Method and device for automatically labeling analogy corpus through computing power of intelligent computing center

Abstract

This invention provides a method and apparatus for automatically annotating analogical corpora using the computing power of an intelligent computing center. The method includes: determining prompt words, which contain analogical corpus pairs as examples and their analogy relationship analysis process; calling a large language model running in the intelligent computing center to analyze the analogy relationship analysis process of the analogical corpus pairs as examples in the prompt words; analyzing the analogy relationships of unannotated analogical corpus pairs; obtaining the analogy relationship analysis results of unannotated analogical corpus pairs; and annotating the analogy relationship types of the analogical corpus pairs based on the analogy relationship analysis results. High-quality analogical corpus pairs can be selected, thereby improving the cognitive ability and analogical reasoning ability of the large language model trained with these high-quality analogical corpus pairs.

Description

Technical Field

[0001] The embodiments of the present invention relate to the fields of computing infrastructure and artificial intelligence technology, and in particular to a method and apparatus for automatically annotating analogical corpora using the computing power of an intelligent computing center. Background Technology

[0002] With the development of artificial intelligence and computing power technologies, the concept of intelligent computing centers has emerged. An "intelligent computing center" refers to a facility that utilizes large-scale heterogeneous computing resources, including general-purpose and intelligent computing power, to provide the necessary computing power, data, and algorithms for artificial intelligence applications (such as the development, training, and inference of deep learning models, including large language models). Intelligent computing centers encompass facilities, hardware, and software, and can provide full-stack capabilities from underlying computing power to top-level application enablement.

[0003] Existing intelligent computing centers require a large amount of corpus to train large language models. The large language models learn the semantic relationships in the corpus to complete specific tasks. However, obtaining high-quality corpus is an urgent problem to be solved. If the corpus used to train the large language model is of low quality, it cannot be guaranteed that the trained large language model will have stable and excellent performance. Summary of the Invention

[0004] This invention provides a method and apparatus for automatically annotating analogy corpora using the computing power of intelligent computing centers. This addresses the problem that existing intelligent computing centers use low-quality corpora for training large language models, which fails to guarantee stable and excellent performance of the trained large language models.

[0005] To solve the above-mentioned technical problems, the present invention is implemented as follows:

[0006] In a first aspect, embodiments of the present invention provide a method for automatically annotating analogical corpora using the computing power of an intelligent computing center, comprising:

[0007] Step S1: Determine the first prompt word, which includes: a first analogy corpus pair as a case, a task and result indicating the similarity analysis of entities in the first analogy corpus pair, and an unlabeled second analogy corpus pair; input the first prompt word into the large language model to obtain the entity similarity analysis result of the second analogy corpus pair;

[0008] Step S2: Determine the second prompt word, which includes: the first analogy corpus pair, the task and result of aligning the sentences in the first analogy corpus pair with structural similarity, and the second analogy corpus pair; input the second prompt word into the large language model to obtain the sentence pairs with structural similarity alignment of the second analogy corpus pair;

[0009] Step S3: Determine the third prompt word, which includes: sentence pairs aligned with the structural similarity of the first analogy corpus, the task and result of performing causal logical similarity analysis on the sentence pairs aligned with the structural similarity of the first analogy corpus, and sentence pairs aligned with the structural similarity of the second analogy corpus; input the third prompt word into the large language model to obtain the causal logical similarity analysis result of the second analogy corpus;

[0010] Step S4: Determine the fourth prompt word, which includes: the first analogy corpus pair and its entity similarity analysis results and causal logic similarity analysis results, the analogy relationship analysis results of the first analogy corpus pair, and the second analogy corpus pair and its causal logic similarity analysis results; input the fourth prompt word into the large language model to obtain the analogy relationship analysis results of the second analogy corpus pair.

[0011] Optionally, the method further includes:

[0012] Step S5: Based on the analysis results of the analogy relations of the second analogy corpus pair, indicate whether the entities are similar in form, whether the first-order relations are similar, and whether the higher-order relations are similar, label the analogy relation type of the second analogy corpus pair.

[0013] Optionally, the analogy relationship type includes at least one of the following:

[0014] The literal similarity type refers to the analogy pairs in which entities, first-order relations, and higher-order relations are all similar.

[0015] A true analogy type, in which the entities in the analogy corpus pairs of the true analogy type are dissimilar, some first-order relations are similar, and higher-order relations are similar;

[0016] The pseudo-analogy type is defined in which, in the analogy pairs of the pseudo-analogy type, entities are dissimilar, first-order relations are similar, and higher-order relations are dissimilar.

[0017] Surface similarity type, in the analogy pairs of the surface similarity type, entities are similar, first-order relations are similar, and higher-order relations are not similar;

[0018] In the analogy pairs of the appearance-only similarity type, the entities are similar, but the first-order relations are dissimilar, and the higher-order relations are dissimilar.

[0019] Anomaly type, in which the entities, first-order relations and higher-order relations in the analogy pairs of the anomaly type are all dissimilar.

[0020] Optionally, the entity includes at least one of the following: background, characters, plot, and common vocabulary, wherein the common vocabulary includes identical words and synonyms.

[0021] Optionally, the first analogy corpus pair is a pseudo-analogy type, in which entities are dissimilar, first-order relations are similar, and higher-order relations are dissimilar.

