Syntactic selection simulation method, syntactic selection simulation system and syntactic selection simulation equipment based on event structure and syntactic frequency interaction

Abstract

The invention discloses a syntactic selection simulation method, system and equipment based on event structure and syntactic frequency interaction. The method belongs to the technical field of dynamic simulation of Chinese syntax generation, and comprises the following steps of: 1, analyzing a question element role relationship of an input event, and generating an event structure vector; 2, calling a syntactic frequency database, and obtaining the use frequency weight of the candidate sentence pattern; 3, calculating the verb semantic distance between the target event and the historical event; 4, dynamically generating a syntactic selection probability through interaction; 5, when the events are the same and verbs are multiplexed, the syntax selection probability is adjusted. Through multi-dimensional interactive modeling of the event structure, the syntactic frequency and the semantic information, more accurate and natural syntactic selection simulation conforming to human language behavior rules is realized, and the intelligent level and simulation authenticity of a language generation system are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of dynamic simulation technology for Chinese syntax generation, and more specifically to a syntactic selection simulation method, system, and device based on the interaction between event structure and syntactic frequency. Background Technology

[0002] When speakers produce a sentence, they first generate an abstract conceptual structure of the event, then project this structure onto the language level for grammatical and phonetic encoding, and finally pronounce it. In recent years, the role of the abstract conceptual structure of events (or event structure) in sentence production has attracted attention. One core argument is that the structural priming effect observed in previous studies may be a downstream effect of event structure activation. In other words, event structure may influence the choice of syntactic structure during the grammatical encoding stage.

[0003] However, existing syntactic selection models have two major drawbacks: first, they do not model the dynamic impact of event structure (thematical role relationships) on syntactic selection; second, they ignore the moderating effect of syntactic usage frequency on event structure activation (e.g., high-frequency "ba" sentences are easily activated, while low-frequency "gei" sentences require strong triggering). The facilitating effect of event similarity on syntactic priming is significantly modulated by sentence frequency (high-frequency / medium-frequency sentences are effective, while low-frequency sentences are ineffective).

[0004] Therefore, how to provide a syntactic selection simulation method, system, and device based on the interaction between event structure and syntactic frequency is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a syntactic selection simulation method, system and device based on the interaction between event structure and syntactic frequency, to solve the technical problems existing in the prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A syntactic selection simulation method based on the interaction between event structure and syntactic frequency includes:

[0008] Step 1: Analyze the role relationships of the input events and generate an event structure vector;

[0009] Step 2: Call the syntax frequency database to obtain the usage frequency weights of candidate sentence patterns;

[0010] Step 3: Calculate the semantic distance between the target event and historical events using verbs;

[0011] Step 4: Dynamically generate syntactic selection probabilities through interactions;

[0012] Step 5: When the events are the same and the verbs are reused, adjust the syntactic selection probability to obtain the final syntactic selection result.

[0013] Furthermore, step 6 is included: generating natural language output based on the final syntactic selection result, and recording this syntactic selection behavior to the historical event log for updating the syntactic frequency database and verb reuse statistics.

[0014] Furthermore, in step 1, the topic role relationship includes at least one of agent, patient, participant, instrument, and location; the event structure vector is generated by mapping each topic role and its semantic features into a numerical vector representation.

[0015] Furthermore, the syntactic frequency database is constructed by statistically analyzing the frequency of use of various sentence patterns under the same or similar event types in a large-scale real corpus, and is indexed and stored according to event category and verb category.

[0016] Furthermore, the calculation of verb semantic distance in step 3 employs vector cosine similarity based on a pre-trained language model, specifically including:

[0017] The input event verbs and verbs in the historical corpus are mapped into high-dimensional semantic vectors respectively;

[0018] Calculate the cosine distance between the semantic vectors of two verbs as the semantic distance between them;

[0019] Semantic distance is converted into a semantic similarity metric for subsequent probabilistic modeling.

[0020] Furthermore, in step 4, the syntactic selection probability is dynamically generated through interaction, and the expression is:

[0021]

[0022] In the formula, P(S) i |E) indicates the choice of sentence structure S under event E. i The probability of S; j F(S) represents all sentence patterns. i (S is a sentence structure) i Historical usage frequency weight; D sem This represents the semantic distance between the current verb and its historical counterpart. This indicates that the j-th candidate sentence structure S is being considered. j Sim(V) represents the semantic distance between the input event verb and the verbs in the historical corpus that support the use of this sentence pattern; e V h V is the current event structure vector. e With historical event structure vector V h Similarity; Indicates the relationship with the j-th sentence pattern Sj The relevant historical event structure vector; α, β, and γ are adjustable model parameters.

