CN-122027705-A - Intelligent gateway for heterogeneous maritime broadcasting protocol

CN122027705ACN 122027705 ACN122027705 ACN 122027705ACN-122027705-A

Abstract

The invention discloses an intelligent gateway for heterogeneous maritime broadcasting protocols, which comprises a data access layer, an intelligent processing layer and a protocol output layer, wherein the intelligent processing layer comprises a natural language processing module, a content association engine, a multi-protocol converter and a guide instruction generator, wherein the natural language processing module is configured to conduct semantic analysis on input maritime safety information and extract key elements comprising time, place and event types, the content association engine is configured to establish association relations among multi-source information aiming at the same maritime event based on the key elements, the multi-protocol converter is configured to generate broadcasting signals conforming to NAVDAT, weather fax and NAVTEX protocols according to the association relations, and the guide instruction generator is configured to generate standardized guide instructions related across protocols. The intelligent gateway for the heterogeneous maritime broadcasting protocol can realize intelligent understanding and automatic association of maritime information, solves the problem that multi-system fusion depends on manual operation, and provides unified, efficient and enhanced information service.

Inventors

ZHENG DEFU
WANG RUI
AN YUTING

Assignees

上海埃威信息科技有限公司

Dates

Publication Date: 20260512
Application Date: 20251223

Claims (10)

1. The intelligent gateway for the heterogeneous maritime broadcasting protocol is characterized by comprising a data access layer, an intelligent processing layer and a protocol output layer, wherein the data access layer is used for receiving and standardizing multi-source heterogeneous data, receiving texts, images and structured data from different information sources, and the protocol output layer is used for modulating and transmitting signals and finally generating and outputting broadcasting signals conforming to three target system systems of NAVDAT, weather fax and NAVTEX; the intelligent processing layer comprises: the natural language processing module is configured to perform semantic analysis on the input maritime safety information and extract key elements including time, place and event types; the content association engine is configured to establish an association relationship among multi-source information aiming at the same maritime event based on the key elements; a multi-protocol converter configured to generate broadcast signals conforming to NAVDAT, weather facsimile and NAVTEX protocols according to the association relationship; A directives generator configured to generate standardized directives associated across protocols.
2. The intelligent gateway of claim 1, wherein the natural language processing module comprises: The element identification sub-module is used for identifying geographic entities, time entities and event entities in the text by adopting a pre-trained language model; The semantic understanding sub-module establishes a logic relationship between key elements based on a deep learning model of an attention mechanism.
3. The intelligent gateway of claim 2, wherein the element recognition submodule adopts a transducer-based pre-training model to optimally train the terms and expressions of the maritime domain as follows: The data preparation comprises the steps of constructing a professional corpus in the maritime field, wherein corpus sources comprise an international maritime organization general function, a historical NAVTEX message database, world meteorological organization meteorological coding data, navigation warning and maritime accident report; preprocessing, namely cleaning and denoising the language materials, and carrying out standardized processing on related expressions related to a coordinate format and a technical term; Continuously pre-training, namely loading a general Chinese BERT model as a basis, continuously pre-training the model by using a maritime corpus with a mask language model as a target, and embedding vocabulary, syntax and knowledge in the maritime field into parameters of the maritime corpus by predicting masked words in the pre-training process; Task trimming, namely performing supervised trimming of a named entity recognition task on the pre-trained model by using a marine text data set marked with time, place and event type labels, and mapping an input sequence into a structured entity label sequence.
4. The intelligent gateway of claim 1, wherein the content correlation engine calculates a composite correlation Score for the multi-source information using the formula Score = α×ss + β×sst + γ×st, ss is semantic similarity, sst is temporal-spatial overlap, st is temporal consistency, α, β, γ are configurable weight parameters, and α+β+γ = 1 is satisfied; And when the Score is more than or equal to 1 and more than or equal to θassoc, automatically establishing the association relation among the multi-source information.
5. The intelligent gateway according to claim 4, wherein the semantic similarity is obtained by calculating cosine similarity by combining a TF-IDF algorithm with a marine dictionary to generate text vectors, the space-time overlapping degree is used for measuring the overlapping degree of events in geography and time, sst= Sgeo × Stime, sgeo is space overlapping degree, stime is time overlapping degree, and the time-efficiency consistency is used for calculating freshness of two pieces of information respectively based on the current time of the system and taking the minimum value of freshness of the two pieces of information.
6. The intelligent gateway of claim 5, wherein the time overlap is obtained by calculating the overlap ratio of two event effective time windows, or using a decay function of Stime = exp (-. DELTA.t/. Tau), where DELTA.t is the difference between key time points and tau is the decay constant.
7. The intelligent gateway according to claim 5, wherein for punctiform positions, the spatial overlap calculates a great circle distance d, then the great circle distance d is converted into a value between intervals 0-1 through a Gaussian attenuation function, sgeo = exp (-d 2 /(σ 2 )), σ is an attenuation radius, and for a region, the spatial overlap calculates an intersection ratio.
8. The intelligent gateway of claim 1, wherein the multi-protocol converter comprises: The NAVDAT protocol adapter encapsulates the structured data into a digital data packet conforming to the ITU-RM.2012 standard, applies forward error correction coding and generates an OFDM signal which can be broadcast in an MF/HF band; The weather facsimile protocol adapter converts the related image into an analog facsimile signal conforming to the WMO standard, and broadcasts the analog facsimile signal in a high-frequency band, including generating a scanning synchronous signal and image modulation; The NAVTEX protocol adapter encodes the structured text alert in accordance with the ITU-RM.2010 standard to generate an FSK modulated signal that is broadcast at an intermediate frequency of 518 kHz.
9. The intelligent gateway of claim 1, wherein the direction instruction format generated by the direction instruction generator comprises a target system, a frequency, a time, and a unique event identifier, wherein the direction instructions are embedded in different protocols as follows: embedding a guide instruction into a message body in a NAVTEX message; Embedding a guide instruction as metadata into a file header in the NAVDAT signal; In a weather facsimile signal, the guidance instructions are embedded in a machine-readable graphic code form in a non-core region of the image.
10. The intelligent gateway of claim 4, wherein the content association engine performs a confidence evaluation on the automatically established association relationship, triggers a manual auditing process when the confidence is lower than a threshold, optimizes association algorithm parameters according to auditing results, and specifically comprises: the comprehensive association Score is mapped into confidence coefficient through an S-shaped function, or the comprehensive association Score, the semantic similarity Ss, the space-time overlap Sst and the time efficiency consistency St are input into a pre-trained confidence coefficient evaluation model to obtain the confidence coefficient; recording a plurality of groups of manual auditing results and corresponding associated feature vectors; Adjusting weight parameters alpha, beta and gamma by a gradient descent method with the aim of minimizing the difference between the manual auditing result and the result predicted based on the associated feature vector; The objective function to minimize the prediction error is the following cross entropy loss function: L= - Σt·log (Score) + (1-T) +log (1-Score) ], t=1 indicating that the manual audit result is correct, and t=0 indicating that the manual audit result is incorrect.

