100results about How to "Improve extraction accuracy" patented technology

The invention provides a non-power-cut test method for mutual inductance parameters of a power transmission line under excitation of different-frequency signals and a related device, and solves the problems of difficulty in accompanying stop and high safety risk of adjacent lines. The method comprises the following steps: selecting one circuit in the same corridor for power failure, connecting the head end with an induced electricity suppression device, and performing three-phase short-circuit grounding at the tail end; injecting a different-frequency sinusoidal current of 45-55Hz into the power failure line; the synchronous measurement device collects head and tail voltage and current of a running line based on a satellite synchronous clock; extracting pilot frequency signals through two-stage notch filtering and Fourier transform; and different-frequency zero-sequence self-impedance and mutual impedance are calculated and converted into power-frequency impedance. According to the invention, accurate test without power failure can be realized, and safety and power supply reliability are guaranteed.

The invention belongs to the technical field of electric power measurement, and discloses a calibration method and system of an intelligent measurement switch and a multi-switch synchronous measurement method, and the method comprises the steps: obtaining voltage and current waveform data of the intelligent measurement switch under a nonlinear load condition, and extracting a frequency spectrum feature matrix of the voltage and current waveform data; constructing an amplitude-phase coupling error model representing a nonlinear incidence relation between amplitude fluctuation and phase deviation, inputting the spectrum characteristic matrix into the amplitude-phase coupling error model, and calculating an amplitude-phase coupling error vector; performing vector decomposition on the amplitude-phase coupling error vector, and separating out an in-phase component error and an orthogonal component error; and generating a calibration compensation coefficient according to the in-phase component error and the orthogonal component error, and performing vector calibration on the intelligent measurement switch, thereby improving the calibration precision of the intelligent measurement switch.

The present application relates to a deep learning image segmentation method and system for ore and core structure analysis, belonging to the technical field of material structure analysis, which solves the technical problems of insufficient multi-scale boundary recognition ability, unstable segmentation of weak boundary area, easy overlap of different mineral phases, insufficient model cross-scene generalization ability, etc. of the existing image segmentation technology. The solution is: a deep learning image segmentation method and system for ore and core structure analysis, which acquires an original image; a multi-scale feature extraction network is constructed based on the encoding end and the decoding end; a material structure prior constraint module is constructed based on the pore connectivity prior, the fracture direction consistency prior and the mineral phase consistency prior; the multi-scale feature extraction network is supervised and trained and self-supervised adapted; the image is segmented; and the image segmentation image is optimized. Compared with the prior art, the present application has the advantages of strong multi-scale boundary recognition ability, stable segmentation, strong generalization ability, wide application range, etc.

This invention discloses a multi-scale icing detection method and system for transmission line insulators under extreme weather conditions. The method includes: dividing a visible light image into pixel blocks and converting them into feature vectors with position encoding; inputting these vectors into an improved Swin Transformer integrating a Dynamic Deep Sample Generation (DDSGM) module; applying DDSGM to enhance features in a four-stage feature extraction phase; additionally fusing polarized light features in the first two stages; and outputting a multi-scale feature map. The multi-scale feature map is then input into a deformable attention unit containing DDSGM, undergoing secondary enhancement and adaptive sampling (fusing physical field information to correct for offset) to obtain enhanced features. A prediction head predicts the icing category, bounding box, and thickness; Hungarian matching is used to associate the target; and an early warning system outputs the results and a warning. This invention is the first to integrate DDSGM, optimize sampling, and combine a multi-physical quantity strategy to improve robustness to low-quality images and transparent ice layers, achieving accurate detection and early warning under extreme weather conditions.

This invention belongs to the field of spectral data processing technology, specifically a spectral data processing system and method based on in-situ data sensing and intelligent analysis. This invention employs in-situ optical imaging, absorption spectroscopy, fluorescence spectroscopy, and PL mapping to acquire real-time three-dimensional spectral data of time, wavelength, and intensity for real-time monitoring of material crystallization, defects, and luminescence properties. It uses an automatic invalid value detection and cleaning algorithm to address data quality issues; it extracts peak position and half-peak width (WHM) characteristic parameters from the spectral data using peak position extraction and WHM calculation methods; it uses a nonlinear optimization algorithm to perform multi-peak Gaussian fitting on the spectral data; and it employs a signal-slot mechanism to achieve real-time synchronization of data presentation and fitting analysis, forming a complete closed loop of in-situ testing, data sensing, processing, analysis, synchronization, and export. This invention achieves real-time acquisition, efficient processing, accurate analysis, and automatic synchronization of spectral data, providing technical support for standardized processing and traceable analysis of spectral data.