[0023] Furthermore, step 5 includes:

[0024] Verb reuse intensity is measured by statistically analyzing the number of times or frequency a verb is used consecutively within a preset time window;

[0025] When the intensity of verb reuse exceeds a preset threshold, the syntactic selection probability of its historical preferred sentence patterns is increased.

[0026] A syntactic selection simulation system based on the interaction between event structure and syntactic frequency includes:

[0027] The event parsing module is used to parse the role relationships of the elements in the input event and generate an event structure vector;

[0028] The frequency retrieval module is used to call the syntactic frequency database to obtain the usage frequency weights of candidate sentence patterns;

[0029] The semantic computation module is used to calculate the verb semantic distance between the target event and historical events;

[0030] The probability modeling module is used to dynamically generate syntactic choice probabilities through interactions.

[0031] The probability adjustment module is used to adjust the syntactic selection probability when the events are the same and the verbs are reused.

[0032] The output module is used to generate natural language output based on the final syntactic selection result and record the syntactic selection behavior to the historical event log, which is used to update the syntactic frequency database and verb reuse statistics.

[0033] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a syntactic selection simulation method based on the interaction between event structure and syntactic frequency.

[0034] As can be seen from the above technical solutions, compared with the prior art, this invention discloses a syntactic selection simulation method, system, and device based on the interaction of event structure and syntactic frequency. By integrating event semantic structure, sentence usage frequency, and verb semantic similarity, and introducing a probability adjustment mechanism under verb reuse, it achieves syntactic selection simulation that is closer to human language cognition. It has the advantages of more realistic modeling, stronger generalization ability, interpretable output, and wide adaptability, and can effectively improve the naturalness and intelligence level of natural language generation. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0036] Figure 1 This is a schematic diagram of the method flow provided by the present invention;

[0037] Figure 2 This is a schematic diagram of the system structure provided by the present invention;

[0038] Figure 3 A diagram illustrating the syntactic selection for event structures. Detailed Implementation

[0039] 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 embodiments of the present invention, and not all embodiments. 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.

[0040] See Figure 1 This invention discloses a syntactic selection simulation method based on the interaction between event structure and syntactic frequency, comprising:

[0041] Step 1: Analyze the role relationships of the input events and generate an event structure vector;

[0042] Step 2: Call the syntax frequency database to obtain the usage frequency weights of candidate sentence patterns;

[0043] Step 3: Calculate the semantic distance between the target event and historical events using verbs;

[0044] Step 4: Dynamically generate syntactic selection probabilities through interactions;

[0045] Step 5: When the events are the same and the verbs are reused, adjust the syntactic selection probability.

[0046] In one specific embodiment, step 6 is also included: generating natural language output based on the final syntactic selection result, and recording this syntactic selection behavior to the historical event log for updating the syntactic frequency database and verb reuse statistics.

[0047] In one specific embodiment, the topic role relationship in step 1 includes at least one of agent, patient, participant, instrument, and location; the event structure vector is generated by mapping each topic role and its semantic features into a numerical vector representation.

[0048] Specifically, the user-described event text "girl breaks vase" is then analyzed to parse the thematic role relationships of the input event, generating an event structure vector, including:

[0049] Extract the verb (break) and the thematic roles (agent: girl; patient: vase);

[0050] Generate event relationship vectors: [verb, agent, patient, life feature].

[0051] In one specific embodiment, the syntax frequency database is constructed by statistically analyzing the frequency of use of various sentence patterns under the same or similar event types in a large-scale real corpus, and is indexed and stored according to event category and verb category.

[0052] Specifically, the syntax frequency database is called to obtain the usage frequency weights of candidate sentence patterns. The syntax frequency weights are statistically graded according to the Chinese corpus: "ba" sentence = high frequency (0.8), "bei" sentence = medium frequency (0.5), and "rang" sentence = low frequency (0.2). The stored frequencies of commonly used Chinese sentence patterns are shown in Table 1.

[0053] Table 1 Frequency of Commonly Used Chinese Sentence Patterns

[0054] Sentence structure Frequency level Event text "ba" sentence high frequency The girl broke the vase. passive voice Intermediate frequency The vase was broken by the girl. Give words and phrases low frequency The girl broke the vase.

[0055] Specifically, this embodiment includes a 2 (priming sentence type: verb predicate sentence / ba sentence) x 3 (event type: different / similar / same) within-subjects factor design. The dependent variable is the percentage of subjects who describe the target image as a ba sentence.

