Case text information extraction method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of big data, and discloses a case text information extraction method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring a case text; detecting an event from the case text to obtain an event of the case text; classifying the events to obtain event categories; extracting arguments related to the event from the case text; classifying the arguments to obtain argument categories; and outputting the event, the event category, the argument and the argument category. By implementing the method provided by the invention, the key event information can be quickly and accurately extracted from the medical text.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and more specifically to a method, device, computer equipment and storage medium for extracting case text information. Background Art

[0002] Information extraction, as one of the key research areas of natural language processing, aims to mine structured event information from unstructured text. This task is almost instinctive for humans, but for artificial intelligence, it is one of the important criteria for measuring its intelligence level.

[0003] In practical applications, especially in the medical and insurance industries, text data is often full of redundant information, while truly valuable data points are relatively rare but crucial. Traditional medical records are usually long, complex in content structure, and contain a lot of professional terms, which makes manual reading and understanding of these documents extremely time-consuming and laborious, seriously restricting the speed of subsequent work. With the rapid development of the medical and insurance industries, the number of medical records is also growing. Simply relying on human interpretation can no longer meet the needs of modern efficient work.

[0004] Therefore, it is necessary to design a new method to quickly and accurately extract key event information from medical texts. Summary of the invention

[0005] The present invention provides a case text information extraction method, device, computer equipment and storage medium to solve the technical problem that the traditional manual reading method is inefficient due to the lengthy, complex and professional nature of the text, and is difficult to meet the growing demand for fast and accurate processing of massive medical records.

[0006] First, a case text information extraction method is provided, including:

[0007] Get case text;

[0008] Detecting events from the case text to obtain events of the case text;

[0009] Classifying the events to obtain event categories;

[0010] Extracting arguments related to the event from the case text;

[0011] Classifying the arguments to obtain argument categories;

[0012] The event, the event category, the argument, and the argument category are output.

[0013] In a second aspect, a case text information extraction device is provided, comprising:

[0014] A text acquisition unit, used for acquiring case text;

[0015] An event detection unit, used for detecting events from the case text to obtain events of the case text;

[0016] An event classification unit, used for classifying the event to obtain an event category;

[0017] An argument extraction unit, used for extracting arguments related to the event from the case text;

[0018] An argument classification unit, used for classifying the arguments to obtain argument categories;

[0019] An output unit is used to output the event, the event category, the argument and the argument category.

[0020] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned case text information extraction method when executing the computer program.

[0021] In a fourth aspect, a storage medium is provided, which stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned case text information extraction method are implemented.

[0022] The above-mentioned case text information extraction method obtains case text containing medical event descriptions from an electronic medical record system or other data sources; uses natural language processing techniques, such as named entity recognition and event extraction models, to identify potential medical events from the text; classifies the detected events and uses pre-trained classification models, such as support vector machines or deep learning models, to classify the events into different categories; extracts event-related arguments from the case text; classifies the extracted arguments according to their types; integrates the detected events, event categories, arguments and argument categories; and outputs structured information, including events, event categories, arguments and their categories, to quickly and accurately extract key event information from medical texts.

[0023] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0025] Figure 1 A schematic diagram of an application scenario of the case text information extraction method provided by an embodiment of the present invention;

[0026] Figure 2 A schematic diagram of a process for extracting case text information provided by an embodiment of the present invention;

[0027] Figure 3 A schematic diagram of a sub-process of a case text information extraction method provided in an embodiment of the present invention;

[0028] Figure 4 A schematic diagram of a sub-process of a case text information extraction method provided in an embodiment of the present invention;

[0029] Figure 5 A schematic diagram of a sub-process of a case text information extraction method provided in an embodiment of the present invention;

[0030] Figure 6 A schematic diagram of a sub-process of a case text information extraction method provided in an embodiment of the present invention;

[0031] Figure 7 A schematic block diagram of a device for extracting case text information provided by an embodiment of the present invention;

