302results about How to "Effective combination" patented technology

The invention discloses a multi-modal document structured processing and knowledge extraction method based on a large language model. The method comprises the following steps: S1, receiving an input multi-modal document; s2, performing preprocessing on the text data in the multi-mode document; s3, performing feature extraction on non-text data in the multi-mode document; s4, performing multi-modal data fusion on the preprocessed text data and the non-text data after feature extraction; s5, performing deep semantic analysis on the fused multi-modal data through a pre-trained improved BERT model, wherein the deep semantic analysis comprises key entity recognition, relation extraction and event detection; s6, based on a deep semantic analysis result, automatically constructing a knowledge graph by using the extracted information; and S7, outputting the data of the knowledge graph in a format for analysis or application. According to the method and the system, structured processing and knowledge extraction of unstructured and multi-modal documents are realized.

The invention discloses a power distribution area equipment monitoring intelligent analysis method, and relates to the technical field of power system equipment monitoring, and the method comprises the steps: collecting and preprocessing the real-time operation data of power distribution area equipment; performing feature extraction on the real-time operation data by using a deep learning model to obtain an operation state feature vector; performing similarity calculation on the running state feature vector and a normal running state feature vector in a historical database to obtain a preliminary abnormal degree quantitative index; performing equipment state evaluation and classification according to the preliminary abnormal degree quantitative index to obtain an equipment state evaluation result; and taking corresponding measures according to an equipment state evaluation result, automatically generating a fault diagnosis report and pushing the fault diagnosis report to a maintenance terminal. The multi-parameter monitoring and deep learning feature extraction technology adopted by the invention improves the accuracy and comprehensiveness of anomaly detection, and can identify some potential fault features which are difficult to find by a traditional method.

The invention relates to the technical field of yarn motion state monitoring, in particular to a yarn motion state monitoring method and system.The yarn motion state monitoring method comprises the steps that firstly, deformation during yarn motion is monitored and converted into tension data, and the tension data are uploaded to a yarn motion state database in real time; a yarn motion video is captured through an industrial camera, the offset and angle between the yarn motion video and a preset track are calculated, whether an offset coefficient exceeds a threshold value or not is judged, and offset abnormal points and related tension data are recorded and uploaded in time; combining acoustic and vibration signals, analyzing and monitoring the breakage condition of the yarn through the composite signal, and recording the breakage tension and the composite signal at the abnormal moment. Besides, the data in the database is analyzed through the Gaussian mixture model, the motion state of the current yarn is accurately obtained, and an efficient and reliable monitoring means is provided for monitoring the motion state of the yarn in the spinning process.

The invention relates to a flame-retardant toughening agent for a polypropylene material and a preparation method of the flame-retardant toughening agent, and belongs to the technical field of macromolecular flame-retardant toughening, and the preparation method comprises the following steps: firstly combining an aromatic compound with melamine, and then combining with a flame-retardant material to obtain a flame-retardant compound; the preparation method comprises the following steps: combining itaconic anhydride, a modified toughening agent, octavinyl silsesquioxane and polypropylene through graft copolymerization to obtain grafted polypropylene; combining the flame-retardant compound with grafted polypropylene to obtain a modified flame-retardant toughening agent; and mixing the modified flame-retardant toughening agent with an antioxidant and a lubricant, and carrying out melt blending extrusion to obtain the flame-retardant toughening agent for the polypropylene material. According to the technical scheme, the aromatic compound, the melamine and the flame-retardant material are combined, so that the flame-retardant property of the polypropylene material is effectively improved; the itaconic anhydride, the modified toughening agent and the octavinyl silsesquioxane are combined with the polypropylene, so that the impact toughness and the mechanical property of the polypropylene material are enhanced, and the comprehensive performance of the flame-retardant toughening agent is generally improved.