This invention discloses a method and system for large-scale urban village extraction based on a token mask mechanism, relating to the technical field of remote sensing geographic information systems. The method includes: acquiring a high-resolution remote sensing image dataset; labeling urban village areas to obtain an urban village labeled dataset; performing data augmentation on the urban village labeled dataset to obtain an augmented urban village labeled dataset; iteratively training a constructed urban village extraction model; setting training parameters; updating the model parameters of the urban village extraction model to obtain a trained urban village extraction model; acquiring a high-resolution remote sensing image of the area to be detected; inputting the trained urban village extraction model; and extracting the urban village areas within the area to be detected. This invention can quickly and accurately identify urban village areas from high-resolution remote sensing images, with high accuracy and efficiency, wide application range, and realizes high-precision and robust large-scale urban village mapping applications.

The present application relates to a method for extracting an acetabular facies lunata model, in particular to a method for extracting an acetabular facies lunata model based on mixed surface fitting, which solves the technical problem of low precision and robustness of the existing method for extracting an acetabular facies lunata model. The present application realizes rough extraction of the acetabular facies lunata model through spherical surface fitting, and then realizes accurate extraction of the acetabular facies lunata model in combination with ellipsoidal surface fitting, which combines the stability of spherical surface fitting and the anatomical fidelity of ellipsoidal surface fitting. In particular, in the case of complex pathologies such as hip dysplasia, the present application can still maintain high extraction accuracy, and improve the accuracy and reliability of clinical diagnosis and surgical planning. Meanwhile, in the iteration process of ellipsoidal surface fitting, an adaptive fault tolerance threshold is introduced, which effectively solves the problems of sensitivity to noise and strong parameter dependence of traditional methods, and has better robustness in processing complex acetabular morphology.

This invention provides a multimodal fusion-based method for correlation analysis of deep-sea environmental data outside the Yangtze River Estuary, belonging to the field of deep-sea environmental analysis technology. This invention achieves accurate data registration by establishing a spatiotemporal standardization model for multi-source data and employing an adaptive spatiotemporal kriging interpolation algorithm. It utilizes a sparse coding marine signal separation algorithm to construct an overcomplete dictionary to separate mixed signals and extract pure features. Based on an attention mechanism, a feature extraction network is constructed to automatically learn deep feature representations of physical, chemical, and biological parameters. A matrix rank deficiency detection algorithm is used to automatically supplement feature compensation vectors to ensure feature integrity. A cross-scale attention mechanism based on wavelet transform is constructed to fuse multi-scale features. A Shapley value interpretability evaluation system is established to quantify the importance of correlated features. This invention solves the technical problem of inaccurate correlated feature extraction during the fusion of multi-source heterogeneous marine environmental data.

The present application belongs to the technical field of fault diagnosis of complex transmission system, and particularly relates to a method for accurately extracting fluctuating speed from vibration signals of integrated transmission and correcting signals, the implementation of the method involves a speed fluctuation amplification unit, a speed fluctuation extraction unit and a speed fluctuation removal unit; the method can accurately estimate speed fluctuation from vibration signals without actually measuring speed signals, and remove speed fluctuation to obtain a stable signal. The present application improves the accuracy of speed fluctuation extraction and the accuracy of speed fluctuation removal by amplifying speed fluctuation for several times, and greatly reduces the time consumption of the algorithm through multiple frequency modulation. The method solves the problem of accurately extracting fluctuating speed from vibration signals of integrated transmission system and removing the influence of speed fluctuation, and provides a low-computing-cost and high-precision method for extracting and removing speed fluctuation.

