Cataloging methods, apparatus, electronic equipment and readable storage media

Abstract

This invention provides a cataloging method, apparatus, electronic device, and readable storage medium. The cataloging system can acquire target text; the target text includes first text, directory data, and second text; based on the first text and directory data, a text sample is determined; based on a mapping function, the text sample and second text are mapped to a matrix set corresponding to a third text and a mask; a parallel structure information group is constructed, which consists of a task example and a fourth text; the parallel structure information group is input into a multimodal model for directory prediction to determine the target directory corresponding to the fourth text. In this way, the cataloging system can construct a parallel structure information group based on the target text and input the parallel structure information group into a multimodal model for directory prediction, thereby improving the efficiency and accuracy of cataloging.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and data management technology, and in particular to a cataloging method, apparatus, electronic device and readable storage medium. Background Technology

[0002] Currently, data assets are characterized by massive volume, multiple sources, heterogeneity, and fragmentation, which greatly increases the workload of data asset managers. Traditional manual cataloging and management methods remain the mainstream for data asset cataloging, but this approach is not only time-consuming and labor-intensive, but also makes it difficult to ensure optimal cataloging results and quality.

[0003] While existing self-supervised learning frameworks can learn rich features, these features often cannot be directly applied to specific downstream tasks and require fine-tuning on labeled datasets for adaptation. However, the fine-tuning process is time-consuming and weakens the model's generalization ability, thus affecting cataloging efficiency. Summary of the Invention

[0004] To overcome the problems existing in related technologies, the present invention provides a cataloging method, apparatus, electronic device, and readable storage medium.

[0005] In a first aspect, the present invention provides a cataloging method, the method comprising:

[0006] Obtain the target text; the target text includes a first text, directory data, and a second text;

[0007] Based on the first text and the directory data, determine the text sample;

[0008] Based on the mapping function, the text sample and the second text are mapped to a set of matrices corresponding to the third text and the mask;

[0009] Construct a parallel structure information group, which consists of a task example and a fourth text;

[0010] The parallel structure information group is input into a multimodal model for directory prediction to determine the target directory corresponding to the fourth text.

[0011] Optionally, obtaining the target text includes:

[0012] Acquire multimodal data information;

[0013] The multimodal data information is converted into a structured data summary;

[0014] The structured data summary is cleaned and integrated to obtain target text that meets preset quality requirements.

[0015] Optionally, the constructed parallel structure information group includes:

[0016] Add a first tag to the task example and a second tag to the fourth text;

[0017] The parallel structure information group is trained based on the first label and the second label.

[0018] Optionally, the method further includes:

[0019] The probability of each feature node corresponding to the first text is extracted using a linear discriminant analysis algorithm.

[0020] Calculate the similarity of each feature node;

[0021] A tree structure is obtained based on the similarity using a hierarchical clustering algorithm, and then the tree structure is segmented.

[0022] Based on the segmented tree structure, the community structure corresponding to the first text is determined;

[0023] Based on the community structure, determine the output text corresponding to the first text.

[0024] Optionally, the method further includes:

[0025] During the model training phase, the matrix set is input into the multimodal model for directory prediction to obtain the output text corresponding to the third text;

[0026] Based on the output text, the target directory corresponding to the mask is extracted.

[0027] Optionally, after obtaining the target text, the following steps are included:

[0028] Based on the features and scene information carried by the target text, a storage method is determined, and the target text is stored based on the storage method.

[0029] In a second aspect, the present invention provides a cataloging apparatus, the apparatus comprising:

[0030] The acquisition module is used to acquire target text; the target text includes first text, directory data, and second text.

[0031] The determining module is used to determine a text sample based on the first text and the directory data;

[0032] The mapping module is used to map the text sample and the second text to a set of matrices corresponding to the third text and the mask based on a mapping function;

[0033] A construction module is used to construct a parallel structure information group, which consists of a task example and a fourth text.

[0034] The determining module is further configured to input the parallel structure information group into the multimodal model for directory prediction, and determine the target directory corresponding to the fourth text.

[0035] Optionally, the acquisition module includes:

[0036] The first acquisition submodule is used to acquire multimodal data information;

[0037] The conversion module is used to convert the multimodal data information into a structured data digest;

[0038] The module is used to clean and integrate the structured data summary to obtain target text that meets preset quality requirements.

[0039] Optionally, the building module further includes:

[0040] Add a module to add a first marker to the task example and a second marker to the fourth text;

[0041] The training module is used to train the parallel structure information group based on the first label and the second label.

[0042] Optionally, the device further includes:

[0043] The extraction module is used to extract the probabilities of each feature node corresponding to the first text through a linear discriminant analysis algorithm.

[0044] A calculation module is used to calculate the similarity of each feature node;

[0045] The segmentation module is used to obtain a tree structure based on the similarity using a hierarchical clustering algorithm, and to segment the tree structure.

[0046] The determination submodule is used to determine the community structure corresponding to the first text based on the segmented tree structure; and to determine the output text corresponding to the first text based on the community structure.

[0047] Optionally, the device further includes:

[0048] A submodule is obtained, which is used to input the matrix set into the multimodal model for directory prediction during the model training phase, and obtain the output text corresponding to the third text;

[0049] The extraction submodule is used to extract the target directory corresponding to the mask based on the output text.

[0050] Optionally, the device further includes:

[0051] The first determining module is used to determine the storage method based on the features and scene information carried by the target text, and to store the target text based on the storage method.

[0052] Thirdly, the present invention provides an electronic device comprising: a processor, a memory, and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the program, implements the cataloging method described in any one of the first aspects above.

[0053] Fourthly, the present invention provides a readable storage medium that, when instructions in the storage medium are executed by a processor of an electronic device, enables the electronic device to perform steps in the cataloging method as described in any of the embodiments of the first aspect above.

