] The embodiment of the present invention is applicable to computer equipment, for example, computers and terminals used to annotate text to be annotated, so as to recommend annotations for the text to be annotated, and the text to be annotated may be online comments, news, and various website materials. In the embodiment of the present invention, the text to be annotated is obtained, and the text to be annotated is preprocessed to obtain a word set, wherein the text to be annotated refers to a data set in a specific field that has not been annotated. When preprocessing the text to be annotated, as an example, each sentence in the text to be annotated can be numbered, the sentences can be split, and the split sentences can be segmented and stop words can be filtered using a segmentation tool to finally obtain a word set. Specifically, the preprocessing of the text to be annotated may include the following steps: regularizing, symbolizing, stemming, word form restoration, word segmentation, part-of-speech tagging, named entity recognition, coreference resolution, collocation extraction, and relationship extraction of the content in the text to be annotated.

[0042] In step S102, based on the word set, a word feature set representing the text to be annotated is obtained through a pre-trained model, and sentence features representing each sentence of the text to be annotated are obtained according to the word feature set;

[0043] In an embodiment of the present invention, a pre-trained model is obtained by training in an unsupervised learning manner using a large-scale corpus, so that the learning effect of the pre-trained model is better. After obtaining the word set through step S101, the word set is input into the pre-trained model, and the word feature set representing the text to be annotated is obtained through the pre-trained model. According to the word feature set, the sentence features representing each sentence of the text to be annotated are obtained. Specifically, the word feature set can be a word feature vector set. After obtaining the word feature set, the word feature vector set is weighted and summed to obtain the sentence features representing each sentence of the text to be annotated, that is, the sentence feature representation. When the word feature vector set is weighted and summed, the weight is set to the TF-IDF (TermFrequency-Inverse Document Frequency) value of the word. For example, the feature representation of one sentence can be a high-dimensional (generally 300 dimensions) feature vector, such as [0.2, 0.12, 0.3, 0.5, ...].

[0044] In step S103, the sentence features are clustered using a preset clustering algorithm to generate pseudo labels corresponding to each category;

[0045] In the embodiment of the present invention, a clustering algorithm is used to cluster each sentence feature of the annotated text to obtain corresponding category labels and keywords. The number of clustered categories can be obtained using empirical values, or a relatively large value can be set. The categories can be merged later (by domain experts) to obtain categories that meet the characteristics of the domain. At this time, the category label does not determine the name of the specific category, but only determines which category or category mark the sentence feature belongs to, which is called a pseudo-label here, for example, category 1, category 2, category 3... or category A, category B, category C, etc.

[0046] In step S104, subject word mining is performed on each category to determine the actual category label and the corresponding subject word set for each category;

[0047] In the embodiment of the present invention, subject word mining is performed on each category, and the actual category label and the corresponding subject word set are determined for each category. Specifically, the preset latent Dirichlet Allocation (LDA) model can be used to determine the subject words, wherein there can be multiple subject words in a category, and the determination of the subject words is determined according to the frequency of occurrence of each word constituting a sentence in each category in the category, and the words with higher frequency of occurrence are used as subject words. The subject words of each category constitute a subject word set, and the number of subject word sets corresponding to different texts is also different. After the subject word set of each category is determined, a word is selected from the subject word set as the actual category label of the category instead of the pseudo label. As an example, assuming that the high-probability subject words mined in a certain category include teacher, teaching, teaching, teaching design, education, etc., if the frequency of the word teacher is higher than the frequency of other subject words in the category, then this category can use teacher as the actual category label of the category.

[0048] In step S105 , annotation suggestions for the text to be annotated are generated according to the actual category label of each sentence and the corresponding keyword set.

[0049] In an embodiment of the present invention, annotation suggestions for the text to be annotated are generated based on the actual category label of each sentence and the corresponding subject word set. The annotation content is the subject word set and the actual category label determined by step S104, and the annotation method can be determined according to different text types, such as frame annotation, point annotation, color annotation, and line annotation. Preferably, at a certain position in the text to be annotated, for example, the upper right corner of the text, the annotation suggestion that best matches the text to be annotated is displayed through a graphical user interface.