The invention discloses an online learning cognitive anomaly student detection method and system based on multi-modal representation, and belongs to the field of anomaly detection. The method comprises the following steps: firstly, acquiring multi-modal data in a student learning process, and constructing a student relation graph; then inputting the multi-modal data of each student into a multi-modal learning representation module pre-trained by an auxiliary classification task to obtain multi-modal embedding corresponding to each student; finally, the multi-mode embedding and student relation graph is input into a cognitive anomaly student detection module which is trained through a comparative learning framework in advance, and cognitive anomaly scores of students corresponding to the target nodes are obtained.According to the method, student cognitive information can be automatically extracted, heavy work of manual processing is avoided, and the learning efficiency is improved. And meanwhile, cognitive anomaly student detection is realized based on multi-view multi-strategy comparison, the cognitive level of students on the platform can be detected on the basis of considering multiple influence factors, and the online learning effect is assisted to be improved.

The invention discloses a method for correcting a laser gas concentration sensor, which is used for measuring gas concentration by using laser with different wavelengths and correcting by combining environmental parameters. The method comprises the following steps: extracting laser absorption characteristics by using a first feedforward neural network, extracting environmental matrix characteristics by using a convolutional neural network, fusing the characteristics through an attention mechanism, and predicting the gas concentration through a second feedforward neural network. The loss function not only considers the difference between the predicted value and the true value, but also comprises the physical constraint between the light intensity value and the theoretical light intensity value. According to the method, the measurement accuracy and the model robustness under different environmental conditions are improved, and the influence of measurement errors and environmental fluctuation is reduced through the combination of multi-wavelength measurement and physical knowledge.

The invention discloses an existing anchoring and shotcreting side slope ecological restoration technology and method with shrub planting as the core, a side slope vegetation hole environment improving device and a hole filling layered structure design are included, and the side slope vegetation hole environment improving device is mainly composed of a conical threaded drill bit, a connecting rod, a motor, a metal connecting hose, a pressure storage tank and an air compressor. The motor provides power for penetration of the gas injection needle head, and the air compressor provides power for injection of the plant nutrient solution; the vegetation hole layered structure comprises a bottom water retention layer, a base fertilizer layer, a nutrient soil layer and a surface covering layer; environment improvement is conducted on the vertical and horizontal directions of the planting holes through an environment improvement device, local dominant shrubs are planted in the planting holes, and local dominant vine plants are planted at the top of the slope to comprehensively and ecologically protect the slope; comprising the following steps: step 1, investigating and surveying plant communities and analyzing slope stability; step 2, slope vegetation hole environment improvement; 3, planting hole construction; and 4, slope top vine plant ecological slope protection construction is carried out.

The invention provides a targeted giant salamander iridovirus nano antibody as well as a preparation method and application thereof. The invention discloses a giant salamander iridovirus targeting nano antibody, and aims to provide a giant salamander iridovirus targeting nano antibody which is low in expression cost, low in difficulty and stable in source, can be effectively combined with giant salamander iridovirus, and specifically realizes immunoassay of the giant salamander iridovirus. The amino acid sequences of the nano antibody are shown as SEQ ID NO.1-6, and the nano antibody belongs to the technical field of biological science.

The invention discloses a drug target combinability prediction method based on variational coding, and belongs to the field of bioinformatics. According to the method, efficient extraction of drug and target features, construction of potential feature space of the drug and target features and prediction of the interaction relationship of the drug and target features are realized through a dual-stage coupled end-to-end neural network. The model is mainly composed of two stages, in the first stage, potential representations of drug and target features are obtained through variational compression and feature extraction, and in the second stage, a drug target action matrix is reconstructed through an up-down sampling comprehensive deep neural network. Through the architecture, the chemical characteristics of the drug and the biological characteristics of the target can be effectively integrated, and the accuracy of interactive relationship prediction is improved. The invention provides a novel and efficient solution for deep analysis of biological big data at the same time.