[0054] In this embodiment of the invention, the cataloging system can acquire target text; the target text includes first text, directory data, and second text; based on the first text and directory data, text samples are determined; based on a mapping function, the text samples and second text are mapped to a matrix set corresponding to the third text and a mask; a parallel structure information group is constructed, which consists of a task example and a fourth text; the parallel structure information group is input into a multimodal model for directory prediction to determine the target directory corresponding to the fourth text. Thus, by first determining the text samples using the first text and directory data in the target text, and then mapping the text samples and second text, a mapped matrix set can be obtained. This matrix set consists of the third text and a mask. Simultaneously, the cataloging system can construct a parallel structure information group containing the task example and the fourth text, that is, inputting the task example and the fourth text into the multimodal model. This not only obtains a more accurate target directory but also improves cataloging efficiency. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a flowchart of a cataloging method provided in an embodiment of the present invention. Figure 1 ;

[0057] Figure 2 This is a flowchart of a cataloging method provided in an embodiment of the present invention. Figure 2 ;

[0058] Figure 3 This is a flowchart of a cataloging method provided in an embodiment of the present invention. Figure 3 ;

[0059] Figure 4 This is a flowchart of the steps of a directory data extraction method provided in an embodiment of the present invention;

[0060] Figure 5 This is a flowchart illustrating the steps involved in constructing a directory data extraction prompt according to an embodiment of the present invention.

[0061] Figure 6 This is a flowchart of a cataloging method provided in an embodiment of the present invention. Figure 4 ;

[0062] Figure 7 This is a schematic diagram of the structure of a cataloging device provided in an embodiment of the present invention;

[0063] Figure 8 This is a structural diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0064] 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 them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0065] Figure 1 This is a flowchart of a cataloging method provided in an embodiment of the present invention. Figure 1 ,like Figure 1 As shown, the method may include:

[0066] Step 101: Obtain the target text; the target text includes the first text, the directory data, and the second text.

[0067] The cataloging method provided in this application embodiment can be applied to a cataloging system. During the cataloging process, the cataloging system can first obtain target text; the target text includes first text, directory data, and second text.

[0068] In this step, the target text can be the text to be processed, and the target text contains the first text, directory data, and the second text.

[0069] For example, in this step, the first text can be the extracted text, such as determining the target directory area to be extracted from the text set, i.e., determining the first text. This is not limited in the embodiments of this application.

[0070] For example, in an article, the content includes: "2. The Big Data Era, 2.1 The Development and Importance of Big Data, XXXXXXXXX, 3. Data Governance in Higher Education, 3.1 The Importance of Data Governance to Higher Education; 3.1.1 The Need to Improve Education Quality; XXXXXX; 3.1.2 The Need to Improve the Scientific Nature of Decision-Making; XXXXXX; 3.1.3 The Need to Improve Management Efficiency; XXXXXX; No limitation is made in the embodiments of this application."

[0071] In this step, the second text can be the text to be predicted, that is, other texts in the text set besides the first text, which is not limited in the embodiments of this application.

[0072] In this step, the directory data can be the already extracted data. For example, the directory data extracted from the first text could be: "'First-level directory': '2. The Big Data Era' {Second-level directory': '2.1 The Development and Importance of Big Data'}; 'First-level directory': '3. Data Governance in Higher Education' {'Second-level directory': '3.1 The Importance of Data Governance to Higher Education' {'Third-level directory': '3.1.1 The Need to Improve Education Quality', '3.1.2 The Need to Improve the Scientific Nature of Decision-Making', '3.1.3 The Need to Improve Management Efficiency' The following table is not limited to the following in the embodiments of this application: {'Secondary directory': '3.2 Analytical Framework for Higher Education Data Governance' {'Third-level directory': '3.2.1 Acquisition and Extraction', '3.2.2 Integration and Analysis', '3.2.3 Interpretation and Prediction'}}; 'First-level directory': '5. The Future of Higher Education Data Governance' {'Secondary directory': '5.1 Application Trends of Big Data in Higher Education', 'Secondary directory': '5.2 Challenges and Problems', 'Secondary directory': '5.3 Future Prospects and Suggestions'}”

[0073] In this step, during the cataloging process, the cataloging system can first obtain the first text and extract the catalog data from the first text, as well as the second text, which is the text to be predicted.

[0074] Furthermore, in this step, when acquiring the target text, the cataloging system can acquire multimodal data information; convert the multimodal data information into a structured data summary; and perform data cleaning and integration on the structured data summary to obtain the target text that meets the preset quality requirements.

[0075] In this step, the multimodal data information can be information from multiple data sources. For example, multimodal data can include text data, image data, audio data, video data, sensor data, biometric data, etc., and is not limited in the embodiments of this application.

[0076] For example, in this application, text data is one of the most common data modalities, including natural language text, labeled text, etc. Text data is commonly used in natural language processing (NLP) tasks, such as sentiment analysis, machine translation, etc.

[0077] For example, image data is another important data modality, including photographs, paintings, etc. Image data plays a crucial role in computer vision (CV) tasks, such as image recognition and image classification.

[0078] For example, audio data includes music, speech, etc. Audio processing has wide applications in fields such as speech recognition and audio classification.

[0079] For example, video data combines the characteristics of images and audio, including video clips, movies, etc. Video data is very important in tasks such as video understanding and video summarization.

[0080] For example, sensor data consists of data collected by various sensors, such as radar, infrared, and accelerometers. This data has wide applications in fields such as the Internet of Things (IoT) and environmental monitoring.

[0081] For example, biometric data includes fingerprints, iris scans, facial features, etc., which play an important role in fields such as biometrics and identity verification.

[0082] In addition, multimodal data can also include other forms of information, such as user behavior data and geographic location data. These multimodal data typically have different characteristics and structures, thus requiring specific technologies and methods for processing and analysis.

[0083] Among them, structured data summaries can be obtained by summarizing and refining structured data, presenting the key information in the data in a concise and clear manner.

[0084] For example, in structured data summaries, key information typically includes the data's attributes, tuples, and the relationships between them. For instance, in structured data like a transcript, attributes might include subject names, scores, etc., while tuples correspond to each student's specific score. Structured data summaries may present this information in concise tables, charts, or text to allow users to quickly understand the overall picture and key details of the data.

[0085] It is important to note that structured data summaries should accurately reflect the key information of the original data, avoiding omissions or misleading information.

[0086] For example, data cleaning and integration can be a process of screening the quality of structured data summaries. The reason for cleaning and integrating data is that the standards of the acquired text data or text information are different, the descriptions are inconsistent, and there is a problem of low overall quality. Therefore, data cleaning is necessary.

