Text problem processing method and device, equipment and medium

Abstract

The invention relates to a text problem processing method and device, equipment and a medium. Obtaining target description information of a target problem in the to-be-processed text; performing feature extraction processing on the target description information based on a feature extraction network to obtain text features of the target description information; processing the text features based on a problem processing network in a pre-trained large language model, and determining initial qualitative information, initial classification information and initial grading information of the target problem; and processing the initial qualitative information, the initial classification information and the initial grading information by utilizing a problem rule model of a reference text corresponding to the to-be-processed text, and sequentially determining target qualitative information, target classification information and target grading information of the target problem. Therefore, automatic qualification, classification and grading of the problems in the text are achieved, the problem processing efficiency and accuracy are improved, and then the management effect of the problems recorded in the text is improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of text processing, and in particular to a method, device, equipment and medium for processing text problems. Background Art

[0002] Texts are usually used to record a variety of issues, so the issues recorded in the text are usually characterized, classified, and graded to achieve management of the issues recorded in the text. For example, audit texts record a variety of audit issues, and the audit issues in the audit texts are characterized, classified, and graded to achieve effective management of the audit issues recorded in the audit texts.

[0003] At present, the work process usually relies on the initial problem finder to manually characterize, classify and grade the problems recorded in the text. However, this problem handling method depends on the experience, ability and attitude of the initial problem finder. The initial problem finder cannot accurately characterize, classify and grade the problems, and the standards are not unified. It often requires high-level and experienced personnel to review, check and confirm. When there are many problems in the text, high-level personnel and the initial problem finder need to communicate, modify, review and confirm repeatedly, resulting in a large workload, long time and low efficiency in the characterization, classification and grading process of text problems. At the same time, there are problems such as manual processing deviation, insufficient depth and inconsistent standards, which will also affect the management effect of the problems recorded in the text. Summary of the invention

[0004] In order to solve the above technical problems, the present disclosure provides a method, device, equipment and medium for processing text problems.

[0005] In a first aspect, the present disclosure provides a method for processing a text problem, comprising:

[0006] Obtain target description information of the target problem in the text to be processed;

[0007] Based on the feature extraction network, feature extraction processing is performed on the target description information to obtain text features of the target description information;

[0008] Based on the question processing network in the pre-trained large language model, the text features are processed to determine the initial qualitative information, initial classification information and initial grading information of the target question;

[0009] The initial qualitative information, the initial classification information and the initial grading information are processed using the question rule model of the reference text corresponding to the text to be processed, and the target qualitative information, target classification information and target grading information of the target question are determined in turn.

[0010] In a second aspect, the present disclosure provides a device for processing text questions, comprising:

[0011] A description information acquisition module is used to obtain target description information of the target problem in the text to be processed;

[0012] A feature extraction module is used to perform feature extraction processing on the target description information based on a feature extraction network to obtain text features of the target description information;

[0013] A first processing module is used to process the text features based on a question processing network in a pre-trained large language model to determine initial qualitative information, initial classification information and initial grading information of the target question;

[0014] The second processing module is used to process the initial qualitative information, the initial classification information and the initial grading information by using the question rule model of the reference text corresponding to the text to be processed, and determine the target qualitative information, target classification information and target grading information of the target question in turn.

[0015] In a third aspect, an embodiment of the present disclosure further provides an electronic device, the device comprising:

[0016] one or more processors;

[0017] a storage device for storing one or more programs,

[0018] When one or more programs are executed by one or more processors, the one or more processors implement the method provided in the first aspect.

[0019] In a fourth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, which implements the method provided in the first aspect when the computer program is executed by a processor.