[0022] Optionally, the first prompt word further includes: first task prompt information, which is used to prompt the large language model to perform similarity analysis on entities in the first analogy corpus pair according to the instructions, and to perform similarity analysis on entities in the second analogy corpus pair, and output the entity similarity analysis results of the second analogy corpus pair;

[0023] And / or, the second prompt word further includes: second task prompt information, which is used to prompt the large language model to perform structural similarity alignment of sentences in the first analogy corpus pair according to the instructions, perform structural similarity alignment of the second analogy corpus pair, and output the structural similarity aligned sentence pair of the second analogy corpus pair;

[0024] And / or, the third prompt word further includes: third task prompt information, which is used to prompt the large language model to perform causal logic similarity analysis on the sentence pairs with structural similarity alignment of the first analogy corpus pair according to the instructions, perform causal logic similarity analysis on the sentence pairs with structural similarity alignment of the second analogy corpus pair, and output the causal logic similarity analysis result of the second analogy corpus pair;

[0025] And / or, the fourth prompt word further includes: fourth task prompt information, which is used to prompt the large language model to infer the analogy relationship analysis result of the first analogy corpus pair according to the first analogy corpus pair and its entity similarity analysis result and causal logic similarity analysis result, and to infer and output the analogy relationship analysis result of the second analogy corpus pair according to the second analogy corpus pair and its causal logic similarity analysis result.

[0026] Secondly, embodiments of the present invention provide an apparatus for automatically annotating analogy corpora using the computing power of an intelligent computing center, comprising:

[0027] The first processing module is used to determine a first prompt word, which includes: a first analogy corpus pair as a case, a task and result indicating the similarity analysis of entities in the first analogy corpus pair, and an unlabeled second analogy corpus pair; the first prompt word is input into a large language model to obtain the entity similarity analysis result of the second analogy corpus pair;

[0028] The second processing module is used to determine a second prompt word, which includes: the first analogy corpus pair, a task and result indicating the structural similarity alignment of sentences in the first analogy corpus pair, and the second analogy corpus pair; and inputting the second prompt word into the large language model to obtain a sentence pair with structural similarity alignment of the second analogy corpus pair.

[0029] The third processing module is used to determine a third prompt word, which includes: a sentence pair with structural similarity alignment of the first analogy corpus pair, a task and result indicating the causal logical similarity analysis of the sentence pair with structural similarity alignment of the first analogy corpus pair, and a sentence pair with structural similarity alignment of the second analogy corpus pair; the third prompt word is input into the large language model to obtain the causal logical similarity analysis result of the second analogy corpus pair;

[0030] The fourth processing module is used to determine the fourth prompt word, which includes: the first analogy corpus pair and its entity similarity analysis results and causal logic similarity analysis results, the analogy relationship analysis results of the first analogy corpus pair, and the second analogy corpus pair and its causal logic similarity analysis results; the fourth prompt word is input into the large language model to obtain the analogy relationship analysis results of the second analogy corpus pair.

[0031] Thirdly, embodiments of the present invention provide an electronic device, including: a processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, it implements the steps of the method for automatically annotating analogy corpora using the computing power of an intelligent computing center as described in the first aspect above.

[0032] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method for automatically annotating analogy corpora using the computing power of an intelligent computing center as described in the first aspect above.

[0033] Fifthly, embodiments of the present invention provide a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the method for automatically annotating analogy corpora using the computing power of an intelligent computing center as described in the first aspect above.

[0034] In this embodiment of the invention, a prompt word is determined. The prompt word includes analogy pairs as examples and their analogy relationship analysis process. A large language model running in an intelligent computing center is invoked to analyze the analogy relationships of unlabeled analogy pairs based on the analogy relationship analysis process of the analogy pairs as examples in the prompt word. The analysis results of the analogy relationships of the unlabeled analogy pairs are obtained. Based on the analysis results, the analogy relationship types of the analogy pairs can be labeled, and high-quality analogy pairs can be screened out. This can improve the cognitive ability and analogical reasoning ability of the large language model trained with these high-quality analogy pairs. Attached Figure Description

[0035] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0036] Figure 1 This is one of the flowcharts illustrating the method for automatically annotating analogical corpora using the computing power of an intelligent computing center according to an embodiment of the present invention;

[0037] Figure 2 This is a second flowchart illustrating the method for automatically annotating analogical corpora using the computing power of an intelligent computing center, according to an embodiment of the present invention.

[0038] Figure 3 This is a schematic diagram of the first prompt word in an embodiment of the present invention;

[0039] Figure 4 This is a schematic diagram of the second prompt word in an embodiment of the present invention;

[0040] Figure 5 This is a schematic diagram of the third prompt word in an embodiment of the present invention;

[0041] Figure 6 This is a schematic diagram of the fourth prompt word in an embodiment of the present invention;

[0042] Figure 7 This is a schematic diagram of the device for automatically annotating analogy corpora using the computing power of an intelligent computing center, according to an embodiment of the present invention.

[0043] Figure 8 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] The technical terms involved in this invention will be briefly explained below.

[0046] The "computing power" mentioned in this invention refers to the ability of computer equipment or computing / data centers to process information. It is the ability of computer hardware and software to work together to perform a certain computing requirement. It is the computing power to achieve the target output by processing information data. It is a new type of productivity that integrates information computing power, network carrying capacity, and data storage capacity. It mainly provides services to society through computing power infrastructure.

[0047] The "computational power" (CP) described in this invention refers to the ability of a data center server to process data and output results. It is a comprehensive indicator of a data center's computing power, encompassing general computing power, supercomputing power, and intelligent computing power. The commonly used unit of measurement is floating-point operations per second (FLOPS, 1EFLOPS = 10^18 FLOPS), with higher values ​​indicating stronger overall computing power. It is estimated that 1EFLOPS is approximately the computing power output of 5 Tianhe-2A supercomputers, 500,000 mainstream server CPUs, or 2 million mainstream laptops. The calculation formula is: CP = CP 通用 +CP 智能 +CP 超级 .