The application discloses a water body extraction method based on combined water body index frequency, comprising the following steps: S1, acquiring remote sensing image data and pre-processing the remote sensing image data; S2, constructing a water body index and constructing an instantaneous water body extraction decision tree model according to the water body index; S3, performing instantaneous water body pixel recognition and extraction according to the instantaneous water body extraction decision tree model, discriminating the water body and preliminarily determining the average water body; S4, extracting the water edge line by using a Canny edge detection method, and determining the average water body extraction result according to the output water body pixel value, the edge detection and the boundary elimination result. The application can realize average water body extraction based on dense long-time sequence images, improve the water body extraction precision, improve the effectiveness of water body extraction in the building shadow and terrain shadow noise environment, resist the interference of the cirrus cloud noise in the single image, and make the large-scale and long-time sequence water body extraction more universal.

This invention discloses a method and system for identifying innovative points in academic papers, relating to the field of artificial intelligence technology. The method first processes the paper abstract through a concept path extraction module, including: structured semantic segmentation, concept pair extraction and verification, constraint relation triple generation, hierarchical verification, and path optimization, to construct a complete set of concept paths for the paper. Then, in a rare path discovery module, the popularity index of each concept path in the global path frequency dictionary is calculated and compared with a threshold to identify rare paths as the scientific innovative points of the paper. This invention effectively suppresses the illusion of language models through a knowledge graph constraint mechanism, significantly improving the accuracy of concept path extraction and its coverage of long-tail concepts, achieving accurate and efficient identification of innovative points in academic papers at the path level.

The present application relates to the technical field of ecological protection, and more particularly to a forest water conservation capacity remote sensing inversion method suitable for Nanling cloudy and rainy areas, aiming at the problem that the existing inversion method cannot overcome the cloud cover and terrain influence, and the precision of forest water conservation capacity is low, the present application obtains optical and microwave remote sensing data, pre-processes the optical and microwave remote sensing data, extracts corresponding optical characteristics and microwave characteristics, weights and fuses the optical characteristics and microwave characteristics, calculates the forest water conservation amount by using a water conservation amount physical model, and generates a Nanling forest water conservation capacity spatial distribution map. The method overcomes the errors caused by weather, terrain shadow and distortion, constitutes a full-weather, high-precision, Nanling special environment-adapted forest water conservation capacity remote sensing inversion system, and realizes accurate quantification of the water conservation capacity of the Nanling special forest ecosystem.

The embodiment of the invention provides a table data extraction method and device based on OCR (Optical Character Recognition) and visual language model collaboration. According to the method, the accurate positioning capability of optical character recognition and the deep semantic understanding capability of a visual language model are creatively fused; a complete cooperative processing framework from image preprocessing to structured data output is constructed, and the limitation of the prior art is effectively overcome. According to the method, the dependence on a fixed template and large-scale annotation data is abandoned, the table type is intelligently judged through adaptive structure analysis, the processing robustness under various abnormal conditions is ensured by utilizing a multi-layer progressive fault-tolerant mechanism, and the generalization adaptability of a system to tables with different formats and different qualities is remarkably improved. Meanwhile, by introducing a comprehensive verification and intelligent error correction process covering formats, semantics and consistency, the output data is not only accurate in the character level, but also reasonable and reliable in the business logic level.

The application is suitable for the technical field of image processing, and provides a battery glue line skeleton extraction method based on an improved ZS thinning algorithm, which comprises the following steps: preprocessing a collected battery glue line image to obtain a binary image; performing improved ZS thinning on the binary image, dividing the image into odd and even subdomains for batch processing, and introducing a constraint condition to iteratively delete redundant pixels to obtain a preliminary skeleton; using a T-shaped 4-neighborhood deletion template to remove T-shaped redundant pixels and realize complete single-pixelization; through 8-chain code tracking, identifying and removing burr branches caused by glue overflow and uneven glue coating, while protecting nodes to obtain a final glue line skeleton. The application solves the problems of skeleton redundancy, over-thinning and burr in existing thinning algorithms, realizes high-precision, single-pixel and burr-free glue line skeleton extraction, and can meet the online detection requirements of battery assembly lines, and has good industrial applicability.

The application provides an aircraft wing skin corrosion depth prediction method based on a GAN and an improved LightGBM algorithm, which comprises the following steps: S1, data enhancement: four fully connected layers are used to extract feature information between columns; S2, an improved condition generator and discriminator are used, and parameters are updated; S3, the LightGBM algorithm is improved, and the aircraft wing skin corrosion depth is predicted. The aircraft wing skin corrosion depth prediction method using the generative adversarial network for data enhancement and the gradient regression model for sample regression can accurately predict small sample data. Meanwhile, the gradient regression model LightGBM algorithm is improved, which is more suitable for the prediction of the aircraft wing skin corrosion depth, and can greatly improve the prediction accuracy of the aircraft wing skin corrosion depth prediction.