[0020] Compared with the prior art, the technical solution provided by the embodiments of the present disclosure has the following advantages:

[0021] A method, device, equipment and medium for processing text problems in the disclosed embodiment include: obtaining target description information of a target problem in a to-be-processed text; performing feature extraction processing on the target description information based on a feature extraction network to obtain text features of the target description information; processing the text features based on a problem processing network in a pre-trained large language model to determine the initial qualitative information, initial classification information and initial grading information of the target problem; using the problem rule model of the reference text corresponding to the to-be-processed text, processing the initial qualitative information, initial classification information and initial grading information, and sequentially determining the target qualitative information, target classification information and target grading information of the target problem. Thus, the target description information in the to-be-processed text recording the problem is sequentially subjected to feature extraction, qualitative analysis, classification processing and grading using a neural network model, and then automatically determined the final qualitative information, classification information and grading information of the problem recorded in the text in combination with the problem rule model of the reference text. Therefore, the automatic qualitative, classification and grading of the problems in the text are realized, the efficiency and accuracy of the problem processing in the text are improved, and the management effect of the problems recorded in the text is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0024] Figure 1 A flowchart of a method for processing text problems provided by an embodiment of the present disclosure;

[0025] Figure 2 A flowchart of another method for processing text problems provided by an embodiment of the present disclosure;

[0026] Figure 3 A schematic diagram of the structure of a device for processing text questions provided by an embodiment of the present disclosure;

[0027] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0028] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features in the embodiments can be combined with each other without conflict.

[0029] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments.

[0030] In order to automatically calibrate the problems in the text and improve the calibration efficiency and accuracy, the following is combined Figure 1 to Figure 2 The method for processing text problems provided in the embodiment of the present disclosure is described. In the embodiment of the present disclosure, the method for processing text problems can be executed by an electronic device. The electronic device can include a device with communication functions such as a tablet computer, a desktop computer, and a laptop computer, and can also include a device simulated by a virtual machine or a simulator.

[0031] Figure 1 A flowchart of a method for processing text problems provided by an embodiment of the present disclosure is shown.

[0032] like Figure 1 As shown, the method for processing the text problem may include the following steps.

[0033] S110, obtaining target description information of the target problem in the text to be processed.

[0034] In this embodiment, after acquiring the text to be processed, the electronic device uses optical character recognition and / or natural language processing to process the text to be processed to obtain target description information containing the target problem, or directly uses the problem description information input by the user as the target description information of the target problem. The electronic device then continues to determine the field of the target problem based on the target description information of the target problem, and then processes the target problem within the field to which it belongs.

[0035] The target problem refers to a problem in the field to which the text to be processed belongs, which is recorded in the text to be processed. Optionally, the field to which the target problem belongs includes but is not limited to the procurement field, the sales field, the engineering quality field, and may also be other field types.

[0036] For example, the text to be processed is an audit text, and the target issues include but are not limited to funding issues and equipment and material issues.

[0037] The target description information refers to the detailed description corresponding to the target problem in the text to be processed.

[0038] S120: Based on the feature extraction network, feature extraction processing is performed on the target description information to obtain text features of the target description information.

[0039] In this embodiment, the electronic device inputs the target description information into the feature extraction network to extract text features from the target description information using the neural network model.

[0040] Optionally, the feature extraction network includes but is not limited to a large language model, and may also be other neural network models with feature extraction capabilities.

[0041] Among them, the specific implementation method of S120 includes but is not limited to the following methods: based on the word embedding extraction network in the feature extraction network, the target description information is processed to obtain the word embedding code corresponding to the target description information; based on the semantic feature extraction network in the feature extraction network, the word embedding code is processed to obtain the semantic features of the target description information; based on the local feature extraction network in the feature extraction network, the word embedding code is processed to obtain the local features of the target description information; based on the feature fusion network in the feature extraction network, the semantic features and local features are weightedly fused to obtain the text features of the target description information.

