CN117454902B

CN117454902B - Report analysis method applied to investigation processing system and marine environment investigation system

Info

Publication number: CN117454902B
Application number: CN202311418921.5A
Authority: CN
Inventors: 柯盛; 张际标; 张鹏
Original assignee: Guangdong Ocean University
Current assignee: Guangdong Ocean University
Priority date: 2023-10-30
Filing date: 2023-10-30
Publication date: 2024-04-05
Anticipated expiration: 2043-10-30
Also published as: CN117454902A

Abstract

The report analysis method applied to the investigation processing system and the marine environment investigation system provided by the embodiment of the invention can provide more comprehensive, accurate and easily understood information by optimizing the initial investigation report text to generate the investigation report optimization text, so that the problem that the traditional technology is difficult to realize high-quality investigation report optimization can be solved, thereby providing help for deep understanding of the target marine ecological detection event and providing guidance for corresponding decisions and actions.

Description

Report analysis method applied to investigation processing system and marine environment investigation system

Technical Field

The invention relates to the technical field of data analysis, in particular to a report analysis method applied to a survey processing system and a marine environment survey system.

Background

Marine environmental surveys are the process of systematically observing, data collecting, and analyzing the marine ecosystem and its surroundings. It covers the aspects of ocean water quality, physical environment, biological diversity, ecological process, influence of human activities on ocean environment, and the like. Marine environmental surveys typically include sampling and monitoring seawater, benthos, plankton, sediment, rock and other related parameters, and the like.

The reasons for conducting marine environmental surveys are as follows: (1) understanding the marine ecosystem: the marine ecosystem is one of the largest and most complex ecosystems on earth, contains a wide range of biodiversity and complex ecological processes, and can acquire data and information about the composition, structure, function and interaction of the marine ecosystem through marine environmental investigation; (2) monitoring environmental changes: marine environment is affected by natural changes and human activities, marine environment investigation provides a means to monitor and evaluate important parameters such as marine water quality, temperature, salinity, oxygen content, ph, etc., which are very important for understanding environmental change trend, coping with climate change, evaluating pollution level, and formulating protective measures; (3) protection of biodiversity: the marine ecological system is a habitat of a plurality of rare or endangered species, the knowledge and protection of the biodiversity in the habitat are of great importance, and specific ecological systems such as coral reefs, seaweed beds, marine protection areas and the like can be researched and monitored through marine environment investigation, the health condition of the specific ecological systems can be evaluated, and corresponding protection measures can be taken; (4) managing fishery resources: marine environmental surveys are critical to sustainable management and protection of fishery resources, which provide data on shoal distribution, quantity, volume, and population dynamics, and support the formulation of fishery management policies, fishing limits, and protective measures; (5) promoting marine pollution control: marine environmental surveys help monitor and evaluate the extent and source of marine pollution, such as chemicals, oil pollution, plastic waste, etc., which can help to formulate and implement pollution control measures to protect marine ecosystems and human health.

In a word, the marine environment investigation provides key data and information for solving a marine ecosystem, monitoring environmental changes, protecting biodiversity, managing resources and preventing and controlling pollution, and provides scientific basis for marine protection and sustainable utilization.

Disclosure of Invention

In order to improve the technical problems in the related art, the invention provides a report analysis method applied to a survey processing system and a marine environment survey system.

In a first aspect, an embodiment of the present invention provides a report analysis method applied to a survey processing system, and applied to a marine environment survey system, the method including:

obtaining an initial investigation report text containing a target marine ecology detection event, wherein the target marine ecology detection event comprises a plurality of marine ecology detection item descriptions;

mining first event text semantics of the target marine ecology detection event based on the initial survey report text, wherein the first event text semantics of the target marine ecology detection event comprise detection item output semantic variables described by each marine ecology detection item in the target marine ecology detection event;

obtaining second event text semantics of the target marine ecology detection event, wherein the second event text semantics of the target marine ecology detection event comprise detection item derived semantic variables described by each marine ecology detection item in the target marine ecology detection event;

According to the semantic fine granularity of the semantic variable output by the detection item described by each marine ecological detection item and the semantic fine granularity of the semantic variable derived by the detection item described by each marine ecological detection item, carrying out semantic aggregation operation on the first event text semantic of the target marine ecological detection event and the second event text semantic of the target marine ecological detection event to obtain the third event text semantic of the target marine ecological detection event;

generating a survey report optimization text of the target marine ecology detection event through third event text semantics of the target marine ecology detection event.

In some aspects, the semantic fine granularity of the semantic variable is output according to the detection item described by each marine ecological detection item, and the semantic fine granularity of the detection item derived semantic variable described by each marine ecological detection item, performing semantic aggregation operation on the first event text semantic of the target marine ecological detection event and the second event text semantic of the target marine ecological detection event, to obtain the third event text semantic of the target marine ecological detection event, including:

according to the semantic fine granularity of the semantic variable output by the detection item described by each marine ecological detection item and the semantic fine granularity of the semantic variable derived by the detection item described by each marine ecological detection item, respectively carrying out regularization aggregation operation and self-adaptive aggregation operation on the first event text semantic of the target marine ecological detection event and the second event text semantic of the target marine ecological detection event to obtain regularization aggregation semantic of the target marine ecological detection event and self-adaptive aggregation semantic of the target marine ecological detection event;

And splicing the regularized aggregation semantics of the target marine ecological detection event and the self-adaptive aggregation semantics of the target marine ecological detection event to obtain third event text semantics.

In some schemes, the semantic fine granularity of the output semantic variable of the detection item of each marine ecological detection item description has a set quantization relation with the content enrichment index of the corresponding marine ecological detection item description in the initial investigation report text; the step of obtaining regularized aggregated semantics of the target marine ecology detection event comprises:

determining the quantized value of each index distribution label in the regularized aggregation index of the first event text semantics and the quantized value of each index distribution label in the regularized aggregation index of the second event text semantics according to the semantic granularity of the semantic variable output by the detection items described by each marine ecological detection item and the semantic granularity of the semantic variable derived by the detection items described by each marine ecological detection item;

when the semantic granularity corresponding to the detection item output semantic variable of any marine ecological detection item description is larger than the semantic granularity of the detection item derivative semantic variable of any marine ecological detection item description, the quantization value of the corresponding index distribution label in the regularized aggregation index corresponding to the first event text semantic is larger than the quantization value of the corresponding index distribution label in the regularized aggregation semantic of the second event text semantic;

And carrying out regular aggregation on the detection item output semantic variables described by each marine ecological detection item and the detection item output semantic variables described by each marine ecological detection item by utilizing the quantized values of each index distribution tag in the regular aggregation index of the first event text semantic and the quantized values of each index distribution tag in the regular aggregation index of the second event text semantic, so as to obtain regular aggregation semantics.

In some aspects, the text mining of the first event text semantics of the target marine ecology detection event based on the initial survey report text comprises:

performing report semantic mining operation on the initial investigation report text, and performing noise attenuation processing on report semantics obtained by mining from the initial investigation report text to obtain a noise attenuation vector about a target marine ecology detection event in the initial investigation report text;

and carrying out semantic transformation on the noise attenuation vector of the target marine ecology detection event, and taking the report semantic obtained by the semantic transformation as the first event text semantic of the target marine ecology detection event.

In some aspects, performing noise attenuation processing on the initial survey report text to obtain a noise attenuation vector for a target marine ecology detection event in the initial survey report text, including:

Performing multi-level report semantic mining operation on the initial investigation report text, and performing multi-level feature compression on mining results obtained by the report semantic mining operation;

when the multi-level feature compression is implemented to the target feature recognition degree, obtaining a semantic relation network of the marine survey report under each feature recognition degree;

and sequentially carrying out feature expansion on the marine survey report semantic relation network under the target feature recognition degree by combining the marine survey report semantic relation network under the feature recognition degree, and taking the marine survey report semantic relation network with different layer orders obtained by the feature expansion as a noise attenuation vector of the target marine ecological detection event.

In some aspects, the noise attenuation vector of the target marine ecology detection event comprises a marine survey report semantic relationship net of different layers;

the semantic transformation is performed on the noise attenuation vector of the target marine ecology detection event, and the report semantic obtained by the semantic transformation is used as the first event text semantic of the target marine ecology detection event, and the method comprises the following steps:

performing feature compression on the ocean survey report semantic relation network of any layer order, and performing feature expansion on the ocean survey report semantic relation network after feature compression;

And taking the ocean survey report semantic relation network subjected to feature expansion again as a first event text semantic of the target ocean ecology detection event under any layer order.

In some aspects, the obtaining the second event text semantics of the target marine ecology detection event comprises:

obtaining potential text semantics generated in the process of mining the first event text semantics of the target marine ecology detection event; the potential text semantics are obtained by compressing the noise attenuation vector of the target marine ecology detection event in the process of mining the text semantics of the first event of the target marine ecology detection event;

generating second event text semantics of the target marine ecology detection event under different levels based on the potential text semantics; the hierarchy corresponding to the second event text semantics is consistent with the hierarchy corresponding to the first event text semantics.