[0056] See Figure 3 In a specific embodiment, 120 key images are included, comprising 90 initiation images and 30 target images. Each initiation image has a corresponding initiation phrase at the top. An additional 120 filler images are included, of which 88 are irrelevant images describing intransitive and existential events, and 32 are memory test question images used to conceal the true purpose. The material is divided into 6 groups based on the level of the independent variable, with each target image appearing only once in each group and corresponding to one of the 6 initiation conditions. The initiation conditions are ordered by a Latin square to ensure that adjacent key trials are not repeated. Simultaneously, two mirrored versions of the initiation and target images are matched to balance the agent's left and right positions on the screen. Furthermore, the lifespan type is controlled so that the agent's lifespan in the target images is each 1 / 3 higher, equal to, or lower than the patient's.

[0057] Specifically, the materials were presented on a computer using E-Prime 3.0 software. Each participant was tested individually in a soundproof room. Instructions were presented first, followed by practice, with six images presented in six trials. In the crucial trial, the starting image was presented, and the participant was asked to read the sentence on the image aloud. Then, the target image was presented, and the participant was asked to describe it in one sentence. The computer automatically recorded the participant's verbal description. The participant pressed the spacebar to start the next trial. The entire process took approximately 25 minutes.

[0058] Specifically, this embodiment also includes a 2 (priming sentence type: verb predicate / passive voice) x 3 (event type: different / similar / same) within-subjects factorial design. The dependent variable is the percentage of subjects who describe the target image as a passive voice.

[0059] Specifically, this embodiment also includes a 2 (priming sentence type: verb predicate / given sentence) x 3 (event type: different / similar / same) within-subjects factor design. The dependent variable is the percentage of subjects who describe the target image as a given sentence.

[0060] In one specific embodiment, the calculation of verb semantic distance in step 3 adopts vector cosine similarity based on a pre-trained language model, specifically including:

[0061] The input event verbs and verbs in the historical corpus are mapped into high-dimensional semantic vectors respectively;

[0062] Calculate the cosine distance between the semantic vectors of two verbs as the semantic distance between them;

[0063] Semantic distance is converted into a semantic similarity metric for subsequent probabilistic modeling.

[0064] This invention comprehensively considers three key factors: event semantic structure (topic roles), sentence usage frequency, and verb semantic similarity. It dynamically calculates syntactic selection probabilities through an interaction model, more closely resembling the cognitive mechanisms in human language production. In particular, the introduction of "verb semantic distance" and "event structure similarity" as moderating variables effectively simulates the psychological process of sentence transfer based on analogy and experience in human language expression, enhancing the model's psychological interpretability.

[0065] By constructing a syntactic frequency database containing historical usage frequencies and introducing a probability adjustment mechanism in verb reuse scenarios, this invention can capture the "priming effect" and "habituation tendency" in language use. When the same verb or similar events occur repeatedly, the system automatically increases the selection probability of its historically preferred sentence patterns, thereby more realistically simulating the formation and reinforcement process of language habits and improving the naturalness and consistency of simulation results.

[0066] In one specific embodiment, step 4 dynamically generates the syntactic selection probability through interaction, and the expression is:

[0067]

[0068] In the formula, P(S) i |E) indicates the choice of sentence structure S under event E. i The probability of S; j F(S) represents all sentence patterns. i (S is a sentence structure) i Historical usage frequency weight; D sem This represents the semantic distance between the current verb and its historical counterpart. This indicates that the j-th candidate sentence structure S is being considered. j Sim(V) represents the semantic distance between the input event verb and the verbs in the historical corpus that support the use of this sentence pattern; e V h V is the current event structure vector. e With historical event structure vector V h Similarity; Indicates the relationship with the j-th sentence pattern S j The relevant historical event structure vector; α, β, and γ are adjustable model parameters.

[0069] This invention employs the aforementioned interaction formula for probabilistic modeling, outputting not only the final sentence selection result but also a quantified selection probability for each candidate sentence. This probabilistic output facilitates subsequent decision control, diversity adjustment, and model debugging, providing technical support for the controllability and interpretability of natural language generation systems.

[0070] In one specific embodiment, step 5 includes:

[0071] Verb reuse intensity is measured by statistically analyzing the number of times or frequency a verb is used consecutively within a preset time window;

[0072] When the intensity of verb reuse exceeds a preset threshold, the syntactic selection probability of its historical preferred sentence patterns is increased.

[0073] This invention achieves more accurate, natural, and human-behavioral syntactic selection simulation through multi-dimensional interactive modeling of event structure, syntactic frequency, and semantic information, significantly improving the intelligence level and simulation realism of language generation systems.