[0048] The "Network Power" (NP) described in this invention refers to the data transmission capability of computing facilities, encompassing comprehensive capabilities including network architecture, network bandwidth, transmission latency, and intelligent management and scheduling. Network power involves network transmission within and between data centers and is a comprehensive indicator of network transmission scheduling capabilities. In this embodiment of the invention, network power is measured using video memory bandwidth.

[0049] The "Storage Power" (SP) described in this invention refers to the comprehensive capabilities of a data center in four aspects: data storage capacity, performance, security and reliability, and green and low-carbon operation. It is a comprehensive indicator for measuring the data storage capacity of a data center, including external storage devices such as storage arrays and internal storage devices within servers. The commonly used unit of measurement for storage capacity is exabytes (EB, 1EB = 2^60 bytes), while the commonly used unit of measurement for performance is the number of read / write operations per second (IOPS / TB). Disaster recovery ratio is an important indicator of security and reliability.

[0050] The "computing infrastructure" described in this invention is a new type of information infrastructure that integrates information computing power, network carrying capacity, and data storage capacity. It can realize centralized computing, storage, transmission, and application of information, and exhibits characteristics such as diversity and ubiquity, intelligence and agility, security and reliability, and green and low carbon emissions.

[0051] The “computing power” mentioned in this invention includes general computing power, intelligent computing power, and supercomputing power.

[0052] The "general-purpose computing power" described in this invention refers to the computing capabilities provided by servers based on CPU (Central Processing Unit) chips, used to support basic general-purpose computing such as cloud computing and edge computing.

[0053] The "intelligent computing power" described in this invention refers to a computing platform deployed on a large scale based on dedicated chips such as GPUs (Graphics Processing Units), FPGAs (Field Programmable Gate Arrays), and ASICs (Application Specific Integrated Circuits) for various innovative artificial intelligence applications, such as natural language processing and machine vision.

[0054] The “supercomputing power” described in this invention mainly refers to the computing power provided by high-performance computing clusters such as supercomputers. It utilizes the centralized computing resources of multiple computer systems working in parallel and uses a dedicated operating system to handle extremely complex or data-intensive problems. It is mainly used for computing in cutting-edge scientific fields, such as planetary simulation, drug molecule design, and gene analysis.

[0055] The "intelligent computing center" described in this invention refers to a facility that utilizes large-scale heterogeneous computing resources, including general-purpose computing power (CPU: Central Processing Unit) and intelligent computing power (GPU: Graphics Processing Unit, FPGA: Field Programmable Gate Array, ASIC: Application Specific Integrated Circuit, etc.), primarily providing the necessary computing power, data, and algorithms for artificial intelligence applications (such as the development, training, and inference of deep learning models). The intelligent computing center encompasses facilities, hardware, and software, and can provide full-stack capabilities from underlying computing power to top-level application enablement.

[0056] The “computing resources” mentioned in this invention refer to the technologies and facilities required for the development of the digital society that have the ability to compute, transmit, store and apply information, including but not limited to computing resources such as CPU (Central Processing Unit) and GPU (Graphics Processing Unit), network resources such as switches and routers, storage resources such as storage arrays and distributed storage, security resources such as firewalls and intrusion detection systems, and supporting and guaranteeing resources such as wind, fire, water and electricity.

[0057] The "large language model" mentioned in this invention refers to a large-scale language model (LLM), which is a language model with a large number of parameters. It is designed to understand and generate human language. It is trained with a large amount of text data and can perform a wide range of tasks, including text summarization, translation, and sentiment analysis.

[0058] In the fields of linguistics and natural language processing, the term "corpus" as used in this invention refers to text or speech data used for language research, analysis, teaching, or technological development. In natural language processing (NLP), corpora are used to train machine learning models, such as language models, text classifiers, and sentiment analyzers. These models learn the statistical regularities and patterns of language by analyzing large amounts of corpora, thereby enabling them to perform various language processing tasks, such as text generation, translation, summarization, and question answering.

[0059] The "prompt" mentioned in this invention refers to a natural language-based prompt or instruction designed to guide a large language model to perform a specific task.

[0060] The "analogy" described in this invention is central to human cognition, allowing us to abstract information and understand new (unfamiliar) situations based on familiar ones. Textual corpora with analogical relationships can be described as textual corpora where entities are dissimilar but their logical structures are similar. The logical structure can include first-order relations and higher-order relations. That is, textual corpora with analogical relationships can also be described as textual corpora where entities are dissimilar, but higher-order relations and some first-order relations are similar.

[0061] The "analogical corpus pair" described in this invention consists of two text corpora that have an analogy relationship. Text corpora with an analogy relationship can be described as text corpora that are dissimilar in entities but similar in logical structure. The logical structure can include first-order relations and higher-order relations. That is, text corpora with an analogy relationship can also be described as text corpora that are dissimilar in entities but similar in higher-order relations and some first-order relations.

[0062] The “entities” mentioned in this invention include objects, people, roles, etc. In this method, the (story) background and plot are also included in the analysis scope of entities.

[0063] The "first-order relationship" described in this invention refers to a simple and direct connection between entities (such as objects, people, etc.). These relationships are usually intuitive and superficial associations, not involving deep causal or logical structures, and mainly describe the spatial, temporal, and interactive relational structures of the entities being described.

[0064] The "higher-order relations" described in this invention refer to more abstract or complex associations that go beyond direct relationships between entities, typically involving cause, effect, purpose, pattern recognition, or deeper logical structures. In other words, higher-order relations explore "why" such direct relationships exist between entities, or how these relationships function within a broader system or context.

[0065] The following example illustrates a text corpus with analogical relationships.

[0066] The text pairs with an analogy relationship include: "Planets in the solar system revolve around the sun" and "Electrons in an atom revolve around the atomic nucleus." The analogies include "planet" and "electron," and "sun" and "atomic nucleus."