In some aspects, the generating second event text semantics of the target marine ecology detection event at different levels based on the latent text semantics includes:

converting the latent text semantics from a first target feature coordinate system to a second target feature coordinate system in the current text unit to obtain a target text mode label;

Obtaining an original text mode label, and updating the original text mode label by using the target text mode label to obtain an updated text mode label;

and generating second event text semantics of the target marine ecology detection event under a target layer level according to the updated text mode label.

In some aspects, the updating the original text mode tag with the target text mode tag to obtain an updated text mode tag includes:

adjusting the original text mode label by utilizing the target text mode label to obtain an adjusted text mode label;

and performing mode checking on the adjusted text mode label, and performing label updating on the text mode label subjected to the mode checking to obtain an updated text mode label.

In some aspects, the generating, according to the updated text mode tag, a second event text semantic of the target marine ecology detection event at a target level includes:

obtaining a text description characterization vector of a text unit before the current text unit, and carrying out moving average processing on the text description characterization vector of the previous text unit by utilizing the updated text mode label;

Performing disturbance transformation on the moving average vector to obtain a text description characterization vector of the current text unit;

taking the text description characterization vector of the current text unit as a text description incoming vector of a next text unit of the current text unit, and obtaining an updated text pattern tag generated by cycling the latent text semantics in the next text unit;

and after the text description characterization vector of the current text unit is subjected to feature expansion in the latter text unit, carrying out moving average processing and disturbance transformation on the text description characterization vector subjected to feature expansion by utilizing the updated text mode label in the latter text unit, so as to obtain the text semantics of the second event of the target marine ecology detection event under the target layer level.

In some aspects, the first event text semantics of the target marine ecology detection event are generated by a survey report text analysis network completing the debugging; the step of obtaining a survey report text analysis network with complete commissioning includes:

obtaining a past investigation report text containing past marine ecology detection events, performing disturbance operation on the past investigation report text, and performing noise attenuation processing on the past investigation report disturbance text containing the past marine ecology detection disturbance events to obtain noise attenuation vectors of the past marine ecology detection disturbance events;

Generating a past noise attenuation text according to the noise attenuation vector of the past marine ecology detection disturbance event;

calculating network debugging cost according to semantic distinction between the past investigation report text and the past noise weakening text, and debugging the investigation report text analysis network by utilizing the calculated network debugging cost until obtaining the investigation report text analysis network which completes debugging.

In some aspects, the second event text semantics of the target marine ecology detection event are generated by a survey report text optimization network that completes the debugging; the step of obtaining a survey report text optimized network with complete commissioning includes:

obtaining a past investigation report text containing past marine ecology detection events, carrying out disturbance operation on the past investigation report text, and loading the past investigation report disturbance text containing the past marine ecology detection disturbance events into the investigation report text optimizing network to obtain a past investigation report optimizing text of the past marine ecology detection disturbance events;

performing context semantic mining on the past survey report optimizing text and the past survey report text respectively to obtain context semantic features of the past survey report optimizing text and context semantic features of the past survey report text;

Generating context debugging cost based on the difference between the context semantic features of the past survey report optimization text and the context semantic features of the past survey report text;

and determining a target debugging cost related to the investigation report text optimization network according to the context debugging cost, and debugging the investigation report text analysis network by utilizing the target debugging cost until obtaining the investigation report text analysis network with completed debugging.

In a second aspect, the invention also provides a marine environment investigation system, comprising a processor and a memory; the processor is in communication with the memory, and the processor is configured to read and execute a computer program from the memory to implement the method described above.

In a third aspect, the present invention also provides a computer-readable storage medium having stored thereon a program which, when executed by a processor, implements the method described above.

By applying the embodiment of the invention, the method for optimizing the initial survey report text to obtain the survey report optimizing text has the following beneficial effects:

(1) Providing comprehensive information: the survey report optimization text is based on semantic aggregation operation, and integrates the first and second event text semantics of the target marine ecology detection event and semantic variables output and derived by each detection item. Thus, more comprehensive and detailed information can be provided, and readers can obtain more deep understanding about the marine ecology condition;

(2) Accurately conveying key findings: through semantic aggregation operations, survey report optimization text can more accurately convey the results and observed features of each test item, revealing correlations and trends between different test items. This helps the reader to accurately understand and evaluate the marine ecology, making a decision with basis;

(3) Clear expression mode: the survey report optimization text is integrated and organized to be presented in a clear, structured manner that allows the reader to more easily read and understand. Related semantic details are refined and accurately revealed through semantic aggregation operation, so that the text is more readable and easier to read;

(4) Targeted suggestions and strategies: by comprehensive analysis and interpretation, survey report optimization text can generate conclusions and suggestions with more insight. These suggestions and strategies can provide guidance for the detected problems or risks and provide an effective direction for improving or protecting the marine environment.

In summary, by optimizing the initial survey report text, generating the survey report optimization text can provide more comprehensive, accurate and easy-to-understand information (which can improve the problem that the conventional technology is difficult to realize high-quality survey report optimization), help readers to know the target marine ecology detection event in depth, and make corresponding decisions and actions.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention.

Fig. 1 is a flow chart of a report analysis method applied to a survey processing system according to an embodiment of the present invention.

Detailed Description

Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with the invention. Rather, they are merely examples of apparatus and methods consistent with aspects of the invention.

It should be noted that the terms "first," "second," and the like in the description of the present invention and the above-described drawings are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order.

The method embodiments provided by the embodiments of the present invention may be implemented in a marine environment investigation system, a computer device or similar computing means. Taking the example of operation on a marine environmental survey system, the marine environmental survey system may include one or more processors (which may include, but are not limited to, a microprocessor MCU or a programmable logic device FPGA or the like processing device) and memory for storing data, and optionally the marine environmental survey system may also include transmission means for communication functions. It will be appreciated by those of ordinary skill in the art that the above-described structure is merely illustrative and is not intended to limit the structure of the marine environmental survey system described above. For example, the marine environmental survey system may also include more or fewer components than those shown above, or have a different configuration than those shown above.

The memory may be used to store a computer program, for example, a software program of application software and a module, such as a computer program corresponding to a report analysis method applied to a survey processing system in an embodiment of the present invention, and the processor executes the computer program stored in the memory to perform various functional applications and data processing, that is, to implement the above-mentioned method. The memory may include high speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid state memory. In some examples, the memory may further include memory remotely located with respect to the processor, the remote memory being connectable to the marine environmental survey system through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The transmission means is used for receiving or transmitting data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider of the marine environmental survey system. In one example, the transmission means comprises a network adapter (Network Interface Controller, simply referred to as NIC) that can be connected to other network devices via a base station to communicate with the internet. In one example, the transmission device may be a Radio Frequency (RF) module, which is used to communicate with the internet wirelessly.

Referring to fig. 1, fig. 1 is a flowchart of a report analysis method applied to a survey processing system according to an embodiment of the present invention, where the method is applied to a marine environment survey system, and further includes steps 101-105.

Step 101, obtaining an initial investigation report text containing a target marine ecology detection event, wherein the target marine ecology detection event comprises a plurality of marine ecology detection item descriptions.

In the embodiment of the invention, the target marine ecology detection event can be a marine benthos diversity evaluation event, a marine fish community health monitoring event, a marine sediment pollution evaluation event, a marine phytoplankton productivity monitoring event, a coral reef whitening event monitoring event and the like, and the following is an introduction of the above-mentioned at least five types of target marine ecology detection events,

1) Marine benthos diversity assessment event: investigation is conducted on a marine region to assess benthic biodiversity in that region and record information on the distribution, density and interactions of different species.

2) Marine fish community health monitoring events: the fish communities in the ocean are monitored by methods such as diving, fish investigation or remote sensing technology, and the conditions of the species composition, the quantity change, the life cycle, the ecological niche and the like are known.

3) Marine sediment pollution assessment event: marine sediment samples were collected and analyzed to assess the pollution level in the area. This may involve detection and concentration measurement of organic contaminants, heavy metals, micro-plastics and the like.

4) Marine phytoplankton productivity monitoring events: phytoplankton in the ocean, including algae and phytoplankton biomass, photosynthesis efficiency, seasonal variation, etc., are monitored using remote sensing technology or on-site sampling to assess the health status of the marine ecosystem.

5) Coral reef whitening event monitoring event: regular investigation is performed in the coral reef area to monitor the occurrence and degree of coral whitening, possibly to collect a coral sample, record the degree of whitening, and compare with previous data to evaluate the pressure and adaptability to the coral reef.

Taking a coral reef whitening event monitoring event as an example, the following is a preliminary investigation report text including the coral reef whitening event monitoring event.