[0087] For example, during the data cleaning process, there may be data classification errors or overlaps; non-standard data descriptions, incorrect attribute entries, or missing key attributes; data descriptions may contain special characters such as "-", "—", " / ", "*", and "x", which are not limited in the embodiments of this application.

[0088] For example, the quality requirements for the data input into the multimodal model need to meet preset quality requirements, such as: the data cannot contain some special characters, or the data cannot contain extra spaces, etc., but this is not limited in the embodiments of this application.

[0089] For example, the target text can be text that meets the filtering criteria, that is, the data format of the target text is uniform and the data quality meets the preset quality requirements, which is not limited in the embodiments of this application.

[0090] In this step, during the acquisition of target text, it is necessary to unify the format of multimodal data with inconsistent formats, and then clean and integrate the data after format unification to obtain target text that meets the preset quality requirements.

[0091] Furthermore, in the embodiments of this application, after obtaining the target text, the cataloging system can also determine the storage method based on the characteristics and scene information carried by the target text, and store the target text based on the storage method.

[0092] In this step, the features carried by the target text can be structured features, unstructured features, or semi-structured features, and are not limited in the embodiments of this application.

[0093] In this step, the scene information can be any different scene information, such as data information in a traffic scene, data information in a library, etc., and is not limited in the embodiments of this application.

[0094] In this step, the storage method can be to store different target texts using the same storage method or to store them using different storage methods; this is not limited in the embodiments of this application.

[0095] In this step, storing the target text can be done by storing the target text in a structured manner, which is not limited in the embodiments of this application.

[0096] In this step, during the storage process, the target text can be stored using the corresponding storage method based on the characteristics and scene information carried by the target text.

[0097] Step 102: Determine the text sample based on the first text and the directory data.

[0098] The cataloging method provided in this application embodiment allows the cataloging system to determine a text sample based on the first text and directory data after obtaining the target text.

[0099] In this step, the text sample can be determined based on the first text and the directory data. For example, the text sample is determined by multiple first texts and multiple directory data, which is not limited in the embodiments of this application.

[0100] In this step, during the cataloging process, the cataloging system can first determine the text sample based on the first text and directory data in the acquired target text.

[0101] Step 103: Based on the mapping function, map the text sample and the second text to the matrix set corresponding to the third text and the mask.

[0102] The cataloging method provided in this application, after determining the text sample based on the first text and the catalog data, allows the cataloging system to map the text sample and the second text to a matrix set corresponding to the third text and the mask based on a mapping function.

[0103] In this step, the mapping function can be a defined function that can perform mapping processing, and is not limited in the embodiments of this application.

[0104] In this step, the third text can be a hint for extracting directory data, for example, it can be the text in the matrix set obtained by mapping the text sample and the second text through a mapping function.

[0105] In this step, the mask can be used to indicate the part that the model should predict, and is not limited in the embodiments of this application.

[0106] In this step, the matrix set corresponding to the third text and the mask can be obtained by mapping the text sample and the second text through a mapping function. This matrix set can include multiple rows and columns, and is not limited in the embodiments of this application.

[0107] In this step, during the cataloging process, the cataloging system can also use a mapping function to map text samples and second text to a matrix set corresponding to third text and mask. This matrix set can include multi-row and multi-column data, and the multi-row and multi-column data is the processed data.

[0108] Step 104: Construct the parallel structure information group, which consists of the task example and the fourth text.

[0109] The cataloging method provided in this application, after mapping text samples and second texts to a matrix set corresponding to third texts and masks based on a mapping function, can also construct a parallel structure information group, which consists of task examples and fourth texts.

[0110] In this step, the parallel structure information group can be obtained by combining two types of data information, which can improve the data processing speed.

[0111] In this step, the task example can be a pre-defined example of the cataloging system. Depending on the different task examples, the target directory obtained during the cataloging process may be different, which is not limited in the embodiments of this application.

[0112] In this step, the fourth text can be the text to be predicted. That is, the fourth text and the task example can be input into the multimodal model to obtain the target directory corresponding to the fourth text.

[0113] Furthermore, in this step, when constructing the parallel structure information group, the cataloging system can add a first tag to the task example and a second tag to the fourth text; and train the parallel structure information group based on the first tag and the second tag.

[0114] In this step, the first label can be a tag added to the sample example. The multimodal model can distinguish the sample example from the text to be predicted based on the first label.

[0115] It should be noted that the second tag can be a label added to the fourth text. The multimodal model can distinguish between the sample examples and the fourth text based on the second tag.

[0116] In this step, during the construction of the parallel structure information group, the cataloging system can label the task examples and the fourth text respectively, so that they can be distinguished according to the labels during the training process.

[0117] Step 105: Input the parallel structure information group into the multimodal model to predict the directory and determine the target directory corresponding to the fourth text.

[0118] The cataloging method provided in this application embodiment, after constructing the parallel structure information group, the cataloging system can also input the parallel structure information group into a multimodal model for directory prediction to determine the target directory corresponding to the fourth text.

[0119] In this step, the multimodal model may include a model that uses a pre-trained model based on Transformer and Conditional Random Fields (CRF) to identify catalog data, but this is not limited in the embodiments of this application.

[0120] In this step, directory prediction can be the process of predicting the text to be predicted.

[0121] The target directory corresponding to the fourth text can be the target directory obtained after predicting the directory of the fourth text using a multimodal model.

[0122] In this step, during the cataloging process, the text to be predicted and the task example are simultaneously input into the multimodal model to obtain the target directory of the text to be predicted.

[0123] Furthermore, during the cataloging process, the cataloging system can also extract the probabilities of each feature node corresponding to the first text through a linear discriminant analysis algorithm; calculate the similarity of each feature node; obtain a tree structure based on the similarity through a hierarchical clustering algorithm, and segment the tree structure; determine the community structure corresponding to the first text based on the segmented tree structure; and determine the output text corresponding to the first text based on the community structure.

[0124] In this step, the implementation process of the linear discriminant analysis algorithm may include: calculating the mean vector of each class, calculating the global mean vector, calculating the inter-class scatter matrix and the intra-class scatter matrix, solving the generalized eigenvalue problem, and projecting the data into a low-dimensional space, etc., which are not limited in the embodiments of this application.

