Biological data entity labeling method and system, electronic equipment and storage medium

Abstract

The invention provides a biological data entity labeling method and system, electronic equipment and a storage medium. The method comprises the steps that a, to-be-labeled biological data is acquired; b, designing an annotation cue word based on the biological data to be annotated; c, splitting the annotated cue word by adopting a cue word splitting method to obtain a plurality of sub cue words; d, inputting the to-be-labeled biological data and the first sub cue word into a pre-trained large language model, and determining first entity labeling data; e, circulating the step d until the entity annotation data of the last sub cue word is obtained; wherein the cue word disassembling method is selected from at least one of a thinking chain technology or a step-by-step questioning method. The method is suitable for entity labeling of the zero-labeling biological data set, professional barriers are remarkably reduced, dependence on domain expert knowledge is reduced, and meanwhile labeling efficiency and accuracy are improved. In addition, dependence on large-scale annotation data is reduced, training and annotation cost is reduced, and the method is flexibly suitable for different biomedical fields.

Description

Technical Field

[0001] The present application relates to the intersection of biology and artificial intelligence, and specifically, to a biological data entity labeling method, system, electronic device and storage medium. Background Art

[0002] With the rapid development of biological research and the continuous growth of biological data, the processing and analysis of biological data sets have become an important task at present. These data sets contain a wealth of professional terms, such as gene sequences, protein structures, disease descriptions, etc. The complexity and professionalism of these data make traditional entity annotation methods face many challenges in practical applications. For example, rule-based and dictionary-based methods require a lot of manual intervention and are difficult to adapt to new entities, resulting in poor flexibility and high annotation costs; although statistical learning and deep learning methods have improved the level of automation, they still require a large amount of annotated data and domain knowledge, and have poor adaptability in new fields; although the method based on pre-trained models has improved performance, it still requires the refinement of domain knowledge and a certain amount of annotated data fine-tuning, and has high computing resource requirements; at the same time, existing entity annotation methods often require a large amount of annotated data for training, which is difficult to achieve in the case of zero-annotated data sets.

[0003] Therefore, current biological data entity annotation methods still need to be improved. Summary of the invention

[0004] The present application aims to solve at least one of the existing problems. To this end, the present application proposes an efficient, accurate and low-cost zero-label biological data entity labeling method.

[0005] Specifically, this application provides the following technical solutions:

[0006] In the first aspect of the present application, the present application proposes a biological data entity labeling method, the method comprising: a. obtaining biological data to be labeled; b. designing labeling prompt words based on the biological data to be labeled; c. using a prompt word decomposition method to split the labeling prompt words to obtain multiple sub-prompt words; d. inputting the biological data to be labeled and the first sub-prompt word into a pre-trained large language model to determine the first entity labeling data; e. looping step d until the entity labeling data of the last sub-prompt word is obtained; wherein the prompt word decomposition method is selected from at least one of the thinking chain technology or the step-by-step questioning method.

[0007] In some examples of this application, this method is suitable for entity annotation of zero-annotated biological data sets, significantly reducing professional barriers, reducing dependence on domain expert knowledge, and improving the efficiency and accuracy of annotation. In addition, it also reduces dependence on large-scale annotated data, reduces training and annotation costs, and is flexibly applicable to different biomedical fields.

[0008] In the second aspect of the present application, the present application proposes a biological data entity annotation system. According to an embodiment of the present application, the system includes: a data acquisition module, used to acquire biological data to be annotated; a prompt word design module, used to design annotation prompt words based on the biological data to be annotated; a sub-prompt word acquisition module, used to adopt a prompt word disassembly method to split the annotation prompt word to obtain multiple sub-prompt words; wherein the prompt word disassembly method is selected from at least one of the thinking chain technology or the step-by-step questioning method; an entity annotation data determination module, used to input the biological data to be annotated and the first sub-prompt word into a pre-trained large language model to determine the first entity annotation data; and the tasks of this module are executed cyclically until the entity annotation data of the last sub-prompt word is obtained.

[0009] In some examples of this application, the system is suitable for entity annotation of zero-annotated biological data sets, significantly reducing professional barriers, reducing dependence on domain expert knowledge, and improving the efficiency and accuracy of annotation. In addition, it also reduces dependence on large-scale annotated data, reduces training and annotation costs, and is flexibly applicable to different biomedical fields.

[0010] In the third aspect of the present application, the present application proposes a computer program product. According to an embodiment of the present application, the computer program product includes: computer instructions; when part or all of the computer instructions are run on a computer, the biological data entity labeling method described in the first aspect of the present application is executed.

[0011] In a fourth aspect of the present application, the present application proposes a computing device. According to an embodiment of the present application, the computing device includes: a processor and a memory; the memory is used to store a computer program; the processor is used to execute the computer program to implement the biological data entity labeling method as described in the first aspect of the present application.

[0012] In a fifth aspect of the present application, the present application proposes a computer-readable storage medium. According to an embodiment of the present application, the computer-readable storage medium stores computer instructions or programs, and when the computer instructions or programs are executed on a computer, the biological data entity annotation method described in the first aspect of the present application is executed.

[0013] In some examples of the present application, the aforementioned computer program product, computing device, and computer-readable storage medium automatically execute the aforementioned biological data entity annotation method through computer instructions, thereby achieving efficient automation and improving annotation efficiency, accuracy, and user experience. Secondly, the characteristics of the instructions make the above method highly consistent and reliable in different application scenarios.