[0067] Another pair of texts with analogical relationships includes: "An employee receives a seemingly harmless attachment containing malware. The malware infiltrates his personal computer and steals his sensitive personal information." and "The citizens of Troy allow access to the Trojan Horse containing Greek soldiers. The Greek soldiers conquer Troy and steal their wealth." The analogical relationships include: "employee" and "citizens of Troy," "attachment" and "Trojan Horse," "malicious software" and "Greek soldiers," "personal computer" and "Troy," and "sensitive personal information" and "wealth."

[0068] To address the issue of low-quality corpora used in existing intelligent computing centers for training large language models, which consequently fails to guarantee stable and excellent performance of the trained models, please refer to... Figure 1 This invention provides a method for automatically annotating analogy corpora using the computing power of an intelligent computing center, comprising:

[0069] Step S1: Determine the first prompt word, which includes: a first analogy corpus pair as a case, a task and result indicating the similarity analysis of entities in the first analogy corpus pair, and an unlabeled second analogy corpus pair; input the first prompt word into the large language model to obtain the entity similarity analysis result of the second analogy corpus pair;

[0070] Please refer to Figure 2 Step S1 can also be called the entity analysis step, which is to perform similarity analysis on the entities of the two analogies in the analogy corpus pair.

[0071] In this embodiment of the invention, the analogy corpus pair consists of two text corpora, that is, the first analogy corpus pair includes two text corpora, and the second analogy corpus pair also includes two text corpora.

[0072] Please refer to Figure 2 In this embodiment of the invention, the second analogy corpus pair can be an analogy corpus pair from an unlabeled analogy dataset, which includes multiple unlabeled analogy corpus pairs. An unlabeled analogy corpus pair refers to an analogy corpus pair that has not been scored, has not had its analogy relationship type labeled, or has not undergone data filtering. The unlabeled analogy dataset may contain analogy corpus pairs with poor analogy ability. Directly using analogy corpus pairs from the unlabeled analogy dataset to train a large language model cannot guarantee that the trained large language model will have stable and excellent performance.

[0073] In this embodiment of the invention, the analogy pairs in the unlabeled analogy dataset can be generated as follows: Obtain the original text corpus, input the original text corpus into a large language model, and obtain analogy pairs generated based on the original text corpus. The analogy pairs include a first text corpus and a second text corpus with an analogy relationship. The first text corpus is obtained by the large language model through information extraction and detail expansion of the original text corpus. The second text corpus is written by the large language model. The second text corpus has a similar logical structure to the original text corpus but dissimilar entities. The original text corpus can be text downloaded from the Internet (Web), or text downloaded from the Internet and processed after data cleaning.

[0074] Of course, in the embodiments of the present invention, it is not excluded that other methods may be used to obtain analogy pairs in the unlabeled analogy dataset.

[0075] Please refer to Figure 3 , Figure 3 This is a schematic diagram of the first prompt word in an embodiment of the present invention. Figure 3 It can be seen that the first cue word includes the first analogy corpus pair as the case study. The first analogy corpus pair includes two text corpora, for example... Figure 3 The first cue word also includes: the Base and Target. It also includes: the task and results of similarity analysis of entities in the first analogy pair, such as... Figure 3 The "Question" in the first analogy corpus indicates the task of performing similarity analysis on the entities in the first analogy corpus pair, such as... Figure 3 The "Answer" in the first analogy corpus refers to the result of the similarity analysis of entities in the first analogy corpus pair, also known as the entity similarity analysis result of the first analogy corpus pair.

[0076] In this embodiment of the application, optionally, the entity includes at least one of the following: background, characters, plot, and common vocabulary, wherein the common vocabulary includes identical words and synonyms. In this embodiment of the invention, the similarity analysis of identical words typically analyzes verbs or nouns in analogy pairs.

[0077] In this embodiment of the application, optionally, the entity similarity analysis result of the first analogy corpus pair can be whether the entities are similar, such as... Figure 3 In this context, False indicates dissimilarity, and True indicates similarity. Common vocabulary can be listed in a list format.

[0078] In this embodiment of the application, optionally, the result of the similarity analysis of entities in the first analogy corpus pair (such as...) Figure 3The "Answer" in the context can also include the entity similarity analysis process and results, such as... Figure 3 The analysis process is as follows: "In 'Base,' the protagonists are a tortoise and a hare. They have a race, and the hare, overconfident, takes a break along the way, ultimately leading to the tortoise's victory. In 'Target,' the protagonists are a thief and a guard. The thief, being fast, always manages to escape, but the guard sets a trap, and the thief is eventually caught. Therefore, the specific backgrounds of the two (the tortoise and the hare race versus the guard catching the thief) are different: fables differ from real events, and the roles and responsibilities of the characters (tortoise and hare, thief and guard) are also different: the tortoise and hare are in a competitive relationship, while the thief and guard are in a capture relationship. However, they share some similarities in their plot development, both being stories about a chase leading to failure. Furthermore, they do not share any common vocabulary."

[0079] In some embodiments, optionally, the first prompt word further includes: a first task prompt information, which is used to prompt the large language model to perform similarity analysis on entities in the first analogy corpus pair according to instructions, perform similarity analysis on entities in the second analogy corpus pair, and output the entity similarity analysis results of the second analogy corpus pair. Through the first task prompt information, the large language model can more accurately obtain the user's intent.

[0080] It should be noted that, Figure 3 The first prompt word in the text does not show the unlabeled second analogy pair.

[0081] Step S2: Determine the second prompt word, which includes: the first analogy corpus pair, the task and result of aligning the sentences in the first analogy corpus pair with structural similarity, and the second analogy corpus pair; input the second prompt word into the large language model to obtain the sentence pairs with structural similarity alignment of the second analogy corpus pair;

[0082] Please refer to Figure 2 Step S2 can also be called the sentence mapping step, which is to compare the sentences in the two analogy corpora in the analogy corpus pair to obtain the sentences with structural similarity in the two analogy corpora.