[0125] For example, the process of calculating the mean vector for each category may include, for each category, calculating the mean vector of all its samples, which represents the center position of that category in the feature space.

[0126] For example, the process of calculating the global mean vector may include calculating the global mean vector of all samples, which represents the central location of the entire dataset in the feature space.

[0127] For example, the inter-class scatter matrix reflects the degree of dispersion between different classes, while the intra-class scatter matrix reflects the degree of dispersion of samples within the same class.

[0128] For example, solving the generalized eigenvalue problem can include finding the optimal projection direction by solving the generalized eigenvalue problems of the between-class scatter matrix and the within-class scatter matrix. This direction maximizes the ratio of the between-class distance to the within-class distance, thereby achieving the best classification performance.

[0129] For example, the process of projecting data into a low-dimensional space may include projecting the original data into a low-dimensional space consisting of the optimal projection direction to obtain the dimensionality-reduced data.

[0130] In this step, in determining the probability of each feature node, we can first determine each feature node, and then determine the corresponding probability of each feature node.

[0131] For example, the probability of each feature node can be any value greater than 0. For example, the probability of a feature node can be 0.8 or 0.6, and is not limited in the embodiments of this application.

[0132] For example, each feature node can be "data governance", "higher education", "improving education quality", etc., and is not limited in the embodiments of this application.

[0133] In this step, the similarity of each feature node can be used to determine the similarity of features corresponding to different levels of directories.

[0134] For example, in the embodiments of this application, the similarity between the feature nodes “Application trend of big data in higher education (from 5.1)” and “Challenges and problems faced (from 5.2)” is 0.2; or the similarity between the feature nodes “Acquisition and extraction (from 3.2.1)” and “Integration and analysis (from 3.2.2)” is 0.6, which is not limited in the embodiments of this application.

[0135] It's important to note that the basic idea behind hierarchical clustering algorithms is to construct a hierarchical structure for the dataset. Each sample is initially represented as a separate cluster, and then clusters are gradually merged into larger clusters (agglomerative clustering) or split into smaller clusters starting from the entire dataset (split clustering). This entire process can be represented as a tree structure, called a clustering tree or dendrogram.

[0136] For example, the tree structure can be a tree structure formed based on multiple node pairs.

[0137] In this step, the community structure corresponding to the first text can be a community structure determined based on multiple node pairs.

[0138] For example, based on multiple node pairs, wherein the node pairs include: "acquisition and extraction; integration and analysis"; "improvement of education quality; improvement of decision-making scientificity"; "enhancing education quality; enhancing decision-making scientificity"; the community structure determined based on the above node pairs may include: "on the importance of data governance (including the relevant content under 3.1)"; "on the operation of data governance (including the relevant content under 3.2)"; "on the future of higher education data governance (including the relevant content under 5.1, 5.2, and 5.3)", etc., which are not limited in the embodiments of this application.

[0139] In this step, the output text corresponding to the first text can be a defined target directory. For example, the output text can be "First-level directory: Inventions on the Importance of Data Governance; Second-level directory: Needs to Improve Education Quality, Needs to Improve the Scientific Nature of Decision Making, Needs to Improve Management Efficiency," etc., which are not limited in the embodiments of this application.

[0140] In this step, based on the cleaned catalog, the cataloging system can also perform final-level classification, which is the finest-grained classification of the document catalog data, for subsequent catalog data management and querying.

[0141] Furthermore, during the cataloging process, the cataloging system can also input the matrix set into the multimodal model for directory prediction during the model training phase to obtain the output text corresponding to the third text; based on the output text, the target directory corresponding to the mask is extracted.

[0142] For example, a cataloging system can extract cue text from target data and determine the output text corresponding to a third text without performing any additional training.

[0143] For example, the target directory corresponding to the mask can be the target directory determined based on the output text corresponding to the third text.

[0144] In this step, the cataloging system can directly input the matrix set into the multimodal model for directory prediction, first obtaining the output text corresponding to the third text; then, based on the output text, extracting the target directory corresponding to the mask.

[0145] In summary, in the embodiments of this application, the cataloging system can obtain target text; the target text includes first text, directory data, and second text; based on the first text and directory data, text samples are determined; based on a mapping function, the text samples and second text are mapped to a matrix set corresponding to the third text and a mask; a parallel structure information group is constructed, which consists of a task example and a fourth text; the parallel structure information group is input into a multimodal model for directory prediction to determine the target directory corresponding to the fourth text. Thus, by first determining the text samples using the first text and directory data in the target text, and then mapping the text samples and second text, a mapped matrix set can be obtained. This matrix set consists of the third text and a mask. Simultaneously, the cataloging system can construct a parallel structure information group containing the task example and the fourth text, that is, inputting the task example and the fourth text into the multimodal model. This not only obtains a more accurate target directory but also improves cataloging efficiency.

[0146] Figure 2 This is a flowchart of a cataloging method provided in an embodiment of the present invention. Figure 2 ,like Figure 2 As shown, the method may include:

[0147] Step 201: Obtain multimodal data information.

[0148] During the cataloging process, the cataloging system can first acquire multimodal data information, which can be information from multiple data sources. For example, multimodal data can include text data, image data, audio data, video data, sensor data, biometric data, etc., which are not limited in the embodiments of this application.

[0149] Step 202: Convert the multimodal data information into a structured data summary.

[0150] The cataloging method provided in this application allows the cataloging system to convert multimodal data information into structured data summaries after acquiring multimodal data information.

[0151] For example, in structured data summaries, key information typically includes the data's attributes, tuples, and the relationships between them. For instance, in structured data like a transcript, attributes might include subject names, scores, etc., while tuples correspond to each student's specific score. Structured data summaries may present this information in concise tables, charts, or text to allow users to quickly understand the overall picture and key details of the data.

[0152] It is important to note that structured data summaries should accurately reflect the key information of the original data, avoiding omissions or misleading information.

[0153] Step 203: Clean and integrate the structured data summary to obtain the target text that meets the preset quality requirements; the target text includes the first text, the directory data, and the second text.

[0154] The cataloging method provided in this application, after converting multimodal data information into a structured data summary, allows the cataloging system to perform data cleaning and integration on the structured data summary to obtain target text that meets preset quality requirements; the target text includes a first text, catalog data, and a second text.