[0032] Figure 8 A schematic block diagram of an event detection unit of a case text information extraction device provided by an embodiment of the present invention;

[0033] Fig. 9 A schematic block diagram of an event classification unit of a case text information extraction device provided in an embodiment of the present invention;

[0034] Fig.10 A schematic block diagram of an argument extraction unit of a case text information extraction device provided in an embodiment of the present invention;

[0035] Fig.11 A schematic block diagram of an argument classification unit of a case text information extraction device provided in an embodiment of the present invention;

[0036] Fig.12 is a schematic diagram of a structure of a computer device in an embodiment of the present invention;

[0037] Fig.13 2 is another schematic diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0040] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0041] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0042] See also Figure 1 and Figure 2 , Figure 1 A schematic diagram of an application scenario of the case text information extraction method provided in an embodiment of the present invention. Figure 2A schematic flow chart of a case text information extraction method provided by an embodiment of the present invention. The case text information extraction method is applied to a server. The server interacts with the terminal for data, and reduces the time cost of manual screening by automatically detecting and classifying events in the text. First, the method uses a deep learning model to extract grammatical and semantic features in the case text, accurately locates valuable event information, and thus filters out irrelevant content. Then, for each detected event, a specially trained model is further used to classify it to ensure the accuracy of the information. In addition, the method can also extract event-related arguments from the text and classify these arguments, thereby enhancing the structuring of the information. By introducing advanced technologies such as attention mechanisms or graph convolutional networks, the argument classification model can better capture the association between arguments and improve the quality of information extraction. Finally, this method not only improves the speed of information processing, but also ensures the comprehensiveness and accuracy of information, greatly promoting the informatization process of the medical and insurance industries. In summary, the embodiment of the present invention effectively solves the problem of low efficiency of traditional methods when facing a large amount of redundant information, and adapts to the needs of modern efficient work.

[0043] Figure 2 FIG. 1 is a flow chart of a method for extracting case text information provided by an embodiment of the present invention. Figure 2 As shown, the method includes the following steps S110 to S160.

[0044] S110, obtaining case text.

[0045] In this embodiment, the case text refers to the text data of the medical record.

[0046] The process of obtaining medical records through electronic medical record systems, manual input or other digital means. This process is the first step of the entire information extraction method, which provides the original data source for subsequent event detection, classification and argument extraction. In order to ensure the effectiveness and accuracy of information extraction, the case text obtained should be as complete and accurate as possible, covering the patient's basic information, diagnosis results, treatment process and prognosis evaluation. By connecting to the hospital information system (HIS), laboratory information system (LIS) or picture archiving and communication system (PACS), case texts can be collected automatically, which improves work efficiency while ensuring data consistency and traceability. This step is crucial to the realization of subsequent automated information processing, which helps to quickly locate and parse valuable medical information, thereby supporting application scenarios such as clinical decision-making and medical insurance claims.

[0047] S120, detecting events from the case text to obtain events of the case text.

[0048] In this embodiment, the events of the case text specifically refer to specific actions or state changes that occur during the medical process, which can be clearly identified and recorded. They can be medical interventions (such as surgery, drug treatment), disease progression, changes in test results, changes in the patient's physiological or psychological state, etc. By detecting and classifying these events, doctors can better understand the patient's disease progression, support clinical decision-making, and promote the effective use of medical resources.

[0049] In one embodiment, see Figure 3 , the above-mentioned step S120 may include steps S121 to S122.

[0050] S121. Extract grammatical and semantic features from the case text to obtain feature extraction results.

[0051] In this embodiment, the feature extraction result refers to key information that helps to identify an event.

[0052] Extract key information that helps identify events from the original case text. This key information includes but is not limited to the subject-verb-object structure of the sentence, verb phrases, noun phrases, and contextual information related to specific medical terms. Through natural language processing technology, especially dependency tree analysis, dependency relationship analysis and other methods, the connection between words can be captured at the sentence level to further understand the deep meaning of the text. In addition, word embedding technology (such as Word2Vec or GloVe) can be used to convert words into vector form for easy processing by machine learning algorithms.