[0014] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 It is a flowchart of a biological data entity annotation method provided by some examples of this application;

[0017] Figure 2 is a flowchart of a biological data entity annotation method provided by other examples of the present application;

[0018] Figure 3 is a schematic diagram of a biological data entity annotation system provided by some examples of this application;

[0019] Figure 4 It is a schematic diagram of an electronic device provided by some examples of this application;

[0020] Figure 5 This is a schematic diagram of the comparison results of F1 scores of the experimental group and the control group provided in some examples of this application. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0022] In this article, unless otherwise specified, the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described here can be implemented in a sequence other than those illustrated or described here. In the present application, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or server comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. In the description of the present application, unless otherwise specified, "multiple" refers to two or more than two.

[0023] This application proposes a biological data entity annotation method, system, electronic device and storage medium. The following describes them in detail:

[0024] Biological Data Entity Annotation Method

[0025] In one aspect of the present application, the present application proposes a method for labeling biological data entities, referring to Figure 1 , the method comprising:

[0026] a. Obtain biological data to be annotated;

[0027] In some examples of the present application, the aforementioned biological data to be annotated is selected from at least one of text data, image data (such as medical images, etc.), audio data (such as medical lectures, audio cases, etc.), tables and structured data (such as experimental results, drug lists, clinical trial data, etc.), preferably text data. The aforementioned text data is selected from at least one of medical literature, biological teaching materials, biotechnology manuals, biological research reports and web pages.

[0028] In some examples of the present application, the aforementioned text data is selected from zero-annotated text data. The so-called "zero-annotated text data" refers to raw text data without manual annotations, comments or labels. These data are usually obtained from public resources, databases, web pages, documents, news reports, etc. They contain rich information, but do not clearly indicate the entities, categories, relationships, etc.

[0029] b. Design annotation prompt words based on the aforementioned biological data to be annotated;

[0030] Analyze the biological data to be annotated obtained in step a. These data may include medical literature, genomic data, experimental data, clinical reports, etc. Design different annotation prompts according to the type of data and application scenario.

[0031] Determine the entity types that need to be annotated in the data to be annotated. For example, in medical literature, you may need to annotate "disease name", "symptoms", "treatment methods", etc.; in genetic data, you may need to annotate "gene name", "mutation type", etc. The entity categories that need to be annotated are different for each data type.

[0032] Design annotation prompts that match the data content according to the characteristics of the data to be annotated. For example, keywords for single entities: "breast cancer", "BRCA1 gene", "chemotherapy drugs", etc.; complete sentences or descriptive phrases: "Early symptoms of breast cancer include lumps, breast pain, and skin changes", "TP53 gene mutations may lead to the occurrence of cancer", etc.; descriptions of entity relationships: "Smoking is closely related to the occurrence of lung cancer", "Cisplatin is a commonly used chemotherapy drug for the treatment of lung cancer", etc.

[0033] In addition to the entities themselves, sometimes you also need to design relationship prompts, which help understand the relationship between entities. For example: "cause": "Gene mutation causes a certain disease"; "related to...": "A certain drug is related to cancer treatment"; "inhibit" or "activate": "A certain drug inhibits tumor growth."

[0034] When designing annotation prompts, the context of the data to be annotated needs to be considered. For example, in medical literature, the same disease may be described in different ways, such as "lung cancer" and "non-small cell lung cancer" can be expressed in different terms. Prompts should try to cover these changes to avoid omissions.

[0035] In the biomedical field, different diseases, symptoms, or drugs may have multiple related terms or variants. Therefore, the prompt words need to cover these variants when they are designed. For example, breast cancer may have multiple expressions, such as "breast cancer", "breast tumor", "breast malignant tumor", etc.

[0036] When designing annotation prompt words, it is also possible to do so in a “template or rule” manner. For example, a fixed structured phrase or sentence format can be designed, such as “[gene name] mutation” or “[drug] treatment of [disease]”.

[0037] Exemplarily, professional terms related to the following categories are extracted:

[0038] Diseases (e.g., cancer, adenoma, tumor, etc.); symptoms (e.g., fever, fatigue, pain, etc.); reactions (e.g., allergic reactions, immune reactions, etc.); drugs (e.g., chemotherapy drugs, penicillin, etc.); genes (e.g., BRCA1, P53, etc.); disease phenomena (e.g., metastasis, remission, etc.); substances (e.g., glucose, proteins, enzymes, etc.);

[0039] Note:

[0040] Exclude pronouns and non-professional terms; if disease-related words such as "cancer", "adenoma" or "tumor" appear, extract them and their related terms; if there are hyphens in the vocabulary (such as "Ataxia-telangiectasia"), treat it as a whole and do not split it; if abbreviations or acronyms appear (such as Ataxia-telangiectasia (AT)), extract the full name and abbreviation separately (such as "Ataxia-telangiectasia" and "AT"); if "and", " / " or "or" connect multiple entities, extract them as a whole; extract all abbreviations in capital letters (regardless of the specific meaning) and treat them as separate entities.

[0041] c. Using the prompt word decomposition method, the aforementioned marked prompt word is split to obtain multiple sub-prompt words;

[0042] In order to improve the accuracy of entity annotation and the efficiency of model processing, the annotation prompt word obtained in step b is disassembled to obtain multiple sub-prompt words. The above prompt word disassembly method is selected from at least one of the thought chain technology or the step-by-step questioning method.

[0043] The term "thinking chain technology" is a strategy to improve the model's capabilities by "decomposing the reasoning process." The core idea is to guide the model to perform logical reasoning step by step, rather than just giving a direct answer. In other words, the thinking chain technology encourages the model to perform a series of intermediate reasoning steps before answering the question, thereby helping the model to derive the final conclusion. In traditional question-answering systems, the model directly gives answers based on the questions, and often fails to make full use of the intermediate steps in the reasoning process. The thinking chain technology requires the model to "show the thinking process," that is, before answering, first list all the reasoning steps, and then derive the final answer based on this. This method imitates the human thinking process, making the model's answers more accurate and reasonable.