[0155] For example, data cleaning and integration can be a process of screening the quality of structured data summaries. The reason for cleaning and integrating data is that the standards of the acquired text data or text information are different, the descriptions are inconsistent, and there is a problem of low overall quality. Therefore, data cleaning is necessary.

[0156] For example, the quality requirements for the data input into the multimodal model need to meet preset quality requirements, such as: the data cannot contain some special characters, or the data cannot contain extra spaces, etc., but this is not limited in the embodiments of this application.

[0157] For example, the target text can be text that meets the filtering criteria, that is, the data format of the target text is uniform and the data quality meets the preset quality requirements, which is not limited in the embodiments of this application.

[0158] Step 204: Determine the text sample based on the first text and the directory data.

[0159] The cataloging method provided in this application, after cleaning and integrating the structured data summary to obtain the target text that meets the preset quality requirements, allows the cataloging system to determine the text sample based on the first text and the directory data.

[0160] In this step, the text sample can be determined based on the first text and the directory data. For example, the text sample is determined by multiple first texts and multiple directory data, which is not limited in the embodiments of this application.

[0161] In this step, during the cataloging process, the cataloging system can first determine the text sample based on the first text and directory data in the acquired target text.

[0162] Step 205: Based on the mapping function, map the text sample and the second text to the matrix set corresponding to the third text and the mask.

[0163] The cataloging method provided in this application, after determining the text sample based on the first text and the catalog data, allows the cataloging system to map the text sample and the second text to a matrix set corresponding to the third text and the mask based on a mapping function.

[0164] In this step, the mapping function can be a defined function that can perform mapping processing, and is not limited in the embodiments of this application.

[0165] In this step, the third text can be a hint for extracting directory data, for example, it can be the text in the matrix set obtained by mapping the text sample and the second text through a mapping function.

[0166] In this step, the mask can be used to indicate the part that the model should predict, and is not limited in the embodiments of this application.

[0167] In this step, the matrix set corresponding to the third text and the mask can be obtained by mapping the text sample and the second text through a mapping function. This matrix set can include multiple rows and columns, and is not limited in the embodiments of this application.

[0168] In this step, during the cataloging process, the cataloging system can also use a mapping function to map text samples and second text to a matrix set corresponding to third text and mask. This matrix set can include multi-row and multi-column data, and the multi-row and multi-column data is the processed data.

[0169] Step 206: Add a first tag to the task example and a second tag to the fourth text.

[0170] The cataloging method provided in this application, after mapping text samples and second texts to a matrix set corresponding to third texts and masks based on a mapping function, allows the cataloging system to add a first tag to the task sample and a second tag to the fourth text.

[0171] In this step, the first label can be a tag added to the sample example. The multimodal model can distinguish the sample example from the text to be predicted based on the first label.

[0172] It should be noted that the second tag can be a label added to the fourth text. The multimodal model can distinguish between the sample examples and the fourth text based on the second tag.

[0173] Step 207: Train the parallel structure information group according to the first label and the second label.

[0174] The cataloging method provided in this application allows the cataloging system to train on the parallel structure information group based on the first and second tags after adding a first tag to the task example and a second tag to the fourth text.

[0175] In this step, during the construction of the parallel structure information group, the cataloging system can label the task examples and the fourth text respectively, so that they can be distinguished according to the labels during the training process.

[0176] Step 208: Input the parallel structure information group into the multimodal model to predict the directory and determine the target directory corresponding to the fourth text.

[0177] The cataloging method provided in this application, after training the parallel structure information group according to the first and second tags, allows the cataloging system to input the parallel structure information group into a multimodal model for directory prediction, thereby determining the target directory corresponding to the fourth text.

[0178] In this step, the multimodal model may include a model that uses a pre-trained model based on Transformer and Conditional Random Fields (CRF) to identify catalog data, but this is not limited in the embodiments of this application.

[0179] In this step, directory prediction can be the process of predicting the text to be predicted.

[0180] The target directory corresponding to the fourth text can be the target directory obtained after predicting the directory of the fourth text using a multimodal model.

[0181] In this step, during the cataloging process, the text to be predicted and the task example are simultaneously input into the multimodal model to obtain the target directory of the text to be predicted.

[0182] Furthermore, during the cataloging process, the cataloging system can also extract the probabilities of each feature node corresponding to the first text through a linear discriminant analysis algorithm; calculate the similarity of each feature node; obtain a tree structure based on the similarity through a hierarchical clustering algorithm, and segment the tree structure; determine the community structure corresponding to the first text based on the segmented tree structure; and determine the output text corresponding to the first text based on the community structure.

[0183] In this step, the implementation process of the linear discriminant analysis algorithm may include: calculating the mean vector of each class, calculating the global mean vector, calculating the inter-class scatter matrix and the intra-class scatter matrix, solving the generalized eigenvalue problem, and projecting the data into a low-dimensional space, etc., which are not limited in the embodiments of this application.

[0184] For example, the process of calculating the mean vector for each category may include, for each category, calculating the mean vector of all its samples, which represents the center position of that category in the feature space.

[0185] For example, the process of calculating the global mean vector may include calculating the global mean vector of all samples, which represents the central location of the entire dataset in the feature space.

[0186] For example, the inter-class scatter matrix reflects the degree of dispersion between different classes, while the intra-class scatter matrix reflects the degree of dispersion of samples within the same class.

[0187] For example, solving the generalized eigenvalue problem can include finding the optimal projection direction by solving the generalized eigenvalue problems of the between-class scatter matrix and the within-class scatter matrix. This direction maximizes the ratio of the between-class distance to the within-class distance, thereby achieving the best classification performance.

[0188] For example, the process of projecting data into a low-dimensional space may include projecting the original data into a low-dimensional space consisting of the optimal projection direction to obtain the dimensionality-reduced data.

[0189] In this step, in determining the probability of each feature node, we can first determine each feature node, and then determine the corresponding probability of each feature node.

[0190] For example, the probability of each feature node can be any value greater than 0. For example, the probability of a feature node can be 0.8 or 0.6, and is not limited in the embodiments of this application.

[0191] For example, each feature node can be "data governance", "higher education", "improving education quality", etc., and is not limited in the embodiments of this application.

[0192] In this step, the similarity of each feature node can be used to determine the similarity of features corresponding to different levels of directories.

