37 results about "Modal particle" patented technology

Compositions for lost circulation materials (LCM) and methods for using same in drilling and/or completing wellbores that help solve lost circulation problems in a wide range of fracture sizes, thereby advantageously eliminating the need for a variety of products for lost circulation in a field at any one time. An unexpected synergy and improved reduction in lost circulation is obtained. The invention provides specific LCM components in specific ratios that are analogs to lost circulation fractures and that yield superior performance in preventing or alleviating lost circulation in drilling and cementing boreholes. The invention compositions have a multi-modal particle size distribution (PSD) which provides a higher concentration of component materials in the same range of two or more fracture widths, thus allowing plugging to occur over a wider range than a single mode or narrow PSD.

The invention discloses a method and a device for providing a voice service. One embodiment of the method for providing the voice service comprises the steps that a voice input signal is acquired; a semantic recognition model trained by the adoption of a machine learning method is utilized to recognize mood and spoken content in the voice input signal, and corresponding mood input information andtext input information are obtained, wherein the mood input information is used for expressing a mood type of the voice input signal; and voice service data query is performed based on the mood inputinformation and the text input information, and voice response information is generated according to the query result. Through the embodiment, mood recognition not relying on a modal particle is realized, the intention of a speaker can be detected more accurately, and the precision of the voice service is improved.

Fine particulate polymer drag reducing agents (DRAs) in bi-modal or multi-modal particle size distributions may be produced simply and efficiently without cryogenic temperatures. The grinding or pulverizing of polymer, e.g. non-porous poly(alpha-olefin) suitable for reducing drag in hydrocarbons may be achieved by the use of at least one liquid grinding aid and at least two grinding processors in series. The blades of the stators of the grinders are of different configuration so that granulated polymer fed to the first processor having relatively larger gaps between blades is ground to an intermediate size which is fed to the second processor having relatively smaller gaps between blades which grinds the polymer to a second, smaller size. A non-limiting example of a suitable liquid grinding aid includes a blend of propylene glycol, water and hexanol. Particulate DRA may be produced at a size of 300 microns or less in only two passes.

The invention discloses a method and system for realizing video and audio driven face animation by combining modal particle characteristics, and the method comprises the steps: constructing a voice feature matrix through the extraction of voice features, carrying out the sampling of the feature matrix through the multilayer convolution operation of a modal particle enhancement training network, and mapping the feature matrix to an intermediate variable of a low-dimensional space; converting the input voice into characters, identifying a mood word from the character content, constructing a onehot vector, and splicing the onehot vector with an intermediate variable to obtain an intermediate variable containing mood word features; carrying out the convolution through a modal particle enhancement training network, and then obtaining an expression AU parameter of a current frame through mapping, wherein the parameter is used for being fitted with an AU parameter generated by video tracking and voice prediction algorithms to serve as a driving parameter of a face model, so expression enhancement is realized. According to the invention, the three-dimensional Avatar model in the virtual scene can be jointly driven by inputting the video content of the face of the user and the audio content of the sound production of the user, and on the basis of real-time driving, the whole and local face animations are enabled to obtain more realistic and vivid expression effects.

The invention discloses a Chinese text feature extracting method with a text mood fusion function. By means of the method, it is achieved that the text feature representation fusing mood features, syntax features and semanteme features is obtained in a lengthened text. The method comprises the steps that firstly, a text word set and a mood word set are constructed, the text word set and the mood word set are transformed into word embedding forms respectively, and corresponding vector models are obtained; secondly, according to the text word embedding represented time step dimensions and feature dimensions, text features are screened, the mood features are fused into the time step dimension of the selected text feature, and the text feature representation which accurately represents the semanteme is obtained. According to the method, the contributions of modal particles to the text semanteme are fully utilized to fuse the mood features, the syntax features and the semanteme features into the text feature representation, and the text feature representation is low in dimension and continuous so that the text semanteme can be better represented, and natural language processing tasks, such as text analysis, language translation and relation extraction, can be better effectively supported.

Provided are methods and systems for generating nanoparticles from an inorganic precursor compound using a hydrothermal process within at least one CSTR or PFR maintained at an elevated temperature and an elevated pressure and a treatment vessel in which this reaction solution can be applied to one or more catalyst substrates. In operation, the reaction solution may be maintained within the CSTR at a substantially constant concentration and within a reaction temperature range for a reaction period sufficient to obtain nanoparticles having a desired average particle size of, for example, less than 10 nm formation and/or deposition. Variations of the basic method and system can provide, for example, the generation of complex particle size distribution profiles, the selective deposition of a multi-modal particle size distribution on a single substrate.