The invention discloses an OPSA-ECMS hydrogen fuel hybrid power unmanned aerial vehicle energy management method, and the method comprises the steps: building an unmanned aerial vehicle hybrid power system model employing a hydrogen fuel cell as a main power supply and a lithium ion battery as an auxiliary power supply, and building an energy management system model. According to the method, the OPSA and the ECMS are combined, the minimum hydrogen consumption and the highest efficiency are taken as optimization objectives, the OPSA is adopted to calculate an optimal reference energy distribution scheme in an off-line manner under a known preset route, and the ECMS is adopted to optimize the reference energy distribution scheme in an on-line manner in the task execution process of the unmanned aerial vehicle, so that the optimal energy distribution of the system is realized. According to the method, OPSA off-line calculation and ECMS on-line optimization are combined, the high energy density of hydrogen fuel and the quick response characteristic of the battery are exerted to the maximum extent, and the overall energy utilization efficiency is improved.

The invention belongs to the field of antennas, and discloses a millimeter wave broadband circularly polarized antenna array and a manufacturing method thereof. The fourth metal layer is used for receiving and transmitting electromagnetic waves based on the sequential rotation feed network; the fourth metal layer is also used for serving as a ground plane of the grounding coplanar waveguide; the third dielectric layer is used for transmitting the received electromagnetic waves to the third metal layer; metal via holes are introduced into the third dielectric layer to realize the performance of the grounding coplanar waveguide; the third metal layer is used for coupling electromagnetic waves to the first radiation layer and the second radiation layer based on slot coupling; the third metal layer is also used for serving as a ground plane of the grounded coplanar waveguide; the first radiation layer and the second radiation layer are respectively used for directionally radiating electromagnetic waves to a free space; metal via holes are introduced into the first radiation layer and the second radiation layer, so that the performance of the magnetoelectric dipole antenna is realized; the millimeter wave antenna meets the requirement of a millimeter wave mobile communication working frequency band, and has the characteristics of large axial ratio bandwidth, easiness in processing, low loss and stable directional diagram.

The invention provides a drug pair interaction prediction method and device based on soft mask double-view learning, and relates to the technical field of drug interaction prediction. The method comprises the following steps: obtaining molecular view data of two to-be-predicted drugs and an interaction relationship type thereof to form a triple, and generating a bipartite graph of a drug pair; inputting the triple and the bipartite graph into an MSMDL encoder, and outputting new node embedding of corresponding drugs; wherein the MSMDL encoder has K layers, the output result of each layer is respectively input into the MSMDL encoder of the next layer and the self-attention graph pooling layer, and graph-level representation is generated through the self-attention graph pooling layer to obtain substructure feature embedding; embedding the substructure features of each layer to serve as input of a common attention mechanism layer, and obtaining importance of interaction between the substructure feature embedding of each layer; and finally, outputting the DDI prediction probability through a decoder. According to the invention, local and global information of the drug is integrated, and the accuracy of drug pair interaction prediction is improved.

The invention discloses a heavy metal curing agent and a curing method of heavy metals in coal gangue, and relates to the field of heavy metal pollution treatment. The heavy metal curing agent comprises a first curing agent, a second curing agent and a third curing agent; wherein the first curing agent comprises silt, fly ash and bentonite, the second curing agent comprises a blend and chitosan loaded on the surface of the blend, the blend comprises quartz sand and calcium carbonate, and the third curing agent comprises a silane compound. The curing agent has an excellent long-term curing effect on heavy metals, and a cured body formed by the curing agent has excellent compressive strength.