[0193] For example, in the embodiments of this application, the similarity between the feature nodes “Application trend of big data in higher education (from 5.1)” and “Challenges and problems faced (from 5.2)” is 0.2; or the similarity between the feature nodes “Acquisition and extraction (from 3.2.1)” and “Integration and analysis (from 3.2.2)” is 0.6, which is not limited in the embodiments of this application.

[0194] It's important to note that the basic idea behind hierarchical clustering algorithms is to construct a hierarchical structure for the dataset. Each sample is initially represented as a separate cluster, and then clusters are gradually merged into larger clusters (agglomerative clustering) or split into smaller clusters starting from the entire dataset (split clustering). This entire process can be represented as a tree structure, called a clustering tree or dendrogram.

[0195] For example, the tree structure can be a tree structure formed based on multiple node pairs.

[0196] In this step, the community structure corresponding to the first text can be a community structure determined based on multiple node pairs.

[0197] For example, based on multiple node pairs, wherein the node pairs include: "acquisition and extraction; integration and analysis"; "improvement of education quality; improvement of decision-making scientificity"; "enhancing education quality; enhancing decision-making scientificity"; the community structure determined based on the above node pairs may include: "on the importance of data governance (including the relevant content under 3.1)"; "on the operation of data governance (including the relevant content under 3.2)"; "on the future of higher education data governance (including the relevant content under 5.1, 5.2, and 5.3)", etc., which are not limited in the embodiments of this application.

[0198] In this step, the output text corresponding to the first text can be a defined target directory. For example, the output text can be "First-level directory: Inventions on the Importance of Data Governance; Second-level directory: Needs to Improve Education Quality, Needs to Improve the Scientific Nature of Decision Making, Needs to Improve Management Efficiency," etc., which are not limited in the embodiments of this application.

[0199] In this step, based on the cleaned catalog, the cataloging system can also perform final-level classification, which is the finest-grained classification of the document catalog data, for subsequent catalog data management and querying.

[0200] Furthermore, during the cataloging process, the cataloging system can also input the matrix set into the multimodal model for directory prediction during the model training phase to obtain the output text corresponding to the third text; based on the output text, the target directory corresponding to the mask is extracted.

[0201] For example, a cataloging system can extract cue text from target data and determine the output text corresponding to a third text without performing any additional training.

[0202] For example, the target directory corresponding to the mask can be the target directory determined based on the output text corresponding to the third text.

[0203] In this step, the cataloging system can directly input the matrix set into the multimodal model for directory prediction, first obtaining the output text corresponding to the third text; then, based on the output text, extracting the target directory corresponding to the mask.

[0204] In summary, in the embodiments of this application, the cataloging system can obtain target text; the target text includes first text, directory data, and second text; based on the first text and directory data, text samples are determined; based on a mapping function, the text samples and second text are mapped to a matrix set corresponding to the third text and a mask; a parallel structure information group is constructed, which consists of a task example and a fourth text; the parallel structure information group is input into a multimodal model for directory prediction to determine the target directory corresponding to the fourth text. Thus, by first determining the text samples using the first text and directory data in the target text, and then mapping the text samples and second text, a mapped matrix set can be obtained. This matrix set consists of the third text and a mask. Simultaneously, the cataloging system can construct a parallel structure information group containing the task example and the fourth text, that is, inputting the task example and the fourth text into the multimodal model. This not only obtains a more accurate target directory but also improves cataloging efficiency.

[0205] Reference Figure 3 The flowchart illustrates the steps of a cataloging method provided in an embodiment of this application. Figure 3 ,like Figure 3 As shown, the method specifically includes steps 301 to 306:

[0206] Step 301, Multimodal data acquisition module.

[0207] The multimodal data acquisition module can collect multimodal data. The system is designed to support access from multiple data sources, is based on a distributed architecture, and employs multi-threaded parallel technology to simultaneously process unstructured data from multiple sources, enabling connections and data capture from different data sources.

[0208] Step 302, Multimodal data storage module.

[0209] The multimodal data storage module focuses on data reliability and accessibility. For multimodal data, an appropriate storage method is selected based on its characteristics and application scenarios. A distributed storage architecture is adopted, storing unstructured data in a scalable manner across multiple nodes, ensuring data security and reliability, and facilitating data access and retrieval by the model.

[0210] Step 303, Multimodal Data Format Unification Module.

[0211] Multimodal data is formatted using semantic understanding. This module extracts information from unstructured or semi-structured multimodal data and transforms it into structured data summaries, which are then produced in a unified text format.

[0212] Step 304: Multimodal data cleaning and integration module.

[0213] The multimodal data acquisition module collected various data and information, but the results suffered from inconsistent data standards, non-standardized descriptions, and low overall quality, necessitating data cleaning. The challenges of data cleaning include: incorrect or overlapping data classification; non-standard data descriptions, incorrect attribute entries, or missing key attributes; and data descriptions containing special characters such as "-", "—", " / ", "*", and "x".

[0214] The research team developed a multimodal data cleaning tool using Natural Language Processing (NLP) technology. This tool intelligently cleans large amounts of data, specifically addressing the challenges of multimodal data cleaning. The tool offers two main functions: intelligent classification identification: it automatically identifies the category using an algorithmic catalog extraction function, and can manually supplement any categories that are not recommended; and intelligent key parameter identification: it identifies key parameters of the multimodal data using algorithms to generate structured data descriptions.

[0215] By using multimodal data cleaning tools, historical data was cleaned, reclassified according to new categories, and data descriptions were identified and processed according to algorithms, thus solving problems such as inconsistent order, special characters, and non-standard descriptions.

[0216] Step 305: Multimodal data catalog generation module.

[0217] Taking the multimodal (Modulated Detection for End-to-End Multi-ModalUnderstanding, MDETR) model as an example, the concept of parallel structure information group is introduced as a prompt engineering, thereby enabling the execution of multiple user-specified tasks without any weight fine-tuning. This method is applicable to all self-supervised learning baseline models.

[0218] MDETR is an NLP-based model for multimodal object detection and segmentation, designed to parse and understand the layout structure in text. It combines traditional text classification and layout analysis techniques to jointly model and predict text and layout information.