Title: initial investigation report for monitoring coral reef whitening event

Date: MM month DD day of YYYY year

Summary:

the report aims to provide initial investigation results of a target marine ecology detection event (coral reef whitening event monitoring). The coral reef health condition of a specific area is evaluated through monitoring the coral reef whitening phenomenon of the specific area, and basic data are provided for protecting and managing the coral reef ecosystem.

Investigation scope and method:

the present investigation selected a coral reef area located near the XX island as a study area. Using the dive survey method, coral samples were collected at different depths and locations and the color, texture, whitening level and correlation to the surrounding environment of the coral were recorded.

Investigation result:

coral whitening phenomenon: part of the coral whitening was observed during the investigation. Based on observations, about 20% of corals showed signs of whitening, manifested as whitening or loss of color of the coral tissue.

Degree and distribution of whitening: the extent of albino coral varies from slight to moderate and is mainly concentrated in areas with shallow water depths. Uneven distribution of coral whitening was observed, with some coral communities being more severely affected.

Environmental parameters: during the investigation, environmental parameters such as water temperature, salinity and illumination were also measured. Preliminary analysis showed that the water temperature was slightly higher than the historical average water temperature, but still within acceptable limits.

Discussion and advice:

whitening cause: whitening may be associated with a recent increase in ocean temperature. Further research requires investigation of other environmental factors such as nutrient content, marine acidification, etc. to fully understand the cause of the whitening event.

Coral reef protection measures: based on preliminary observations, it is suggested to enhance coral reef protection and management measures, including measures to limit excessive diving, reduce pollutant emissions, and improve public awareness.

Conclusion:

the initial investigation provides valuable preliminary data for the target marine ecology detection event (coral reef whitening event monitoring). Further research and monitoring will help to gain insight into the mechanism of occurrence of coral whitening events and provide a more effective protection strategy to maintain the health and sustainability of the marine ecosystem.

Signature:

XXX (research team responsible)

On the basis of the above-mentioned preliminary survey report text, the following are some possible marine ecology detection item descriptions: coral species and abundance: according to the investigation result, the distribution and the quantity of different coral species are described in detail, including main hard coral and soft coral; whitening degree evaluation: quantitatively evaluating the degree of whitened corals, wherein scales or indexes can be used for representing the whitening degrees of different corals, and comparing the whitening degrees with the previous investigation data; coral physical health assessment: evaluating the health condition of coral physique by observing and recording the parameters of the texture, swelling degree, damage degree and the like of coral; and (3) monitoring coral fish shoal falling: investigation and recording of the species and quantity of fish symbiotic to the coral reef to understand the integrity and complexity of the coral reef ecosystem; and (3) monitoring water quality parameters: measuring and recording key water quality parameters such as water temperature, salinity, dissolved oxygen, illumination intensity and the like to reveal the influence of the factors on the health of the coral reef; coral biodiversity assessment: biological sampling and species identification are carried out on a coral reef area, the coral biodiversity of the area is estimated, and the species composition and abundance are analyzed; coral disease monitoring: whether coral is affected by the disease is observed and recorded, including the type, incidence and spread of the disease, etc. These marine ecology detection item descriptions can be further refined and manipulated for specific application in practical investigation.

Step 102, mining the first event text semantics of the target marine ecology detection event based on the initial survey report text, wherein the first event text semantics of the target marine ecology detection event comprises the detection item output semantic variables described by each marine ecology detection item in the target marine ecology detection event.

In the embodiment of the present invention, step 102 is based on the initial survey report text, and the semantic variable is output by mining the first event text semantic of the target marine ecology detection event, where the first event text semantic includes the detection item description of each marine ecology detection item.

In other words, in the first event text semantics, the related semantic information is extracted for each marine ecology detection item description through semantic parsing or processing technology to represent the result or output of the detection item. These semantic variables may be values, states, indices, proportions, etc. for describing and quantifying the observed characteristics or parameters of the marine ecology detection term.

For example, in coral benthos diversity assessment this target marine ecological detection event, semantic variables like the following may be extracted: benthos diversity index: the degree of diversity of benthic communities in the region can be represented by numerical values; main species abundance: describing the number or relative abundance of the most common or important benthic species; species distribution pattern: representing the spatial distribution pattern of different benthic species in the region, such as aggregative, discrete, etc. By mining and extracting the text semantics of the first event, the output semantic variables of the detection items described by each marine ecological detection item can be obtained, and a foundation is provided for further semantic aggregation operation and investigation report optimization.

Step 103, obtaining second event text semantics of the target marine ecology detection event, wherein the second event text semantics of the target marine ecology detection event comprise detection item derived semantic variables described by each marine ecology detection item in the target marine ecology detection event.

In the embodiment of the present invention, step 103 is to obtain the second event text semantics of the target marine ecology detection event, where the detection term derived semantic variables described by each marine ecology detection term are included. In this step, for each marine ecology detection item description, relevant semantic information is further mined and extracted to obtain further derived semantic variables about the detection item. These derived semantic variables can provide a deeper understanding and fine-grained description, further enriching the semantic information of the target marine ecology detection event.

For example, in the second event text semantics of coral benthos diversity assessment, the following test item derived semantic variables may be extracted: species abundance profile: shows the trend of the relative abundance of different benthic species with depth or distance; species symbiotic relationship: describing the existence and degree of interaction relations between benthonic organisms, such as symbiosis, predation and the like; species richness index: and (5) calculating the comprehensive index reflecting the quality of the coral reef habitat by combining the types and the quantity of benthos. By mining and extracting the text semantics of the second event, the detection item derived semantic variables described by each marine ecological detection item can be obtained, and the understanding of the target marine ecological detection event is further enriched. These semantic variables will play an important role in the subsequent semantic aggregation operations and survey report optimization.

Step 104, according to the semantic granularity of the semantic variable output by the detection item described by each marine ecological detection item and the semantic granularity of the semantic variable derived by the detection item described by each marine ecological detection item, performing semantic aggregation operation on the first event text semantic of the target marine ecological detection event and the second event text semantic of the target marine ecological detection event to obtain the third event text semantic of the target marine ecological detection event.

In the embodiment of the invention, step 104 is to perform semantic aggregation operation on the first event text semantic and the second event text semantic of the target marine ecology detection event based on the semantic fine granularity of the detection item output semantic variable and the detection item derived semantic variable described by each marine ecology detection item, so as to obtain the third event text semantic of the target marine ecology detection event.

In the step, semantic information of different detection items is integrated by combining the first event text semantic and the second event text semantic, so that more comprehensive, accurate and detailed description is obtained. This process may include the following operations: merging similar semantic variables: the same or similar semantic variables are integrated, and repeated or redundant information is eliminated, so that more concise and accurate representation is achieved; establishing a semantic relation: according to the association or the dependency relationship between the detection items, semantic connection is established, so that the third event text semantics can accurately reflect the relationship between different detection items. For example, correlating benthic diversity index with species abundance profile to illustrate how it affects species distribution patterns; introduction of semantic enhancement: and introducing proper modifier words or descriptive words according to the output of the detection items and derived semantic variables so as to more accurately describe the target marine ecology detection event. For example, in describing benthos diversity, word vocabulary such as richness, diversity index, and category composition, etc. may be used to enhance semantic expression. Through the semantic aggregation operations, the obtained third event text semantics can more comprehensively and accurately present the characteristics and the results of the target marine ecology detection event. This provides a basis for subsequent survey report optimization and helps the reader to better understand and assess marine ecology.

And 105, generating a survey report optimization text of the target marine ecology detection event through third event text semantics of the target marine ecology detection event.

In an embodiment of the present invention, step 105 is to generate survey report optimized text by third event text semantics of the target marine ecology detection event.

In this step, the third event text semantics of the obtained target marine ecology detection event are utilized to integrate and organize the same so as to generate an optimized investigation report text. The survey report will provide an exhaustive description and analysis of the targeted marine ecology detection event based on more comprehensive, accurate and detailed semantic information.

When generating the survey report optimization text, the following may be included: overview and overview: providing description of the overall situation and the outline of the target marine ecology detection event, including important information such as a research area, a detection method, a time range and the like; examination item results overview: summarizing the output and derived semantic variables of each marine ecological detection item, presenting main observation results and trends in a concise and brief manner, and giving out overall evaluation of marine ecological conditions; deep analysis and interpretation: and carrying out detailed analysis and explanation on semantic fine granularity results of each marine ecological detection item, including description of observed characteristics, trends, relevance, abnormal conditions and the like. These analyses may provide a deeper understanding and insight based upon semantic information mined and aggregated in the earlier steps; conclusion and advice: based on the comprehensive analysis of the target marine ecology detection event, conclusive statements are given, including assessment of marine ecology conditions, identification of problems or risks, and advice to improve or protect the marine ecology environment. The survey report optimization text generated by step 105 will provide a clear, accurate, and easy to understand presentation that will enable the reader to better understand the key findings and conclusions of the targeted marine ecology detection event.