The invention provides a standing bell jar welding method and a welding robot. The standing bell jar welding method comprises the steps that image collection is conducted on a to-be-welded workpiece; according to the single-frame welding line image, extracting welding line center line coordinates and welding line widths corresponding to all center line points; splicing a plurality of frames of images by adopting a weighted fusion mode to obtain a full-circle weld joint panoramic expansion view; welding seam areas with different widths are classified and positioned according to threshold values, corresponding area masks are generated, the masks are fused with the whole-circle welding seam panoramic expansion graph to achieve color coding visual marking, and a welding seam width marking graph is output; segmenting the welding seam width marking graph according to a preset welding seam width threshold value to obtain segmented graphs; and the image coordinate system in the segmented graph is compared with the workpiece coordinate system of the workpiece to be welded, and the welding speed is controlled according to the welding seam width corresponding to the segmented graph. According to the method, the influence of image distortion is reduced, the welding quality is improved by combining the welding seam width, and the wide application prospect is achieved.

The present application relates to the technical field of image data processing, and particularly relates to a carbon slide edge detection optimization method based on cascade filtering. The present application extracts local mean, local median and local variance from the carbon slide gray image pixel by pixel, and corrects abnormal statistics after three layers of verification. The pixel uncertainty descriptor composed of the variance term, mean and median of local gray level quantifies and distinguishes the contribution proportion of Gaussian noise and impulse noise. The quantile feature working point is extracted from the uncertainty descriptor field, and the pixel-level hybrid weight is generated by the serial cascade link composed of the sorting filter, bilateral filter and anisotropic diffusion filter through Sigmoid mapping, so as to perform the cascade recursion in the convex combination fusion mode. The adaptive edge judgment threshold is constructed by combining the local gradient kurtosis and the local gradient absolute median difference with the centralization reference of the median, and the suspicious false edge is filtered through the connected component length threshold, and the final edge map is output after the secondary judgment of the corrected threshold.

The invention discloses an environment survey report structured information extraction method based on a large language model, and the method comprises the steps: obtaining a PDF file of an environment survey report, constructing an environment survey report chapter tree, and carrying out the splitting processing, and obtaining a lightweight environment survey report; introducing a field layer, a fragment layer and a dialogue layer, constructing a hierarchical Prompt large language model, and extracting the lightweight environment survey report to obtain a plurality of candidate results; and performing confidence comprehensive score calculation on the plurality of candidate results, performing screening, and performing adaptive optimization on the screening result to realize extraction of the structured information of the environmental survey report. According to the method, invalid Token calculation can be reduced, the model calling cost is reduced, and meanwhile, the extraction precision of the key field is remarkably improved. The environmental survey report structured information extraction method based on the large language model can be widely applied to the technical field of artificial intelligence.

The invention relates to a pseudo-random high-frequency quarter sinusoidal signal injection method and system for permanent magnet synchronous motor sensorless control, and belongs to the technical field of motor control. The method aims at solving the problems that noise is remarkable, torque ripple is aggravated and rotor position estimation precision is limited due to signal injection during zero-low-speed operation of a traditional high-frequency injection method. According to the technical scheme, the method comprises the following steps: constructing a pseudo-random high-frequency windowing quarter sine injection signal to suppress noise and torque ripple; high-frequency response current containing rotor position information is obtained through Clark transformation and Park transformation; designing a layered second-order generalized integrator band-pass filtering method to accurately extract position signals; and a cosine delay elimination normalization demodulation method is provided to eliminate the influence of the delay effect. According to the invention, synchronous improvement of noise suppression, torque ripple attenuation and rotor position estimation precision is realized, and the system stability and reliability are significantly enhanced.