[0053] S122, inputting the feature extraction result into an event detection model to perform event detection to obtain events of the case text.

[0054] In this embodiment, the feature extraction results obtained from step S121 are processed using a pre-trained event detection model. The model is usually built based on a deep learning framework, such as a recurrent neural network (RNN), a long short-term memory network (LSTM), or a convolutional neural network (CNN). These models can learn complex patterns and structures in texts, thereby effectively identifying potential event triggers. In particular, the model proposed in the present invention can not only independently identify events, but also be combined with other NLP tasks (such as named entity recognition NER and relation extraction RE) to jointly improve the accuracy and recall of event detection.

[0055] The event detection model refers to the one obtained by extracting corresponding grammatical and semantic features from a case text dataset containing events and using it as a sample set to train a deep learning model.

[0056] The event detection model is trained using a large number of case text datasets containing known events. During the training process, the model learns how to extract grammatical and semantic features related to events from the text. For example, in the given text "Patient A had an appendectomy in 2024", "appendectomy" is an event trigger word. By manually annotating a large number of such cases, the model can gradually learn to automatically identify similar event trigger words.

[0057] S130: Classify the event to obtain an event category.

[0058] In this embodiment, the event category refers to the classification result of the detected event according to certain standards or dimensions. These standards or dimensions can be the nature of the event (such as medical behavior, diagnosis results, changes in condition), time attributes (such as immediate events, continuous states), participant roles (such as doctors, patients), geographic locations, etc. Event categories are intended to provide more refined and structured data support for subsequent information processing and analysis. For example, in a medical scenario, event categories can be subdivided into surgical operations, drug treatments, laboratory test results, worsening or improvement of the condition, etc.

[0059] In one embodiment, see Figure 4 , the above-mentioned step S130 may include steps S131 to S132.

[0060] S131. Extract features of the event to obtain event features.

[0061] In this embodiment, event features refer to key features that are useful for classification extracted from detected events; these features include but are not limited to event trigger words, contextual information of the event, time information of the event, location information, participant information, etc. Through natural language processing technology, especially entity recognition, relationship extraction and other methods, this information can be accurately extracted from the text. In addition, the text can also be converted into numerical features using technologies such as word vectors and sentence vectors to facilitate processing by machine learning models.

[0062] S132: Input the event features into an event classification model for classification to obtain an event category.

[0063] The event features obtained from step S131 are processed using a pre-trained event classification model. The model can be built based on a variety of machine learning algorithms, such as support vector machines (SVM), decision trees, random forests, or deep learning models (such as DNN, CNN, RNN, etc.). These models can learn the mapping relationship between different types of event features and predefined event categories, thereby achieving high-precision event classification.

[0064] In this embodiment, the event classification model is obtained by training a machine learning model using features extracted from several events as a sample set using remote supervision or semi-supervised learning.

[0065] The event classification model is trained by a large number of data sets containing known events and their categories. During the training process, the model can not only use fully annotated data sets, but also use remote supervision or semi-supervised learning methods to use large-scale unlabeled or partially annotated data to enhance the learning ability of the model. This method can significantly reduce the workload of manual annotation while improving the generalization ability and adaptability of the model. For example, by introducing domain knowledge or external resources (such as medical knowledge graphs), the model can better understand the semantics of events, thereby improving classification accuracy.

[0066] S140. Extracting arguments related to the event from the case text.

[0067] In this embodiment, event-related arguments refer to entities or concepts that play an important role in a specific event or have clear semantics. They can be names (such as patient names), times (such as the specific date of the event), places (such as the place where the event occurred), actions (such as operations or activities performed), causes (the reasons for the event), etc. For example, in the sentence "Patient A underwent appendectomy in 2024", "A" is the argument of the patient's identity, "2024" is the time argument, and "appendectomy" is the specific medical operation argument. Correctly identifying these arguments is crucial to understanding the nature of the event, analyzing the influencing factors of the event, and supporting subsequent information retrieval and decision-making processes.

