30 results about "Expressive communication ability" patented technology

The invention discloses a speech and language function impairment grade converter based on an ICF framework, which includes the following function evaluation indicators: auditory comprehension (%); visual comprehension (%); right brain function (%); Subdivided into word naming (%), simple paraphrase (%), word paraphrase (%), two-syllable word duration 2cvT (s), two-syllable word fundamental frequency 2cvF0 (Hz), sentence paraphrase (%), Sentence duration (s), sentence fundamental frequency (Hz), serial speech (%), oral description (%), and reading (%); writing (%); body language (%), enabling precise language function for patients assessment to facilitate subsequent treatment of the patient.

The invention discloses a compression method and platform of a pre-trained language model based on knowledge distillation. The method first designs a universal knowledge distillation strategy for feature transfer, and in the process of distilling the knowledge of the teacher model to the student model, the The feature map of each layer of the student model is close to the characteristics of the teacher, focusing on the expressive ability of small samples in the middle layer of the teacher model, and using these features to guide the student model; then using the self-attention distribution of the teacher model to detect the semantic and The syntactic ability builds a cross-knowledge distillation method based on self-attention; finally, in order to improve the learning quality of the learning model in the early stage of training and the generalization ability in the later stage of training, a linear migration strategy based on the Bernoulli probability distribution is designed and gradually completed from Knowledge transfer with teacher-to-student feature maps and self-attention distributions. Through the present invention, the multi-task-oriented pre-training language model is automatically compressed, and the compression efficiency of the language model is improved.

A spoken language evaluation method and device, comprising: obtaining a speech material to be evaluated and obtaining an evaluation dimension from a plurality of evaluation dimensions, and extracting evaluation features of the speech material to be evaluated according to the evaluation dimension; Score, get the spoken language evaluation value, and score the evaluation features according to the pre-stored oral language norm model, and get the public average level value for the speech material to be evaluated; output the evaluation result of the speech material to be evaluated according to the spoken language evaluation value and the public average level value . The spoken language evaluation method and device provided by the present invention use the seven dimensions of pronunciation and intonation ability, vocabulary use ability, grammar use ability, language logic ability, as well as the degree of theme concentration, the degree of expressiveness in a single sentence and the degree of completeness of expression to evaluate the learners' actual oral ability. Comprehensive evaluation and precise guidance provide comprehensive, practical and effective help for learners to improve their oral expression ability in various real situations.

The invention discloses a gait recognition system and method based on deep learning of a self-attention mechanism, belonging to the field of gait recognition. The present invention proposes to use the attention mechanism on the original feature map. By learning a matrix between 0 and 1 with the same size as the original feature, the original feature is denoised, and the salient features in the picture are selected to reduce the noise in the feature map. The noise of the classification loss and the comparison verification loss are organically combined to punish the features in combination with the loss function, which not only uses the identity information of the target, but also uses the different relationships between the targets to increase the discrimination between different features; The most important limb features in gait are added to the original depth features as prior knowledge. This combination can not only use the target global body shape features, but also correct the features that are not conducive to classification brought about by the clothes transformation learned by the global features, and enhance the original depth features. The expressive ability of deep features enhances the expressive ability of features from different dimensions.

The invention provides an intelligent robot oriented multimodal emotion data interaction method. The method comprises a step of receiving and analyzing multimodal interactive input information so as to obtain an intention parameter and a user emotional parameter inputted by a user, a step of generating text reply information corresponding to the multimodal interactive input information with the combination with the intention parameter and the emotional parameter and generating a speech emotional parameter with the combination of the text reply information, context interaction data and the user emotional parameter, and a step of carrying out synthesis to obtain speech information according to the text reply information and the speech emotional parameter and outputting the speech information. Through the intelligent robot oriented multimodal emotion data interaction method of the embodiment of the invention, when an intelligent robot interacts with the user, the sound effects of outputted TTS speech are different in different emotional states, the emotional expression ability of the robot is improved greatly, and thus the intelligent degree of the robot is improved greatly.

The present invention provides a video human behavior recognition method based on salient trajectory space information, which redefines the salience of the trajectory in the video, effectively removes the background and the trajectory of the non-moving parts of the human body in the video, and leaves the foreground with significant motion. These trajectories have smaller error and stronger expressive ability; in addition, this method distinguishes the moving parts and interactive objects of different human body parts, and uses the spatial and semantic relationship between them through multi-core learning to improve The recognition effect of the algorithm.

The invention discloses an emotional dialogue generation method and system based on interactive fusion, comprising: S100: receiving dialogue sentence text; S200: an emotion classifier identifying the emotion category contained in the dialogue sentence text and expressing it as a vector; S300: an encoder Encode the dialogue sentence text into a context vector; S400: the fine-grained interaction module fuses the emotion category representation vector and the context vector of the dialogue sentence text to generate an interactive vector; S500: the decoder uses the interactive vector to decode and generate a reply; S600: output reply. The invention can solve the problems of weak emotion perception ability and insufficient emotion expression ability of the existing man-machine dialogue system.

The present application provides a method for generating POI copy, which belongs to the field of computer technology, and is used to solve the problems in the POI copy generation method in the prior art that the copy does not have a strong ability to express POI and the expression accuracy is not high. The method includes: acquiring POI information for generating copy; predicting the subject tendency of the POI information, and determining the subject heading of the POI information; generating the copy of the POI to which the POI information belongs according to the determined subject heading. The POI copy generation method disclosed in the present application first predicts the subject tendency of the POI information, obtains the subject words, and uses the subject words as the input text for generating the copy, narrows the data range of the input text for generating the copy, and makes the generated copy The input text is more targeted, which improves the accuracy and expressiveness of the generated copywriting.