[0050] After obtaining the text to be annotated, the embodiment of the present invention preprocesses the text to be annotated to obtain a word set, and based on the word set, obtains a word feature set representing the text to be annotated through a pre-trained model, and obtains sentence features representing each sentence of the text to be annotated according to the word feature set, clusters the sentence features using a preset clustering algorithm to generate pseudo labels corresponding to each category, performs keyword mining on each category, determines the actual category label and the corresponding keyword set for each category, and generates accurate and reliable annotation suggestions according to the actual category label of each sentence and the corresponding keyword set, thereby improving the accuracy of text annotation.

[0051] Embodiment 2:

[0052] Figure 2 This is the implementation process of the text annotation recommendation method provided by the second embodiment of the present invention. For the convenience of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0053] In step S201, a modification and / or merging request for actual category labels is received from a domain expert, and based on the request, the actual category labels of the corresponding categories are modified and / or the actual category labels of the corresponding categories and the corresponding keyword sets are merged to obtain updated actual category labels and corresponding keyword sets.

[0054] In an embodiment of the present invention, after subject word mining is performed on each category and the actual category label and the corresponding subject word set are determined for each category, before generating annotation suggestions for the text to be annotated based on the actual category label of each sentence and the corresponding subject word set, a modification and / or merging request for the actual category label is received from a domain expert, and based on the request, the actual category label of the corresponding category is modified and / or the actual category label of the corresponding category and the corresponding subject word set are merged to obtain an updated actual category label and a corresponding subject word set. Among them, the domain expert is a professional who is familiar with a certain field, has comprehensive professional knowledge or skilled skills, and can more accurately analyze and understand the content of the annotated text, thereby determining more accurate subject words for the text. Therefore, the introduction of domain experts can not only introduce professional knowledge as an aid to the annotation system, but also check whether the category label of the text in the field is appropriate.

[0055] In step S202, the updated actual category label and the corresponding theme word set are input into the pre-trained model, so that the pre-trained model is trained according to the updated actual category label and the corresponding theme word set, thereby updating the pre-trained model.

[0056] In an embodiment of the present invention, the modified and / or merged actual category labels and corresponding keyword sets obtained in step S201 are fed back to the pre-trained model, so that the pre-trained model is trained according to the updated actual category labels and corresponding keyword sets. After multiple training updates and an increase in annotation data, the pre-trained model in the general field gradually adapts to the text characteristics of the specific professional field, further improving the accuracy of the annotation suggestion recommendation.

[0057] Embodiment three:

[0058] Figure 3 This is the implementation process of the text annotation recommendation method provided in the third embodiment of the present invention. For the convenience of explanation, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0059] In step S301, the actual category label determined for each category and the corresponding subject word set are stored in a preset category-subject word database;

[0060] In an embodiment of the present invention, the category-theme database is used to store and manage the actual category labels and corresponding theme word sets determined for each category. After theme word mining is performed on each category and the actual category labels and corresponding theme word sets are determined for each category, before generating annotation suggestions for the text to be annotated based on the actual category labels and corresponding theme word sets of each sentence, the actual category labels and corresponding theme word sets determined for each category are stored in a pre-set category-theme database to achieve efficient management of multiple actual category labels and theme words corresponding to multiple texts.

[0061] In step S302, the corresponding category labels and subject word sets in the category-subject word database are updated according to the annotation results input by the annotator, and the updated category labels and subject word sets are input into the pre-trained model, so that the pre-trained model is trained according to the updated category labels and subject word sets to update the pre-trained model.

[0062] In an embodiment of the present invention, after generating annotation suggestions for the text to be annotated based on the category label and the corresponding subject word set of each sentence, the corresponding category label and subject word set in the category-subject word database are updated according to the annotation results input by the annotator, and the updated category label and subject word set are input into the pre-training model, so that the pre-training model is trained according to the updated category label and subject word set to update the pre-training model. Specifically, the updated actual category label and the corresponding subject word set are fed back to the pre-training model, and the weights of the corresponding word features are adjusted based on the updated actual category label and the corresponding subject word set, and the corresponding feature representation is modified based on the adjusted weights. By fine-tuning the weights of the word features, the text feature representation can be updated in a timely manner, and the updated text features will be processed by the clustering algorithm to obtain better clustering results.

