The invention provides a method for selecting positive sample data in source domain data to extend training data of a target domain by fusing semantic difference and label difference of sentences in the source domain and the target domain, so as to achieve the purpose of enhancing named entity recognition performance of the target domain Based on a conventional Bi-LSTM+CRF model, in order to fusesemantic differences and label differences of sentences in a source domain and a target domain, semantic difference and label difference are introduced through state representation and reward settingin reinforcement learning; therefore, the trained decision network can select sentences having positive influence on the named entity recognition performance of the target domain in the data of the source domain, expand the training data of the target domain, solve the problem of insufficient training data of the target domain, and improve the named entity recognition performance of the target domain at the same time

Semantic and label difference fused named entity identification field self-adaption method

The invention discloses an adaptive decision tree fall detection method and system, and belongs to the technical field of human body behavior recognition and judgment. The method comprises the following steps: 1, acquiring three-axis acceleration and three-axis angular velocity data of falling and non-falling actions of a human body, and performing screening; 2, calculating resultant accelerationand resultant angular acceleration, dividing a training set, substituting a test set and a verification set into a TSFRESH library to calculate features, and screening and deleting useless features; 3, selecting preliminary important features by using a random forest sieve; 4, establishing a decision tree model for training and verification, and testing a result; 5, continuing to obtain a new sample, repeating the steps 2 to 4, and updating the decision tree model. According to the invention, an accurate tumble judgment result can be obtained through a decision tree algorithm with a small calculation amount; after a certain number of samples are collected, the decision tree model is updated, so the judgment precision of the algorithm can be further improved.

Adaptive decision tree fall detection method and system

The invention relates to a text classification method and system, electronic equipment and a computer readable storage medium, and the method comprises the steps: adding an input layer of a BERT pre-training model, enabling the BERT pre-training model to participate in the training, and carrying out the classification and recognition of a to-be-classified text based on a classification model obtained after training, wherein the added input layer is a feature representation layer for assisting classification and recognition. By adding the input layer of the BERT pre-training model, the reference characteristic quantity during model text classification and recognition is increased, so that the text classification accuracy can be improved. Besides, the structure of the BERT pre-training modelis not changed, so that the BERT model does not need to be improved by a Google company and the like requesting to provide the BERT pre-training model, namely, the method is not limited by a basic model providing company, and the problem of inconvenience does not exist.

Text classification method and system, electronic equipment and computer readable storage medium

The invention relates to the technical field of data processing, and relates to an epidemic situation label data processing method and system. The method comprises the steps of: extracting tracking behavior characteristics of a tracking information sequence recorded by each user terminal; screening out a first tracking behavior characteristic of a epidemic situation related user and a second tracking behavior characteristic of a to-be-tested user; comparing the second tracking behavior characteristics of the to-be-tested user with the first tracking behavior characteristics of each epidemic situation related user; and determining characteristic attributes of the to-be-tested user under the different epidemic situation influence type labels, thereby sending corresponding epidemic situationlabel information to a user terminal of the to-be-tested user according to the characteristic attributes of the to-be-tested user under the different epidemic situation influence type labels for prompting. According to the invention, differentiated epidemic situation label analysis can be realized for each individual in a crowd, so that rapid and accurate prompt that the infectious disease epidemic situation may be affected is realized, and compared with a traditional scheme, the method and the system can effectively provide the accuracy degree of epidemic situation influence prompt.

Epidemic situation label data processing method and system

The invention discloses a traditional Chinese medicine diagnosis and treatment knowledge graph automatic construction method based on deep learning. The traditional Chinese medicine diagnosis and treatment knowledge graph automatic construction method comprises the steps: constructing an initialized literature medical record corpus, carrying out sentence segmentation and word segmentation on a medical record, and marking a theory-law-prescription-medicine entity in the medical record; predicting the entity through a bidirectional LSTM, and automatically extracting the entity from traditional Chinese medicine literature medical records through a deep learning model; and clustering similar entities appearing in the same medical record to form an entity group, then forming a triple accordingto a predefined relationship between entities, and constructing a knowledge graph. According to the invention, the relationship between traditional Chinese medicine diagnosis and treatment concepts ispredefined; construction of the knowledge graph is converted into a traditional Chinese medicine diagnosis and treatment named entity recognition task; and entities are automatically extracted from traditional Chinese medicine literature medical records through a deep learning model, and the entities are clustered to form an entity set, so that the many-to-many problem between traditional Chinesemedicine diagnosis and treatment concepts is solved, and the famous and old traditional Chinese medicine diagnosis and treatment thought in the medical records is completely displayed.