The invention belongs to the technical field of controllable preparation of high-entropy nano materials and electrolyzed water catalysts, and particularly discloses a preparation method for controllable preparation of a two-dimensional high-entropy sulfide efficient electrolyzed water catalyst based on a two-dimensional high-entropy hydroxide template. The preparation method comprises the following steps: respectively adding at least five metal precursors into deionized water, then adding ammonium fluoride and urea, carrying out uniform ultrasonic mixing, and carrying out a stirring reaction to obtain a template two-dimensional high-entropy hydroxide; and placing the obtained two-dimensional high-entropy hydroxide template in the middle position of a single-temperature-zone tubular furnace, continuously introducing argon into a quartz tube, and carrying out vulcanization reaction in an argon atmosphere to obtain the target product, namely the two-dimensional high-entropy sulfide. By utilizing the synthesis method of the two-dimensional high-entropy sulfide, the problems of high synthesis temperature, non-uniform element distribution, unstable two-dimensional structure and the like of the two-dimensional high-entropy sulfide can be solved, the phenomena of phase separation, segregation and the like in high-temperature synthesis are avoided, the production cost and the process difficulty are reduced, and the yield of the two-dimensional high-entropy sulfide is improved. A series of high-performance two-dimensional high-entropy sulfide electrolyzed water catalysts are synthesized.

The invention discloses an offline writer identification method and device, electronic equipment and a medium. The method comprises the following steps: acquiring handwritten word image data; converting the handwritten image into a grayscale image, performing binarization processing on the grayscale image, adjusting the size of the processed image to a preset size, and constructing a training set; constructing a handwriting recognition model, and training the handwriting recognition model by adopting the training set; wherein the handwriting recognition model comprises a local modeling module, an identical branch and a global module, the local modeling module is used for capturing local features, the identical branch is used for extracting global features, and the global module is used for integrating the local features and the global features; and inputting a handwritten image to be recognized into the trained handwriting recognition model to obtain a result of a final classification layer of the model, performing softmax probability calculation, and taking a writer with the maximum probability as a recognition result. According to the method, robust feature representation can be effectively extracted from limited handwriting, and the recognition precision far higher than that of other methods is achieved.

The invention discloses a short pulse laser assisted continuous laser welding device and method. The method comprises the steps that materials are laid; a welding area is divided, and the whole welding area is divided into small areas; performing feature extraction on the small regions; constructing a speed mapping table according to the extracted different features; and planning a laser scanning path according to the speed mapping table. The short-pulse laser-assisted continuous laser welding device and method have the beneficial effects that near-infrared wavelength laser (such as 1064nm) is selected in the whole welding process, and the characteristics of high reflectivity of titanium alloy to the wavelength, high absorptivity of thermoplastic resin, relatively low absorptivity of a resin part in a thermosetting composite material and poor absorptivity of carbon fibers to the wavelength are utilized; laser energy is mainly concentrated on the thermoplastic resin layer, and local heating is achieved. The method comprises the following steps: dividing a region into small regions, extracting features, constructing a speed mapping table, controlling laser energy input, and reducing heat affected zones of the titanium alloy and the thermosetting composite material.

The invention provides a label-free magnetic self-assembly nanoprobe, a construction method thereof and application of the label-free magnetic self-assembly nanoprobe in detection of PrPC protein. The label-free magnetic self-assembly nanoprobe disclosed by the invention comprises a magnetic Fe3O4 / Fe2O3 (at) Au heteroplasmon nanorod array and a sulfhydrylation aptamer nucleic acid chain which is combined with the magnetic Fe3O4 / Fe2O3 (at) Au heteroplasmon nanorod array through an Au-S bond; a biosensor prepared from the label-free magnetic self-assembly nanoprobe is good in reproducibility and stability; the RSD value of multiple times of detection (intraday / day) is 1.55%; the storage time of multiple groups of sensors is 0-15 days, and the change of response signal values does not exceed 4%; the biosensor has a relatively low detection limit (1pg / mL), a relatively wide linear range (1 * 10 < pg / mL >-1 * 10 < 5 > pg / mL) and relatively good sensitivity, has very good application in detection of PrPC, and has very good practicability.