[0068] In one embodiment, see Figure 5 , the above-mentioned step S140 may include steps S141 to S142.

[0069] S141. Extract boundary information and context information from the case text to obtain an extraction result.

[0070] In this embodiment, the extraction result refers to identifying information segments that may contain arguments from the text.

[0071] Boundary information refers to the starting and ending positions of each potential argument in the text, while context information includes the words, phrases or sentences around the argument, which is crucial for correctly identifying and understanding the argument. By combining natural language processing technologies such as part-of-speech tagging and dependency analysis, the system can effectively capture this information and lay a solid foundation for subsequent argument extraction.

[0072] S142. Input the extraction results into an argument extraction model to extract arguments to obtain arguments related to the event.

[0073] In this embodiment, after completing the extraction of boundary information and context information, the next task is to use a specially designed argument extraction model to conduct an in-depth analysis of this information, so as to accurately identify the specific arguments of the event. The argument extraction model used here is usually built based on a deep learning framework, for example, it is fine-tuned through pre-trained models such as BERT to adapt to the argument extraction needs of specific fields. When training the model, a large number of labeled case text data sets are used. These data sets contain not only event information, but also labeled argument boundaries and context information to ensure that the model can learn how to effectively identify the correct arguments from complex texts.

[0074] Among them, the argument extraction model is obtained by extracting corresponding boundary information and context information from several case text data sets containing arguments and using them as sample sets to train a deep learning model.

[0075] Specifically, first, a high-quality training dataset needs to be prepared. This dataset consists of multiple case texts containing arguments, and each text is annotated with arguments related to specific events. For example, in the sentence "Patient A underwent appendectomy in 2024", the three arguments are: "A" (patient), "2024" (time), and "appendectomy" (operation). These arguments have been manually annotated in the dataset to ensure that the position and type of each argument are accurate.

[0076] After the annotated data is prepared, the next step is to extract the boundary information and context information of the argument from the text. Boundary information refers to the exact position of the argument in the text, that is, the start and end position index. Context information includes the language environment around the argument, which helps the model better understand the semantics and role of the argument. For example, in the above example, the boundary information of "A" may be (3, 6), and its context information may include several words before and after, such as "patient" and "at".

[0077] With boundary information and context information, these data can be used as sample sets to train deep learning models. The present invention chooses to use the BERT (Bidirectional Encoder Representations from Transformers) model for training because the BERT model has shown strong capabilities in processing natural language tasks, especially in understanding context.

[0078] During the training process, the model learns how to predict argument boundaries based on contextual information. Specifically, the training task is defined as an entity recognition task, where each argument is considered a target entity. The model gradually improves its recognition accuracy of argument boundaries through a large number of iterative learning.

[0079] In order to ensure the generalization ability of the model, the training data set contains at least 5,000 manually annotated case texts. These texts cover various types of medical events to ensure that the model can work effectively in different scenarios. The pre-trained BERT model is used as the basis, and then fine-tuned for the argument extraction task. During the fine-tuning process, the model outputs a label on each token to indicate whether the token is part of a certain argument. Commonly used loss functions include the cross-entropy loss function, which is used to measure the difference between the model's predicted label and the true label. Commonly used evaluation indicators include precision, recall, and F1 score, which can help us evaluate the performance of the model on the argument extraction task.

[0080] After sufficient training, the model will be able to accurately extract arguments from new case texts. These arguments can be used in various subsequent application scenarios, such as automatic medical record summarization, medical information retrieval, clinical decision support, etc.

[0081] In summary, the argument extraction model can effectively extract event-related arguments from case texts through training with a large number of annotated data sets, providing strong support for automated processing in the medical field.

[0082] S150, classifying the arguments to obtain argument categories.