[0042] Specifically, the electronic device first preprocesses the target description information, for example, performs noise processing, error correction processing, and format standardization processing on the target description information to obtain the preprocessed target description information. Then, the electronic device inputs the preprocessed target description information into the word embedding extraction network in the feature extraction network. The word embedding extraction network generates a corresponding embedding vector for the preprocessed target description information to obtain a word embedding code. Then, the electronic device inputs the word vector code into the semantic feature extraction network and the local feature extraction network in the feature extraction network respectively to obtain the contextual semantic information of the word as a semantic feature, and obtains keywords or phrases as local features. Then, the electronic device inputs the semantic features and local features into the feature fusion network in the feature extraction network to use the feature fusion network to weight the semantic features and local features according to their respective weights and then sum them up, thereby inputting the text features of the target description information.

[0043] Optionally, the word embedding extraction network includes but is not limited to a pre-trained deep learning model (Bidirectional Encoder Representations from Transformers, BERT), and may also be other forms of network models.

[0044] Optionally, the semantic feature extraction network includes but is not limited to Long Short Term Memory networks (LSTM), and may also be other forms of network models.

[0045] Optionally, the local feature extraction network includes but is not limited to a convolutional neural network (CNN), and may also be other forms of network models.

[0046] Optionally, the feature fusion network includes but is not limited to a term frequency-inverse document frequency (TF-IDF) model, and may also be other forms of network models.

[0047] S130, based on the question processing network in the pre-trained large language model, classify the text features to determine the initial qualitative information, initial classification information and initial grading information of the target question.

[0048] In this embodiment, the electronic device continues to use the question processing network in the pre-trained large language model to perform qualitative, categorical and grade analysis on the target question, thereby obtaining initial qualitative information, initial classification information and initial grading information of the target question.

[0049] The specific implementation method of S130 includes but is not limited to the following method: based on the qualitative template and classification and grading template corresponding to the problem processing network, the text features are processed to determine the initial qualitative information, initial classification information and initial grading information of the target problem.

[0050] Among them, the qualitative template can be understood as the qualitative system and framework of the problem, and the classification and grading template can be understood as the classification and grading database of the problem. For example, if the target problem is an audit problem, then the qualitative template is specifically the qualitative system and framework of the audit problem, and the classification and grading template is specifically the classification and grading database of the audit problem.

[0051] Therefore, after automatically extracting features from the target description information in the text to be processed that records the problem using the feature extraction network, the problem processing network in the pre-trained large language model is used to automatically perform qualitative, classified, and graded analysis on the target problem, so as to efficiently and accurately determine the initial qualitative information, initial classification information, and initial grading information of the target problem.

[0052] S140, using the question rule model of the reference text corresponding to the text to be processed, the initial qualitative information, the initial classification information and the initial grading information are processed to determine the target qualitative information, the target classification information and the target grading information of the target question in turn.

[0053] In order to improve the accuracy of qualitative, classified and graded analysis of the target problem, the electronic device calls the question rule model of the reference text corresponding to the text to be processed, and then matches the initial qualitative information, initial classification information and initial grading information with the qualitative information, classification information and grading information contained in the question rule model, so as to accurately perform the final qualitative, classified and graded analysis of the target problem.

[0054] Among them, the specific implementation method of S140 includes but is not limited to the following method: determine in turn whether the problem rule model contains candidate qualitative information, candidate classification information and candidate classification information that match the initial qualitative information, initial classification information and initial classification information of the target problem respectively; if contained, the candidate qualitative information, candidate classification information and candidate classification information are used as the target qualitative information, target classification information and target classification information of the target problem in turn; if not contained, the initial qualitative information, initial classification information and initial classification information are used as the target qualitative information, target classification information and target classification information of the target problem in turn, and the initial qualitative information, initial classification information and initial classification information are added to the problem rule model to form a new problem rule model.

[0055] Among them, the problem rule model can be understood as a system qualitative classification and grading rule model.