Description

Intelligent gateway for heterogeneous maritime broadcasting protocol Technical Field The present invention relates to a gateway, and more particularly, to an intelligent gateway for heterogeneous maritime broadcasting protocols. Background With the development of maritime communication technology, a coexistence situation of multiple heterogeneous broadcasting systems is formed. NAVTEX system is broadcast in 518kHz Medium Frequency (MF) with text coding protocol, weather facsimile system is broadcast in several High Frequency (HF) bands with analog image protocol, NAVDAT system is used as new generation digital broadcasting system, and can be broadcast in medium frequency (500 kHz) and several high frequency bands with high speed. The systems have significant differences in protocol formats, data structures and transmission characteristics, resulting in information splitting, complex operations and resource waste. The prior art lacks solutions that enable deep understanding of information content and implementation of cross-protocol intelligent associations. Disclosure of Invention The technical problem to be solved by the invention is to provide the intelligent gateway for the heterogeneous maritime broadcasting protocol, which can realize intelligent understanding and automatic association of maritime information, solve the problem that the multi-system fusion depends on manual operation, and provide unified, efficient and enhanced information service. The intelligent processing layer comprises a natural language processing module, a content association engine, a multiprotocol converter and a guide instruction generator, wherein the natural language processing module is used for carrying out semantic analysis on input maritime safety information and extracting key elements comprising time, place and event types, the content association engine is used for establishing association relations among different format information based on the key elements, the multiprotocol converter is used for generating broadcast signals conforming to NAVDAT, fax and NAVTEX protocols according to the association relations, and the guide instruction generator is used for generating standardized guide instructions related to the cross-protocol. Further, the natural language processing module comprises an element recognition sub-module for recognizing geographic entities, time entities and event entities in the text by adopting a pre-trained language model, and a semantic understanding sub-module for establishing a logic relationship among key elements based on a deep learning model of an attention mechanism. The element recognition submodule adopts a pre-training model based on a transducer to perform optimized training on the professional terms and expression modes of the maritime field, wherein data preparation comprises the steps of constructing a professional corpus of the maritime field, a corpus source comprising an international maritime organization general function, historical NAVTEX message database, world meteorological organization weather coding data, navigation warning and maritime accident report, preprocessing, cleaning and denoising the materials, performing standardized processing on related expressions of coordinate formats and the professional terms, continuing pre-training, loading a general Chinese BERT model as a basis, using the maritime corpus to perform continuous pre-training on the model by taking a mask language model as a target, embedding the vocabulary, the syntax and the knowledge of the maritime field into parameters of the maritime corpus through predicting the masked words in the pre-training process, and performing task fine adjustment, and performing supervised fine adjustment of a named entity recognition task on the pre-trained model by using the maritime text data set marked with time, location and event type labels, and mapping an input sequence into a structured entity label sequence. Further, the content association engine calculates the comprehensive association Score of the multi-source information by adopting the following formula, wherein score=alpha×ss+beta×sst+gamma×St, ss is semantic similarity, sst is space-time overlapping degree, st is time-efficiency consistency, alpha, beta and gamma are configurable weight parameters and satisfy alpha+beta+gamma=1, the content association engine sets an association threshold value θassoc, and when 1 is more than or equal to Score and more than or equal to θassoc, the association relation among the multi-source information is automatically established. Further, the semantic similarity is obtained by calculating cosine similarity through combining a TF-IDF algorithm with a marine dictionary to generate text vectors, the space-time overlapping degree is used for measuring the overlapping degree of events in geography and time, sst= Sgeo × Stime, sgeo is the space overlapping degree, stime is the time overlapping degree, the aging consistency