Traditional Chinese medicine diagnosis and treatment knowledge graph automatic construction method based on deep learning

The invention discloses an activity analysis method for hypertensive elder people based on multi-modal attention fusion. Multi-modal data features are fused, and attention representation of all modaldata is extracted based on an attention mechanism. Meanwhile, based on pulse waves and an electrocardio sensor, the method measures the real-time blood pressure of the hypertension elder people, calculates the variability of the blood pressure, analyzes the correlation between real-time activities and the blood pressure, and obtains daily activities suitable for the hypertension elder people. According to the analysis method, the attention mechanism is introduced based on a convolutional neural network model and a long-term and short-term memory neural network combination model, multi-modal feature extraction and fusion are realized, and the activity recognition precision can be effectively improved. Besides, the method analyzes the association between the real-time blood pressure and thedaily activities, and provides a basis for recommending reasonable daily activities for the hypertension elder people.

Activity analysis method for hypertensive elder people based on multi-modal attention fusion

The invention provides a traditional Chinese medicine planting resource named entity identification method. The method comprises the following steps: S1, acquiring literatures of the traditional Chinese medicine planting resources; S2, labeling the traditional Chinese medicinal material planting resource literature according to a certain rule, and splitting the labeled literature into text sentences; S3, respectively searching a word vector and a character vector corresponding to each text sentence one by one, and training a GRU-CRF model by utilizing the word vectors and the character vectors; and S4, carrying out named entity identification on unknown traditional Chinese medicinal material planting resource literatures by utilizing the trained GRU-CRF model. According to the traditionalChinese medicine planting resource named entity identification method, by constructing the GRU-CRF model, the named entity of the traditional Chinese medicine planting resource literature can be automatically recognized, the recognition accuracy and recognition efficiency are greatly improved, the manual recognition time expenditure is reduced, and the method can be expanded to other named entitycategories such as the soil mineral content and the moisture content.

Traditional Chinese medicine planting resource named entity identification method

The invention belongs to the technical field of three-dimensional gait analysis and discloses a gait classification and quantification method based on the Mahalanobis distance. The gait classificationand quantification method comprises the steps that by taking a normal joint angle as a reference template, an abnormal index dist of an i joint angle of a sample j is calculated, abnormal indexes of all the joint angles of samples are calculated, the abnormal indexes dist ...dist of all the joint angles of the sample j are combined into vectors to serve as abnormal index vectorsdist of the sample j, wherein dist =dist ...dist ; relevance among all the joint angles is learned to obtain a Mahalanobis matrix sigma, the Mahalanobis distance between the abnormal index vectors dist of the samples and a reference gait vector of normal people is calculated, and thus overall gait abnormal indexes of the samples are obtained; and the abnormal index vectors dist of the sample j serve as feature vectors of the samples, the Mahalanobis matrix sigma is combined, and an SVM classifier is used for classifying sample gaits.

Gait classification and quantification method based on Mahalanobis distance

The invention discloses a device and method for urine feature digital acquisition in Tibetan medicine. The device comprises a hollow cuboid case, wherein a first light source and a second light sourceare arranged at the edge parts of the tops of two parallel side surfaces of an inner cavity of the case; a third light source is arranged on the other side edge of the case; a first camera is arranged on the side edge parallel to the side edge provided with the third light source; a second camera, a third camera and a fourth camera in sequential arrangement are arranged at the top surface of theinner cavity of the case; a urine container is arranged in the middle position of the bottom of the inner cavity of the case. The method comprises the steps of putting the container filled with urineinto the right center of the bottom of the case; shooting the steam, color and foam features at the hot urine stage; shooting the precipitate and suspended matter features at the warm urine stage; shooting the stirred whirly state features at the cool urine stage; transmitting shot images/videos to a PC through a USB. The recognition efficiency can be improved; the urine feature recognition objectification can be realized.

Device and method for urine feature digital acquisition in Tibetan medicine

The invention discloses an ECG and inertial sensing data combined fall detection method and system. The method comprises the following steps: equipment comprising ECG and inertial sensing units is used for acquiring fall data and daily action data in the daily activity process of a human body; the collected fall data and daily action data are segmented according to a same time period, inertial sensing data characteristics and ECG data characteristics are extracted respectively, normalization is performed, and normalized sample data characteristics are obtained; the normalized sample data characteristics are subjected to dimension reduction by a principal component analysis algorithm to obtain dimension reduction data characteristics, the dimension reduction data characteristics are classified by a support vector machine algorithm, and a support vector machine classification model is obtained; and newly collected data are input into the support vector machine classification model, and fall detection results are obtained. Changes of the human fall process are reflected from two different dimensions of human physiological activity and posture, the defect of inaccurate data of a singletype is avoided, and accuracy of fall detection is improved.

Li Qiaoqin

Papers

Activity analysis method for hypertensive elder people based on multi-modal attention fusion

Traditional Chinese medicine planting resource named entity identification method

Gait classification and quantification method based on Mahalanobis distance

Device and method for urine feature digital acquisition in Tibetan medicine

ECG and inertial sensing data combined fall detection method and system