[0044] For example, suppose the question is: "If a patient is diagnosed with breast cancer, what should her treatment options include?"

[0045] Thinking chain reasoning steps:

[0046] Step 1: "Breast cancer" refers to a malignant tumor that originates from breast tissue;

[0047] Step 2: "Breast cancer" can be divided into different subtypes (e.g. hormone receptor positive, HER2 positive, etc.) according to its type and stage;

[0048] Step 3: For breast cancer treatment, firstly, it is necessary to evaluate the stage of cancer (such as stage I, II, III or IV) and whether there is lymph node metastasis;

[0049] Step 4: Depending on the stage and subtype, treatment options usually include surgery, radiotherapy, chemotherapy, targeted therapy or endocrine therapy;

[0050] Early-stage breast cancer may be treated with surgery to remove the tumor;

[0051] If the cancer has spread (for example, to the lymph nodes or other organs), chemotherapy or targeted therapy may be needed;

[0052] Final Steps: Therefore, the treatment plan for breast cancer will be customized according to the patient's specific condition and may include surgery, radiotherapy, chemotherapy, targeted therapy, endocrine therapy and other steps.

[0053] Output: Treatment options for breast cancer usually include surgery, radiation therapy, chemotherapy, targeted therapy or endocrine therapy. The specific treatment method depends on the type and stage of cancer and the individual situation of the patient.

[0054] The inventors of this application have been able to reduce the probability of the model outputting erroneous results and significantly improve the accuracy and efficiency of annotation by applying the mind chain technology to entity annotation of complex biological data.

[0055] The term "step-by-step questioning" is a strategy for breaking down a complex task into multiple independent and relatively simple questions. Each sub-question focuses on a certain aspect or sub-task of the problem, allowing the model to solve the problem step by step and eventually get a comprehensive answer. The core idea of ​​step-by-step questioning is to break down a complex problem into multiple small questions, each of which answers a part or stage of the problem. In this way, the model can focus on simple sub-tasks, reduce the chance of errors, and gradually deduce the complete answer.

[0056] For example, let's assume the question is: "A patient presents with a high fever, cough, and shortness of breath. What is the likely diagnosis?"

[0057] Steps of the step-by-step questioning method:

[0058] First question: Does the patient have symptoms of an upper respiratory tract infection?

[0059] A: Yes, manifested as cough and shortness of breath.

[0060] The second question: Does the patient have fever symptoms?

[0061] Answer: Yes, high fever (maybe body temperature over 38.5℃).

[0062] Third question: Do these symptoms suggest viral or bacterial pneumonia?

[0063] A: These symptoms (especially fever and shortness of breath) are often associated with pneumonia and may be manifestations of bacterial or viral pneumonia.

[0064] The fourth question: Has the patient been in contact with pneumonia patients or people at high risk of respiratory infections?

[0065] A: Assuming the patient has a history of relevant contact, this increases the likelihood of pneumonia.

[0066] Final Answer: Based on the patient's symptoms and medical history, the most likely diagnosis is pneumonia, which may be caused by bacteria or a virus.

[0067] Output:

[0068] The patient's symptoms (high fever, cough, and shortness of breath) may suggest pneumonia. The specific cause (bacterial or viral) requires further clinical examination and testing.

[0069] The inventors of the present application applied the step-by-step questioning method to the entity annotation of complex biological data to ensure the accuracy and detail of each entity annotation; effectively improve the flexibility and adaptability of the model; the model can flexibly adjust the order and content of questions for different biological data types or annotation requirements to cope with diverse annotation scenarios; it helps to reduce the risk of model errors, and by gradually guiding the model to complete the annotation task, it avoids directly skipping key steps or deriving erroneous results, thereby improving the reliability and accuracy of the annotation.

[0070] In some examples of the present application, this step further includes logically sorting or importance sorting the aforementioned multiple sub-prompt words. By logically or importance sorting the sub-prompt words, the accuracy of entity annotation can be effectively improved and the adaptability to different biological data can be increased.

[0071] d. inputting the aforementioned biological data to be annotated and the first sub-prompt word into the pre-trained large language model to determine the first entity annotated data;

[0072] In some examples of the present application, the aforementioned pre-trained large language model is a natural language processing model based on deep learning, such as BERT, GPT or its variants. In a preferred example of the present application, the aforementioned pre-trained model is selected from GPT-4. The aforementioned pre-trained model can come from a public open source model, such as Hugging Face Model Hub, etc., or it can be obtained by fine-tuning an open source model.

[0073] e. Repeat step d until the entity annotation data of the last sub-prompt word is obtained.

[0074] The loop step d includes: inputting the Nth entity annotation data, the biological data to be annotated, and the N+1th sub-prompt word into the pre-trained large language model to determine the N+1th entity annotation data, wherein N is an integer not less than 1. For example, inputting the first entity annotation data, the biological data to be annotated, and the second sub-prompt word into the pre-trained large language model to determine the second entity annotation data.

[0075] Through multiple rounds of iterations and gradual refinement of the annotation results, we ensure the accuracy, completeness, and coverage of the annotated data, especially when facing complex biomedical data. Each round of annotation not only improves the model's understanding of the data, but also helps correct errors and fill in omissions, ultimately obtaining high-quality entity annotation data.