[0083] In this embodiment, argument classification refers to further subdividing the arguments extracted from the text into specific types in order to more accurately understand and apply the argument information. For example, in the sentence "Patient A underwent appendectomy in 2024", the three arguments "A", "2024" and "appendectomy" are classified as "name", "time" and "surgical method" respectively.

[0084] In one embodiment, see Figure 6 , the above-mentioned step S150 may include steps S151 to S152.

[0085] S151. Extract the features of the argument to obtain argument features.

[0086] In this embodiment, before classifying the arguments, it is first necessary to extract useful features from the extracted arguments. These features may include but are not limited to:

[0087] Morphological features: the lexical form of the argument itself, such as roots, affixes, etc.

[0088] Contextual features: information about the words surrounding the argument, such as the preceding and following words and their parts of speech.

[0089] Position feature: The position of the argument in the sentence, such as the beginning, middle or end of the sentence.

[0090] Syntactic features: the syntactic role of the argument in the sentence, such as subject, object, etc.

[0091] Semantic features: semantic information represented by the argument, such as time, place, person, etc.

[0092] S152. Input the argument features into an argument classification model for classification to obtain an argument category.

[0093] In this embodiment, after the argument features are extracted, these features are then input into a pre-trained argument classification model to obtain the category of the argument. The argument classification model is obtained by extracting the corresponding features from data with argument categories and then using it as a sample set to train a machine learning model. Specifically, the data set for model training contains a large number of argument instances, each with a correct category label. Through such training, the model can learn the mapping relationship between different feature combinations and argument categories.

[0094] The argument classification model is obtained by extracting corresponding features from data with argument categories and using them as sample sets to train a machine learning model.

[0095] An attention mechanism or a graph convolutional network is set in the argument classification model.

[0096] In this embodiment, in order to improve the accuracy of argument classification, the present invention introduces methods such as attention mechanism or graph convolutional network (GCN) in the argument classification model. These methods can effectively utilize the dependency relationship and context information between arguments, thereby enhancing the classification ability of the model.

[0097] Attention mechanism: By calculating the importance weight of each feature, the model pays more attention to those features that have an important impact on the classification results during classification.

[0098] Graph Convolutional Network: By constructing a relationship graph between arguments, the complex dependencies between arguments are captured by using information transfer between nodes.

[0099] In addition, this embodiment also uses transfer learning or domain adaptation methods to apply models trained in other fields or tasks to argument classification to reduce the reliance on large-scale labeled data. This method can not only improve the generalization ability of the model, but also improve the performance of the model when data is scarce.

[0100] In order to train the argument classification model, the BERT model was used to train 5,000 case texts with argument category annotations. The specific steps are as follows:

[0101] Prepare a dataset of 5,000 case texts with argument category annotations. Use the BERT model to encode each argument and extract its feature vector. Input the extracted feature vector into a classifier for training. The classifier can be a multi-classifier, conditional random field (CRF), or neural network. Evaluate the performance of the model through methods such as cross-validation to ensure its accuracy and robustness in argument classification tasks.

[0102] S160, output the event, the event category, the argument, and the argument category.

[0103] The event, the event category, the argument, and the argument category are output to a terminal for display.

[0104] By introducing transfer learning and domain adaptation methods, the model can achieve good performance with a small amount of labeled data, thus significantly reducing the need for manual labeling. The multi-stage model design allows each stage to focus on specific tasks, such as entity recognition, argument extraction, and argument classification. This staged approach helps improve the overall accuracy and generalization ability of the model. Using pre-trained models and transfer learning techniques, the model can quickly adapt to new tasks and domains on small-scale labeled datasets, reducing dependence on large-scale labeled data. The design of the multi-stage model allows each component to be developed and optimized independently, improving development efficiency while also reducing overall construction costs.

[0105] The multi-stage event extraction module is not limited to a single business scenario, but can be flexibly applied to multiple business areas, such as underwriting claims auxiliary entry, intelligent decision-making, etc. Through modular design, the model can be easily integrated into different business systems, providing a powerful support tool for business personnel. The automated event extraction function significantly improves the speed and accuracy of data entry and reduces the workload of business personnel. In the insurance industry, the construction and optimization of the automatic underwriting process is crucial. The multi-stage event extraction module can quickly and accurately extract key information and accelerate the development and deployment of the automatic underwriting system.