[0056] Specifically, first, the electronic device matches the initial qualitative information with the qualitative information contained in the question rule model. If the question rule model contains qualitative information that matches the initial qualitative information, it is output and used as the target qualitative information. If the question rule model does not contain qualitative information that matches the initial qualitative information, the initial qualitative information is used as the target qualitative information of the target question, and the initial qualitative information is added to the question rule model to update the question rule model; then, the electronic device classifies the target question into a first-level classification according to the field to which the target description information of the target question belongs, and then gradually subdivides it into a second-level classification, and finally further subdivides it into a third-level classification, so as to obtain the target classification information of the target question. If the first-level classification does not belong to it, a first-level classification is automatically created according to the field to which it belongs. Similarly, if the second-level classification does not belong to it, it is automatically added to the second-level classification. A secondary classification is automatically created after the last category of the primary classification. Similarly, if the third-level classification does not belong to the third-level classification, a third-level classification is automatically created after the last category of the third-level classification, and the automatically created classification is used to update the problem rule model; then, the electronic device continues to classify the target problem, first classifying the target problem to the first level, then gradually subdividing it into the second level, and finally further subdividing it into the third level, so as to obtain the target classification information of the target problem. If the first level does not belong to the first level, the second level is automatically created according to the field to which it belongs. Similarly, if the second level does not belong to the second level, the second level is automatically created after the last level of the second level. Similarly, if the third level does not belong to the third level, the third level is automatically created after the last level of the second level, and the automatically created level is used to update the problem rule model.

[0057] In this way, after using the large language model to initially conduct qualitative, classified and graded analysis on the target question, the question rule model of the reference text is also combined to further conduct qualitative, classified and graded analysis on the target question, thereby improving the efficiency and accuracy of problem processing in the text.

[0058] A method for processing text problems in an embodiment of the present disclosure includes: obtaining target description information of a target problem in a to-be-processed text; performing feature extraction processing on the target description information based on a feature extraction network to obtain text features of the target description information; processing the text features based on a problem processing network in a pre-trained large language model to determine the initial qualitative information, initial classification information, and initial grading information of the target problem; using a problem rule model of a reference text corresponding to the to-be-processed text, processing the initial qualitative information, initial classification information, and initial grading information, and sequentially determining the target qualitative information, target classification information, and target grading information of the target problem. Thus, the target description information in the to-be-processed text recording the problem is sequentially subjected to feature extraction, qualitative analysis, classification processing, and grading using a neural network model, and then automatically determined the final qualitative information, classification information, and grading information of the problem recorded in the text in combination with the problem rule model of the reference text. Therefore, the automatic qualitative, classification, and grading of the problems in the text are realized, the efficiency and accuracy of the problem processing in the text are improved, and thus the management effect of the problems recorded in the text is improved.

[0059] In some embodiments of the present disclosure, after determining the target classification of the target problem, the electronic device can also generate a qualitative classification classification table according to actual needs. Specifically, the electronic device obtains the requirement items for generating the table, and then generates a qualitative classification classification table according to the target classification of the target problem according to the requirement items.

[0060] Optionally, the requirement items include but are not limited to project dimensions, processing entity dimensions, etc.

[0061] In this way, for the target classification of the target problem, the qualitative classification grading table can be flexibly determined according to actual needs, and the automatic generation of the qualitative classification grading table is realized.

[0062] In another embodiment of the present disclosure, before executing S110 , a hierarchical model of the reference text may be generated.

[0063] Figure 2 A flowchart of another method for processing text problems provided by an embodiment of the present disclosure is shown.

[0064] like Figure 2 As shown, the method for processing the text problem may specifically include the following steps.

[0065] S210: Obtain reference description information of the reference question in the reference text.

[0066] In this embodiment, the electronic device collects a large amount of reference texts and processes the reference texts using optical character recognition and / or natural language processing to obtain reference description information containing reference questions, or directly uses the question description information input by the user as the reference description information of the reference question.

[0067] The reference question refers to the question in the field of the reference text recorded in the reference text. The reference text and the text to be processed belong to the same field. Similar to the target question, the field to which the reference question belongs includes but is not limited to the procurement field, sales field, engineering quality field, and can also be other field types.