[0076] In some examples of the present application, the aforementioned method further includes: using a regular matching algorithm to standardize the entity labeling results of the aforementioned pre-trained large language model. Through the standardization operation, it is avoided that the ambiguous or confusingly formatted data generated during the labeling process affects the next round of labeling. The clarity of the labeling is ensured, and the stability and reliability of the model when processing complex data are enhanced.

[0077] In some examples of the present application, the aforementioned method further includes: reviewing the entity annotation data of the aforementioned last sub-prompt word to ensure that the final annotation result has high quality and high credibility.

[0078] In some examples of the present application, the aforementioned method further includes: evaluating the aforementioned pre-trained model based on the audit result, and ensuring the effectiveness and accuracy of the model in practical applications by evaluating the performance of the pre-trained model.

[0079] In other examples of this application, reference Figure 2 , the method includes: collecting data to be extracted (such as zero-annotated literature abstracts) and selecting a large language model; combining the thinking chain technology and the step-by-step questioning method, formulating prompt words for the data to be extracted based on the entity data extraction framework; in the entity data extraction framework, inputting prompt word 1 (same as the "first sub-prompt word") and the data to be extracted (source text) into the selected large language model to obtain answer 1 (same as the "first entity annotated data");

[0080] Optionally, an algorithm (such as a regular matching algorithm, etc.) is used to standardize the above answer 1 to obtain a candidate word 1;

[0081] Input the above output result or candidate word 1, prompt word 2 (same as “second sub-prompt word”) and source text into the selected large language model to obtain answer 2 (same as “second entity annotation data”);

[0082] Optionally, an algorithm (such as a regular matching algorithm, etc.) is used to standardize the aforementioned answer 2 to obtain a candidate word 2;

[0083] Repeat the above steps once or multiple times to obtain the entity annotation result of the source text;

[0084] Optionally, manually review the entity annotation results of the source text;

[0085] Optionally, the large language model is further evaluated based on manual review results.

[0086] The aforementioned biological data entity labeling method reduces the reliance on domain knowledge experts and a large amount of labeled data, and can achieve accurate labeling in the absence of specially labeled data. At the same time, through the thinking chain technology, the labeling accuracy and interpretability of the model are improved, solving the "black box" problem in traditional methods. In addition, zero-sample learning effectively reduces the cost of data labeling and model training, and reduces the demand for computing resources. In terms of domain adaptability, this method can flexibly respond to different biomedical fields and emerging entities, ensuring long-term stability and effectiveness.

[0087] Biological Data Entity Annotation System

[0088] On the other hand, this application proposes a biological data entity annotation system, referring to Figure 3 The system includes: a data acquisition module 100, a prompt word acquisition module 200, a sub-prompt word acquisition module 300 and an entity annotation data determination module 400.

[0089] 100 modules, used to obtain biological data to be annotated.

[0090] In some examples of the present application, the aforementioned biological data to be annotated is selected from at least one of text data, image data, audio data, table and structured data, preferably text data. The aforementioned text data is selected from at least one of medical literature, biological teaching materials, biotechnology manuals, biological research reports and web pages.

[0091] In some examples of the present application, the aforementioned text data is selected from zero-annotated text data.

[0092] Module 200 is used to design annotation prompt words based on the aforementioned biological data to be annotated.

[0093] Module 300 is used to use a prompt word decomposition method to decompose the aforementioned marked prompt word to obtain multiple sub-prompt words; wherein the aforementioned prompt word decomposition method is selected from at least one of the thinking chain technology or the step-by-step questioning method.

[0094] In some examples of the present application, the module further includes: performing logical sorting or importance sorting on the aforementioned multiple sub-prompt words.

[0095] Module 400 is used to input the aforementioned biological data to be annotated and the first sub-prompt word into the pre-trained large language model to determine the first entity annotation data; and execute the tasks of this module in a loop until the entity annotation data of the last sub-prompt word is obtained.

[0096] In some examples of the present application, the aforementioned pre-trained large language model is a natural language processing model based on deep learning, such as BERT, GPT or their variants.

[0097] In some examples of the present application, the tasks of this module are executed cyclically, including: inputting the aforementioned Nth entity annotation data, the aforementioned biological data to be annotated, and the N+1th sub-prompt word into the pre-trained large language model, and determining the N+1th entity annotation data, where N is an integer not less than 1.

[0098] In some examples of the present application, this module further includes: using a regular matching algorithm to standardize the entity labeling results of the aforementioned pre-trained large language model.

[0099] In some examples of the present application, this module further includes: reviewing the entity annotation data of the aforementioned last sub-prompt word.

[0100] In some examples of the present application, this module further includes: evaluating the aforementioned pre-trained model based on the audit results.

[0101] In some examples of the present application, the aforementioned modules are connected, and the aforementioned connection includes a physical connection or a network connection.

[0102] Those skilled in the art will appreciate that the features and advantages described above for the biological data entity labeling method are also applicable to the above system and will not be described in detail here.

[0103] It should be understood that the system embodiment and the method embodiment may correspond to each other, and similar descriptions may refer to the method embodiment. To avoid repetition, they will not be described here. Specifically, Figure 3 The system shown can execute the above-mentioned embodiment of the biological data entity labeling method, and the operations and / or functions performed by each module in the system correspond to the method embodiment, which will not be described here for brevity.

