An information transmission system and method based on a semantic model

Abstract

The application relates to the technical field of industrial intelligent manufacturing, and discloses an information transmission system and method based on a semantic model, which comprises a semantic model information acquisition module, a word meaning association information construction module, an information transmission path analysis module, a path quantitative comprehensive evaluation module and a semantic model information output module, can recognize conversation information of a user by executing a running instruction of the semantic model, can set word meaning association graph nodes according to word meanings, can form conversation information extraction paths based on each word meaning node in the word meaning association graph, can determine candidate paths of information transmission, can perform index quantitative evaluation on the conversation information extraction paths and the candidate paths to determine an optimal path, can extract code information sequences of the optimal path to complete extraction of the conversation information, and can determine an optimal path of output information to complete output of the information, so that the efficiency and processing capacity of information transmission are effectively improved through quantitative analysis on the conversation information extraction paths and the candidate paths.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial intelligent manufacturing, more particularly to an information transmission system and method based on a semantic model. BACKGROUND

[0002] In the early stage of the development of communication technology, the core goal of information transmission is to achieve bit-level accurate transmission, that is, to ensure that the original data is completely consistent at the sending end and the receiving end, and the semantic connotation of the information is not interpreted, and the transmission process is completely disconnected from the semantic value of the information, resulting in a large amount of redundant data consuming bandwidth. With the development of mobile Internet, Internet of Things and other technologies, information transmission has entered the era of massive data + complex scenarios, and the global data volume has grown exponentially, with Internet of Things devices, high-definition videos and other technologies becoming the main sources of data. The traditional full-quantity transmission mode is difficult to cope with the bandwidth pressure, and the sensitivity of information semantic value in different scenarios is significantly different.

[0003] With the rapid development of industrial intelligent manufacturing technology, the efficiency and accuracy of the information transmission system have become key factors restricting the level of intelligent manufacturing. Semantic communication takes understanding first and transmission second as the core, and compresses redundant data through a semantic model to only transmit key features.

[0004] However, traditional information transmission systems often rely on fixed transmission protocols and formats, making it difficult to adapt to complex and changing user conversation information, resulting in low information transmission efficiency and easy information loss or misunderstanding. At the same time, traditional information transmission systems rely on fixed transmission paths and cannot flexibly adjust transmission strategies according to the dynamic changes in user conversation content, thus limiting the real-time and accuracy of information transmission. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, the present application provides an information transmission system based on a semantic model to solve the problems existing in the background art.

[0006] The present application provides the following technical solution: an information transmission system based on a semantic model, comprising: a semantic model information acquisition module, a word meaning association information construction module, an information transmission path analysis module, a path quantization comprehensive evaluation module, and a semantic model information output module;

[0007] The semantic model information acquisition module includes a semantic model running layer and a conversation information recognition layer, which recognizes user conversation information by executing the running instructions of the semantic model, and the conversation information includes historical dialogue and current dialogue;

[0008] The word meaning association information construction module includes a word meaning node matching layer and an edge relationship construction layer, which sets word meaning association graph nodes according to word meaning and connects word meaning association graph nodes to form the edges of the word meaning association graph according to word meaning relationship.

[0009] The information transmission path analysis module forms a conversation information extraction path based on each word sense node in the word sense association graph, and determines a candidate path of information transmission;

[0010] The path quantitative comprehensive evaluation module is used for quantitative evaluation of the conversation information extraction path and the candidate path, judges the optimal path based on the evaluation result, and real-time acquires the encoding information of each word sense node in the optimal path and extracts the encoding information sequence.

[0011] The semantic model information output module completes the extraction of the conversation information based on the encoding information sequence corresponding to the real-time updated optimal path, and determines the optimal path of the output information to complete the output of the information.

[0012] Preferably, the semantic model information acquisition module includes the specific contents of the semantic model running layer and the conversation information recognition layer as follows:

[0013] The semantic model running layer executes the running instruction of the semantic model to connect the multi-source conversation entrance, comprehensively captures the historical conversation and the current conversation of the user, classifies and stores the historical conversation according to the time stamp and the conversation scene label, and real-time accesses the current conversation in the form of streaming data.