[0219] The MDETR model learns cross-modal representations through a unified text multimodal Transformer, trains the model using multimodal contrastive learning, achieves information transformation through multi-head self-attention and feedforward neural network layers, and combines residual connections and normalization to improve training performance.

[0220] The model preparation phase of MDETR can be divided into two main steps: pre-training. In the pre-training phase, the model is trained on large-scale unlabeled text and layout data to learn the correlation between text and layout. It employs a multi-task learning framework that simultaneously predicts text classification and layout structure to capture the complex relationships between text and layout.

[0221] Step 306: Directory data output module.

[0222] After the directory data is extracted and cleaned, a directory data catalog is finally formed. NLP technology is used to extract the directory data, identify entities in the document directory data, remove decorative content, and obtain the final level classification, which is also the base name of the document.

[0223] During the directory extraction phase, information from each level of the directory is processed. vp1 y vp2 y vp3 ...A hybrid model, HybridModels, is used to achieve directory data recognition based on a pre-trained Transformer model combined with Conditional Random Fields (CRF). The Transformer and CRF combine the advantages of multiple technologies. The Transformer learns high-quality language representations through large-scale pre-training, and then automatically extracts features from documents through fine-tuning. A CRF layer is added to model and process the features extracted by the Transformer, particularly handling the transition relationships between features, which improves the accuracy of directory data recognition. The identified entities are cleaned, such as personnel, channels, and methods, removing decorative content such as redundant descriptive text and symbols. Based on the cleaned directory data, a final-level classification is performed—the finest-grained classification of the document directory data—for subsequent directory data management and retrieval.

[0224] Reference Figure 4 The flowchart illustrates the steps of a directory data extraction method provided in an embodiment of this application, as shown below. Figure 4 As shown, the method specifically includes steps 401 to 404:

[0225] Step 401: Input the target text.

[0226] The input data consists of three parts: x i Extracted text, and y i The extracted directory data is used for model learning, x q The corresponding text to be predicted.

[0227] Step 402: Construct a directory data extraction Prompt.

[0228] Define input / output examples Here are some examples of input and output text extraction, where x i It is the input text, y i This is extracted directory data. Given a new text x to be predicted. q The goal is to predict the corresponding y. q .

[0229] Define a function F that takes an input sample S and the text to be predicted x as input. q Mapped to new text and mask: [x vp ,m]=F(S,x q ), where x vp This is a hint for extracting directory data; the mask 'm' defines the part that the model should predict.

[0230] The goal of the model is to respond to the prompt x vp Output y without performing any additional training vp And extract the corresponding part to get y q .

[0231] Step 403: Construct a parallel structure information matrix.

[0232] Construct a parallel structure information group consisting of task input / output examples and new text to be predicted. The task input / output examples and the new text to be predicted are labeled to facilitate model differentiation.

[0233] Step 404: Input the data into the model for catalog prediction.

[0234] Inputting a set of parallel structure information into the MDETR model, the model observes x i and x vp Based on its knowledge and language comprehension capabilities acquired during pre-training, the model will attempt to understand the prompt and generate corresponding results. The model identifies and generates catalog data extraction prompts. vp Then according to x vp Output y vp1 y vp2 y vp3 ..., and extract the part y corresponding to the mask m. q .

[0235] Reference Figure 5 This document illustrates a flowchart of the steps involved in constructing a directory data extraction prompt according to an embodiment of this application. Figure 5 As shown, the method specifically includes steps 501 to 504:

[0236] Step 501: Define input and output samples S.

[0237] Define input / output examples Here are some examples of input and output text extraction, where x i It is the input text, y i This is extracted directory data. Given a new text x to be predicted. q The goal is to predict the corresponding y. q .

[0238] Step 502: Given the text to be predicted x q Output the prediction result y q .

[0239] Cataloging systems can use models to generate output catalog prediction results based on the input text to be predicted.

[0240] Step 503: Define function F, input / output sample S and text to be predicted x. q Mapped to new text and mask.

[0241] Define a function F that takes an input sample S and the text to be predicted x as input. q Mapped to new text and mask: [x vp ,m]=F(S,x q ), where x vp This is a hint for extracting directory data; the mask 'm' defines the part that the model should predict.

[0242] The goal of the model is to respond to the prompt x vp Output y without performing any additional training vp And extract the corresponding part to get y q .

[0243] Step 504: Extract directory data and construct the Prompt.

[0244] Reference Figure 6 The flowchart illustrates the steps of a cataloging method provided in an embodiment of this application. Figure 4 ,like Figure 6 As shown, the method specifically includes steps 601 to 605:

[0245] Step 601: Input sample.

[0246] Based on the cleaned directory, a final-level classification is performed, which is also the finest-grained classification of the document directory data, for subsequent directory data management and querying.

[0247] The cataloging system allows you to input samples first, which will be used for subsequent processing.

[0248] Step 602: Extract the probability corresponding to the feature node using the LDA algorithm.

[0249] After inputting samples, the cataloging system can also extract the probabilities corresponding to feature nodes using the LDA algorithm.

[0250] Step 603: Calculate the similarity of feature nodes.

[0251] After extracting the probabilities corresponding to feature nodes using the LDA algorithm, the cataloging system can also calculate the similarity of feature nodes.

[0252] Step 604: Form a tree structure and divide it using a hierarchical clustering algorithm to obtain the community structure.

[0253] After calculating the similarity of feature nodes, the cataloging system can also form a tree structure and segment it through hierarchical clustering algorithms to obtain the community structure.

[0254] Step 605: Determine the output sample based on the community structure.

[0255] After forming a tree structure and dividing it using a hierarchical clustering algorithm to obtain the community structure, the cataloging system can also determine the output samples based on the community structure.

[0256] Figure 7 This is a schematic diagram of the structure of a cataloging device provided in an embodiment of the present invention, as shown below. Figure 7 As shown, the device may specifically include:

[0257] The acquisition module 701 is used to acquire target text; the target text includes first text, directory data, and second text;

[0258] The determining module 702 is used to determine a text sample based on the first text and the directory data;

[0259] The mapping module 703 is used to map the text sample and the second text to a set of matrices corresponding to the third text and the mask based on a mapping function;

[0260] Module 704 is used to construct a parallel structure information group, which consists of a task example and a fourth text.