[0063] In an embodiment of the present invention, the actual category label and the corresponding subject word set determined for each category are stored in a pre-set category-subject word database, and the corresponding category label and subject word set in the category-subject word database are updated according to the annotation results input by the annotator, and the updated category label and subject word set are input into the pre-trained model, so that the pre-trained model is trained according to the updated category label and subject word set. After multiple model updates and training, the semantic feature representation gradually converges from the open field to the specific field text set, thereby making the annotation suggestions more accurate and reliable.

[0064] Embodiment 4:

[0065] Figure 4 The structure of the text annotation recommendation device provided by the fourth embodiment of the present invention is shown. For the convenience of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0066] A preprocessing unit 41 is used to obtain the text to be annotated and preprocess the text to be annotated to obtain a word set;

[0067] A feature conversion unit 42 is used to obtain a word feature set representing the text to be annotated based on the word set through a pre-trained model, and obtain sentence features representing each sentence of the text to be annotated according to the word feature set;

[0068] A clustering unit 43, used to cluster sentence features using a preset clustering algorithm to generate pseudo labels corresponding to each category;

[0069] A subject word mining unit 44 is used to perform subject word mining on each category and determine an actual category label and a corresponding subject word set for each category; and

[0070] The annotation suggestion generating unit 45 is used to generate annotation suggestions for the text to be annotated according to the actual category label of each sentence and the corresponding theme word set.

[0071] Preferably, in one embodiment, the text annotation recommendation device may further include an optimization unit for receiving modification and / or merging requests for actual category labels from domain experts, and based on the modification and / or merging requests, modifying the actual category labels of the corresponding categories, and / or merging the actual category labels of the corresponding categories and the corresponding keyword sets, to obtain updated actual category labels and corresponding keyword sets.

[0072] In the embodiment of the present invention, each unit of the text annotation recommendation device can be implemented by a corresponding hardware or software unit, and each unit can be an independent software or hardware unit, or can be integrated into a software or hardware unit, which is not intended to limit the present invention. The specific implementation of each unit can refer to the description of the aforementioned method embodiment, which will not be repeated here.

[0073] Embodiment five:

[0074] Figure 5 The structure of the computing device provided by the fifth embodiment of the present invention is shown. For the convenience of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0075] The computing device 5 of the embodiment of the present invention includes a processor 50, a memory 51, and a computer program 52 stored in the memory 51 and executable on the processor 50. When the processor 50 executes the computer program 52, the steps in the above-mentioned various method embodiments are implemented, for example Figure 1 Alternatively, when the processor 50 executes the computer program 52, the functions of each unit in the above-mentioned device embodiment are realized, for example Figure 4 The functions of the units 41 to 45 are shown.

[0076] The computing device of the embodiment of the present invention may be a personal computer, mobile device, etc. for text annotation. The steps implemented when the processor 50 in the computing device 5 executes the computer program 52 to implement the above method can refer to the description of the above method embodiment, which will not be repeated here.

[0077] Embodiment six:

[0078] In an embodiment of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the above method embodiment are implemented, for example, Figure 1 Alternatively, when the computer program is executed by a processor, the functions of each unit in the above-mentioned device embodiment are realized, for example Figure 4 The functions of the units 41 to 45 are shown.

[0079] After obtaining the text to be annotated, the embodiment of the present invention first preprocesses the text to be annotated to obtain a word set, and based on the word set, obtains a word feature set representing the text to be annotated through a pre-trained model, and obtains sentence features representing each sentence of the text to be annotated according to the word feature set, clusters the sentence features using a preset clustering algorithm to generate pseudo labels corresponding to each category, performs keyword mining on each category, determines the actual category label and the corresponding keyword set for each category, and generates annotation suggestions for the text to be annotated according to the actual category label of each sentence and the corresponding keyword set, thereby generating accurate and reliable annotation suggestions for reference by annotators, thereby improving the accuracy of text annotation by annotators.

[0080] The computer-readable storage medium of the embodiment of the present invention may include any entity or device, recording medium that can carry computer program code, for example, ROM / RAM, magnetic disk, optical disk, flash memory and other memories.