The invention discloses a power-interruption-free gear-shifting hub motor two-gear transmission of a duplex planetary compound gear train. The power-interruption-free gear-shifting hub motor two-gear transmission is mainly composed of a driving motor; a first planetary gear train; a second planetary gear train; a controllable one-way clutch; a clutch; a central support shaft; and a transmission shell and the like. The first planetary gear train is a duplex planet compound planetary gear train with small characteristic parameters. An output shaft of the driving motor is connected with the first sun gear; the first right planet carrier is connected with the second sun gear; the second planet carrier is connected with wheels; the controllable one-way clutch is used for connecting the first gear ring and a transmission shell, and a one-way locking state, a two-way locking state or a two-way overrunning state can be achieved. The clutch is used for connecting the drive motor output shaft and the first left planet carrier. According to the hub motor two-gear transmission, seven different running states including first-gear driving, second-gear driving, neutral gear, reversing, first-gear regenerative braking, second-gear regenerative braking and low-speed sliding friction braking can be achieved through the control clutch and the controllable one-way clutch.

The invention provides a control method and system for glass laser marking, and relates to the field of intelligent control or adjusting systems in industrial machining.The method comprises the steps that three-dimensional height data of all machining points on the working face of a machining table is obtained, and a three-dimensional height data set is formed; generating a laser head height compensation data set of each machining position of the machining table based on the three-dimensional height distribution data of the overall machining table; acquiring information of to-be-marked glass and geometric data of a to-be-marked area on the to-be-marked glass; acquiring height compensation data of the marking area according to the to-be-marked glass information and the geometric data of the marking area, and generating a marking area path plan; according to the scheme, three-dimensional height data acquisition of a machining table is taken as a core, a marking area path plan for controlling operation of a marking machine is obtained on the basis of the data, a marking area path is adjusted through a height compensation data set, a final area path scheme is generated, and the movement of a laser head is dynamically adjusted by adopting an intelligent control technology.

The invention provides a reverse structure and fluid migration integrated experimental device and method.The reverse structure and fluid migration integrated experimental device comprises an experimental box, the experimental box is a rectangular box, and an experimental model for simulating a stratum structure is arranged in the experimental box; the push-pull power mechanism is connected with the experiment box and is used for pushing and pulling the experiment model in the experiment box so as to stretch or extrude the simulated stratum; and the fluid injection mechanism is used for injecting dyeing fluid from the side direction of the experimental model. According to the device and the method, the extrusion structure deformation process can be simulated on the basis of a stretching simulation result of uniform stress transmission, fluid migration simulation is completed under the condition that an elastic polymer used in a stretching simulation experiment is not damaged, and effective combination of reversal structure simulation of stretching first and then extrusion and fluid migration simulation is achieved.

The invention relates to the technical field of artificial intelligence and the field of financial science and technology, and discloses a spoken text generation method, which comprises the following steps of: segmenting a written text into a plurality of sentences, and extracting text features of each sentence by using a pre-training language model; and based on a context attention mechanism, generating a context token in combination with the text feature and the spoken instruction of each sentence. And projecting the context token and the text feature of each sentence to a unified feature space through a linear projector to generate a content token. And inputting the text feature of each sentence into the autoregression language model, and generating a spoken text in combination with the content token. Through a context attention mechanism, association and semantic information between sentences can be more effectively combined, and the naturalness and fluency of the spoken text are improved. Text features and context tokens are processed through a linear projector, so that the coherence and consistency of generated texts are ensured. The problem that text generation is not coherent is solved, and the naturalness and accuracy of the spoken text are improved.