[0261] The determining module is further configured to input the parallel structure information group into the multimodal model for directory prediction, and determine the target directory corresponding to the fourth text.

[0262] This invention provides a cataloging device that acquires target text. The target text includes first text, directory data, and second text. Based on the first text and directory data, text samples are determined. Using a mapping function, the text samples and second text are mapped to a matrix set corresponding to a third text and a mask. A parallel structure information group is constructed, consisting of a task example and a fourth text. The parallel structure information group is input into a multimodal model for directory prediction to determine the target directory corresponding to the fourth text. In this way, by first determining the text samples using the first text and directory data in the target text, and then mapping the text samples and second text, a mapped matrix set can be obtained. This matrix set consists of the third text and a mask. Simultaneously, the cataloging system can construct a parallel structure information group containing the task example and the fourth text. By inputting the task example and the fourth text into the multimodal model, more accurate target directories can be obtained, while also improving cataloging efficiency.

[0263] Optionally, the acquisition module 701 includes:

[0264] The first acquisition submodule is used to acquire multimodal data information;

[0265] The conversion module is used to convert the multimodal data information into a structured data digest;

[0266] The module is used to clean and integrate the structured data summary to obtain target text that meets preset quality requirements.

[0267] Optionally, the device further includes:

[0268] Add a module to add a first marker to the task example and a second marker to the fourth text;

[0269] The training module is used to train the parallel structure information group based on the first label and the second label.

[0270] The extraction module is used to extract the probabilities of each feature node corresponding to the first text through a linear discriminant analysis algorithm.

[0271] A calculation module is used to calculate the similarity of each feature node;

[0272] The segmentation module is used to obtain a tree structure based on the similarity using a hierarchical clustering algorithm, and to segment the tree structure.

[0273] The determination submodule is used to determine the community structure corresponding to the first text based on the segmented tree structure; and to determine the output text corresponding to the first text based on the community structure.

[0274] Optionally, the device further includes:

[0275] A submodule is obtained, which is used to input the matrix set into the multimodal model for directory prediction during the model training phase, and obtain the output text corresponding to the third text;

[0276] The extraction submodule is used to extract the target directory corresponding to the mask based on the output text.

[0277] Optionally, the device further includes:

[0278] The first determining module is used to determine the storage method based on the features and scene information carried by the target text, and to store the target text based on the storage method.

[0279] The present invention also provides an electronic device, see [link to relevant documentation]. Figure 8 It includes: a processor 801, a memory 802, and a computer program 8021 stored in the memory and executable on the processor, wherein the processor executes the program to implement the cataloging method of the foregoing embodiments.

[0280] The present invention also provides a readable storage medium, wherein when the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is able to perform the cataloging method of the foregoing embodiments.

[0281] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0282] The algorithms and cataloging provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, this invention is not directed to any particular programming language. It should be understood that the contents of the invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the invention.

[0283] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0284] Similarly, it should be understood that, in order to simplify the invention and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof. However, this disclosure should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the following claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into this detailed description, wherein each claim itself is a separate embodiment of the invention.

[0285] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0286] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components in the sorting device according to the present invention. The present invention can also be implemented as a device or apparatus program for performing part or all of the methods described herein. Such a program implementing the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0287] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0288] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0289] It should be noted that all actions involving the acquisition of signals, information, or data in this application are carried out in compliance with the relevant data protection laws and policies of the country where the application is located, and with the authorization granted by the owner of the relevant device.

[0290] 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 scope of protection of the present invention.

[0291] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A cataloging method, characterized in that, Applied to a cataloging system, the method includes: Obtain the target text; the target text includes a first text, directory data, and a second text; the first text is the extracted text; the second text is the text to be predicted; Based on the first text and the directory data, determine the text sample; Based on the mapping function, the text sample and the second text are mapped to a set of matrices corresponding to the third text and the mask; Construct a parallel structure information group, which consists of a task example and a fourth text; The parallel structure information group is input into a multimodal model for directory prediction to determine the target directory corresponding to the fourth text.

2. The method according to claim 1, characterized in that, The acquisition of the target text includes: Acquire multimodal data information; The multimodal data information is converted into a structured data summary; The structured data summary is cleaned and integrated to obtain target text that meets preset quality requirements.

3. The method according to claim 1, characterized in that, The constructed parallel structure information group includes: Add a first tag to the task example and a second tag to the fourth text; The parallel structure information group is trained based on the first label and the second label.

4. The method according to claim 1, characterized in that, The method further includes: The probability of each feature node corresponding to the first text is extracted using a linear discriminant analysis algorithm. Calculate the similarity of each feature node; A tree structure is obtained based on the similarity using a hierarchical clustering algorithm, and then the tree structure is segmented. Based on the segmented tree structure, the community structure corresponding to the first text is determined; Based on the community structure, determine the output text corresponding to the first text.

5. The method according to claim 1, characterized in that, The method further includes: During the model training phase, the matrix set is input into the multimodal model for directory prediction to obtain the output text corresponding to the third text; Based on the output text, the target directory corresponding to the mask is extracted.

6. The method according to claim 1, characterized in that, After obtaining the target text, the method further includes: Based on the features and scene information carried by the target text, a storage method is determined, and the target text is stored based on the storage method.

7. A cataloging device, characterized in that, The device includes: The acquisition module is used to acquire target text; the target text includes first text, directory data, and second text; the first text is the extracted text; the second text is the text to be predicted. The determining module is used to determine a text sample based on the first text and the directory data; The mapping module is used to map the text sample and the second text to a set of matrices corresponding to the third text and the mask based on a mapping function; A construction module is used to construct a parallel structure information group, which consists of a task example and a fourth text. The determining module is further configured to input the parallel structure information group into the multimodal model for directory prediction, and determine the target directory corresponding to the fourth text.

8. The apparatus according to claim 7, characterized in that, The acquisition module includes: The first acquisition submodule is used to acquire multimodal data information; The conversion module is used to convert the multimodal data information into a structured data digest; The module is used to clean and integrate the structured data summary to obtain target text that meets preset quality requirements.

9. An electronic device, characterized in that, include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the cataloging method as described in any one of claims 1-6.

10. A readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device is able to perform the cataloging method according to any one of claims 1-6.

Images