[0068] S220. Generate a qualitative template for the reference question based on the qualitative information included in the reference description information, and generate a classification and grading template for the reference question based on the classification information and grading information included in the reference description information.

[0069] In this embodiment, the specific implementation method of generating a qualitative template of a reference question based on the qualitative information included in the reference description information includes but is not limited to the following method: using a pre-generated case-based reasoning algorithm to process the qualitative information and generate a qualitative template of the reference question.

[0070] The reference description information is text information that contains a specific description of the reference question.

[0071] Among them, the case-based reasoning algorithm refers to a neural network model that solves the current problem by retrieving historical examples.

[0072] Among them, the qualitative template of reference questions can be understood as a qualitative system and framework of questions, such as the qualitative system and framework of audit questions.

[0073] In this embodiment, a specific implementation method for generating a classification and grading template for reference questions based on the classification information and grading information contained in the reference description information includes but is not limited to the following method: clustering the multiple classification items contained in the classification information and the multiple levels contained in the grading information respectively to obtain a classification and grading template for reference questions containing classification item clusters and graded clusters.

[0074] Among them, the classification item refers to the type of reference question, and the level refers to the importance level of the reference question.

[0075] Among them, the classification and grading template can be understood as a problem classification and grading database, such as a problem classification and grading database for audit issues.

[0076] Specifically, the electronic device calls a preset clustering algorithm to cluster the multiple classification items contained in the classification information and the multiple levels contained in the hierarchical information. The specific process is to first classify the classification information into a first-level classification according to the field corresponding to the reference description information to which the classification information belongs, and then gradually subdivide it into a second-level classification, and finally further subdivide it into a third-level classification. If the first-level classification cannot be classified, a first-level classification is automatically created according to the field to which it belongs. Similarly, if the second-level classification cannot be classified, a second-level classification is automatically created after the last category of the second-level classification. Similarly, if the third-level classification cannot be classified, a third-level classification is automatically created after the last category of the third-level classification. In this way, the clustering process is completed, classification item clusters and hierarchical clusters are generated, and the classification item clusters and hierarchical clusters are combined to form a classification and grading template for reference questions.

[0077] S230: Using the pre-acquired decision-making reasoning algorithm, the qualitative template and the classification and grading template are processed to generate a question rule model of the reference text.

[0078] Among them, the decision reasoning algorithm refers to the reasoning mechanism that simulates the human decision-making process, which relies on a set of predefined rule sets to make judgments and inferences.

[0079] In this way, based on the qualitative information, classification information and grading information of the reference question, a qualitative template and a classification and grading template are first generated, and then a question rule model is generated, so that when a new question is obtained, the question rule model is called to qualitatively, classify and grade the new question.

[0080] S240: Obtain target description information of the target problem in the text to be processed.

[0081] S250: Based on the feature extraction network, feature extraction processing is performed on the target description information to obtain text features of the target description information.

[0082] S260, based on the question processing network in the pre-trained large language model, process the text features to determine the initial qualitative information, initial classification information and initial grading information of the target question.

[0083] S270, using the question rule model of the reference text corresponding to the text to be processed, the initial qualitative information, the initial classification information and the initial grading information are processed to determine the target qualitative information, the target classification information and the target grading information of the target question in turn.

[0084] Among them, S240~S270 are similar to S110~S140, and will not be described in detail here.

[0085] The present disclosure also provides a text problem processing device for implementing the above-mentioned text problem processing method. Figure 3In the embodiment of the present disclosure, the device for processing the text question may be an electronic device. The electronic device may include a tablet computer, a desktop computer, a laptop computer, or other device with a communication function, or may include a device simulated by a virtual machine or a simulator.

[0086] Figure 3 A schematic diagram of the structure of a text question processing device provided by an embodiment of the present disclosure is shown.