[0104] The system of the embodiment of the present application is described above from the perspective of the functional module in conjunction with the accompanying drawings. It should be understood that the functional module can be implemented in hardware form, can be implemented by instructions in software form, and can also be implemented by a combination of hardware and software modules. Specifically, the steps of the method embodiment in the embodiment of the present application can be completed by the hardware integrated logic circuit and / or software form instructions in the processor, and the steps of the method disclosed in the embodiment of the present application can be directly embodied as a hardware decoding processor to perform, or a combination of hardware and software modules in the decoding processor to perform. Optionally, the software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, an electrically erasable programmable memory, a register, etc. The storage medium is located in a memory, and the processor reads the information in the memory, and completes the steps in the above method embodiment in conjunction with its hardware.

[0105] Computer program product, computing device, and computer readable storage medium

[0106] In another aspect, the present application proposes a computer program product, a computing device, and a computer-readable storage medium. Based on the aforementioned computer program product, computing device, or computer-readable storage medium, the aforementioned biological data entity labeling method is executed.

[0107] This embodiment takes electronic equipment as an example for detailed description.

[0108] The so-called electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers. Computing devices may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are intended to be examples only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0109] like Figure 4 As shown, the electronic device 500 includes a computing unit 501, which can perform various appropriate actions and processes according to a computer program stored in a ROM (Read-Only Memory) 502 or a computer program loaded from a storage unit 508 to a RAM (Random Access Memory) 503. In the RAM 503, various programs and data required for the operation of the device 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An I / O (Input / Output) interface 505 is also connected to the bus 504.

[0110] A number of components in the device 500 are connected to the I / O interface 505, including: an input unit 506, such as a keyboard, a mouse, etc.; an output unit 507, such as various types of displays, speakers, etc.; a storage unit 508, such as a disk, an optical disk, etc.; and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the device 500 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0111] The computing unit 501 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a CPU (Central Processing Unit), a GPU (Graphic Processing Units), various dedicated AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, a DSP (Digital Signal Processor), and any appropriate processor, controller, microcontroller, etc. The computing unit 501 performs the various methods and processes described above, such as a biological data entity annotation method. For example, in some embodiments, the biological data entity annotation method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 508. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 500 via the ROM 502 and / or the communication unit 509. When the computer program is loaded into the RAM 503 and executed by the computing unit 501, one or more steps of the method described above may be performed. Alternatively, in other embodiments, the computing unit 501 may be configured to execute the aforementioned biological data entity labeling method in any other appropriate manner (eg, by means of firmware).

[0112] In the present application, the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, which can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program may be printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or processing in other suitable ways as necessary, and then stored in a computer memory. The various computer-readable storage media described herein may represent one or more devices and / or other machine-readable storage media for storing information. The term "machine-readable storage medium" may include, but is not limited to, wireless channels and various other media capable of storing, containing, and / or carrying instructions and / or data.

[0113] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0114] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0115] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0116] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0117] Embodiments of the present invention will be described in more detail below, examples of which are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0118] Example 1: Entity Annotation Application

[0119] This embodiment takes part of the content of an English document as an example to perform entity annotation to verify the feasibility of the present application solution.

[0120] Pre-trained large language model: GPT-4;

[0121] Original text to be extracted:

[0122] "text":"Ataxia-telangiectasia(A-T)is a recessive multi-system disorder caused by mutations in the ATM gene at 11q22-q23(ref.3).The risk of cancer,especially lymphoid neoplasias,is substantially elevated in A-T patients and has long been associated with chromosomal instability.By analysing tumour DNA from patients with sporadic T-cell prolymphocytic leukaemia(T-PLL),a rare clonal malignancy with similarities to a mature T-cell leukaemia seen in A-T,we demonstrate a high frequency of ATM mutations in T-PLL.In marked contrast to the ATM mutation pattern in A-T,the most frequent nucleotide changes in this leukaemia were missense mutations.";

[0123] Prompt 1:

[0124] """Assuming you are an entity extraction model.You will receive sentences related to biology and medicine.You need to complete two tasks.Task 1:Extract as many professional terms related to diseases,symptoms,reactions,drugs,genes,disease phenomena,and substances as possible.

[0125] Note:

[0126] Do not include pronouns.

[0127] If terms such as“cancer,”“adenoma,”or“tumor”appear,extract them orany related disease terms.

[0128] If hyphens appear,treat the entire phrase as a single entity;do notsplit the words on either side of the hyphen.

[0129] If there is an abbreviation with an explanation,return theabbreviation separately.For example,Ataxia-telangiectasia(A-T)should bereturned as Ataxia-telangiectasia and A-T.

[0130] If“and,”“ / ”,or“or”connect multiple entities,treat them as a singleterm and extract them together.

[0131] If uppercase letters appear as an abbreviation or acronym,regardlessof what it means,extract them as a separate word.

[0132] Return a list of terms or sub-terms that meet these criteria.

[0133] Task 2:Identify all the acronym words composed of uppercase lettersin the text.

[0134] Return a list of all entities from both Task 1and Task 2,for example:['a”,'b'].If no terms meet the criteria,return‘[]’.

[0135] text:"{text}"

[0136] Answer 1: "['Ataxia-telangiectasia','A-T','recessive multi-systemdisorder','mutations','ATM gene','cancer','lymphoid neoplasias','chromosomalinstability','tumour DNA','sporadic T-cell prolymphocytic leukaemia','T-PLL','rare clonal malignancy','mature T-cell leukaemia','high frequency','ATMmutations','nucleotide changes','missense mutations','A-T','T-PLL','ATM']",

[0137] Prompt 2:

[0138] """

[0139] You will be given a text and a list of keywords.Please understand thecontent of the text,then analyze each keyword individually to determine itsmeaning in the text.Assess whether it refers to a disease name,a disease-related symptom,or an abbreviation of a disease.Finally,provide a summarylist of the terms that meet these criteria,such as[‘a’,‘b’].If no terms meetthe criteria,return‘[]’.Note that if the termincludes“cancer”“tumor”or“adenoma”return the disease names mentioned in the text.If the text onlymentions“cancer,”“adenoma,”or“tumor”without specifying a disease name,returnonly“cancer,”“adenoma,”or“tumor.”Descriptive terms and descriptions ofdisease-related abnormalities are also considered diseases.If there areabbreviations among the candidate terms,determine their specific meanings inthe text and evaluate whether they meet the criteria.If they do,return theabbreviation without expanding it into its full form.