[0106] In summary, this method not only achieves higher accuracy and generalization ability in technology, but also provides broad application prospects in business. By reducing the workload of manual annotation and lowering construction costs, this model has brought significant efficiency improvements to multiple business scenarios, accelerated the automation of business processes, and provided strong support for the digital transformation of enterprises.

[0107] The above-mentioned case text information extraction method obtains case text containing medical event descriptions from an electronic medical record system or other data sources; uses natural language processing techniques, such as named entity recognition and event extraction models, to identify potential medical events from the text; classifies the detected events and uses pre-trained classification models, such as support vector machines or deep learning models, to classify the events into different categories; extracts event-related arguments from the case text; classifies the extracted arguments according to their types; integrates the detected events, event categories, arguments and argument categories; and outputs structured information, including events, event categories, arguments and their categories, to quickly and accurately extract key event information from medical texts.

[0108] Figure 7 FIG. 3 is a schematic block diagram of a case text information extraction device 300 provided in an embodiment of the present invention. Figure 7 As shown, corresponding to the above case text information extraction method, the present invention also provides a case text information extraction device 300. The case text information extraction device 300 includes a unit for executing the above case text information extraction method, and the device can be configured in a server. Figure 7 The case text information extraction device 300 includes a text acquisition unit 301, an event detection unit 302, an event classification unit 303, an argument extraction unit 304, an argument classification unit 305 and an output unit 306.

[0109] A text acquisition unit 301 is used to acquire a case text; an event detection unit 302 is used to detect events from the case text to obtain events of the case text; an event classification unit 303 is used to classify the events to obtain event categories; an argument extraction unit 304 is used to extract arguments related to the event from the case text; an argument classification unit 305 is used to classify the arguments to obtain argument categories; an output unit 306 is used to output the event, the event category, the arguments and the argument category.

[0110] In one embodiment, if Figure 8 As shown, the event detection unit 302 includes a text feature extraction subunit 3021 and a detection subunit 3022 .

[0111] The text feature extraction subunit 3021 is used to extract grammatical and semantic features from the case text to obtain a feature extraction result; the detection subunit 3022 is used to input the feature extraction result into an event detection model to perform event detection to obtain events of the case text;

[0112] The event detection model refers to the one obtained by extracting corresponding grammatical and semantic features from a case text dataset containing events and using it as a sample set to train a deep learning model.

[0113] In one embodiment, if Fig. 9 As shown, the event classification unit 303 includes an event feature extraction subunit 3031 and a first classification subunit 3032 .

[0114] The event feature extraction subunit 3031 is used to extract features from the event to obtain event features; the first classification subunit 3032 is used to input the event features into an event classification model for classification to obtain event categories.

[0115] In one embodiment, if Fig.10 As shown, the argument extraction unit 304 includes an information extraction subunit 3041 and an argument extraction subunit 3042 .

[0116] The information extraction subunit 3041 is used to extract boundary information and context information from the case text to obtain an extraction result; the argument extraction subunit 3042 is used to input the extraction result into the argument extraction model to extract arguments to obtain arguments related to the event;

[0117] Among them, the argument extraction model is obtained by extracting corresponding boundary information and context information from several case text data sets containing arguments and using them as sample sets to train a deep learning model.

[0118] In one embodiment, if Fig.11 As shown, the argument classification unit 305 includes an argument feature extraction subunit 3051 and an argument category determination subunit 3052 .

[0119] The argument feature extraction subunit 3051 is used to extract the feature of the argument to obtain the argument feature; the argument category determination subunit 3052 is used to input the argument feature into the argument classification model for classification to obtain the argument category;

[0120] The argument classification model is obtained by extracting corresponding features from data with argument categories and using them as sample sets to train a machine learning model.