[0014] The conversation information recognition layer adopts the natural language processing technology to perform semantic understanding on the user conversation content, recognizes the word sense and the word sense relationship contained in the user conversation content, extracts the user conversation word sense using the natural language processing technology for the current conversation of the user, and recognizes the word sense relationship based on the historical conversation of the user using the natural language processing technology.

[0015] Preferably, the word sense association information construction module includes the specific contents of the word sense node matching layer and the edge relationship construction layer as follows:

[0016] The word sense node matching layer extracts the word sense unit sequence contained in the user conversation content, the word sense unit sequence represents a sequence constructed according to the order of the word sense appearing in the user conversation content, and performs standardization processing on the word sense unit sequence to construct the node of the word sense association graph, wherein the standardization processing of the word sense unit includes stem extraction, morphological restoration and stop word processing.

[0017] The edge relationship construction layer connects the word sense association graph nodes based on the order of the word sense appearing, constructs the dynamic edge of the word sense association graph according to the direction of the word sense association graph node, and the dynamic edge of the word sense association graph represents the directed edge composed of the word sense association graph nodes connected according to the word sense order.

[0018] Preferably, the specific content of the information transmission path analysis module is as follows:

[0019] The word meaning association information construction module connects word meaning association graph nodes in accordance with word meaning sequence to form a directed edge, forming a conversation information extraction path;

[0020] The nodes in the conversation information extraction path are labeled, and equivalent semantic replacement units of different nodes are obtained. Based on the initially formed conversation information extraction path, the equivalent semantic replacement units are used as nodes to generate multiple parallel candidate paths.

[0021] Preferably, the path quantitative comprehensive evaluation module has the following specific content:

[0022] The transmission time of the conversation information extraction path and the candidate path is quantitatively analyzed to obtain the quantitative analysis result of the transmission time of the conversation information extraction path and the candidate path.

[0023] The semantic fidelity of the conversation information extraction path and the candidate path is quantitatively analyzed to obtain the quantitative analysis result of the semantic fidelity of the conversation information extraction path and the candidate path.

[0024] The transmission time and semantic fidelity of the conversation information extraction path and the candidate path are weighted according to the preset weight value, and the path with the highest weighted score is selected as the optimal information transmission path.

[0025] Preferably, the specific content of the transmission time quantitative analysis of the conversation information extraction path and the candidate path is as follows:

[0026] Each node in the conversation information extraction path is simultaneously sent a node processing instruction through a semantic model, and the node receiving time, node reaction time, node processing time, and node transmission delay time of each node in the conversation information extraction path are recorded.

[0027] The total running time of each node in the conversation information extraction path is counted, and the maximum node processing time in the conversation information extraction path is obtained by maximum analysis of the node processing time of each node in the conversation information extraction path according to the total running time. The maximum node processing time represents the transmission time of the conversation information extraction path.

[0028] The transmission time of each node in the candidate path is counted, and the transmission time represents the node replacement time, node receiving time, node reaction time, node processing time, and node transmission delay time of each node in the candidate path. The maximum node processing time of each path in the candidate path is obtained, and the maximum node processing time represents the transmission time of the conversation information extraction path.

[0029] The transmission time of the conversation information extraction path and the transmission time of the candidate path are combined to obtain the quantitative analysis result of the transmission time of the conversation information extraction path and the candidate path.

[0030] Preferably, the specific content of the semantic fidelity quantitative analysis of the conversation information extraction path and the candidate path is as follows:

[0031] The conversation information extraction path represents a path formed according to the original semantics, and the semantic fidelity of the conversation information extraction path is 1;

[0032] Each candidate path is compared with the conversation information extraction path, and the semantic fidelity of each candidate path and the conversation information extraction path is calculated, the semantic fidelity representing the similarity of the node vectors in each candidate path and the conversation information extraction path, and the specific value being obtained by cosine similarity;

[0033] The semantic fidelity of the conversation information extraction path and the semantic fidelity of the conversation information extraction path are combined to obtain the quantitative analysis result of the semantic fidelity of the conversation information extraction path and the candidate path.