It can be seen that, by applying the steps 101-105, optimizing the initial survey report text to obtain the survey report optimizing text has the following advantages:

In some possible embodiments, in step 104, the outputting, according to the detection items described by the marine ecological detection items, the semantic fine granularity of the semantic variable and the semantic fine granularity of the detection item derived semantic variable described by the marine ecological detection items, performing semantic aggregation operation on the first event text semantic of the target marine ecological detection event and the second event text semantic of the target marine ecological detection event to obtain the third event text semantic of the target marine ecological detection event may include step 1041 and step 1042.

Step 1041, performing regularization aggregation operation and self-adaptive aggregation operation on the first event text semantics of the target marine ecological detection event and the second event text semantics of the target marine ecological detection event respectively according to the semantic fine granularity of the detection item output semantic variable described by each marine ecological detection item and the semantic fine granularity of the detection item derivative semantic variable described by each marine ecological detection item, so as to obtain the regularization aggregation semantics of the target marine ecological detection event and the self-adaptive aggregation semantics of the target marine ecological detection event.

Step 1042, splicing the regularized aggregation semantics of the target marine ecology detection event and the self-adaptive aggregation semantics of the target marine ecology detection event to obtain a third event text semantics.

In the possible embodiments described above, steps 1041 and 1042 provide further operations regarding the text semantics of the target marine ecology detection event.

Step 1041 firstly uses the semantic fine granularity of the detection item output semantic variable and the detection item derived semantic variable described by each marine ecological detection item to respectively perform regularized aggregation operation and adaptive aggregation operation on the first event text semantic and the second event text semantic of the target marine ecological detection event. Regularized aggregation operations refer to the explicit regularized combination of multiple semantic variables, meaning that semantic variables are weighted, fused or combined using a pre-determined rule or method to generate regularized aggregated semantics. This aggregation may be implemented based on expert knowledge or predefined algorithms. The adaptive aggregation operation is to aggregate semantic variables implicitly, and automatically adjust and optimize the combination mode of the semantic variables by learning and adapting to the characteristics of the data. Thus, the aggregation strategy can be flexibly changed according to specific situations so as to better express the semantics of the target marine ecology detection event.

Step 1042, splicing the regularized aggregation semantics and the adaptive aggregation semantics to obtain third event text semantics. This means that the semantic results of the two different aggregation manners obtained in step 1041 are integrated to generate a third event text semantic comprehensively reflecting the target marine ecology detection event.

It can be seen that step 1041 and step 1042 process semantic variables through explicit aggregation and implicit aggregation operations, thereby providing richer, accurate and comprehensive text semantic information of the target marine ecology detection event.

In some examples, the semantic fine granularity of the test item output semantic variable of each marine ecology test item description has a set quantitative relationship (such as a positive correlation relationship) with the content enrichment index of the corresponding marine ecology test item description in the initial survey report text. Based on this, the step of obtaining the regularized aggregated semantics of the target marine ecology detection event in step 1041 includes steps 10411-10413.

Step 10411, determining a quantization value of each index distribution label in the regularized aggregation index of the first event text semantics and a quantization value of each index distribution label in the regularized aggregation index of the second event text semantics according to the semantic granularity of the semantic variable output by the detection item described by each marine ecological detection item and the semantic granularity of the semantic variable derived by the detection item described by each marine ecological detection item; wherein, the index distribution label can be understood as index parameter bits.

Step 10412, when the semantic granularity corresponding to the detection item output semantic variable of any marine ecological detection item description is greater than the semantic granularity of the detection item derived semantic variable of any marine ecological detection item description, the quantization value of the corresponding index distribution tag in the regularized aggregation index corresponding to the first event text semantic is greater than the quantization value of the corresponding index distribution tag in the regularized aggregation semantic of the second event text semantic.

Step 10413, performing regularization aggregation on the detection item output semantic variables described by the marine ecological detection items and the detection item output semantic variables described by the marine ecological detection items by using the quantized values of the index distribution labels in the regularized aggregation index of the first event text semantic and the quantized values of the index distribution labels in the regularized aggregation index of the second event text semantic to obtain regularized aggregation semantics.

In the above-described embodiment, the specific operation of obtaining the regularized aggregation semantics of the target marine ecology detection event in step 1041 is described in detail in steps 10411 to 10413.

Step 10411 first sets a quantitative relationship between the semantic fine granularity of the semantic variable according to the test item output semantic variable of each marine ecological test item description and the content enrichment index of the corresponding marine ecological test item description in the initial survey report text. This means that the higher the semantic fine granularity of the semantic variable, the higher the quantization value of the corresponding index distribution label. Similarly, when the detected item derived semantic variable is involved, the semantic granularity is lower, and the quantization value of the corresponding index distribution label is also lower.

Step 10412 compares the quantized value of each index distribution label in the regularized aggregation index corresponding to the first event text semantic with the quantized value of the corresponding index distribution label in the regularized aggregation semantic of the second event text semantic. If the semantic granularity of the detected item output semantic variable of any marine ecological detected item is larger than the semantic granularity of the corresponding detected item derived semantic variable, the quantized value of the corresponding index distribution label in the regularized aggregation index corresponding to the first event text semantic is larger than the quantized value of the corresponding index distribution label in the regularized aggregation semantic of the second event text semantic.

Step 10413 uses the quantized value of each index distribution label in the regularized aggregation index of the first event text semantic and the quantized value of each index distribution label in the regularized aggregation index of the second event text semantic to perform regularized aggregation operation on the detection item output semantic variable and the detection item derived semantic variable described by each marine ecological detection item. This means that the semantic variables are weighted, fused or combined according to the quantized values to get a regularized aggregated semantic.

It can be seen that, in step 10411, the quantized values of the index distribution labels are determined by setting the quantized relationships, in step 10412, the quantized value magnitude relationships of the two events in the regularized aggregation index are compared, and in step 10413, the regularized aggregation operation is performed on the semantic variables by using the quantized values, so as to obtain the regularized aggregation semantics of the target marine ecological detection event. In the whole process, the regularized aggregation operation is explicit, and weighting and fusion are performed based on the set quantization relation and the quantization value of the index distribution label.

In some design considerations, the operations of steps 10411 to 10413 may be described in connection with the monitoring of a coral reef whitening event. The coral reef whitening is a phenomenon that coral algae symbionts are disturbed due to environmental changes and the like, and coral tissues lose pigment and appear white. In this example, two events will be considered: coral physiological health status and ambient seawater temperature.

In step 10411, a quantitative relationship between the semantic fine granularity of the test item output semantic variable of each test item description and the content enrichment index in the initial survey report text is determined. It is assumed that the output semantic variable of the detection item of the coral physiological health state has a high semantic fine granularity and the content enrichment index thereof is also high, so that the quantized value thereof is set to 5. The output semantic variable of the detection item of the ambient seawater temperature has lower semantic granularity, the content enrichment index is lower, and the quantized value is set to be 2.

Next, in step 10412, the normalized aggregate index quantification value (5) of the coral physiological health state and the normalized aggregate index quantification value (2) of the ambient sea water temperature are compared. Because the coral physiological health state has higher semantic fine granularity and related content richness, the quantized value in the regularized aggregation index corresponding to the first event text semantic is larger than that in the regularized aggregation semantic of the second event text semantic.

In step 10413, a regularization aggregation operation is performed on the output semantic variable of the detection item of the coral physiological health state and the output semantic variable of the detection item of the surrounding sea water temperature by using the regularization aggregation index quantized value (5) of the first event text semantic and the regularization aggregation index quantized value (2) of the second event text semantic. Because of the higher importance of coral physiological health, it will take up more weight in the final regularized aggregated semantics.

By way of this example, it is demonstrated how the regularized aggregation operations and quantized values are used to determine the weights and fusion of different semantic variables in a target marine ecology detection event according to a set quantization relationship. Therefore, the importance of the physiological health state of the coral on the event can be reflected more accurately by the regularized aggregation semantics in the coral reef whitening event monitoring scene.

In some possible embodiments, the mining of the first event text semantics of the target marine ecology detection event based on the initial survey report text in step 102 may include step 1021 and step 1022.

And 1021, carrying out report semantic mining operation on the initial investigation report text, and carrying out noise attenuation processing on report semantics obtained by mining from the initial investigation report text to obtain a noise attenuation vector about a target marine ecology detection event in the initial investigation report text.

And 1022, carrying out semantic transformation on the noise attenuation vector of the target marine ecology detection event, and taking the report semantic obtained by the semantic transformation as the first event text semantic of the target marine ecology detection event.