[0081] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A text annotation recommendation method, characterized in that: The method comprises: Acquire a text to be annotated, and preprocess the text to be annotated to obtain a word set; Based on the word set, a word feature set representing the text to be annotated is obtained through a pre-trained model, and after the word feature set is obtained, the word feature set is weighted summed to obtain sentence features representing each sentence of the text to be annotated; Clustering the sentence features using a preset clustering algorithm to generate pseudo labels corresponding to each category; Performing keyword mining on each of the categories to determine an actual category label and a corresponding keyword set for each category; storing the actual category labels and corresponding subject word sets determined for each category in a pre-set category-subject word database; Generate annotation suggestions for the text to be annotated according to the actual category label of each sentence and the corresponding theme word set; According to the annotation results input by the annotator, the corresponding category labels and subject word sets in the category-subject word database are updated; The updated category labels and subject word sets in the category-subject word database are input into the pre-trained model so that the pre-trained model is trained according to the updated category labels and subject word sets to update the pre-trained model, wherein the updating process of the pre-trained model includes: the pre-trained model adjusts the weights of corresponding word features according to the updated category labels and subject word sets, and modifies the corresponding feature representations based on the adjusted weights.

2. The text annotation recommendation method according to claim 1, characterized in that: After performing keyword mining on each category and determining the actual category label and the corresponding keyword set for each category, and before generating the annotation suggestion for the text to be annotated according to the actual category label and the corresponding keyword set of each sentence, the method includes: Receive modification and / or merging requests for the actual category labels from domain experts, and based on the modification and / or merging requests, modify the actual category labels of the corresponding categories, and / or merge the actual category labels of the corresponding categories and the corresponding subject word sets to obtain updated actual category labels and corresponding subject word sets.

3. The text annotation recommendation method according to claim 2, characterized in that: After the steps of modifying the actual category label of the corresponding category and / or merging the actual category label of the corresponding category and the corresponding subject word set, the method further includes: The updated actual category label and the corresponding theme word set are input into the pre-trained model so that the pre-trained model is trained according to the updated actual category label and the corresponding theme word set to update the pre-trained model.

4. The text annotation recommendation method according to claim 1, characterized in that: The pre-training model is trained by unsupervised learning.

5. A text annotation recommendation device, characterized in that: The device comprises: A preprocessing unit, used for obtaining a text to be annotated, and preprocessing the text to be annotated to obtain a word set; A feature conversion unit is used to obtain a word feature set representing the text to be annotated based on the word set through a pre-trained model, and after obtaining the word feature set, perform weighted summation on the word feature set to obtain sentence features representing each sentence of the text to be annotated; A clustering unit, used to cluster the sentence features using a preset clustering algorithm to generate pseudo labels corresponding to each category; A subject word mining unit, used to perform subject word mining on each of the categories, and determine an actual category label and a corresponding subject word set for each category; and A tagging suggestion generating unit, used for generating tagging suggestions for the text to be tagged according to the actual category label of each sentence and the corresponding theme word set; wherein, after subject word mining is performed on each of the categories, the actual category label and the corresponding subject word set are determined for each category, and before the annotation suggestion of the text to be annotated is generated according to the actual category label and the corresponding subject word set of each sentence, the actual category label and the corresponding subject word set determined for each category are stored in a preset category-subject word database; After generating annotation suggestions for the text to be annotated according to the category label and the corresponding subject word set of each sentence, the corresponding category label and subject word set in the category-subject word database are updated according to the annotation results input by the annotator; the updated category label and subject word set in the category-subject word database are input into the pre-trained model, so that the pre-trained model is trained according to the updated category label and subject word set to update the pre-trained model, wherein the updating process of the pre-trained model includes: the pre-trained model adjusts the weights of the corresponding word features according to the updated category label and subject word set, and modifies the corresponding feature representation based on the adjusted weights.

6. The text annotation recommendation device according to claim 5, characterized in that: Also includes: The optimization unit is used to receive the modification and / or merging request of the actual category label from the domain expert, and based on the modification and / or merging request, modify the actual category label of the corresponding category, and / or merge the actual category label of the corresponding category and the corresponding subject word set to obtain an updated actual category label and a corresponding subject word set.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the text annotation recommendation method according to any one of claims 1 to 4 are implemented.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the text annotation recommendation method according to any one of claims 1 to 4 are implemented.

Images