[0034] Preferably, the specific content of the path quantitative comprehensive evaluation module is as follows:

[0035] The conversation information extraction path and the candidate path are quantitatively evaluated based on the preset index weight to obtain the quantitative evaluation result of the conversation information extraction path and each candidate path;

[0036] The maximum value of the quantitative evaluation result is screened, the path corresponding to the maximum value is determined as the optimal path, and the node code information sequence is extracted according to the optimal path to complete the extraction of the conversation information.

[0037] Preferably, the semantic model information output module, based on the code information sequence corresponding to the real-time updated optimal path, completes the extraction of the conversation information, and according to the extracted conversation information, obtains output information through a preset output instruction, and repeats the step of obtaining the optimal path to complete the output of the information.

[0038] An information transmission method based on a semantic model, comprising the following steps:

[0039] Step S01: Recognize the conversation information of the user by executing the running instruction of the semantic model, the conversation information including historical dialogue and current dialogue;

[0040] Step S02: Set the word sense association graph node according to the word sense, and connect the word sense association graph nodes according to the word sense relationship to form the edge of the word sense association graph;

[0041] Step S03: Form the conversation information extraction path based on each word sense node in the word sense association graph, and determine the candidate path of information transmission;

[0042] Step S04: index quantization evaluation is carried out on the session information extraction path and the candidate path, the optimal path is judged based on the evaluation result, the encoding information of each word sense node in the optimal path is obtained in real time, and the encoding information sequence is extracted;

[0043] Step S05: based on the encoding information sequence corresponding to the real-time updated optimal path, the session information is extracted, and the optimal path of the output information is determined to complete the output of the information.

[0044] The technical effects and advantages of the present application are as follows:

[0045] The present application realizes efficient and accurate transmission of information through the mutual cooperation of the semantic model information acquisition module, the word sense association information construction module, the information transmission path analysis module, the path quantization comprehensive evaluation module and the semantic model information output module.

[0046] By constructing the word sense association graph, the relationship between the word senses is effectively revealed, which provides strong support for the analysis of the information transmission path; and based on the word sense association graph, the session information extraction path is formed, the candidate path is obtained according to the replaceability of the word senses, and multiple choices are provided for subsequent evaluation; the optimal path is accurately selected through quantitative analysis, which ensures the efficiency and accuracy of information transmission;

[0047] By compressing redundant data of the semantic model, only the key features are transmitted, the bandwidth pressure is significantly reduced, the information transmission efficiency is improved, the construction of multiple paths adapts to the complex and variable user session information, and the flexibility and real-time performance of information transmission are effectively improved through dynamic adjustment of the transmission strategy. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 It is a structural schematic diagram of an information transmission system based on a semantic model.

[0049] Figure 2 It is a flowchart of an information transmission method based on a semantic model. DETAILED DESCRIPTION

[0050] The technical solutions in the present application will be described in detail below with reference to the drawings in the present application, and the forms of each structure described in the following embodiments are only examples, and the information transmission system and transmission method based on the semantic model involved in the present application are not limited to the forms of each structure described in the following embodiments. All other embodiments obtained by those skilled in the art without making creative efforts fall within the scope of protection of the present application.

[0051] As Figure 1As shown, the application provides a semantic model-based information transmission system, comprising: a semantic model information acquisition module, a word sense associated information construction module, an information transmission path analysis module, a path quantitative comprehensive evaluation module, and a semantic model information output module;

[0052] The semantic model information acquisition module comprises a semantic model running layer and a conversation information recognition layer, which identifies the conversation information of the user by executing the running instructions of the semantic model, and the conversation information comprises historical dialogues and current dialogues.

[0053] The word sense associated information construction module comprises a word sense node matching layer and an edge relationship construction layer, which sets the word sense associated graph nodes according to the word sense, and connects the word sense associated graph nodes to form the edges of the word sense associated graph according to the word sense relationship.

[0054] The information transmission path analysis module forms the conversation information extraction path based on each word sense node in the word sense associated graph, and determines the candidate path of information transmission.

[0055] The path quantitative comprehensive evaluation module is used for index quantitative evaluation of the conversation information extraction path and the candidate path, judges the optimal path based on the evaluation result, and real-time acquires the encoding information of each word sense node in the optimal path and extracts the encoding information sequence.

[0056] The semantic model information output module completes the extraction of the conversation information based on the encoding information sequence corresponding to the real-time updated optimal path, and determines the optimal path of the output information to complete the output of the information.