[0083] Please refer to Figure 4 , Figure 4 This is a schematic diagram of the second prompt word in an embodiment of the present invention. Figure 4 It can be seen that the second cue word includes the first analogy corpus pair used as an example. The first analogy corpus pair includes two text corpora, for example... Figure 4 The Base and Target in the first analogy corpus. The second cue word also includes: instructions on the task and results of aligning sentences in the first analogy corpus pair based on structural similarity, such as... Figure 4The "Question" in the first analogy corpus refers to the task of aligning the sentences in the first analogy corpus pair based on structural similarity, such as... Figure 4 The Answer in the first analogy corpus is the result of aligning the sentences in the first analogy corpus based on structural similarity, also known as the sentence pair aligned with the structural similarity of the first analogy corpus.

[0084] It should be noted that, in this embodiment of the invention, when performing structural similarity alignment on analogy pairs, it is necessary to ensure that all sentences in the two text corpora in the analogy pairs participate in structural similarity alignment. If a sentence in one text corpus does not have a corresponding sentence with structural similarity alignment in the other text corpus, a prompt message indicating that there is no sentence with structural similarity alignment is returned.

[0085] Optionally, the sentence described in this embodiment of the invention refers to a sentence containing only one period.

[0086] In some embodiments, optionally, the second prompt word further includes: a second task prompt information, which is used to prompt the large language model to perform structural similarity alignment of sentences in the first analogy corpus according to the instructions, perform structural similarity alignment of the second analogy corpus, and output the structurally similar sentence pair of the second analogy corpus. Through the second task prompt information, the large language model can more accurately obtain the user's intent.

[0087] Step S3: Determine the third prompt word, which includes: sentence pairs aligned with the structural similarity of the first analogy corpus, the task and result of performing causal logical similarity analysis on the sentence pairs aligned with the structural similarity of the first analogy corpus, and sentence pairs aligned with the structural similarity of the second analogy corpus; input the third prompt word into the large language model to obtain the causal logical similarity analysis result of the second analogy corpus;

[0088] Please refer to Figure 2 Step S3 can also be called the relation alignment step, which is to perform causal logic similarity analysis on each sentence pair of the analogy corpus.

[0089] Please refer to Figure 5 , Figure 5 This is a schematic diagram of the third prompt word in an embodiment of the present invention. Figure 5 It can be seen that the third cue word includes sentence pairs that are structurally similar to the first analogy corpus pairs. The third cue word also includes: the task and results of performing causal logical similarity analysis on the sentence pairs that are structurally similar to the first analogy corpus pairs, such as... Figure 5The "Question" in the text refers to the task of performing causal logical similarity analysis on sentence pairs aligned with the structural similarity of the first analogy corpus, such as... Figure 5 The "Answer" in the text refers to the result of causal logical similarity analysis on the sentence pairs aligned with the structural similarity of the first analogy corpus, also known as the causal logical similarity analysis result of the first analogy corpus.

[0090] In this embodiment of the invention, optionally, the causal logic similarity analysis results of the first analogy corpus pair may include the analysis process and the analysis results, such as... Figure 5 As shown, the analysis process is as follows:

[0091] "1. Because one side (Target) contains 'none', it is classified as the irrelevant group."

[0092] 2. The rabbit in Base is confident of winning because of its speed, just as the thief in Target is able to escape because of his speed. Both are examples of confidence or success based on speed advantage, and are therefore grouped into a similar group.

[0093] 3. Because one side (Target) contains 'none', it is classified as an irrelevant group.

[0094] 4. The description in Base is that the rabbit stops to rest because of complacency, while the description in Target is that the guards use tricks to capture the thief. The characters described in the two are not structurally corresponding and there is no logical correlation between their contents, so they are classified as unrelated groups.

[0095] 5. Because Base describes the tortoise continuing its journey while the hare rests, while Target describes the thief being knocked unconscious by a carriage during his escape, both depict the main characters losing consciousness, but the reasons are different: the hare rests because it thinks the tortoise is too slow and it won't hurt to rest for a while, while the thief didn't intend to stop, but was knocked unconscious by an unexpected carriage during his escape, not because he was confident in his abilities and deliberately lay down to be caught by the guards, so they are classified as dissimilar groups.

[0096] 6. The ending of Base is that the rabbit wakes up to find it has failed, just like the thief in Target wakes up to find himself surrounded by guards. Both depict the unfavorable situation faced by the protagonist after regaining consciousness from a coma or sleep, but the reasons are different: the rabbit's failure stems from subjective overconfidence, while the thief's failure is due to objective accidents. Therefore, they are grouped into the "dissimilar" category.

[0097] In some embodiments, optionally, the third prompt word further includes: a third task prompt information, which is used to prompt the large language model to perform causal logic similarity analysis on the sentence pairs of the first analogy corpus that are aligned with the structural similarity of the first analogy corpus, and to perform causal logic similarity analysis on the sentence pairs of the second analogy corpus that are aligned with the structural similarity of the second analogy corpus, and to output the causal logic similarity analysis result of the second analogy corpus. Through the third task prompt information, the large language model can more accurately obtain the user's intent.

[0098] Step S4: Determine the fourth prompt word, which includes: the first analogy corpus pair and its entity similarity analysis results and causal logic similarity analysis results, the analogy relationship analysis results of the first analogy corpus pair, and the second analogy corpus pair and its causal logic similarity analysis results; input the fourth prompt word into the large language model to obtain the analogy relationship analysis results of the second analogy corpus pair.