[0140] Note:

[0141] 症状或描述性术语属于注释范围,“遗传遗传模式”也属于此范围。“缺乏性症状”属于疾病范围。“异常”属于注释范围,“遗传模式”、“某些类型的疾病”、“遗传疾病”和“遗传现象”也属于此范围。

[0142] 句子:“{文本}”

[0143] 判断列表:“{答案1}”

[0144] Answer 2: ['Ataxia-telangiectasia', 'A-T','sporadic T-cell prolymphocytic leukaemia', 'T-PLL', 'lymphoid neoplasias','mature T-cell leukaemia', 'cancer']",

[0145] Prompt 3:

[0146] 你是一个疾病提取机器人。你将收到一段源文本和一份待筛选的术语列表。你需要:

[0147] 任务1:确定列表中的术语是否存在于文本中。

[0148] Task 2:Evaluate each termindividually to determine if it meets thecriteria for disease classification.The criteria for inclusion are broad andinclude:

[0149] Disease names

[0150] Descriptive terms related to diseases

[0151] Genetic inheritance patterns

[0152] Absence of gender-specific symptoms

[0153] Abnormalities

[0154] Certain types of diseases

[0155] Genetic diseases

[0156] Genetic phenomena

[0157] Genetic patterns

[0158] Outcome-oriented terms(e.g.,premature aging)

[0159] Terms that should not be classified as diseases include:

[0160] Variants not directly related to specific diseases,such as missensemutations

[0161] Gene names,like PAX6

[0162] Concepts,such as disease susceptibility

[0163] Protein names

[0164] Genetic phenomena not directly related to the occurrence of specificdiseases

[0165] Names of biological structures

[0166] Finally,provide a summary list of the terms that meet these criteria,such as[‘a’,‘b’].If no terms meet the criteria,return‘[]’.

[0167] Sentence:"{text}"

[0168] Judgment list:"{answer2}"

[0169] Answer 3: Task 1: Determine if the terms exist in the text.\nAll the terms in the judgment list exist in the provided sentence.\n\nTask 2: Evaluate each term for disease classification.\n\n1. **Ataxia-telangiectasia**: This is a known disease characterized by a combination of ataxia and telangiectasia; thus, it meets the criteria for classification.\n2. **A-T**: This is an acronym for Ataxia-telangiectasia, which refers to the same disease. It meets the criteria for classification.\n3. **sporadic T-cell prolymphocytic leukaemia**: This is a specific type of cancer with regards to T-cells and qualifies as a disease.\n4. **T-PLL**: This is an acronym for sporadic T-cell prolymphocytic leukaemia, thus qualifying as a disease.\n5. **lymphoid neoplasias**: This term refers to cancers of the lymphoid tissue and qualifies as a disease category.\n6. **mature T-cell leukaemia**: This is a specific type of cancer involving mature T-cells and qualifies as a disease.\n7.**cancer**:This is a broad term encompassing various diseasescharacterized by abnormal cell growth,thus qualifying.\n\n###Summary List\nAll terms meet the criteria for disease classification.\n\n**Summary:**\n['Ataxia-telangiectasia','A-T','sporadic T-cell prolymphocytic leukaemia','T-PLL','lymphoid neoplasias','mature T-cell leukaemia','cancer']".

[0170] Extraction output: ['Ataxia-telangiectasia','A-T','sporadic T-cellprolymphocytic leukaemia','T-PLL','lymphoid neoplasias','mature T-cellleukaemia','cancer']

[0171] In prompt words 1, 2, and 3, a step-by-step questioning method is used to gradually add qualifiers and require a large language model to check the output content.

[0172] In prompt word 3, a thought chain prompt word is used, requiring the large language model to first interpret the word and then make a judgment.

[0173] Finally, according to the output of the large language model, a regular matching algorithm is used to obtain the result output.

[0174] Example 2: Comparison of different entity labeling methods

[0175] This example uses a public data set to verify the advantages of the present application solution over the prior art. The details are as follows:

[0176] Public dataset sources: Gene dataset (https: / / github.com / spyysalo / bc2gm-corpus); Chemical dataset (https: / / ftp.ncbi.nlm.nih.gov / pub / lu / BC5CDR / ); Disease dataset (https: / / www.ncbi.nlm.nih.gov / CBBresearch / Dogan / DISEASE / );

[0177] Control group: Questions were asked directly according to the annotation rules of the dataset, without using step-by-step questions and thinking chains;

[0178] Original text to be extracted:

[0179] "text":"Ataxia-telangiectasia(AT) is a recessive multi-systemdisorder caused by mutations in the ATM gene at 11q22-q23(ref.3). The risk of cancer, especially lymphoid neoplasias, is substantially elevated in A-Tpatients and has long been associated with chromosomal instability. Byanalysing tumour DNA from patients with sporadic T-cell prolymphocyticleukaemia(T-PLL), a rare clonal malignancy with similarities to a mature T-cell leukaemia seen in AT, we demonstrate a high frequency of ATM mutations in T-PLL.In marked contrast to the ATM mutation pattern in AT, the mostfrequent nucleotide changes in this leukaemia were missense mutations.";

[0180] Tips:

[0181] """

[0182] You will receive a sentence related to biology.Based on the contentof the sentence and your existing knowledge,please analyze each term todetermine its meaning in the biological field and assess whether it meets thefollowing criteria:

[0183] "What to Annotate:

[0184] Annotate all Specific Disease mentions

[0185] Atextual string referring to a disease name may refer to a SpecificDisease,or a Disease Class.Disease mentions that could be described as afamily of many Specific Diseases were annotated with an annotation categorycalled Disease Class.The annotation category Specific Disease was used forthose mentions which could be linked to one specific definition that does notinclude further categorization.