The invention provides a preparation method of a catalyst for synthesizing indole, which comprises the following steps: sequentially impregnating silicon dioxide in a manganese salt solution, a first alkali solution, a silver nitrate solution containing an auxiliary agent M and a second alkali solution by adopting an impregnation method, standing for 0-72 hours at the temperature of 0-50 DEG C during each impregnation, and drying for 0-72 hours at the drying temperature of 20-290 DEG C after each standing, the auxiliary agent M is any one or more of aluminum nitrate, zirconium nitrate, lanthanum nitrate and cerium nitrate. Manganese dioxide is used as an anchor point, so that silver oxide is effectively dispersed and anchored on the surface of a silicon dioxide carrier modified with manganese dioxide; and finally, a nitrate auxiliary agent with the atomic radius larger than that of silver atoms is added and used for inhibiting the rate of increasing the particle size of the silver particles due to agglomeration of the silver particles caused by high-temperature sintering. The preparation process does not involve treatment at 300 DEG C or above and roasting treatment, and the preparation method is simple in step.

The invention provides a discrete element method for simulating fragmentation of granular materials in any form, and belongs to the technical field of computational mechanics. The method comprises the following steps: scanning a particle model through a three-dimensional image to obtain particle surface information, and constructing a geometric structure of the particle through a triangular surface unit; establishing a periodic hexagonal topological grid of the particles by adopting a lattice point method, and obtaining equivalent stiffness and equivalent Poisson's ratio through a strain energy equivalence principle; introducing a breaking criterion at the interface of the topological unit, calculating a stress field in the topological structure of the unit, and analyzing and judging whether the unit is broken or not according to a stress state; the crushed particles directly participate in calculation of the next stage, and multi-stage crushing of the particles is achieved. Uniform construction of particles in any form can be realized; the crushing process of the discrete material under the action of an external load can be accurately reproduced, and the reduction degree is high; the flexible selection of the crushing criterion can be suitable for a plurality of engineering fields; and the crushed particles continue to participate in calculation, and multi-stage crushing of the particles is achieved.

The invention provides application of nano CDs as a fluorescent probe in fluorescence detection of sugammadex sodium and a fluorescence detection method of sugammadex sodium, and relates to the technical field of drug analysis and detection. According to the invention, carboxyl in sugammadex sodium and nitrogen atoms in nano CDs interact with each other, the fluorescence intensity of nano CDs is enhanced through intramolecular charge transfer, sugammadex sodium and nano CDs can be effectively combined through a static combination effect, and nano CDs have high sensitivity and selectivity to sugammadex sodium; the method is suitable for detecting sugammadex sodium. By adopting a fluorescence analysis method, the high-sensitivity, high-selectivity, simple, rapid and accurate quantitative detection of the sugammadex can be realized according to the fluorescence enhancement efficiency of the sugammadex relative to the nano CDs after the sugammadex is added and a standard curve of the concentration of the sugammadex and the fluorescence enhancement efficiency. The fluorescence detection method provided by the invention is low in detection cost.

The invention provides a knowledge graph completion method and system fusing entity description information and graph attention, relates to the technical field of knowledge graph completion, and aims to solve the problems that in the prior art, only triple information in a knowledge graph is considered, complete combination of entity description information and graph structure information of the knowledge graph is ignored, and the knowledge graph completion efficiency is low. And a large-scale knowledge graph is difficult to be accurately and completely complemented. The method comprises the following steps: encoding obtained entity description information to obtain an entity description vector; fusing the entity description vector and the initialized representation vector of the entity; inputting the fused entity vectors into a graph attention network model, and aggregating neighbor information; and inputting the aggregated entity vectors into a decoder to obtain the prediction probability of the triple, and complementing the missing knowledge graph according to the prediction probability. According to the method, the semantic information and the graph structure information of the entity are effectively combined together, and the efficiency and the accuracy of the model in a knowledge graph completion task are remarkably improved.

The invention relates to the technical field of variable subtitles, and particularly provides a multi-variable subtitle method for adaptive balanced perception. The method comprises the following steps: obtaining training set visual features of changed subtitles; constructing a vocabulary and a tag file; in the first stage of the encoder, image features are obtained; in the second stage of the encoder, the obtained common characteristics are as follows; in the third stage of the encoder, extracting difference features of the image; splicing the two groups of image features, and constructing a comparison expression of two differences; finally, modeling the overall difference and the original image respectively; in the first stage of a decoder, grammar correlation is established for words in text data one by one through a text dependence model; in the second stage of the decoder, a group of subtitle information describing image differences is obtained, according to grammar rules and context information, the method ensures that sentences are complete and smooth, visual information and language information are effectively combined, and the accuracy of semantic information is improved.