[0074] On the other hand, see Figure 2 This invention also discloses a syntactic selection simulation system based on the interaction between event structure and syntactic frequency, comprising:

[0075] The event parsing module is used to parse the role relationships of the elements in the input event and generate an event structure vector;

[0076] The frequency retrieval module is used to call the syntactic frequency database to obtain the usage frequency weights of candidate sentence patterns;

[0077] The semantic computation module is used to calculate the verb semantic distance between the target event and historical events;

[0078] The probability modeling module is used to dynamically generate syntactic choice probabilities through interactions.

[0079] The probability adjustment module is used to adjust the syntactic selection probability when the events are the same and the verbs are reused.

[0080] The output module is used to generate natural language output based on the final syntactic selection result and record the syntactic selection behavior to the historical event log, which is used to update the syntactic frequency database and verb reuse statistics.

[0081] On the other hand, embodiments of the present invention also disclose an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a syntactic selection simulation method based on the interaction between event structure and syntactic frequency.

[0082] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0083] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A syntactic selection simulation method based on the interaction between event structure and syntactic frequency, characterized in that, include: Step 1: Analyze the role relationships of the input events and generate an event structure vector; Step 2: Call the syntax frequency database to obtain the usage frequency weights of candidate sentence patterns; Step 3: Calculate the semantic distance between the target event and historical events using verbs; Step 4: Dynamically generate syntactic selection probabilities through interactions; Step 5: When the events are the same and the verbs are reused, adjust the syntactic selection probability.

2. The syntactic selection simulation method based on the interaction between event structure and syntactic frequency as described in claim 1, characterized in that, It also includes step 6: generating natural language output based on the final syntactic selection result, and recording this syntactic selection behavior to the historical event log for updating the syntactic frequency database and verb reuse statistics.

3. The syntactic selection simulation method based on the interaction between event structure and syntactic frequency as described in claim 1, characterized in that, In step 1, the thematic role relationships include at least one of agent, patient, participant, instrument, and location; the event structure vector is generated by mapping each thematic role and its semantic features into a numerical vector representation.

4. The syntactic selection simulation method based on the interaction between event structure and syntactic frequency as described in claim 1, characterized in that, The syntax frequency database is constructed by statistically analyzing the frequency of use of various sentence patterns under the same or similar event types in a large-scale real corpus, and is indexed and stored according to event category and verb category.

5. The syntactic selection simulation method based on the interaction between event structure and syntactic frequency according to claim 1, characterized in that, In step 3, the calculation of verb semantic distance uses vector cosine similarity based on a pre-trained language model, specifically including: The input event verbs and verbs in the historical corpus are mapped into high-dimensional semantic vectors respectively; Calculate the cosine distance between the semantic vectors of two verbs as the semantic distance between them; Semantic distance is converted into a semantic similarity metric for subsequent probabilistic modeling.

6. The syntactic selection simulation method based on the interaction between event structure and syntactic frequency according to claim 1, characterized in that, In step 4, the syntactic selection probability is dynamically generated through interaction, and the expression is: In the formula, P(S) i |E) indicates the choice of sentence structure S under event E. i The probability of S; j F(S) represents all sentence patterns. i (S is a sentence structure) i Historical usage frequency weight; D sem This represents the semantic distance between the current verb and its historical counterpart. This indicates that the j-th candidate sentence structure S is being considered. j Sim(V) represents the semantic distance between the input event verb and the verbs in the historical corpus that support the use of this sentence pattern; e V h V is the current event structure vector. e With historical event structure vector V h Similarity; Indicates the relationship with the j-th sentence pattern S j The relevant historical event structure vector; α, β, and γ are adjustable model parameters.

7. The syntactic selection simulation method based on the interaction between event structure and syntactic frequency according to claim 1, characterized in that, Step 5 includes: Verb reuse intensity is measured by statistically analyzing the number of times or frequency a verb is used consecutively within a preset time window; When the intensity of verb reuse exceeds a preset threshold, the syntactic selection probability of its historical preferred sentence patterns is increased.

8. A syntactic selection simulation system based on the interaction between event structure and syntactic frequency, characterized in that, include: The event parsing module is used to parse the role relationships of the elements in the input event and generate an event structure vector; The frequency retrieval module is used to call the syntactic frequency database to obtain the usage frequency weights of candidate sentence patterns; The semantic computation module is used to calculate the verb semantic distance between the target event and historical events; The probability modeling module is used to dynamically generate syntactic choice probabilities through interactions. The probability adjustment module is used to adjust the syntactic selection probability when the events are the same and the verbs are reused. The output module is used to generate natural language output based on the final syntactic selection result and record the syntactic selection behavior to the historical event log, which is used to update the syntactic frequency database and verb reuse statistics.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a syntactic selection simulation method based on the interaction between event structure and syntactic frequency as described in any one of claims 1 to 7.

Images