[0186] e.g.Diastrophic dysplasia is an autosomal recessive diseasecharacterized by short stature,very short limbs and joint problems thatrestrict mobility.

[0187] Annotate:"Diastrophic dysplasia"as Specific Disease category and"autosomal recessive disease"as Disease Class category

[0188] Annotate contiguous text strings.

[0189] Atextual string may refer to two or more separate diseasementions.Such mentions are annotated with the Composite Mention category.

[0190] e.g.The text phrase"Duchenne and Becker muscular dystrophy"refers totwo separate diseases.If this phrase is separated into two strings:"Duchenne"and"Becker muscular dystrophy",it results in information loss,because theword"Duchenne"on its own is not a disease mention.

[0191] Annotate disease mentions that are used as Modifiers for otherconcepts

[0192] A textual string may refer to a disease name,but it may modify a nounphrase,or not be a noun phrase and this is better expressed with the Modifierannotation category.

[0193] e.g.:Although this mutation was initially detected in four of33colorectal cancer families analysed from eastern England,more extensiveanalysis has reduced the frequency to four of52English HNPCC kindredsanalysed.

[0194] Annotate:"colorectal cancer"as Modifier category and"HNPCC"asModifier category.

[0195] Annotate duplicate mentions.

[0196] For each sentence in the PubMed abstract and title,the locations ofall disease mentions are marked,including duplicates within the samesentence.

[0197] Annotate minimum necessary span of text.

[0198] The minimum span of text necessary to include all the tokensexpressing the most specific form of the disease is preferred.

[0199] e.g.,in the case of"insulin-dependent diabetes mellitus",the diseasemention including the whole phrase was preferred over its substrings such as"diabetes mellitus"or"diabetes".

[0200] Annotate all synonymous mentions.

[0201] Abbreviation definitions such as"Huntington disease"("HD")areseparated into two annotated mentions.

[0202] 2:What NOT to Annotate:

[0203] Do not annotate organism names.

[0204] Organism names such as"human"were excluded from the preferredmention.Viruses,bacteria,and other organism names were not annotated unlessit was clear from the context that the disease caused by these organisms isdiscussed.

[0205] e.g.Studies of biopsied tissue for the presence of Epstein-Barr virusand cytomegalovirus were negative.

[0206] In this case:"Epstein-Barr virus"and"cytomegalovirus"are annotated asSpecific Disease category.

[0207] Do not annotate gender.

[0208] Tokens such as"male"and"female"were only included if theyspecifically identified a new form of the disease,

[0209] e.g."male breast cancer".

[0210] Do not annotate overlapping mentions.

[0211] For example,the phrase"von Hippel-Lindau(VHL)disease"was annotated asone single disease mention,Specific Disease category.

[0212] Do not annotate general terms.

[0213] Very general terms such as:disease,syndrome,deficiency,complications,abnormalities,etc.were excluded.However,the terms cancer and tumor wereretained.

[0214] Do not annotate references to biological processes.

[0215] e.g.Terms corresponding to biological processes such as"tumorigenesis"or"cancerogenesis".

[0216] Do not annotate disease mentions interrupted by nested mentions.

[0217] Basically,do not break the contiguous text rule.

[0218] e.g.WT1 dysfunction is implicated in both neoplastic(Wilms tumor,mesothelioma,leukemia,and breast cancer)and nonneoplastic(glomerulosclerosis)disease.

[0219] In this example,the list of all disease mentions includes:"neoplasticdisease"and"nonneoplastic disease"in addition to"Wilms tumor","mesothelioma","leukemia","breast cancer"and"glomerulosclerosis".However,theywere not annotated in our corpus,because other tokens break up the phrases.

[0220] Examples:

[0221] "Insulin-dependent diabetes mellitus",Specific Disease,prefer thewhole string

[0222] CDH1 mutations predispose to early onset"colorectal cancer"

[0223] "early onset"may or may not be part of the disease mention,depends on1.If ether is a UMLS concept specifying this as a separate form of disease,and 2.If the annotator believes it should be included.

[0224] Human"X-linked recessive disorder"

[0225] Her fresh serum lacked complement-mediated bactericidal activityagainst Neisseria gonorrhoeae.

[0226] "Huntington disease"("HD"),the long form and the short formconstitute two separate mentions

[0227] This report represents the first cases of"C7 deficiency"associatedwith infectious complications

[0228] "infectious complications"is too general a term;C7 deficiency howeveris a Specific Disease.

[0229] "colorectal,endometrial,and ovarian cancers"is considered oneComposite Mention of several Specific Diseases.

[0230] WT1 dysfunction is implicated in both neoplastic("Wilms tumor","mesothelioma","leukemias",and"breast cancer")and nonneoplastic("glomerulosclerosis")disease.

[0231] If we list all disease mentions in this sentence we have thefollowing:Neoplastic disease,nonneoplastic disease,wilms tumor,mesothelioma,leukemias,breast cancer,glomerulosclerosis.However,"neoplastic disease"or"nonneoplastic disease"are not annotated because these mentions areintercepted with other tokens.

