English sentence generation method based on continuation theory and intelligent system thereof

Abstract

The invention discloses an English sentence generation method based on a continuation theory and an intelligent system thereof. The method belongs to the cross technical field of artificial intelligence and computer-aided language learning. The technical problems that an existing language generation model lacks bilingual learning theory guidance, so that generated sentences are low in communication validity, and the deep semantic relation cannot be intelligently reflected are solved. According to the technical scheme, the method comprises the steps of receiving verbs and context parameters input by a user; calling a large language model, and activating a mapping network containing nine adversities in three dimensions of materialization, rationalization and dialectization; a multi-modal optimization mechanism is adopted and comprises a corpus-based adversity position adjustment sub-module, an error prediction-cognitive enhancement cycle sub-module, a semantic association enhancement sub-module and a weight distribution optimization sub-module, and collaborative screening and optimization are carried out on adversity combinations; and outputting the English sentence with the enhanced communication function. The method is mainly used for core engine construction of intelligent writing assistance, a self-adaptive language teaching system and cross-cultural communication software.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of artificial intelligence and language education technology, specifically involving an English sentence generation method and system based on the "continuation theory" in second language acquisition theory (IPC classification number: G06F40 / 205). This invention addresses the technical shortcomings of existing AI-generated sentences in terms of communicative validity and cognitive adaptability by constructing a verb-oriented multi-adverbial matrix and combining it with a multimodal optimization mechanism. It achieves intelligent generation of English sentences that are grammatically correct, semantically reasonable, and communicatively appropriate. The technical solution of this invention is mainly applied to the sentence generation engine of intelligent writing assistance systems (G09B5 / 06), adaptive language teaching platforms, and cross-cultural communication software. Background Technology

[0002] Current English sentence generation technology has three main limitations: (1) Lack of theoretical guidance: For example, the pure statistical generation method used in patent CN110765032A relies solely on the probability prediction of the language model, resulting in sentences that, although grammatically correct, lack deep logical connections. (2) Insufficient contextual completion: For example, the rule template matching technology used in patent CN112926178B can only achieve the association between verbs and a single type of adverbial, and cannot meet the context's needs for adverbials of multiple dimensions such as cause, concession, and condition; (3) Cognitive matching bias: For example, the large language model scheme based on patent CN114580557A does not take into account the cognitive development stage characteristics of second language learners, resulting in a mismatch between the generated content and the user's language level.

[0003] In-depth analysis reveals that the fundamental flaw in existing technologies lies in their failure to achieve a triadic integration of second language acquisition theory, cognitive science, and artificial intelligence technology. On the one hand, traditional methods treat adverbs as dispensable modifiers rather than explicit markers of thought processes. On the other hand, existing systems lack the technical pathway to transform cognition and thought into computable, multi-dimensional models. This directly results in significant deficiencies in the generated sentences' applicability (accuracy of action description) and communicativeness (rigor of logical argumentation). Summary of the Invention

[0004] This invention addresses the shortcomings of existing technologies by proposing a method and intelligent system for generating English sentences based on continuation of arguments. Its core innovation lies in establishing a three-dimensional dynamic mapping relationship between verbs and nine types of adverbs, and designing a four-fold optimization mechanism to achieve intelligent generation from simple action descriptions to complex logical expressions. The specific technical solution includes three key stages: (1) In the three-dimensional adverbial matrix construction stage, the system simultaneously activates the adverbial network of the specific dimension (time, place, manner), rational dimension (cause, result, purpose), and dialectical dimension (condition, concession, comparison) according to the input verb and context parameters; (2) In the multimodal optimization stage of the coze workflow, the following four mechanisms are adopted: ① Adverbial position probability correction based on Longman Corpus, ② Error prediction-cognitive reinforcement loop using Python language plugin, ③ Semantic association enhancement guided by 12,000 prompt words, and ④ Collocation weight allocation driven by 2,000 verb knowledge base. These four mechanisms work together to ensure the quality of generation. (3) Cognitive adaptation output stage: By adjusting the complexity level of adverbial phrases (beginner, intermediate, advanced), the output sentence is accurately matched with the user's language cognitive level.