[0057] In this embodiment, it needs to be specifically pointed out that the specific content of the semantic model information acquisition module comprising the semantic model running layer and the conversation information recognition layer is as follows:

[0058] The semantic model running layer executes the running instructions of the semantic model to interface multiple source conversation entrances, comprehensively captures the historical dialogues and current dialogues of the user, classifies and stores the historical dialogues according to the time stamp and dialogue scene label, and real-time accesses the current dialogues in the form of streaming data.

[0059] The conversation information recognition layer adopts natural language processing technology to perform semantic understanding on the user conversation content, identifies the word sense and word sense relationship contained in the user conversation content, extracts the user conversation word sense using natural language processing technology for the current dialogue of the user, and identifies the word sense relationship based on the historical conversation of the user using natural language processing technology.

[0060] In this embodiment, it needs to be specifically pointed out that the specific content of the word sense associated information construction module comprising the word sense node matching layer and the edge relationship construction layer is as follows:

[0061] The word sense node matching layer extracts a word sense unit sequence contained in the user session content, the word sense unit sequence represents a sequence constructed according to the order of appearance of word senses in the user session content, and the word sense unit sequence is standardized to construct nodes of the word sense association graph, wherein the standardization of the word sense unit includes stem extraction, lemmatization and stop word processing;

[0062] The edge relationship construction layer connects the word sense association graph nodes based on the order of appearance of the word senses, and constructs dynamic edges of the word sense association graph according to the direction of the word sense association graph nodes, the dynamic edges of the word sense association graph represent directed edges formed by connecting the word sense association graph nodes according to the order of the word senses, and the directed edges dynamically change according to the order of the word sense association graph nodes in the user session information.

[0063] In this embodiment, the specific content of the information transmission path analysis module is as follows:

[0064] Based on the directed edges formed by connecting the word sense association graph nodes according to the order of the word senses in the word sense association information construction module, a session information extraction path is formed, the session information extraction path represents a reference path, when a candidate path appears, the candidate path is quantitatively analyzed with the reference path, and then the optimal path can be obtained;

[0065] The nodes in the session information extraction path are labeled, and equivalent semantic replacement units of different nodes are obtained, based on the initially formed session information extraction path, the equivalent semantic replacement units are used as nodes to generate a plurality of parallel candidate paths.

[0066] In this embodiment, the specific content of the path quantization comprehensive evaluation module is as follows:

[0067] The transmission time of the session information extraction path and the candidate path is quantitatively analyzed, and the quantization analysis result of the transmission time of the session information extraction path and the candidate path is obtained;

[0068] The semantic fidelity of the session information extraction path and the candidate path is quantitatively analyzed, and the quantization analysis result of the semantic fidelity of the session information extraction path and the candidate path is obtained;

[0069] According to the preset weight value, the transmission time and the semantic fidelity of the session information extraction path and the candidate path are weighted calculated, and the path with the highest weighted score is selected as the optimal information transmission path.

[0070] In this embodiment, the specific content of the transmission time quantization analysis of the session information extraction path and the candidate path is as follows:

[0071] The node processing instructions are simultaneously sent to each node in the session information extraction path through the semantic model, and the node receiving time, node reaction time, node processing time and node transmission delay time of each node in the session information extraction path are recorded;

[0072] The total running time of each node in the session information extraction path is counted, and the maximum node processing time in the session information extraction path is obtained by maximum analysis of the node processing time of each node in the session information extraction path according to the total running time ;

[0073] The transmission time of each node in the candidate path is counted, the transmission time representing the node replacement time consumption, node receiving time, node reaction time, node processing time and node transmission delay time of each node in the candidate path, and the maximum node processing time of each path in the candidate path is obtained , wherein i represents the candidate path number, i = 0, 1, 2,..., n;

[0074] The transmission time of the session information extraction path and the transmission time of the candidate path are combined to obtain the quantitative analysis result of the transmission time of the session information extraction path and the candidate path, which represents the maximum node processing time of each path.