[0087] like Figure 3 As shown, the text question processing device 300 may include:

[0088] A description information acquisition module 310 is used to acquire target description information of a target problem in a text to be processed;

[0089] A feature extraction module 320 is used to perform feature extraction processing on the target description information based on a feature extraction network to obtain text features of the target description information;

[0090] A first processing module 330 is used to process the text features based on the question processing network in the pre-trained large language model to determine the initial qualitative information, initial classification information and initial classification information of the target question;

[0091] The second processing module 340 is used to process the initial qualitative information, the initial classification information and the initial grading information by using the question rule model of the reference text corresponding to the text to be processed, and determine the target qualitative information, target classification information and target grading information of the target question in turn.

[0092] A device for processing text problems in an embodiment of the present disclosure includes: obtaining target description information of a target problem in a to-be-processed text; performing feature extraction processing on the target description information based on a feature extraction network to obtain text features of the target description information; processing the text features based on a problem processing network in a pre-trained large language model to determine the initial qualitative information, initial classification information, and initial grading information of the target problem; using a problem rule model of a reference text corresponding to the to-be-processed text, processing the initial qualitative information, initial classification information, and initial grading information, and sequentially determining the target qualitative information, target classification information, and target grading information of the target problem. Thus, the neural network model is used to sequentially perform feature extraction, qualitative analysis, classification processing, and grading on the target description information in the to-be-processed text recording the problem, and then combined with the problem rule model of the reference text, the final qualitative information, classification information, and grading information of the problem recorded in the text are automatically determined. Therefore, the automatic qualitative, classification, and grading of the problems in the text are realized, the efficiency and accuracy of the problem processing in the text are improved, and then the management effect of the problems recorded in the text is improved.

[0093] In some embodiments of the present disclosure, the feature extraction module 320 includes:

[0094] A word embedding extraction unit, used to process the target description information based on the word embedding extraction network in the feature extraction network to obtain a word embedding code corresponding to the target description information;

[0095] A semantic extraction unit, configured to process the word embedding code based on a semantic feature extraction network in the feature extraction network to obtain semantic features of the target description information;

[0096] A local feature extraction unit, used to process the word embedding code based on the local feature extraction network in the feature extraction network to obtain local features of the target description information;

[0097] A weighted fusion unit is used to perform weighted fusion processing on the semantic features and the local features based on the feature fusion network in the feature extraction network to obtain the text features of the target description information.

[0098] In some embodiments of the present disclosure, the first processing module 330 is specifically configured to:

[0099] Based on the qualitative template and the classification and grading template corresponding to the question processing network, the text features are processed to determine the initial qualitative information, the initial classification information and the initial grading information of the target question.

[0100] In some embodiments of the present disclosure, the second processing module 340 includes:

[0101] A judging unit, configured to judge in sequence whether the question rule model contains candidate qualitative information, candidate classification information and candidate classification information that respectively match the initial qualitative information, the initial classification information and the initial classification information of the target question;

[0102] A first determining unit is configured to, if included, use the candidate qualitative information, the candidate classification information, and the candidate grading information as target qualitative information, target classification information, and target grading information of the target problem in sequence;

[0103] The second determination unit is used for, if not included, using the initial qualitative information, the initial classification information and the initial grading information as the target qualitative information, the target classification information and the target grading information of the target problem in turn, and adding the initial qualitative information, the initial classification information and the initial grading information to the problem rule model to form a new problem rule model.

[0104] In some embodiments of the present disclosure, the device further includes:

[0105] A reference description information acquisition module is used to acquire reference description information of reference questions in reference texts;

[0106] A first generating module, configured to generate a qualitative template for the reference question according to the qualitative information included in the reference description information, and to generate a classification and grading template for the reference question according to the classification information and grading information included in the reference description information;

[0107] The second generation module is used to process the qualitative template and the classification and grading template using a pre-acquired decision-making inference algorithm to generate a question rule model of the reference text.

[0108] In some embodiments of the present disclosure, the first generating module is specifically used to:

[0109] The qualitative information is processed using a pre-generated case-based reasoning algorithm to generate a qualitative template for the reference question.