This invention discloses a hybrid lunar surface segmentation method, system, device, and storage medium for photogrammetric point clouds in lunar rover exploration areas, belonging to the field of deep space exploration technology. The method includes: denoising and extracting initial lunar surface points from the raw scattered point cloud collected by the lunar rover in its exploration area; refining the extracted initial lunar surface points and labeling them as first lunar surface points; dividing all first lunar surface points into sub-blocks and obtaining terrain scene information for each sub-block; extracting and labeling second lunar surface points from the denoised scattered point cloud based on adaptive slope; extracting third lunar surface points from all unlabeled points based on an irregular triangular network, and labeling other points in the original scattered point cloud as non-lunar surface points; and correcting misclassified non-lunar surface points based on connected component segmentation and slope statistics, achieving accurate segmentation of lunar and non-lunar surface points. This invention can robustly filter out non-lunar surface points while preserving local lunar micro-topographical points.

The invention provides a KAN network time series data processing method and device based on modal decomposition and a related product, and relates to the technical field of data processing. According to the method, through modal decomposition, the extraction precision of the inherent characteristics of the time series data of the financial data center and the business system is improved; through a KAN network architecture, an activation function can be learned to strengthen specialized recognition capability for different fluctuation modes, and the unique design of the KAN enables the interpretability of the model to be obviously improved; by ingeniously combining a signal processing theory and a novel network architecture, fusion features are input into a pre-trained infrastructure performance index calculation model, and infrastructure performance indexes are output, so that trend analysis and abnormity early warning can be performed on the infrastructure performance indexes such as network delay and resource utilization rate, and the infrastructure performance indexes can be obtained. The accuracy and timeliness of financial infrastructure performance monitoring are effectively improved, and technical guarantee is provided for stable operation of a financial information system.

The application discloses a noise estimation method of an industrial camera image, comprising the following steps: step 1, collecting multiple reference images of a uniform light source under different brightness and different temperatures in a lens-free scene; step 2, calculating an amplitude graph of the image, and constructing a reference noise lookup table through the amplitude graph; step 3, calculating a direction graph of a current image, and calculating a normalized angle graph, performing threshold segmentation and screening based on the normalized angle graph, and obtaining a texture feature set; step 4, obtaining a shielding area of noise evaluation of the current image through a union set of a pixel set of an edge feature of the current image, a pixel set of an ultra-low amplitude and the texture feature set, and obtaining an effective noise evaluation area; and step 5, performing noise evaluation on the noise evaluation area to obtain a noise value. The application provides a noise estimation method of an industrial camera image based on fusion of a physical imaging prior and multi-dimensional feature screening, and the method has high robustness.

This invention relates to a flood disaster assessment method based on the fusion of remote sensing imagery and GIS data, belonging to the field of flood disaster monitoring and assessment technology. This method solves the problems of insufficient timeliness, limited coverage, and the inability of a single data source to comprehensively reflect the disaster situation in traditional assessment methods. Its technical solution mainly includes utilizing GIS data to provide spatial semantic background support, employing an improved UMNet++ model to accurately extract water body information before and after the flood from Sentinel-2 remote sensing imagery, and achieving quantitative assessment and hierarchical visualization of disaster severity through multi-source data fusion and spatial overlay analysis. The beneficial effects of this invention are that it significantly improves the accuracy and comprehensiveness of flood disaster assessment, enabling rapid and scientific disaster response and decision support.

The application discloses a solar meridian extraction method, device, equipment, medium and product, relates to the technical field of atmospheric polarized light navigation, and comprises the following steps: acquiring an original polarization image of a target area; in a Cartesian coordinate system, determining a linear polarization degree of an effective data area according to the original polarization image; performing coordinate system conversion on the linear polarization degree and determining an optimal symmetry axis based on a mirror symmetry principle; extracting a polarization angle corresponding to the effective data area based on the linear polarization degree of the effective data area; performing complex conversion and coordinate system conversion on the polarization angle corresponding to the effective data area to determine an antisymmetric polarization angle; extracting a skip line estimation value by using an antisymmetric mode according to the antisymmetric polarization angle; and fusing the optimal symmetry axis and the skip line estimation value through time series filtering to obtain a solar meridian. The application can improve the extraction precision of the solar meridian, thereby improving the precision of heading estimation.

This invention relates to the field of terminology database technology in the construction engineering field, and discloses a method and system for constructing and applying a terminology database in the construction engineering field. The method includes preprocessing unstructured data to obtain structured construction engineering text data; extracting structured construction engineering text data based on a joint extraction architecture to obtain structured terminology data and construction engineering triple information; establishing a standard database and a terminology database; constructing a terminology-standard two-layer knowledge graph based on the inherent relationship between the standard database and the terminology database and the construction engineering triple information; and responding to user queries by using a hybrid decision-making mechanism based on rule-based reasoning and graph neural network learning to call the standard database, terminology database, and terminology-standard two-layer knowledge graph to obtain query results. This invention realizes an intelligent closed loop for the entire lifecycle of construction engineering terminology from extraction and management to application, solving the problems of difficult acquisition, complex management, and inefficient application of industry terminology knowledge.