[0075] In this embodiment, it needs to be specifically pointed out that the specific content of the quantitative analysis of the semantic fidelity of the session information extraction path and the candidate path is as follows:

[0076] The session information extraction path represents the path formed according to the original semantics, which is the reference path, and the semantic fidelity of the session information extraction path is 1;

[0077] Each candidate path is compared with the session information extraction path to calculate the semantic fidelity of each candidate path and the session information extraction path, which represents the similarity of the node vectors in each candidate path and the session information extraction path, and the specific numerical value is obtained through the cosine similarity;

[0078] The candidate path is obtained by replacing the nodes in the session information extraction path, and it is known that the number of nodes in the session information extraction path and the candidate path is consistent. The directed edge formed between adjacent nodes in the session information extraction path is represented by a vector, which is represented as The directed edge formed between adjacent nodes in the candidate path is represented by a vector, which is represented as , m = 1, 2, 3,..., M, wherein m represents the node vector number, and M+1 represents the total number of nodes;

[0079] The semantic fidelity of the corresponding nodes in the session information extraction path and the candidate path is calculated by cosine similarity calculation, and the calculation formula is , which represents the ratio of the dot product of the corresponding node vectors in the session information extraction path and the candidate path to the modulus, is a conventional cosine similarity calculation formula, and the value range is [-1, 1], and the larger the value is, the higher the similarity is;

[0080] The semantic fidelity of the corresponding nodes in the session information extraction path and the candidate path is summed and averaged to obtain the semantic fidelity of the candidate path, and the calculation formula is:

[0081] The semantic fidelity of the session information extraction path and the semantic fidelity of the session information extraction path are combined to obtain the quantitative analysis result of the semantic fidelity of the session information extraction path and the candidate path.

[0082] In this embodiment, it needs to be specifically pointed out that the specific content of the path quantitative comprehensive evaluation module is as follows:

[0083] The session information extraction path and the candidate path are quantitatively evaluated based on the preset index weight to obtain the quantitative evaluation result of the session information extraction path and each candidate path;

[0084] The transmission time of the session information extraction path and each candidate path is normalized to obtain a dimensionless transmission time normalization index T, and when the weight of the transmission time is , the weight of the semantic fidelity is , the session information extraction path and each candidate path are quantitatively evaluated, and the evaluation formula is: , wherein D represents the quantitative evaluation result,

[0085] When the transmission time normalization index T of the session information extraction path is 0.4, the semantic fidelity P of the session information extraction path is a fixed value 1, and when , , the quantitative evaluation result of the session information extraction path is: .

[0086] The maximum value of the quantitative evaluation result is selected, the path corresponding to the maximum value is judged as the optimal path, and the node coding information sequence is extracted according to the optimal path to complete the extraction of the session information.

[0087] In this embodiment, it needs to be specifically pointed out that the semantic model information output module, based on the coding information sequence corresponding to the real-time updated optimal path, completes the extraction of the session information, and according to the extracted session information, the output information is obtained through the preset output instruction, and the acquisition step of the optimal path is repeated to complete the output of the information.​​

[0088] As Figure 2 shown in the embodiment, it needs to be specifically pointed out that a semantic model-based information transmission method includes the following steps:

[0089] Step S01: Identify the user's session information by executing the running instructions of the semantic model, the session information including historical dialogues and current dialogues;

[0090] Step S02: Set the word sense association graph node according to the word sense, and connect the word sense association graph node according to the word sense relationship to form the edge of the word sense association graph;

[0091] Step S03: Form a session information extraction path based on each word sense node in the word sense association graph, and determine a candidate path for information transmission;

[0092] Step S04: Quantitative evaluation of session information extraction path and candidate path, based on the evaluation results to determine the optimal path, and real-time acquisition of the encoding information of each word sense node in the optimal path, extraction of the encoding information sequence;

[0093] Step S05: Based on the real-time updated encoding information sequence corresponding to the optimal path, complete the extraction of the session information, and determine the optimal path of the output information to complete the output of the information.