In step 1021, a report semantic mining operation is performed on the initial survey report text. This means that key information and semantics related to the target marine ecology detection event are identified and extracted through text analysis and natural language processing techniques. Such critical information may include marine organism type, environmental parameters, event descriptions, etc. Then, noise attenuation processing is performed on the report semantics mined from the initial survey report text. Noise attenuation refers to removing or reducing interference, uncorrelated or low quality information that may be present to improve semantic accuracy and reliability. Through these two steps, a noise attenuation vector is obtained for the target marine ecology detection event in the initial survey report text.

Next, in step 1022, the noise attenuation vector of the target marine ecology detection event is semantically transformed. Semantic transformation refers to transforming noise attenuation vectors into report semantics that have more semantic meaning and expressive power. This may be accomplished using semantic models, word embedding techniques, or other text representation methods. Through semantic transformation, the noise attenuation vector is converted into a more accurate and rich semantic expression form to serve as the text semantic of the first event of the target marine ecology detection event.

In summary, steps 1021 and 1022 are key sub-steps for mining the first event text semantics of the target marine ecology detection event in step 102. They involve mining of report semantics, noise attenuation processing, and semantic transformation operations aimed at extracting accurate and expressive semantic information about target events from the initial survey report text.

Further, steps 1021 and 1022 may be described in detail in connection with the monitoring of a coral reef whitening event.

For step 1021, the report semantic mining and noise reduction process, there is an initial survey report text in which observations and data about the event are recorded during coral reef whitening event monitoring. In step 1021, a report semantic mining operation is first performed to identify key information related to the target event. This may include coral species, ocean temperature, water quality parameters, etc.

Assume that the following key information is extracted from the initial survey report text: coral species: brain coral, scallop coral, and deer coral; ocean temperature: rising to 30 ℃; water quality parameters: high nitrogen content and low salinity.

Next, noise attenuation processing will be performed on report semantics mined from the initial survey report text. This means that there is a need to remove or reduce interference, irrelevant or low quality information that may be present to improve semantic accuracy and reliability.

It is assumed that after the noise attenuation process, the following noise attenuation vectors are obtained: coral species: brain coral, scallop coral; ocean temperature: rising to 30 ℃; water quality parameters: high nitrogen content.

Key information related to the coral reef whitening event is mined from the initial survey report text, and noise attenuation processing is performed to reduce unnecessary or low quality information, via step 1021.

In step 1022, the noise attenuation vector of the target marine ecology detection event is subjected to semantic transformation so as to have more expressive power and semantic meaning.

It is assumed that a pre-trained semantic model is used for the semantic transformation operation. Inputting the noise attenuation vector into the model to obtain the following report semantics after semantic transformation: coral species: brain coral, scallop coral; ocean temperature: raising to 30 ℃; water quality parameters: there is a higher nitrogen content.

The noise attenuation vector is semantically transformed using the semantic model, converting it to more accurate and expressive report semantics, via step 1022. In this example, the coral species, ocean temperature and water quality parameters are semantically transformed and more abundant and accurate information is obtained.

In summary, through steps 1021 and 1022, relevant report semantics are extracted from the initial survey report text of coral reef whitening event monitoring, and noise attenuation processing and semantic transformation are performed to obtain more accurate, reliable and expressive first event text semantics.

In some possible embodiments, the noise attenuation processing performed on the initial survey report text in step 1021 to obtain a noise attenuation vector for the target marine ecology detection event in the initial survey report text may include steps 10211-10213.

Step 10211, performing multi-level report semantic mining operation on the initial investigation report text, and performing multi-level feature compression on mining results obtained by the report semantic mining operation;

step 10212, when the multi-level feature compression is implemented to the target feature recognition degree, obtaining a semantic relation network of the marine survey report under each feature recognition degree.

Step 10213, combining the semantic relation network of the marine survey report under each feature recognition degree to sequentially perform feature expansion on the semantic relation network of the marine survey report under the target feature recognition degree, and taking the semantic relation network of the marine survey report with different layer orders obtained by the feature expansion as a noise attenuation vector of the target marine ecological detection event.

Steps 10211 to 10213 are sub-steps of noise attenuation processing of the initial survey report text in step 1021 to obtain noise attenuation vectors for the target marine ecology detection event.

In step 10211, a multi-level report semantic mining operation is performed on the initial survey report text. This means that semantic information related to the target event is gradually extracted from low-level text features to high-level semantic concepts through a plurality of levels of text analysis and natural language processing techniques. And then, performing multi-level feature compression on the mining result obtained by the report semantic mining operation. This means that features extracted from different levels of semantic information are compressed and combined to reduce the dimensions of the features and preserve the most representative information. This can improve the conciseness and expressive power of the noise attenuation vector.

In step 10212, a semantic relationship network of the marine survey report under each feature recognition degree is obtained according to the requirement of implementing multi-level feature compression to the target feature recognition degree. This means that a semantic relationship network is constructed from the compressed features, where nodes represent features and edges represent semantic relationships between them. Through the setting of a plurality of feature recognition degrees, the complexity and the information density of the generated semantic relation network can be controlled so as to adapt to different analysis requirements and accuracy requirements.

In step 10213, the feature expansion operation is sequentially performed in combination with the semantic relation network of the marine survey report under each feature recognition degree. This means that the features are expanded layer by layer according to the semantic relation network generated previously, and higher-level semantic information and association relations are acquired. And then, taking the sea investigation report semantic relation network of different layers obtained by feature expansion as a noise attenuation vector of the target marine ecological detection event.

In summary, through steps 10211 to 10213, the initial survey report text is subjected to multi-level report semantic mining, feature compression, semantic relation net generation, and feature expansion operations, thereby generating noise attenuation vectors for the target marine ecology detection event. These steps enable the removal of noise while mining semantic information and the acquisition of richer and more accurate feature expression forms using semantic relationship networks.

Steps 10211 to 10213 may be described in detail below in connection with a monitoring scenario of a coral reef whitening event.

For step 10211, there is an initial survey report text containing observations and data regarding coral reef whitening in the coral reef whitening event monitor. In step 10211, a multi-level report semantic mining operation is performed, and multi-level feature compression is performed on the mining result.

Assume the following actions are taken:

first level report semantic mining: extracting basic information such as coral types, ocean temperatures, observation time and the like from an initial survey report text;

second level report semantic mining: further mining relevant environmental parameters such as illumination intensity, water flow speed, salinity and the like according to the basic information;

third level report semantic mining: based on the first two-stage mining results, higher-level semantic analysis is performed, and causal relationships, influence factors and the like related to the coral whitening event are extracted.

And then, performing multi-level feature compression on the mining result obtained by the report semantic mining operation. This means that features extracted from different levels of semantic information are compressed and combined to reduce the dimensions of the features and preserve the most representative information.

It is assumed that after feature compression, the following features are obtained:

feature 1: coral species (brain coral, scallop coral);

feature 2: the ocean temperature rises;

feature 3: the water flow speed is increased;

feature 4: the illumination intensity decreases.

Through step 10211, multi-level report semantic mining is performed from the initial survey report text, and multi-level feature compression is utilized to obtain the features as part of the noise attenuation vector.

In step 10212, a semantic relationship network of marine survey reports under each feature recognition level is generated according to the requirement of implementing multi-level feature compression to the target feature recognition level.

Assume that two feature recognition levels are set: basic feature recognition degree and advanced feature recognition degree.

For basic feature recognition, a simplified semantic relationship network is constructed in which nodes represent features and edges represent semantic relationships between them. For example, there may be the following relationship:

feature 1 has causal relationship with feature 2: coral species are associated with an increase in ocean temperature;

feature 3 has an inverse relationship with feature 4: the increase in water flow velocity results in a decrease in light intensity.

For advanced feature recognition, a more complex semantic relationship network is further constructed, including relationships between more features, environmental impact, and the like. For example, the following relationship may be added: there is an interaction relationship between feature 1 and feature 3: the particular coral species is more sensitive to changes in water flow rate.

Through step 10212, a more complex semantic relationship network can be further generated according to the semantic relationship network under the feature recognition degree after feature compression, so as to obtain more abundant information.

At the basic feature recognition level, a causal relationship between coral species and the rise in ocean temperature may be observed. It can be predicted that some corals are susceptible to whitening in high temperature environments by this relationship.

More semantic relationships can be added under advanced feature recognition, such as correlation between coral whitening and increased water flow velocity, reduced light intensity, and sensitivity of particular coral species to water flow velocity changes.

In step 10213, feature expansion is performed according to the semantic relation network under different feature recognition degrees, and a noise attenuation vector is constructed.

For the semantic relation network under the basic feature recognition degree, the features of the coral whitening event can be further expanded according to the existing causal relation and interaction relation. For example, characteristics such as marine acidification and water quality change associated with coral whitening may be added.

For the semantic relation network under the advanced feature recognition degree, the features of the coral whitening event can be further expanded, including other environmental factors, artificial interference and the like. For example, characteristics such as fishery activities and coastal development related to coral whitening may be added.