This application discloses a method, system, device, and storage medium for generating project documents. The method involves acquiring code data, parsing the code data to determine its structural information, and then mapping this structural information to corresponding positions within the code data. Based on the structural information, the code data is split into several code segments. A trained large language model is used to generate corresponding first descriptive text. Based on the first descriptive text for each code segment, the project document corresponding to the code data is obtained. This application generates project documents corresponding to code data automatically, reducing manual costs. Using a large language model to extract information from the code data and obtain the first descriptive text improves the efficiency of project document generation, reduces the probability of errors, and enhances the quality and usability of the project documents. The technical solution of this application can be widely applied in the field of code processing technology.

The application discloses a patent technology function extraction method for technical literature, analyzes the features of patent technology terms and function terms in the field of high-end equipment, and improves the accuracy of technology term and function term extraction in Chinese patents. The application constructs a deep learning model for technology term and function term extraction, combines the sentence pattern rules of patents, constructs the heuristic features of technology terms, positions the function sentences by constructing a function term feature dictionary, accelerates the training speed of the model, and improves the extraction accuracy. In order to reduce the cost of manual sample labeling and avoid the model overfitting phenomenon caused by too small data set, a self-training algorithm is used to realize weak supervision learning of the model. The technology terms most similar to the theme of the patent text are selected from the word clustering, the cosine similarity is used to combine the similar semantic terms, and a technology function matrix is constructed.

This invention relates to the field of aircraft assembly technology, and more particularly to an aircraft door repair method based on measured data. It includes: processing the original point cloud data using filtering and denoising algorithms, and filling in sharp boundary missing areas using point cloud interpolation algorithms to enhance point cloud quality; identifying the geometric boundary features of the door and fuselage frame using a feature-aware boundary extraction algorithm; calculating the door repair amount based on the minimum clearance principle, and outputting a precise door repair line. This invention achieves accurate geometric identification and repair trajectory generation for thin-walled, complex curved surface door structures. This method effectively solves the geometric deviation problem caused by manufacturing errors and structural deformation in door repair, significantly improving the automation level, assembly accuracy, and processing efficiency of aircraft door repair, and providing highly reliable data support and technical path for digital manufacturing in aviation assembly.

The application belongs to the technical field of unmanned driving, and specifically discloses a mine area road waterlogging detection method based on fusion of vision and laser radar, which comprises the following steps: acquiring image data and performing waterlogging segmentation; acquiring point cloud data, removing non-waterlogging point cloud to obtain a first segmentation result, and performing segmentation according to waterlogging semantic segmentation labeling to obtain a second segmentation result, and fusing to obtain a waterlogging recognition result; judging whether the image confidence degree is qualified or not; extracting ground point cloud in the point cloud data and modeling to obtain a fitting plane, projecting the edge contour pixel data to the fitting plane to construct a point cloud convex hull contour, and recognizing the point cloud convex hull contour to obtain the waterlogging recognition result; and has the following advantages: the multi-sensor fusion sensing mode can effectively deal with the problems existing in the single-sensor sensing, and has higher real-time performance, robustness and high confidence degree.

The present application belongs to the technical field of digital publication processing, and provides a multi-modal PDF content purification and intelligent proofreading system and method based on visual perception enhancement, which comprises a visual perception module, a semantic understanding module and an intelligent decision module; the visual perception module eliminates physical layer interference by generating multi-resolution images, extracting fine structure features and locating invisible characters; the semantic understanding module constructs a knowledge graph and a DST-Transformer model to realize character recognition, semantic coding and image-text association repair; the intelligent decision module outputs a purified document through weighted confidence calculation, dynamic threshold adjustment and cross-platform verification; the method corresponds to the three module steps, realizes full-process processing from physical layer analysis to logical layer reconstruction and then to intelligent output, improves the PDF content purification accuracy and cross-platform compatibility, and is suitable for electronic document deep purification and proofreading.