[0110] In some embodiments of the present disclosure, the first generating module is specifically used to:

[0111] Clustering is performed on the multiple classification items contained in the classification information and the multiple levels contained in the grading information respectively, so as to obtain a classification and grading template of the reference question including classification item clusters and grading clusters.

[0112] It should be noted that Figure 3 The text question processing device 300 shown can execute Figure 1 to Figure 2 The various steps in the method embodiment shown in the figure are implemented Figure 1 to Figure 2 The various processes and effects in the method embodiment shown are not described in detail here.

[0113] Figure 4 A schematic structural diagram of an electronic device provided by an embodiment of the present disclosure is shown.

[0114] like Figure 4 As shown, the electronic device may include a processor 401 and a memory 402 storing computer program instructions.

[0115] Specifically, the processor 401 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0116] The memory 402 may include a large capacity memory for information or instructions. By way of example and not limitation, the memory 402 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a tape, or a universal serial bus (USB) drive or a combination of two or more of these. Where appropriate, the memory 402 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 402 may be inside or outside the integrated gateway device. In a particular embodiment, the memory 402 is a non-volatile solid-state memory. In a particular embodiment, the memory 402 includes a read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (Electrically Erasable Programmable ROM, EEPROM), an electrically rewritable ROM (EAROM) or a flash memory, or a combination of two or more of these.

[0117] The processor 401 reads and executes the computer program instructions stored in the memory 402 to perform the steps of the text problem processing method provided in the embodiment of the present disclosure.

[0118] In one example, the electronic device may further include a transceiver 403 and a bus 404. Figure 4 As shown, the processor 401, the memory 402 and the transceiver 403 are connected via a bus 404 and communicate with each other.

[0119] The bus 404 includes hardware, software, or both. For example, but not limitation, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industrial Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a Memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses or a combination of two or more of these. Where appropriate, the bus 404 may include one or more buses. Although embodiments of the present application describe and illustrate a particular bus, the present application contemplates any suitable bus or interconnect.

[0120] The following is an embodiment of a computer-readable storage medium provided in an embodiment of the present disclosure. The computer-readable storage medium and the method for processing text problems in the above-mentioned embodiments belong to the same inventive concept. For details not described in detail in the embodiment of the computer-readable storage medium, please refer to the embodiment of the method for processing text problems in the above-mentioned embodiments.

[0121] This embodiment provides a storage medium containing computer executable instructions. When the computer executable instructions are executed by a computer processor, they are used to execute a method for processing a text problem. The method includes:

[0122] Obtain target description information of the target problem in the text to be processed;

[0123] Based on the feature extraction network, feature extraction processing is performed on the target description information to obtain text features of the target description information;

[0124] Based on the question processing network in the pre-trained large language model, the text features are processed to determine the initial qualitative information, initial classification information and initial grading information of the target question;

[0125] The initial qualitative information, the initial classification information and the initial grading information are processed using the question rule model of the reference text corresponding to the text to be processed, and the target qualitative information, target classification information and target grading information of the target question are determined in turn.

[0126] Of course, the storage medium containing computer executable instructions provided by the embodiment of the present disclosure is not limited to the above method operations, and the computer executable instructions can also execute related operations in the text problem processing method provided by any embodiment of the present disclosure.

[0127] Through the above description of the implementation methods, the technical personnel in the relevant field can clearly understand that the present disclosure can be implemented by means of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present disclosure can be essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer cloud platform (which can be a personal computer, server, or network cloud platform, etc.) to execute the text problem processing method provided by each embodiment of the present disclosure.

[0128] Note that the above are only preferred embodiments of the present disclosure and the technical principles used. Those skilled in the art will understand that the present disclosure is not limited to the specific embodiments herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present disclosure. Therefore, although the present disclosure is described in more detail through the above embodiments, the present disclosure is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present disclosure, and the scope of the present disclosure is determined by the scope of the appended claims.