[0005] The significant technological advancement of this invention lies in its decomposition of the "contextual gap-filling hypothesis" into a three-dimensional, nine-category adverbial gap-filling system. Test data shows that, compared to traditional methods, the generated sentences exhibit significant improvements in three dimensions: specificity (BLEU-4 score improvement of 42%), rationality (expert score improvement of 35%), and dialectical reasoning (logical completeness improvement of 38%). More importantly, this technology effectively cultivates learners' multidimensional thinking abilities through the adverbial gap-filling process, achieving a deep integration of the technological solution and educational goals.

Claims

1. A method for generating English sentences based on continuation of arguments, characterized in that... This paper combines the continuation of the field of second language acquisition with natural language processing technology. By constructing a three-dimensional mapping relationship between verbs and nine types of adverbs, it achieves intelligent generation of English sentences, specifically including the following steps: (1) Receive verbs and initial context parameters input by the user, wherein the context parameters may include tense information, subject features and communicative intent tags; (2) Call the Large Language Model (LLM) to activate the adverbial mapping network with three dimensions based on the input verb and context parameters: ① Specificity dimension: Automatically associates adverbs of time, place, and manner used to describe the details of the action; ②Rationalization dimension: Intelligent matching is used to explain the adverbial clauses of cause, result, and purpose in explaining the behavioral logic; ③ Dialectical dimension: Autonomously generate conditional adverbs, concessive adverbs, and comparative adverbs to express complex semantics; (3) A multimodal optimization mechanism is used to collaboratively optimize the initially generated adverbial phrase combinations in the coze workflow, including: ① The corpus-based adverbial position adjustment submodule adjusts the linear order of adverbials in sentences according to the frequency distribution of various adverbials in the Longman English Spoken and Written Corpus; ② Error prediction - cognitive reinforcement loop submodule, which implements the consistency and rationality verification mechanism of adverbial combination through Python language plugin, automatically triggers the regeneration process when an adverbial logical conflict is detected; ③ Semantic association enhancement submodule: Loads a prompt dictionary containing 12,000 words of adverbial usage rules to enhance the semantic association between adverbials and verbs; ④ The weight allocation optimization submodule calls the pre-built knowledge base of 2,000 verb usages and adjusts the adverbial weight coefficients according to the common collocation patterns of verbs; (4) Output the English sentence optimized by the multimodal filtering mechanism. While maintaining grammatical correctness, the sentence also has the specificity of action description, the rationality of behavioral logic and the dialectical nature of semantic expression.

2. The intelligent system for generating English sentences according to claim 1, characterized in that... The construction of the adverbial mapping network specifically includes: (1) Establish a dynamic association matrix between verbs and nine types of adverbial phrases, with different activation conditions for each type of adverbial phrase: ① The activation threshold for adverbs of time, place, and manner is set to 0.5-0.6; ② When the adverbial phrases of cause and effect form a causal pair, their activation difference does not exceed 0.2; ③ Conditional and concessive adverbs adopt an adversarial activation mechanism; if there is no semantic contradiction or logical conflict, they should be used together. (2) Implement an adverbial selection algorithm based on the central meaning of the verb: ① Determine the core meaning of a verb through semantic disambiguation; ② Calculate the semantic similarity between verbs and various adverbial phrase types, and filter out candidate adverbial phrases that meet the similarity criteria; ③ Construct an adverbial distribution model with the verb as the origin in the word vector space; (3) Implement differentiated processing strategies for adverbial phrases of different dimensions: ① The specific dimension adverbial is assigned a processing weight of 0.4-0.5; ② The rationalization dimension adverbial is assigned a processing weight of 0.3-0.35; ③ The dialectical dimension adverbial is assigned a processing weight of 0.2-0.25.