[0094] In the embodiment, it needs to be specifically pointed out that the difference between the present embodiment and the prior art is mainly that the present embodiment realizes efficient and accurate transmission of information by cooperating with the semantic model information acquisition module, the word sense association information construction module, the information transmission path analysis module, the path quantitative comprehensive evaluation module and the semantic model information output module;

[0095] By constructing the word sense association graph, the relationship between the word senses is effectively revealed, which provides strong support for the analysis of the information transmission path; and based on the word sense association graph, the session information extraction path is formed, the candidate path is obtained according to the replaceability of the word sense, which provides a variety of choices for subsequent evaluation; the optimal path is accurately selected through quantitative analysis, which ensures the efficiency and accuracy of information transmission;

[0096] By compressing redundant data through the semantic model, only the key features are transmitted, which significantly reduces the bandwidth pressure, improves the information transmission efficiency, and the construction of multiple paths adapts to the complex and variable user session information, through dynamic adjustment of the transmission strategy, effectively improves the flexibility and real-time performance of information transmission.

[0097] Finally: the above only for the preferred embodiment of the present application, and not for limiting the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application, should be included in the protection scope of the present application.

[0098] The above description is only specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A semantic model based information transmission system, characterized by: The application relates to a semantic model information acquisition module, a word meaning association information construction module, an information transmission path analysis module, a path quantitative comprehensive evaluation module and a semantic model information output module. The semantic model information acquisition module comprises a semantic model running layer and a conversation information identification layer, conversation information of a user is identified by executing a running instruction of a semantic model, and the conversation information comprises historical dialogue and current dialogue; The word meaning association information construction module comprises a word meaning node matching layer and an edge relationship construction layer, word meaning association graph nodes are set according to word meanings, and edges of the word meaning association graph are formed by connecting the word meaning association graph nodes according to word meaning relationships; The information transmission path analysis module forms conversation information extraction paths based on the word meaning nodes in the word meaning association graph, and determines candidate paths of information transmission; The path quantitative comprehensive evaluation module is used for performing index quantitative evaluation on the conversation information extraction paths and the candidate paths, judging an optimal path based on an evaluation result, and acquiring encoding information of each word meaning node in the optimal path in real time and extracting an encoding information sequence; The semantic model information output module completes extraction of conversation information based on the encoding information sequence corresponding to the optimal path updated in real time, and determines an optimal path of output information to complete information output. The semantic model information acquisition module comprises a semantic model running layer and a conversation information identification layer, conversation information of a user is identified by executing a running instruction of a semantic model, and the conversation information comprises historical dialogue and current dialogue; 2. The information transmission system based on semantic model according to claim 1, characterized in that: The word meaning association information construction module comprises a word meaning node matching layer and an edge relationship construction layer, word meaning association graph nodes are set according to word meanings, and edges of the word meaning association graph are formed by connecting the word meaning association graph nodes according to word meaning relationships; The information transmission path analysis module forms conversation information extraction paths based on the word meaning nodes in the word meaning association graph, and determines candidate paths of information transmission; The path quantitative comprehensive evaluation module is used for performing index quantitative evaluation on the conversation information extraction paths and the candidate paths, judging an optimal path based on an evaluation result, and acquiring encoding information of each word meaning node in the optimal path in real time and extracting an encoding information sequence; 3. The information transmission system based on semantic model according to claim 1, characterized in that: The semantic model information output module completes extraction of conversation information based on the encoding information sequence corresponding to the optimal path updated in real time, and determines an optimal path of output information to complete information output. The semantic model information acquisition module comprises a semantic model running layer and a conversation information identification layer, conversation information of a user is identified by executing a running instruction of a semantic model, and the conversation information comprises historical dialogue and current dialogue; The word meaning association information construction module comprises a word meaning node matching layer and an edge relationship construction layer, word meaning association graph nodes are set according to word meanings, and edges of the word meaning association graph are formed by connecting the word meaning association graph nodes according to word meaning relationships; 4. The information transmission system based on semantic model according to claim 1, characterized in that: The information transmission path analysis module forms conversation information extraction paths based on the word meaning nodes in the word meaning association graph, and determines candidate paths of information transmission; The path quantitative comprehensive evaluation module is used for performing index quantitative evaluation on the conversation information extraction paths and the candidate paths, judging an optimal path based on an evaluation result, and acquiring encoding information of each word meaning node in the optimal path in real time and extracting an encoding information sequence; The semantic model information output module completes extraction of conversation information based on the encoding information sequence corresponding to the optimal path updated in real time, and determines an optimal path of output information to complete information output. The nodes in the conversation information extraction path are marked, and equivalent semantic replacement units of different nodes are obtained respectively, and based on the initial formed conversation information extraction path, the equivalent semantic replacement units are taken as nodes to generate a plurality of parallel candidate paths.