Claims

1. A method for processing text problems, characterized in that: include: Obtain target description information of the target problem in the text to be processed; Based on the feature extraction network, feature extraction processing is performed on the target description information to obtain text features of the target description information; Based on the question processing network in the pre-trained large language model, the text features are processed to determine the initial qualitative information, initial classification information and initial grading information of the target question; The initial qualitative information, the initial classification information and the initial grading information are processed using the question rule model of the reference text corresponding to the text to be processed, and the target qualitative information, target classification information and target grading information of the target question are determined in turn.

2. The method according to claim 1, characterized in that The feature extraction network is used to perform feature extraction processing on the target description information to obtain text features of the target description information, including: Based on the word embedding extraction network in the feature extraction network, the target description information is processed to obtain a word embedding code corresponding to the target description information; Based on the semantic feature extraction network in the feature extraction network, the word embedding code is processed to obtain the semantic features of the target description information; Based on a local feature extraction network in the feature extraction network, the word embedding code is processed to obtain local features of the target description information; Based on the feature fusion network in the feature extraction network, the semantic features and the local features are subjected to weighted fusion processing to obtain text features of the target description information.

3. The method according to claim 1, characterized in that The question processing network based on the pre-trained large language model processes the text features to determine the initial qualitative information, initial classification information and initial grading information of the target question, including: Based on the qualitative template and the classification and grading template corresponding to the question processing network, the text features are processed to determine the initial qualitative information, the initial classification information and the initial grading information of the target question.

4. The method according to claim 1, characterized in that: The method of using the question rule model of the reference text corresponding to the to-be-processed text to process the initial qualitative information, the initial classification information and the initial grading information to sequentially determine the target qualitative information, the target classification information and the target grading information of the target question includes: Determining in sequence whether the question rule model contains candidate qualitative information, candidate classification information, and candidate classification information that respectively match the initial qualitative information, the initial classification information, and the initial classification information of the target question; If included, the candidate qualitative information, the candidate classification information and the candidate grading information are used as the target qualitative information, the target classification information and the target grading information of the target problem in sequence; If not included, the initial qualitative information, the initial classification information and the initial grading information are used as the target qualitative information, target classification information and target grading information of the target problem respectively, and the initial qualitative information, the initial classification information and the initial grading information are added to the problem rule model to form a new problem rule model.

5. The method according to claim 1, characterized in that Also includes: Get the reference description information of the reference question in the reference text; Generating a qualitative template for the reference question based on the qualitative information included in the reference description information, and generating a classification and grading template for the reference question based on the classification information and grading information included in the reference description information; The qualitative template and the classification and grading template are processed using a pre-acquired decision-making inference algorithm to generate a question rule model of the reference text.

6. The method according to claim 5, characterized in that The step of generating a qualitative template for the reference question according to the qualitative information included in the reference description information includes: The qualitative information is processed using a pre-generated case-based reasoning algorithm to generate a qualitative template for the reference question.

7. The method according to claim 5, characterized in that The step of generating a classification and grading template for the reference question according to the classification information and grading information included in the reference description information includes: Clustering is performed on the multiple classification items contained in the classification information and the multiple levels contained in the grading information respectively, so as to obtain a classification and grading template of the reference question including classification item clusters and grading clusters.

8. A device for processing text problems, comprising: A description information acquisition module is used to obtain target description information of the target problem in the text to be processed; A feature extraction module is used to perform feature extraction processing on the target description information based on a feature extraction network to obtain text features of the target description information; A first processing module is used to process the text features based on a question processing network in a pre-trained large language model to determine initial qualitative information, initial classification information and initial grading information of the target question; The second processing module is used to process the initial qualitative information, the initial classification information and the initial grading information by using the question rule model of the reference text corresponding to the text to be processed, and determine the target qualitative information, target classification information and target grading information of the target question in turn.

9. An electronic device, characterized in that: include: processor; A memory for storing executable instructions; The processor is used to read the executable instructions from the memory and execute the executable instructions to implement the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the method according to any one of claims 1 to 7.

Images