Through step 10213, feature expansion is performed on the semantic relation network of the marine survey report under the target feature recognition degree by using the semantic relation network under each feature recognition degree, and a noise attenuation vector is generated. These noise attenuation vectors will provide more comprehensive, accurate information for monitoring and analyzing coral reef whitening events.

In some examples, the noise attenuation vector of the target marine ecology detection event includes a network of marine survey report semantics of different layers. Based on this, the semantic transformation is performed on the noise attenuation vector of the target marine ecology detection event in step 1022, and the report semantic obtained by the semantic transformation is used as the first event text semantic of the target marine ecology detection event, which includes step 10221 and step 10222.

Step 10221, performing feature compression on the ocean survey report semantic relation network of any layer, and performing feature expansion on the ocean survey report semantic relation network after feature compression.

Step 10222, using the ocean survey report semantic relation network subjected to feature expansion again as a first event text semantic of the target ocean ecology detection event under any layer order.

In step 10221, a feature compression operation is performed on the marine survey report semantic relationship network for each level. This means that features extracted from the semantic relationship network are compressed and combined using appropriate techniques and methods to reduce the dimensions of the features and preserve the most representative information. After the characteristics are compressed, a compressed sea investigation report semantic relation network is obtained, wherein nodes represent the characteristics, and edges represent semantic relations among the characteristics. And then, carrying out feature expansion again on the semantic relation network after the feature compression.

In step 10222, the post-feature-expansion marine survey report semantic relationship net is used as the first event text semantic for the target marine ecology detection event at any level. After the feature expansion, richer semantic information and association relations are obtained. Such semantic information may be used to describe the first event text semantics of the target marine ecology detection event, including various features, causal relationships, interactions, etc. of the event.

Through step 10221 and step 10222, the feature compression and the feature expansion operation are performed again on the semantic relation network of the marine survey report of each level, so that more concise and accurate semantic information is extracted and used as the first event text semantic of the target marine ecology detection event. This helps to further analyze and understand the target event, providing a more comprehensive basis for subsequent processing and decision making.

In some examples, steps 10221 and 10222 may be presented in connection with a monitoring scenario of a coral reef whitening event.

In step 10221, the sea survey report semantic relation network is subjected to feature compression and feature expansion again.

In coral reef whitening event monitoring, a multi-level marine survey report semantic relationship network has been generated. Feature compression will now be performed for the semantic relationship network of one of the levels and then feature expansion will be performed again.

Assume as an example that a semantic relationship net under advanced feature recognition is selected.

Step 10221a: feature compression

In the characteristic compression stage, the characteristic compression is carried out on the semantic relation network of the marine survey report under the advanced characteristic recognition degree by using proper technology and method. This includes reducing the dimensions of the features while retaining the most representative information.

Examples:

the following features are extracted from the semantic relation network of the advanced feature recognition degree:

feature 1: coral species (brain coral, scallop coral);

feature 2: the ocean temperature rises;

feature 3: the water flow speed is increased;

feature 4: the illumination intensity is reduced;

feature 5: and (5) acidizing the sea.

These features are then compressed by a specific algorithm or method to reduce the number of features and extract the most important, representative information.

Step 10221b: again feature extension

And in the secondary feature expansion stage, carrying out feature expansion on the semantic relation network after feature compression. This means that more features and semantic relationships will be added to enrich the semantic information.

Features can be further expanded based on the feature-compressed semantic relationship net, for example:

feature 6: degree of coral whitening (mild, moderate, severe);

Feature 7: covering ratio of coral;

feature 8: a change in water quality;

feature 9: jamming (e.g., travel activity, fishing).

At the same time, more semantic relations, such as causal relation between coral whitening degree and ocean temperature rise, and interaction relation between coral coverage and illumination intensity reduction, are added.

Through step 10221, feature compression and secondary feature expansion are carried out on the semantic relation network of the marine survey report under the advanced feature recognition degree, and more concise and accurate semantic information is extracted from the semantic relation network.

In step 10222, the post-feature-expansion marine survey report semantic relationship net is used as the first event text semantic of the target marine ecology detection event under the advanced feature recognition degree. This means that the various features, causal relationships and interactions of the event are described using a semantic relationship network after feature expansion. For example, in coral whitening event monitoring, this semantic relationship network may be used to describe the following information:

causal relationship between coral whitening degree and ocean temperature rise: when the ocean temperature rises, different coral species are prone to whitening to different degrees;

interaction relationship between coral coverage and water quality change: the water quality change can affect the coral coverage rate, which in turn affects the water quality;

Correlation of human interference with coral whitening: traveling activities and fishing, etc. can increase the risk of coral whitening.

By using the sea survey report semantic relation network after the secondary feature expansion, the first event text semantic of the coral reef whitening event can be more comprehensively described, and the relation and influence among various features are revealed.

In some possible embodiments, the obtaining of the second event text semantics of the target marine ecology detection event in step 103 includes step 1031 and step 1032.

Step 1031, obtaining potential text semantics generated in the process of mining the text semantics of the first event of the target marine ecology detection event; the potential text semantics are obtained by compressing the noise attenuation vector of the target marine ecology detection event in the process of mining the text semantics of the first event of the target marine ecology detection event.

In a possible embodiment, to obtain the second event text semantics of the target marine ecology detection event, it is first necessary to obtain the potential text semantics generated in the process of mining the first event text semantics. These latent text semantics are obtained by feature-compressing the noise attenuation vector of the target marine ecology detection event. Specifically, when the first event text semantics of the target event are analyzed, a feature compression method is adopted to reduce the dimension of the noise attenuation vector, and the most important and representative information is extracted. The vector of feature compression thus obtained is referred to as latent text semantics.

Step 1032, generating second event text semantics of the target marine ecology detection event under different levels based on the latent text semantics; the hierarchy corresponding to the second event text semantics is consistent with the hierarchy corresponding to the first event text semantics.

In step 1032, second event text semantics of the target marine ecology detection event at different levels are generated based on the obtained latent text semantics. Wherein, the corresponding layer level of each second event text semantic is consistent with the corresponding layer level of one first event text semantic.

In other words, a set of second event text semantic hierarchies corresponding to the first event text semantic hierarchies are generated from the latent text semantics. The purpose of this is to reveal its semantic information from multiple perspectives in order to more fully describe the target event at different levels.

Through step 1031 and step 1032, second event text semantics of the target marine ecology detection event are generated using the feature compressed latent text semantics. The method can provide further semantic description on the target event, so that each layer of the event can be more comprehensively and accurately understood, and a richer information basis is provided for deep analysis and decision-making.

Steps 1031 and 1032 may be described in connection with a coral reef whitening event monitoring scenario.

Step 1031 is a process of obtaining a latent text semantic, in which a first event text semantic of a target event has been obtained in coral reef whitening event monitoring, and key information such as coral whitening degree, ocean temperature rise, water quality change, etc. is described. Step 1031 will now be performed to obtain latent text semantics.

In analyzing the text semantics of the first event, noise attenuation vectors are used to represent the various features of the coral reef whitening event. These characteristics include the degree of coral whitening, ocean temperature variation, water quality parameters, etc. Then, a feature compression operation is performed to compress the features into a smaller number of most important and representative features.

Assume that the following underlying text semantics are obtained by feature compression:

potential feature 1: degree of whitening (mild, moderate, severe);

potential feature 2: temperature change (rise, stability, fluctuation);

latent feature 3: water quality (good, general, poor).

Through step 1031, potential text semantics of the coral reef whitening event are obtained, and the characteristics are extracted after characteristic compression.

Step 1032 is a process of generating second event text semantics. In step 1032, based on the obtained latent text semantics, second event text semantics of the target marine ecology detection event at a different level will be generated. The level of each second event text semantic corresponds to a level of one first event text semantic.

Based on the obtained latent text semantics, a plurality of levels of second event text semantics may be generated. For example:

second event text semantic level 1: degree of coral whitening (mild, moderate, severe);

second event text semantic level 2: temperature change (rise, stability, fluctuation);

second event text semantic level 3: water quality (good, general, poor).

In this way, various aspects of the coral reef whitening event can be described in more detail at different levels. Each level provides a semantic description of a particular feature from the extent of coral whitening to temperature change and water quality.

Through step 1032, second event text semantics of the target coral reef whitening event at different levels are generated. Such a description enables analysis and understanding of the event from multiple angles.

In some alternative embodiments, the generating of the second event text semantics of the target marine ecology detection event at a different level based on the latent text semantics in step 1032 includes steps 10321-1033.

Step 10321, converting the latent text semantic from the first target feature coordinate system to the second target feature coordinate system in the current text unit to obtain the target text mode label.