5. The information transmission system based on semantic model according to claim 1, characterized in that: The specific content of the path quantitative comprehensive evaluation module is as follows: The transmission time of the conversation information extraction path and the candidate path is quantitatively analyzed, and the quantitative analysis result of the transmission time of the conversation information extraction path and the candidate path is obtained. The semantic fidelity of the conversation information extraction path and the candidate path is quantitatively analyzed, and the quantitative analysis result of the semantic fidelity of the conversation information extraction path and the candidate path is obtained. The transmission time and the semantic fidelity of the conversation information extraction path and the candidate path are weighted according to the preset weight value, and the path with the highest weighted score is selected as the optimal information transmission path.

6. The information transmission system based on semantic model according to claim 5, characterized in that: The specific content of the transmission time quantitative analysis of the conversation information extraction path and the candidate path is as follows: The nodes in the conversation information extraction path are marked, and equivalent semantic replacement units of different nodes are obtained respectively, and based on the initial formed conversation information extraction path, the equivalent semantic replacement units are taken as nodes to generate a plurality of parallel candidate paths. The specific content of the transmission time quantitative analysis of the conversation information extraction path and the candidate path is as follows: The nodes in the conversation information extraction path are marked, and equivalent semantic replacement units of different nodes are obtained respectively, and based on the initial formed conversation information extraction path, the equivalent semantic replacement units are taken as nodes to generate a plurality of parallel candidate paths. The specific content of the transmission time quantitative analysis of the conversation information extraction path and the candidate path is as follows:

7. The information transmission system based on semantic model according to claim 5, characterized in that: The specific content of the semantic fidelity quantitative analysis of the conversation information extraction path and the candidate path is as follows: The conversation information extraction path represents the path formed according to the original semantics, so the semantic fidelity of the conversation information extraction path is 1. Each candidate path is compared with the conversation information extraction path to calculate the semantic fidelity of each candidate path and the conversation information extraction path, which represents the similarity of the node vectors in each candidate path and the conversation information extraction path, and the specific value is obtained by cosine similarity. The semantic fidelity of the conversation information extraction path and the semantic fidelity of the conversation information extraction path are combined to obtain the quantitative analysis result of the semantic fidelity of the conversation information extraction path and the candidate path.

8. The information transmission system based on semantic model according to claim 1, characterized in that: The specific content of the path quantitative comprehensive evaluation module is as follows: Based on the preset index weight, the conversation information extraction path and the candidate path are quantitatively evaluated, and the quantitative evaluation result of the conversation information extraction path and each candidate path is obtained. The maximum value screening is performed on the quantized evaluation result, the path corresponding to the maximum value is determined as the optimal path, and the node coding information sequence is extracted according to the optimal path, so as to complete the extraction of the conversation information.

9. The information transmission system based on semantic model according to claim 1, characterized in that: The semantic model information output module extracts the conversation information based on the coding information sequence corresponding to the real-time updated optimal path, acquires output information according to the extracted conversation information through a preset output instruction, and repeatedly performs the optimal path acquisition step to complete the information output.

10. A method for information transmission based on semantic model, using the information transmission system based on semantic model according to any one of claims 1-9. The method comprises the following steps: Step S01: identifying conversation information of a user by executing a running instruction of a semantic model, wherein the conversation information comprises historical dialogue and current dialogue; Step S02: setting a word sense association graph node according to a word sense, and connecting word sense association graph nodes according to a word sense relationship to form edges of a word sense association graph; Step S03: forming a conversation information extraction path based on each word sense node in the word sense association graph, and determining a candidate path of information transmission; Step S04: performing index quantization evaluation on the conversation information extraction path and the candidate path, determining an optimal path based on an evaluation result, and real-time acquiring coding information of each word sense node in the optimal path to extract a coding information sequence; Step S05: extracting conversation information based on a coding information sequence corresponding to a real-time updated optimal path, and determining an optimal path of output information to complete information output.

Images