[0099] Please refer to Figure 2 Step S4 can also be called the analogy summarization step, which is to determine the analogy relationship analysis results of the analogy corpus pairs based on the entity similarity analysis results and the causal logic similarity analysis results of the analogy corpus pairs.

[0100] In some embodiments, optionally, the analogy analysis results include whether the entities are similar in form, whether the first-order relations are similar, and whether the higher-order relations are similar.

[0101] Please refer to Figure 6 , Figure 6 This is a schematic diagram of the fourth prompt word in an embodiment of the present invention. Figure 6 It can be seen that the fourth prompt word includes the first analogy corpus pair, the entity similarity analysis results of the first analogy corpus pair, the causal logic similarity analysis results, and the analogy relation analysis results of the first analogy corpus pair. The entity similarity analysis results of the first analogy corpus pair are... Figure 6 The results of the causal logic similarity analysis, which state that "Base and Target have different story backgrounds (False), different characters (False), and similar plots (True)," indicate that they are similar. Figure 6 The expression “len(similar groups) = 1, len(dissimilar groups) = 2, len(irrelevant groups) = 3” represents the result of the analogy relationship analysis of the first analogy corpus pairs. Figure 6 ""entities":"dissimilar","one-order relations":"similar","higher-order relations":"dissimilar"" in .

[0102] In some embodiments, optionally, the fourth prompt word further includes: a fourth task prompt information, which is used to prompt the large language model to infer the analogy relationship analysis result of the first analogy corpus pair based on the first analogy corpus pair and its entity similarity analysis result and causal logic similarity analysis result, and to infer and output the analogy relationship analysis result of the second analogy corpus pair based on the second analogy corpus pair and its causal logic similarity analysis result. Through the fourth task prompt information, the large language model can more accurately obtain the user's intent.

[0103] In this embodiment of the invention, a prompt word is determined. The prompt word includes analogy pairs as examples and their analogy relationship analysis process. A large language model running in an intelligent computing center is invoked to analyze the analogy relationships of unlabeled analogy pairs based on the analogy relationship analysis process of the analogy pairs as examples in the prompt word. The analysis results of the analogy relationships of the unlabeled analogy pairs are obtained. Based on the analysis results, the analogy relationship types of the analogy pairs can be labeled, and high-quality analogy pairs can be screened out. This can improve the cognitive ability and analogical reasoning ability of the large language model trained with these high-quality analogy pairs.

[0104] In this embodiment of the invention, the execution order of steps S1 and S2 is not limited. Step S1 can be executed first, followed by step S2; or step S2 can be executed first, followed by step S1; or steps S1 and S2 can be executed simultaneously. Step S3 needs to be executed after step S2, and step S4 needs to be executed after steps S1 and S3.

[0105] In some embodiments, optionally, the method further includes:

[0106] Step S5: Based on the analysis results of the analogy relations of the second analogy corpus pair, indicate whether the entities are similar in form, whether the first-order relations are similar, and whether the higher-order relations are similar, label the analogy relation type of the second analogy corpus pair.

[0107] In some embodiments, optionally, while outputting the analogy relation analysis results of the second analogy corpus pair, the large language model may also label and output the analogy relation type of the second analogy corpus pair based on whether the entities are similar in form, whether the first-order relations are similar, and whether the higher-order relations are similar, as indicated in the analogy relation analysis results of the second analogy corpus pair.

[0108] In some embodiments, optionally, the analogy relationship type includes at least one of the following:

[0109] The literal similarity type refers to the analogy pairs in which entities, first-order relations, and higher-order relations are all similar.

[0110] A true analogy type, in which the entities in the analogy corpus pairs of the true analogy type are dissimilar, some first-order relations are similar, and higher-order relations are similar;

[0111] The pseudo-analogy type is defined in which, in the analogy pairs of the pseudo-analogy type, entities are dissimilar, first-order relations are similar, and higher-order relations are dissimilar.

[0112] Surface similarity type, in the analogy pairs of the surface similarity type, entities are similar, first-order relations are similar, and higher-order relations are not similar;

[0113] In the analogy pairs of the appearance-only similarity type, the entities are similar, but the first-order relations are dissimilar, and the higher-order relations are dissimilar.

[0114] Anomaly type, in which the entities, first-order relations and higher-order relations in the analogy pairs of the anomaly type are all dissimilar.

[0115] In some embodiments, optionally, the first analogy corpus pair is of pseudo-analogy type, in which the entities are dissimilar, the first-order relations are similar, and the higher-order relations are dissimilar. Since the pseudo-analogy corpus pair includes a more comprehensive range of cases, it is preferred to use the pseudo-analogy corpus as an example.

[0116] The computing power of the intelligent computing center in this embodiment of the invention may include GPUs, etc., which have high computing power. This allows it to process a large amount of analogical corpus as described above, resulting in a large number of labeled analogical corpus pairs for training large language models. This can further improve the cognitive ability and analogical reasoning ability of the large language models trained with these analogical corpus pairs.

[0117] Please refer to Figure 7 This invention also provides a device 10 for automatically annotating analogy corpora using the computing power of an intelligent computing center, comprising:

[0118] The first processing module 11 is used to determine a first prompt word, which includes: a first analogy corpus pair as a case, a task and result indicating the similarity analysis of entities in the first analogy corpus pair, and an unlabeled second analogy corpus pair; inputting the first prompt word into a large language model to obtain the entity similarity analysis result of the second analogy corpus pair;

[0119] The second processing module 12 is used to determine a second prompt word, which includes: the first analogy corpus pair, a task and result indicating the structural similarity alignment of sentences in the first analogy corpus pair, and the second analogy corpus pair; and inputting the second prompt word into the large language model to obtain a sentence pair with structural similarity alignment of the second analogy corpus pair.