[0121] It should be noted that technicians in the relevant field can clearly understand that the specific implementation process of the above-mentioned case text information extraction device 300 and each unit can refer to the corresponding description in the aforementioned method embodiment, and for the convenience and brevity of description, it will not be repeated here.

[0122] Each module in the above-mentioned case text information extraction device 300 can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module above.

[0123] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.12 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client via a network connection. When the computer program is executed by the processor, it implements the functions or steps on the server side of a case text information extraction method.

[0124] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Fig.13 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a client side of a case text information extraction method

[0125] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0126] Acquire case text; detect events from the case text to obtain events of the case text; classify the events to obtain event categories; extract arguments related to the events from the case text; classify the arguments to obtain argument categories; output the events, the event categories, the arguments, and the argument categories.

[0127] In one embodiment, a storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0128] Acquire case text; detect events from the case text to obtain events of the case text; classify the events to obtain event categories; extract arguments related to the events from the case text; classify the arguments to obtain argument categories; output the events, the event categories, the arguments, and the argument categories.

[0129] It should be noted that the functions or steps that can be implemented by the above storage medium or computer device can be correspondingly referred to the relevant descriptions on the server side and the client side in the aforementioned method embodiment. In order to avoid repetition, they will not be described one by one here.

[0130] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0131] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0132] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

[0133] The non-Company software tools or components appearing in the embodiments of this application are merely examples and do not represent actual use.

Claims

1. A case text information extraction method, characterized in that: include: Get case text; Detecting events from the case text to obtain events of the case text; Classifying the events to obtain event categories; Extracting arguments related to the event from the case text; Classifying the arguments to obtain argument categories; The event, the event category, the argument, and the argument category are output.

2. The case text information extraction method according to claim 1, characterized in that: The detecting of events from the case text to obtain events of the case text includes: Extracting grammatical and semantic features from the case text to obtain a feature extraction result; Inputting the feature extraction results into an event detection model to perform event detection to obtain events of the case text; The event detection model refers to the one obtained by extracting corresponding grammatical and semantic features from a case text dataset containing events and using it as a sample set to train a deep learning model.

3. The case text information extraction method according to claim 1, characterized in that: The event is classified to obtain an event category, including: Extracting features from the event to obtain event features; The event features are input into an event classification model for classification to obtain event categories.

4. The method for extracting case text information according to claim 3, characterized in that: The event classification model is obtained by using features extracted from several events as a sample set to train a machine learning model using a remote supervision or semi-supervised learning method.

5. The method for extracting case text information according to claim 1, characterized in that: The step of extracting arguments related to the event from the case text includes: Extracting boundary information and context information from the case text to obtain an extraction result; Inputting the extraction results into an argument extraction model to extract arguments, so as to obtain arguments related to the event; Among them, the argument extraction model is obtained by extracting corresponding boundary information and context information from several case text data sets containing arguments and using them as sample sets to train a deep learning model.

6. The method for extracting case text information according to claim 1, characterized in that: The classifying the arguments to obtain argument categories includes: Extracting features of the argument to obtain argument features; Inputting the argument features into an argument classification model for classification to obtain argument categories; The argument classification model is obtained by extracting corresponding features from data with argument categories and using them as sample sets to train a machine learning model.

7. The method for extracting case text information according to claim 6, characterized in that: An attention mechanism or a graph convolutional network is set in the argument classification model.

8. A case text information extraction device, characterized in that: include: A text acquisition unit, used for acquiring case text; An event detection unit, used for detecting events from the case text to obtain events of the case text; An event classification unit, used for classifying the event to obtain an event category; An argument extraction unit, used for extracting arguments related to the event from the case text; An argument classification unit, used for classifying the arguments to obtain argument categories; An output unit is used to output the event, the event category, the argument and the argument category.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the case text information extraction method as described in any one of claims 1 to 7 are implemented.

10. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the case text information extraction method as described in any one of claims 1 to 7 are implemented.

Images