The invention relates to the technical field of building detection, and discloses a building detection device based on a communication technology, the building detection device comprises a base capable of electrically walking, the top of the base is fixedly connected with a control box with an opening in the side face, a box door is hinged to the opening, and the top of the control box is fixedly connected with a take-off frame; the take-off frame is provided with an outdoor detection mechanism used for detecting an outdoor high-altitude building. According to the building detection equipment based on the communication technology, indoor and outdoor detection work of a building is effectively combined, the convenience and comprehensiveness of the detection device are improved, an unmanned aerial vehicle in the outdoor detection mechanism can drive a first camera to take off, so that detection work can be conveniently carried out on positions such as outdoor high-rise walls, and the detection efficiency is improved. According to the anti-falling unmanned aerial vehicle, anti-falling treatment can be conducted on the unmanned aerial vehicle losing stability and falling rapidly through the anti-falling assembly arranged in the shell, and the anti-falling performance of the unmanned aerial vehicle can be further improved through the arrangement of the auxiliary assembly.

The invention belongs to the technical field of sprayed concrete, and particularly relates to rapid early-strength self-repairing sprayed concrete and a construction method thereof. The sprayed concrete comprises a base layer, an interface layer and a functional layer which are sequentially arranged, wherein the functional layer comprises a rapid early strength system and a composite microcapsule system; the composite microcapsule system comprises an inner-layer microcapsule and an outer-layer microcapsule, the inner-layer microcapsule comprises bisphenol A epoxy resin and modified aliphatic amine, and the outer-layer microcapsule comprises modified polyurethane; the rapid early strength system comprises a cementing material, an additive system and a fiber component, the interface layer comprises organic components, namely acrylate emulsion and epoxy resin, and inorganic components, namely nano silicon dioxide and nano aluminum oxide. According to the rapid early strength type self-repairing sprayed concrete and the construction method thereof, a synergistic mechanism of double microcapsules and a rapid early strength system is provided, and synchronous improvement of the self-repairing performance and the early strength of the concrete is achieved through complex interface regulation and accurate design of functional components.

The invention discloses an artificial intelligence platform interaction method based on a large language model and an intelligent agent. The method mainly comprises three parts of artificial intelligence platform script language development, intelligent agent construction and text conversion operation. The method comprises the following steps: firstly, developing a scripting language of an artificial intelligence platform, and defining and packaging mouse and keyboard operations such as dragging, pulling, adding, deleting and modifying nodes and connecting lines of the artificial intelligence platform into the scripting language in order to realize the interaction capability of subsequent text conversion operation; thirdly, semantic analysis is conducted on the platform requirements expressed by the natural language, corresponding operation steps and function tools needing to be called in each step are inferred, and the function tools are executed; and finally, the user reviews and corrects a text conversion operation result, and feeds back a correction result to the background database. According to the method, an intelligent interaction method of a text conversion operation user and a platform is realized; and the intelligent interaction method for completing the text conversion operation realizes effective enabling of a leading-edge large language model technology on a traditional artificial intelligence platform.

The invention belongs to the field of text-video retrieval, and particularly relates to a subtitle generation-based multi-granularity semantic alignment text-video retrieval method, which comprises the following steps of: establishing an SAEC model, training the SAEC model, and carrying out text-video matching retrieval through the trained SAEC model. According to the method, a new model SAEC is provided, the model significantly improves the retrieval accuracy from the text to the video by matching local details in the video and understanding the global semantics of the video, and meanwhile, the model not only can understand the direct content of the video, but also can grasp a more complex situation by jointly generating the subtitles and the video.