[0120] The third processing module 13 is used to determine a third prompt word, which includes: a sentence pair with structural similarity alignment of the first analogy corpus pair, a task and result indicating the causal logical similarity analysis of the sentence pair with structural similarity alignment of the first analogy corpus pair, and a sentence pair with structural similarity alignment of the second analogy corpus pair; the third prompt word is input into the large language model to obtain the causal logical similarity analysis result of the second analogy corpus pair;

[0121] The fourth processing module 14 is used to determine a fourth prompt word, which includes: the first analogy corpus pair and its entity similarity analysis results and causal logic similarity analysis results, the analogy relationship analysis results of the first analogy corpus pair, and the second analogy corpus pair and its causal logic similarity analysis results; the fourth prompt word is input into the large language model to obtain the analogy relationship analysis results of the second analogy corpus pair.

[0122] In this embodiment of the invention, a prompt word is determined. The prompt word includes analogy pairs as examples and their analogy relationship analysis process. A large language model running in an intelligent computing center is invoked to analyze the analogy relationships of unlabeled analogy pairs based on the analogy relationship analysis process of the analogy pairs as examples in the prompt word. The analysis results of the analogy relationships of the unlabeled analogy pairs are obtained. Based on the analysis results, the analogy relationship types of the analogy pairs can be labeled, and high-quality analogy pairs can be screened out. This can improve the cognitive ability and analogical reasoning ability of the large language model trained with these high-quality analogy pairs.

[0123] In some embodiments, optionally, it also includes:

[0124] The annotation module is used to annotate the analogy relationship type of the second analogy corpus pair based on whether the entities are similar in form, whether the first-order relations are similar, and whether the higher-order relations are similar, as indicated in the analogy relationship analysis results of the second analogy corpus pair.

[0125] In some embodiments, optionally, the analogy relationship type includes at least one of the following:

[0126] The literal similarity type refers to the analogy pairs in which entities, first-order relations, and higher-order relations are all similar.

[0127] A true analogy type, in which the entities in the analogy corpus pairs of the true analogy type are dissimilar, some first-order relations are similar, and higher-order relations are similar;

[0128] The pseudo-analogy type is defined in which, in the analogy pairs of the pseudo-analogy type, entities are dissimilar, first-order relations are similar, and higher-order relations are dissimilar.

[0129] Surface similarity type, in the analogy pairs of the surface similarity type, entities are similar, first-order relations are similar, and higher-order relations are not similar;

[0130] In the analogy pairs of the appearance-only similarity type, the entities are similar, but the first-order relations are dissimilar, and the higher-order relations are dissimilar.

[0131] Anomaly type, in which the entities, first-order relations and higher-order relations in the analogy pairs of the anomaly type are all dissimilar.

[0132] In some embodiments, optionally, the entity includes at least one of the following: background, characters, plot, and common vocabulary, wherein the common vocabulary includes identical words and synonyms.

[0133] In some embodiments, optionally, the first analogy corpus pair is a pseudo-analogy type, in which the entities are dissimilar, the first-order relations are similar, and the higher-order relations are dissimilar.

[0134] In some embodiments, optionally, the first prompt word further includes: a first task prompt information, which is used to prompt the large language model to perform similarity analysis on entities in the first analogy corpus pair according to the instructions, perform similarity analysis on entities in the second analogy corpus pair, and output the entity similarity analysis result of the second analogy corpus pair;

[0135] And / or, the second prompt word further includes: second task prompt information, which is used to prompt the large language model to perform structural similarity alignment of sentences in the first analogy corpus pair according to the instructions, perform structural similarity alignment of the second analogy corpus pair, and output the structural similarity aligned sentence pair of the second analogy corpus pair;

[0136] And / or, the third prompt word further includes: third task prompt information, which is used to prompt the large language model to perform causal logic similarity analysis on the sentence pairs with structural similarity alignment of the first analogy corpus pair according to the instructions, perform causal logic similarity analysis on the sentence pairs with structural similarity alignment of the second analogy corpus pair, and output the causal logic similarity analysis result of the second analogy corpus pair;

[0137] And / or, the fourth prompt word further includes: fourth task prompt information, which is used to prompt the large language model to infer the analogy relationship analysis result of the first analogy corpus pair according to the first analogy corpus pair and its entity similarity analysis result and causal logic similarity analysis result, and to infer and output the analogy relationship analysis result of the second analogy corpus pair according to the second analogy corpus pair and its causal logic similarity analysis result.

[0138] Please refer to Figure 8 The present invention also provides an electronic device 20, including a processor 21, a memory 22, and a computer program stored in the memory 22 and executable on the processor 21. When the computer program is executed by the processor 21, it implements the various processes of the above-described method embodiment for automatically annotating analogy corpora using the computing power of an intelligent computing center, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0139] This invention also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the various processes of the above-described method embodiment for automatically annotating analogy corpora using the computing power of an intelligent computing center, and achieves the same technical effect. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0140] This application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the above-described... Figure 1 The various processes of the method embodiment shown, which automatically annotates analog corpora using the computing power of the intelligent computing center, can achieve the same technical effect. To avoid repetition, they will not be described again here.

[0141] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0142] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0143] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of the present invention.