[0232] Tumorigenesis and cancerogenesis are not annotated as diseases inthis corpus.These terms correspond to the process by which normal cells aretransformed into cancer cells.

[0233] "autosomal recessive disease"refers to a whole family of diseaseswhich can be passed down through families because of their geneticcharacteristics"Disease Class category

[0234] Large phenotypic variability has been observed,with convulsivedisorders,motor retardation and mental retardation being the most abundantmanifestations.

[0235] "convulsive disorders","motor retardation"and"mental retardation"areexamples of the grey area terms which may be classified as disease mentions,but do not correspond to one Specific Disease,therefore they are annotated asDisease Class category

[0236] Borjeson-Forssman-Lehmann syndrome(BFLS)is a syndromal X-linkedmental retardation.

[0237] In this case,"X-linked mental retardation syndrome"is a family ofdiseases,and"Borjeson-Forssman-Lehmann syndrome"is a Specific Disease of thatfamily.

[0238] Acute meningococcal pericarditis"Constitutes a disease mention and,exists as a separate concept in UMLS,however

[0239] Acute Neisseria infection"May or may not include"acute"in the mentionname.

[0240] Classical galactosemia"Includes"classical",because it corresponds toa particular form of the disease.

[0241] Inherited spinocerebellar ataxia"May or may not include"inherited",depends on annotator

[0242] Adenomatous polyps of the colon and rectum,-Composite Mention

[0243] Fibroepithelial or epithelial hyperplasias,-Composite Mention

[0244] Stage II or stage III colorectal cancer"Composite Mention

[0245] "

[0246] Do not modify the terms in the noun list.Finally,summarize byreturning a list of terms that meet the criteria,such as['a','b'].If no terms meet the criteria,return "[]".

[0247] Sentence:"{text}"

[0248] """

[0249] Extract output: ['Ataxia-telangiectasia','A-T','sporadic T-cellprolymphocytic leukaemia','T-PLL','lymphoid neoplasias','mature T-cellleukaemia']

[0250] Experimental group: see the detailed description of the embodiments of this application;

[0251] The F1 score is used to evaluate different schemes. The results are shown in Table 1. Figure 5 As shown, compared with the control group, the predicted F1 scores of the experimental group were improved by 0.27 (gene dataset), 0.223 (chemical dataset), and 0.262 (disease dataset), respectively.

[0252] Table 1

[0253] F1-score Control group Experimental Group Gene dataset 0.593 0.863 Chemical Datasets 0.649 0.872 Disease datasets 0.531 0.793

[0254] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0255] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in the field can change, modify, replace and modify the above embodiments within the scope of the present application without departing from the principles and purpose of the present application.

Claims

1. A biological data entity labeling method, characterized in that: include: a. Obtain biological data to be annotated; b. Designing annotation prompt words based on the biological data to be annotated; c. Using a prompt word decomposition method, the marked prompt word is split to obtain multiple sub-prompt words; d. inputting the biological data to be annotated and the first sub-prompt word into the pre-trained large language model to determine the first entity annotated data; e. Repeat step d until the entity annotation data of the last sub-prompt word is obtained; Wherein, the prompt word disassembly method is selected from at least one of the thought chain technique or the step-by-step questioning method.

2. The method according to claim 1, characterized in that The biological data to be annotated is selected from at least one of text data, image data, audio data, table and structured data, preferably text data.

3. The method according to claim 2, characterized in that The text data is selected from at least one of medical literature, biological teaching materials, biotechnology manuals, biological research reports and web pages.

4. The method according to claim 3, characterized in that The text data is selected from zero-annotated text data.

5. The method according to claim 4, characterized in that Step c further includes performing a logical sort or an importance sort on the multiple sub-prompt words.

6. The method according to claim 1, characterized in that The loop step d includes: inputting the Nth entity annotation data, the biological data to be annotated, and the N+1th sub-prompt word into a pre-trained large language model to determine the N+1th entity annotation data, wherein N is an integer not less than 1; Optionally, the method further comprises: using a regular matching algorithm to standardize the entity labeling results of the pre-trained large language model; Optionally, the method further comprises: reviewing entity annotation data of the last sub-prompt word; Optionally, the method further comprises: evaluating the pre-trained model based on the audit results.

7. The method according to any one of claims 1 to 6, characterized in that: The pre-trained large language model is a natural language processing model based on deep learning, such as BERT, GPT or their variants.

8. A biological data entity annotation system, characterized in that: include: A data acquisition module, used to acquire biological data to be annotated; A prompt word design module, used for designing a labeling prompt word based on the biological data to be labeled; a sub-prompt word acquisition module, used to split the marked prompt word by a prompt word disassembly method to obtain a plurality of sub-prompt words; wherein the prompt word disassembly method is selected from at least one of a thought chain technique or a step-by-step questioning method; The entity annotation data determination module is used to input the biological data to be annotated and the first sub-prompt word into the pre-trained large language model to determine the first entity annotation data; input the biological data to be annotated, the second sub-prompt word and the first entity annotation data into the pre-trained large language model to determine the second entity annotation data; and execute the tasks of this module in a loop until the entity annotation data of the last sub-prompt word is obtained.

9. A computing device, characterized in that include: Processor and memory; The memory is used to store computer programs; The processor is used to execute the computer program to implement the biological data entity labeling method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions or programs, and when the computer instructions or programs are executed on a computer, the biological data entity labeling method according to any one of claims 1 to 7 is executed.

Images