The invention belongs to the field of semantic recognition and discloses a semantic recognition method and a system. Acquiring speech information to be recognized; inputting the speech information tobe recognized into the recognition model, removing the non-keywords and repetitive contents that do not affect the semantics in the speech information to be recognized, and generating the target speech; the target speech is semantically understood and the result of semantic recognition is obtained. The invention filters non-keywords such as modal particles and repetitive contents in the speech information to be recognized by the recognition model, solves the problem that too much redundant information in user sentences leads to semantic understanding errors, so as to improve the accuracy of semantic recognition, in order to more accurately understand the true intentions of users, better serve users, improve the user experience.

A modified zircon-based refractory material comprising TiO2 and Y2O3 having a density in certain embodiments higher than 4.00 g cm-3, and method of making such zircon materials. The process involves the use of zircon particles having a multi-modal particle distribution curve in making the green body. The material exhibits an improved performance in creep rate. The invention is particularly useful for making zircon-based isopipes for fusion down-draw forming of glass sheet materials.

The invention discloses an intelligent advertisement identifying method and especially relates to a solving method for identifying an advertisement from mass information. The method comprises the following steps: establishing a word stock and a disabled word stock, wherein the disabled word stock contains some adverbs and modal particles with higher probability of occurrence; selecting some samples including the advertisement and common information; respectively extracting the characteristics of the advertisement and common information; calculating two-classified characteristic probability according to the bayesian algorithm and generating a model; continuously optimizing the model during a use process, thereby increasing the judging accuracy and recall rate of the model for the advertisement; if the probability of the judged advertisement is higher than the information probability, judging the information as the advertisement.

The invention discloses a digital building presentation system based on voice interaction, which comprises a voice feature extraction module, a voice model and mode matching module, and a control module, wherein the voice feature extraction module extracts a voice feature sequence changing over time from a voice waveform; the voice model and mode matching module matches and compares unknown voice features with the voice models in the case of recognition, the distance between a feature vector sequence of the unknown voice and each sound production module is calculated, and syntax and semantic analysis is carried out on a recognition result; and the control module performs corresponding house type presentation control on a computer system according to a result outputted by the voice model and mode matching module. According to the digital building presentation system based on voice interaction, environmental adaptability is strong, the recognition rate is high, modal particles in the case of continuous voice recognition can be removed, control on building presentation can be carried out, and the full view of the building can be presented in real time through voice interaction.

The invention discloses a broadcast signal prejudging and preprocessing method. The method includes the steps that a speech segment is intercepted, and the segment with the speech and the segment without the speech are separated; a speech time domain sequence containing the segment with the speech is subjected to framing and sliding window adding to obtain a speech time domain sequence, the additive noise is removed, and a speech time domain sequence after additive noise removal is obtained; the speech time domain sequence is filtered to obtain a speech time domain sequence after impulse noise removal; a recirculating network neural model is established, the segment with the speech after impulse noise removal is screened, and a first speech segment containing standard Chinese speeches only and a second speech segment containing both the standard Chinese speeches and the background music are screened out; and the background music of the second speech segment is separated from the standard Chinese speeches. A pure broadcast recording file for eliminating the current noise, the ambient sound, speaker modal particle and the background music is obtained, thus the accuracy of results obtained by keyword recognition is improved, and the mis-recognition rate is decreased.

An abrasive compact having at least a tri-modal particle size distribution, and a binder phase, define a plurality of interstices. The binder phase is distributed in the interstices to form binder pools that correspond substantially in average size to that of an ultrahard polycrystalline composite material having a monomodal particle size distribution and substantially the same overall average particle grain size.

The embodiments of the invention provide an emotion audio determining method, an emotion display method, a method for converting text into voice, and related devices. The emotion audio determining method comprises the steps that to-be-processed emotion data are obtained; emotion dimensions and emoticons are determined; from the to-be-processed emotion data, corresponding emotion feature words andemotion modal particles corresponding to the emotion feature words are determined specific to all the emotion dimensions; the emotion feature words and/or the emotion modal particles are mapped with the corresponding emoticons, and the corresponding relationship between the emotion feature words and/or the emotion modal particles and the emoticons is formed; and based on the corresponding relationship, matched audio is configured for the emoticons, wherein the audio is pronunciations of the emoticons or sound effects consistent with ideographical expressions of the emoticons. Through the emotion audio determining method, the technical problem of how to enable data emotional expression to be attached to actual emotional communication is solved, the expression effect of the emoticons is improved, and the more personific and more natural emotional experience is realized.