Claims

1. A method for automatically annotating analogical corpora using the computing power of an intelligent computing center, characterized in that, include: Step S1: Determine the first prompt word, which includes: a first analogy corpus pair as a case, a task and result indicating the similarity analysis of entities in the first analogy corpus pair, and an unlabeled second analogy corpus pair; input the first prompt word into the large language model to obtain the entity similarity analysis result of the second analogy corpus pair; Step S2: Determine the second prompt word, which includes: the first analogy corpus pair, the task and result of aligning the sentences in the first analogy corpus pair with structural similarity, and the second analogy corpus pair; input the second prompt word into the large language model to obtain the sentence pairs with structural similarity alignment of the second analogy corpus pair; Step S3: Determine the third prompt word, which includes: sentence pairs aligned with the structural similarity of the first analogy corpus, the task and result of performing causal logical similarity analysis on the sentence pairs aligned with the structural similarity of the first analogy corpus, and sentence pairs aligned with the structural similarity of the second analogy corpus; input the third prompt word into the large language model to obtain the causal logical similarity analysis result of the second analogy corpus; Step S4: Determine the fourth prompt word, which includes: the first analogy corpus pair and its entity similarity analysis results and causal logic similarity analysis results, the analogy relationship analysis results of the first analogy corpus pair, and the second analogy corpus pair and its causal logic similarity analysis results; input the fourth prompt word into the large language model to obtain the analogy relationship analysis results of the second analogy corpus pair.

2. The method according to claim 1, characterized in that, Also includes: Step S5: Based on the analysis results of the analogy relations of the second analogy corpus pair, indicate whether the entities are similar in form, whether the first-order relations are similar, and whether the higher-order relations are similar, label the analogy relation type of the second analogy corpus pair.

3. The method according to claim 2, characterized in that, The analogy relationship type includes at least one of the following: The literal similarity type refers to the analogy pairs in which entities, first-order relations, and higher-order relations are all similar. A true analogy type, in which the entities in the analogy corpus pairs of the true analogy type are dissimilar, some first-order relations are similar, and higher-order relations are similar; The pseudo-analogy type is defined in which, in the analogy pairs of the pseudo-analogy type, entities are dissimilar, first-order relations are similar, and higher-order relations are dissimilar. Surface similarity type, in the analogy pairs of the surface similarity type, entities are similar, first-order relations are similar, and higher-order relations are not similar; In the analogy pairs of the appearance-only similarity type, the entities are similar, but the first-order relations are dissimilar, and the higher-order relations are dissimilar. Anomaly type, in which the entities, first-order relations and higher-order relations in the analogy pairs of the anomaly type are all dissimilar.

4. The method according to claim 1, characterized in that, The entity includes at least one of the following: background, characters, plot, and common vocabulary, wherein the common vocabulary includes identical words and synonyms.

5. The method according to claim 1, characterized in that, The first analogy corpus pair is of pseudo-analogy type. In the pseudo-analogy type of analogy corpus pair, the entities are not similar, the first-order relations are similar, and the higher-order relations are not similar.

6. The method according to claim 1, characterized in that, The first prompt word further includes: first task prompt information, which is used to prompt the large language model to perform similarity analysis on entities in the first analogy corpus pair according to the instructions, and to perform similarity analysis on entities in the second analogy corpus pair, and output the entity similarity analysis results of the second analogy corpus pair; And / or, the second prompt word further includes: second task prompt information, which is used to prompt the large language model to perform structural similarity alignment of sentences in the first analogy corpus pair according to the instructions, perform structural similarity alignment of the second analogy corpus pair, and output the structural similarity aligned sentence pair of the second analogy corpus pair; And / or, the third prompt word further includes: third task prompt information, which is used to prompt the large language model to perform causal logic similarity analysis on the sentence pairs with structural similarity alignment of the first analogy corpus pair according to the instructions, perform causal logic similarity analysis on the sentence pairs with structural similarity alignment of the second analogy corpus pair, and output the causal logic similarity analysis result of the second analogy corpus pair; And / or, the fourth prompt word further includes: fourth task prompt information, which is used to prompt the large language model to infer the analogy relationship analysis result of the first analogy corpus pair according to the first analogy corpus pair and its entity similarity analysis result and causal logic similarity analysis result, and to infer and output the analogy relationship analysis result of the second analogy corpus pair according to the second analogy corpus pair and its causal logic similarity analysis result.

7. A device for automatically annotating analogy corpora using the computing power of an intelligent computing center, characterized in that, include: The first processing module is used to determine a first prompt word, which includes: a first analogy corpus pair as a case, a task and result indicating the similarity analysis of entities in the first analogy corpus pair, and an unlabeled second analogy corpus pair; the first prompt word is input into a large language model to obtain the entity similarity analysis result of the second analogy corpus pair; The second processing module is used to determine a second prompt word, which includes: the first analogy corpus pair, a task and result indicating the structural similarity alignment of sentences in the first analogy corpus pair, and the second analogy corpus pair; and inputting the second prompt word into the large language model to obtain a sentence pair with structural similarity alignment of the second analogy corpus pair. The third processing module is used to determine a third prompt word, which includes: a sentence pair with structural similarity alignment of the first analogy corpus pair, a task and result indicating the causal logical similarity analysis of the sentence pair with structural similarity alignment of the first analogy corpus pair, and a sentence pair with structural similarity alignment of the second analogy corpus pair; the third prompt word is input into the large language model to obtain the causal logical similarity analysis result of the second analogy corpus pair; The fourth processing module is used to determine the fourth prompt word, which includes: the first analogy corpus pair and its entity similarity analysis results and causal logic similarity analysis results, the analogy relationship analysis results of the first analogy corpus pair, and the second analogy corpus pair and its causal logic similarity analysis results; the fourth prompt word is input into the large language model to obtain the analogy relationship analysis results of the second analogy corpus pair.

8. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the method for automatically annotating analogy corpora using the computing power of an intelligent computing center as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method for automatically annotating analogy corpora using the computing power of an intelligent computing center as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The method includes computer instructions that, when executed by a processor, implement the steps of the method for automatically annotating analogy corpora using the computing power of an intelligent computing center as described in any one of claims 1 to 6.

Images