In an alternative embodiment, to generate the second event text semantics of the target marine ecology detection event at a different level, the latent text semantics need to be converted from the first target feature coordinate system to the second target feature coordinate system. This conversion process may be implemented by step 10321. Specifically, features in the latent text semantics are mapped from a first target feature coordinate system to a second target feature coordinate system during the conversion process. For example, in coral reef whitening event monitoring, there may be multiple target feature coordinate systems such as the extent of coral whitening, temperature changes, and water quality conditions. Step 10321 is intended to ensure accurate and consistent conversion between different target feature coordinate systems in order to better represent the second event text semantics of the target marine ecology detection event.

Step 10322, obtaining an original text mode label, and updating the original text mode label by using the target text mode label to obtain an updated text mode label.

In step 10322, the original text mode tag is obtained and updated with the target text mode tag to obtain an updated text mode tag. The text mode tag is used to specify the output style or format of the text. In this case, the original text pattern tag is updated with the target text pattern tag to match the second event text semantics of the target marine ecology detection event. By using updated text mode tags, it can be ensured that the generated second event text semantics are consistent with the target in terms of output style and format, thereby making the description more accurate and easy to understand.

Step 10323, generating a second event text semantic of the target marine ecology detection event under the target hierarchy according to the updated text mode label.

In step 10323, second event text semantics of the target marine ecology detection event at the target level are generated from the updated text pattern tags. In other words, based on the updated text mode tags, second event text semantics corresponding to the target level are generated. The purpose of this is to translate the event description into the specific semantic representation required for the target hierarchy to meet the needs of analysis, communication and decision making.

Through steps 10321 to 10323, the generation of the second event text semantics of the target marine ecology detection event can be completed, and the accuracy and consistency of the second event text semantics under different layers can be ensured. This provides a more specific, targeted semantic description to support in-depth analysis and decision-making of coral reef whitening events. Updating the text mode tab ensures that the output results are consistent with the target, providing a text style that is more suited to the particular needs.

In some alternative embodiments, the updating the original text mode label with the target text mode label in step 10322, to obtain an updated text mode label, includes: adjusting the original text mode label by utilizing the target text mode label to obtain an adjusted text mode label; and performing mode checking on the adjusted text mode label, and performing label updating on the text mode label subjected to the mode checking to obtain an updated text mode label.

In step 10322, the original text mode label is adjusted using the target text mode label and an adjusted text mode label is obtained. And then, carrying out mode checking on the adjusted text mode label, and updating the checked text mode label to obtain an updated text mode label.

The concrete explanation is as follows:

adjusting the original text mode label by using the target text mode label: in this step, the original text mode label is adjusted using the target text mode label. The original text mode label may be based on an initial setting or a default setting, while the target text mode label reflects the desired output style or format. By applying the target text mode tags to the original tags, they can be tailored to the target requirements;

pattern checking is carried out on the adjusted text pattern label: in this step, a pattern check is performed on the adjusted text pattern tag. The pattern check is intended to ensure that the text pattern tag complies with a predetermined pattern or specification, such as a grammar, format or convention. By performing a pattern check, it can be verified whether the adjusted text pattern tag satisfies a specific rule or criterion;

and (3) carrying out label updating on the text mode label after the mode checking: after the pattern check is completed, if the text pattern tag is found to be out of specification or contract, it is updated. This means that the adjusted text mode label is modified or revised to meet the requirements and expectations. By completing the label updating, the consistency of the text mode label and the required output style is ensured.

Through the series operation of the steps, the original text mode label can be adjusted by using the target text mode label, and the text mode label is updated after the mode check. In this way updated text pattern tags are obtained which ensure that the generated second event text semantics conform to specific output style, format or specification requirements.

In some alternative embodiments, the generating, in step 10323, second event text semantics of the target marine ecology detection event at a target level from the updated text pattern tag includes steps 103231-103234.

Step 103231, obtaining a text description token vector of a text unit before the current text unit, and performing a moving average processing on the text description token vector of the previous text unit by using the updated text mode label.

In this step, a text description token vector of a text unit preceding the current text unit is obtained, and the text description token vector of the preceding text unit is subjected to a moving average process using the updated text pattern tag. The moving average process (convolution process) is a technique of smoothing data to reduce noise or fluctuation by calculating an average value over a certain time window in a data sequence. In this embodiment, a moving average process is applied to the text description token vector of the previous text unit to reduce the fluctuation variation inside it, making it more stable and reliable.

And 103232, performing disturbance transformation on the moving average vector to obtain the text description characterization vector of the current text unit.

In this step, the moving average vector is perturbed to obtain a text description token vector for the current text unit. The disturbance transformation is an operation of introducing randomness, and the diversity and the richness of the data are increased by carrying out tiny random offset or transformation on the data. In this embodiment, a degree of randomness is introduced by perturbing the moving average vector, thereby increasing the diversity of the text description token vector.

Step 103233, using the text description token vector of the current text unit as the text description incoming vector of a next text unit of the current text unit, and obtaining an updated text pattern tag that loops the latent text semantics in the next text unit to generate.

In this step, the text description token vector for the current text unit is taken as the text description incoming vector for the next text unit and looped to generate an updated text pattern tag for the latent text semantic. By entering the text description token vector for the current text unit into the next text unit and performing a looping operation, an updated text pattern tag reflecting the underlying text semantics in the next text unit may be generated. This process allows for gradual adjustment and updating of text pattern tags to better reflect the characteristics and semantics of the target marine ecology detection event.

And 103234, after the text description characterization vector of the current text unit is subjected to feature expansion in the next text unit, performing moving average processing and disturbance transformation on the text description characterization vector subjected to feature expansion by using the updated text mode label in the next text unit to obtain a second event text semantic of the target marine ecological detection event under a target layer level.

In the step, the text description characterization vector of the current text unit is subjected to feature expansion, and the text description characterization vector after feature expansion is subjected to moving average processing and disturbance transformation by utilizing the updated text mode label in the next text unit, so that second event text semantics under the target layer level are generated.

Feature expansion is an operation of adding dimensions or expanding the original feature vector to obtain a richer feature representation. In this embodiment, the text description token vector for the current text unit is feature extended to provide more descriptive information.

The feature-expanded text description token vector is then subjected to a moving average process and a perturbation transformation using the updated text pattern tags in the next text unit. These operations aim to further smooth the data and introduce randomness to improve the accuracy and diversity of the second event text semantics at the target level of generation.

In some alternative embodiments, the first event text semantics of the target marine ecology detection event are generated by a survey report text analysis network completing the debugging, based on which the step of obtaining a survey report text analysis network completing the debugging comprises steps 201-203.

Step 201, obtaining a past investigation report text containing past marine ecology detection events, performing disturbance operation on the past investigation report text, and performing noise attenuation processing on the past investigation report disturbance text containing the past marine ecology detection disturbance events to obtain noise attenuation vectors of the past marine ecology detection disturbance events.

In this step, past investigation report texts (sample investigation report texts) containing past marine ecology detection events (sample marine ecology detection events) are acquired, and disturbance operations are performed on these texts. The perturbation operation refers to a tiny random transformation or modification of text to introduce a degree of noise and diversity. Noise attenuation vectors are generated by perturbing the past survey report text containing past marine ecology detection events. This vector reflects the text content after adding noise or random variations to the original text.

And 202, generating a past noise attenuation text according to the noise attenuation vector of the past marine ecology detection disturbance event.

In this step, past noise-reduced text is generated based on the noise-reduced vector. The past noise-reduced text refers to text data that is disturbed by noise by applying a noise-reduced vector to the past survey report text. This process can help simulate the presence of noise or imperfect text data in the real scene, increasing the robustness and generalization ability of the model.

And 203, calculating network debugging cost according to semantic distinction between the past investigation report text and the past noise weakening text, and debugging the investigation report text analysis network by utilizing the calculated network debugging cost until obtaining the investigation report text analysis network after completing debugging.

In this step, the network debugging cost is calculated according to the semantic distinction between the past investigation report text and the past noise weakening text, and the investigation report text analysis network is debugged by using the calculated network debugging cost until the investigation report text analysis network with completed debugging is obtained.

First, the semantic distinction between past survey report text and past noise-attenuated text is used to measure the performance or error rate of the network. By comparing the differences between the original text and the disturbed text, the behaviour of the network in handling noise or variations can be evaluated.

And then, according to the calculated network debugging cost, debugging the debugging report text analysis network. Debugging refers to the process of optimizing and improving a network to reduce semantic differences, increase performance, or reduce error rates. The parameters or structure of the network are continuously adjusted until the expected network performance and accuracy are achieved, thereby obtaining a survey report text analysis network with the debugging completed.

Through the continuous operation of steps 201 to 203 described above, data for training and optimizing the survey report text analysis network can be generated using past survey report text and noise attenuation, ensuring that the network can accurately handle marine ecology detection events in a truly noisy environment.

In other possible embodiments, the second event text semantics of the target marine ecology detection event are generated by a survey report text optimization network that completes the debugging. Based on this, the step of obtaining a survey report text optimized network with complete commissioning includes steps 301-304.