Drag reduction of hydrocarbon fluids flowing through pipelines of various lengths is improved by polyolefin drag reducer dispersions or dispersions using bi- or multi-modal particle size distributions. Drag reducers having larger particle sizes dissolve more slowly than drag reducers having smaller particle sizes. By using at least bi-modal particle size distributions drag reduction can be distributed more uniformly over the length of the pipeline where smaller sized particles dissolve sooner or earlier in the pipeline and larger sized particles dissolve later or further along the pipeline.

Fine particulate polymer drag reducing agent (DRA) in bi-modal or multi-modal particle size distributions may be produced simply and efficiently without cryogenic temperatures. The grinding or pulverizing of polymer, such as poly(alpha-olefin) suitable for reducing drag in flowing hydrocarbons may be achieved by the use of at least one liquid grinding aid and at least two grinding processors in series. The impellers of the grinders are of different openness so that granulated polymer fed to the first processor having a relatively more open impeller is ground to an intermediate size which is fed to the second processor having a relatively more closed impeller which grinds the polymer to a second, smaller size. A non-limiting example of a suitable liquid grinding aid includes a blend of propylene glycol, water and hexanol. Particulate DRA may be produced at a size of about 300 microns or less in only two passes.

The invention discloses a fuzzy fitness value interactive evolutionary optimization method used for indoor wall clock design. The fuzzy fitness value interactive evolutionary optimization method used for indoor wall clock design comprises the specific steps that: before the evolution is started, a system provides a design environment for a user so as to randomly generate initial evolutionary species; a person carries out single value evaluation for an individual through an interactive interface, and the system records the individual evaluation time; the system estimates an individual fitness value evaluated by the user based on the individual evaluation time and the single value evaluation value; the system clusters the species based on individual phenotype similarity and distributes unevaluated individual fussy fitness values; an individual selection adaptive value is built through the central value of the individual fussy fitness values and the individual phenotype similarity, and individual selection is finished by adopting a league method. The fuzzy fitness value interactive evolutionary optimization method used for indoor wall clock design efficiently reduces operation burden of designers as the designers only need to evaluate a small number of product types according to a single value; the system realizes estimation of the individual fuzzy fitness value evaluated by the user automatically at the background; and in comparison with the traditional fussy fitness value giving, inputting of modal particle is unnecessary and a half of the operation quantity is saved.

The invention discloses a deep learning-based garbage text filtering method. The method comprises the steps of filter character data at first, removing unnecessary symbols, spaces and modal particles,classifying according to different data types in the garbage text, marking and distinguishing character data and graph data without changing sequence and position of the two types of data, convertinggraph data into character data through a deep learning algorithm, wherein data conversion is integral to the deep learning method, comparing with forbidden words in a cloud server through the deep learning algorithm by combining original character data so as to obtain a garbage text, wherein text comparison represents important promotion of the deep learning method and can realize effective, deepinterception and prompt. The existing text filtering method cannot screen out the garbage text consisting both character data and graph data. The method solves the above problem, and uses the deep learning algorithm in garbage text processing, thereby improving screening accuracy.

The invention discloses a teaching knowledge base construction method, which comprises the following steps of S1, knowledge base framework construction: constructing a knowledge base module frameworkwhich comprises a vocabulary module, a case production module, a case analysis module and a simulation pronunciation module, S2, vocabulary module construction, wherein the vocabulary module is composed of a basic vocabulary submodule, a mood vocabulary submodule, an emotion vocabulary submodule, a tone vocabulary submodule and a sentence segmentation module. According to the invention, two or more than two sub-modules are extracted from the basic vocabulary sub-module, the mood vocabulary sub-module, the emotion vocabulary sub-module, the tone vocabulary sub-module, the sentence segmentationmodule and the punctuation sub-module to intelligently generate and combine cases, so that continuous cases can be intelligently formed, and the case storage space can be saved; the emotion of a speaker during pronunciation can be simulated, and teaching is performed in combination with the emotion, so that a learner can accurately master the profound language charm of Chinese.