Step 301, obtaining a past investigation report text containing a past marine ecology detection disturbance event, and loading the past investigation report disturbance text containing the past marine ecology detection disturbance event into the investigation report text optimization network to obtain a past investigation report optimization text of the past marine ecology detection disturbance event when the past investigation report text is subjected to disturbance operation.

In this step, past survey report text containing past marine ecology detection events is obtained and the disturbance text is loaded into a survey report text optimization network. Through this operation, the past survey report optimization text is generated. The past investigation report optimizing text refers to text data processed by the investigation report text optimizing network. Improved and optimized past text data is obtained by inputting the past survey report text into an optimization network and applying the optimization capabilities of the network.

And 302, performing context semantic mining on the past investigation report optimization text and the past investigation report text respectively to obtain the context semantic features of the past investigation report optimization text and the context semantic features of the past investigation report text.

In this step, the past survey report optimization text and the past survey report text are context semantic mined to obtain their respective context semantic features. Contextual semantic mining refers to extracting and capturing context information and semantic associations from text. Contextual semantic features related to text content may be extracted and obtained by analyzing the context, semantic relationships, and related features in the text.

Step 303, generating a context debugging cost based on the difference between the context semantic features of the past survey report optimization text and the context semantic features of the past survey report text.

In this step, a contextual debugging cost is calculated based on the difference between the contextual semantic features of the past survey report optimization text and the past survey report text. The context commissioning cost is to measure the performance or error rate of the network by comparing the difference between the context semantic features of the past survey report optimization text and the past survey report text. By quantifying the differences between the context semantic features, the performance of the network in processing the context information can be evaluated.

And step 304, determining a target debugging cost related to the investigation report text optimization network according to the context debugging cost, and debugging the investigation report text analysis network by utilizing the target debugging cost until obtaining the investigation report text analysis network with completed debugging.

In this step, a target debugging cost for the survey report text optimization network is determined based on the contextual debugging cost, and the survey report text analysis network is debugged using the target debugging cost. The target debugging cost refers to a target value for optimizing the network, which is determined according to the context debugging cost. By setting the target debugging cost, the training and optimizing process of the network can be guided to improve the performance and accuracy of the network.

Through the continuous operation of steps 301 through 304 described above, past survey report text and a survey report text optimization network can be utilized to generate optimized second event text semantics to improve analysis and understanding of marine ecology detection events.

An example of the application of steps 301-304 in a coral reef whitening event monitoring scenario is described below:

in step 301, past survey report texts containing past coral reef whitening events are obtained, and disturbance operations are performed on these past survey report texts. For example, a series of investigation report texts of past coral reef whitening events are collected, then disturbance operation is carried out on some reports in the investigation report texts, different observation conditions or parameter settings are introduced, and a past investigation report disturbance text containing the disturbance of the past coral reef whitening events is generated;

in step 302, context semantic mining is performed on the past survey report optimization text and the past survey report text, and their context semantic features are extracted. For example, through natural language processing technology, semantic analysis is carried out on the past investigation report optimization text and the past investigation report text, and the characteristics such as context information, keywords, semantic similarity and the like of the past investigation report optimization text and the past investigation report text are obtained;

In step 303, a context debug cost is generated based on the difference between the contextual semantic features of the past survey report optimization text and the contextual semantic features of the past survey report text. For example, comparing the contextual semantic features of the past survey report optimization text and the past survey report text, and calculating a difference metric between them as a contextual debugging cost;

in step 304, a target debug cost for the survey report text optimization network is determined based on the contextual debug cost, and the survey report text analysis network is debugged using the target debug cost until a complete debug survey report text analysis network is obtained. For example, according to the result of the context debugging cost, setting an optimization target of the investigation report text optimization network, and debugging the network by using a corresponding optimization algorithm until the expected debugging effect is achieved, so that the investigation report text of the coral reef whitening event can be accurately identified and analyzed.

Through implementation of the steps, the investigation report text analysis network can be effectively optimized, the monitoring capability of the coral reef whitening event is improved, and valuable information is provided for further research and protection work.

Further, there is also provided a computer-readable storage medium having stored thereon a program which, when executed by a processor, implements the above-described method.

In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other manners. The apparatus and method embodiments described above are merely illustrative, for example, flow diagrams and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

In addition, functional modules in the embodiments of the present invention may be integrated together to form a single part, or each module may exist alone, or two or more modules may be integrated to form a single part.

The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a network device, or the like) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes. It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A report analysis method for use in a survey processing system, the method comprising:

generating a survey report optimization text of the target marine ecology detection event through third event text semantics of the target marine ecology detection event;

the second event text semantic of the target marine ecology detection event is obtained, which comprises the following steps:

generating second event text semantics of the target marine ecology detection event under different levels based on the potential text semantics; the hierarchy corresponding to the text semantics of the second event is consistent with the hierarchy corresponding to the text semantics of the first event;

Wherein the generating, based on the latent text semantics, second event text semantics of the target marine ecology detection event at different levels comprises: converting the latent text semantics from a first target feature coordinate system to a second target feature coordinate system in the current text unit to obtain a target text mode label; obtaining an original text mode label, and updating the original text mode label by using the target text mode label to obtain an updated text mode label; generating a second event text semantic of the target marine ecology detection event under a target hierarchy according to the updated text mode label;

the updating the original text mode label by using the target text mode label to obtain an updated text mode label includes: adjusting the original text mode label by utilizing the target text mode label to obtain an adjusted text mode label; performing mode checking on the adjusted text mode label, and performing label updating on the text mode label subjected to the mode checking to obtain an updated text mode label;

the generating, according to the updated text mode tag, a second event text semantic of the target marine ecology detection event under a target layer level includes: obtaining a text description characterization vector of a text unit before the current text unit, and carrying out moving average processing on the text description characterization vector of the previous text unit by utilizing the updated text mode label; performing disturbance transformation on the moving average vector to obtain a text description characterization vector of the current text unit; taking the text description characterization vector of the current text unit as a text description incoming vector of a next text unit of the current text unit, and obtaining an updated text pattern tag generated by cycling the latent text semantics in the next text unit; and after the text description characterization vector of the current text unit is subjected to feature expansion in the latter text unit, carrying out moving average processing and disturbance transformation on the text description characterization vector subjected to feature expansion by utilizing the updated text mode label in the latter text unit, so as to obtain the text semantics of the second event of the target marine ecology detection event under the target layer level.

2. The method of claim 1, wherein outputting the semantic fine granularity of the semantic variable according to the detection item description of each marine ecology detection item and deriving the semantic fine granularity of the semantic variable according to the detection item description of each marine ecology detection item, performing semantic aggregation operation on the first event text semantic of the target marine ecology detection event and the second event text semantic of the target marine ecology detection event to obtain the third event text semantic of the target marine ecology detection event, and comprising:

3. The method of claim 2, wherein the semantic fine granularity of the test item output semantic variable of each marine ecology test item description has a set quantitative relationship with the content enrichment index of the corresponding marine ecology test item description in the initial survey report text; the step of obtaining regularized aggregated semantics of the target marine ecology detection event comprises:

4. The method of claim 1, wherein the text mining the first event text semantic of the target marine ecology detection event based on the initial survey report text comprises:

5. The method of claim 4, wherein noise attenuation processing the initial survey report text to obtain a noise attenuation vector for a target marine ecology detection event in the initial survey report text, comprising:

sequentially carrying out feature expansion on the ocean survey report semantic relation network under the target feature recognition degree by combining the ocean survey report semantic relation network under the feature recognition degree, and taking the ocean survey report semantic relation network with different layer orders obtained by the feature expansion as a noise attenuation vector of the target ocean ecological detection event;

the noise attenuation vector of the target marine ecology detection event comprises a marine investigation report semantic relation network of different layers; the semantic transformation is performed on the noise attenuation vector of the target marine ecology detection event, and the report semantic obtained by the semantic transformation is used as the first event text semantic of the target marine ecology detection event, and the method comprises the following steps: performing feature compression on the ocean survey report semantic relation network of any layer order, and performing feature expansion on the ocean survey report semantic relation network after feature compression; and taking the ocean survey report semantic relation network subjected to feature expansion again as a first event text semantic of the target ocean ecology detection event under any layer order.

6. The method of claim 1, wherein the first event text semantics of the target marine ecology detection event are generated by a survey report text analysis network that completes a debugging; the step of obtaining a survey report text analysis network with complete commissioning includes:

7. The method of claim 1, wherein the second event text semantics of the target marine ecology detection event are generated by a survey report text optimization network that completes debugging; the step of obtaining a survey report text optimized network with complete commissioning includes:

8. A marine environmental survey system comprising a processor and a memory; the processor is communicatively connected to the memory, the processor being configured to read a computer program from the memory and execute the computer program to implement the method of any of claims 1-7.

9. A computer readable storage medium, characterized in that a program is stored thereon, which program, when being executed by a processor, implements the method of any of claims 1-7.