7results about How to "Highlight the importance" patented technology

The invention discloses a text processing method and device, electronic equipment and a storage medium, and belongs to the technical field of natural language process.The method comprises the steps that sequence features and word features of a to-be-processed text segment after the to-be-processed text segment is input into a BERT model are obtained, position codes of the to-be-processed text segment are obtained, the word features and the position codes are fused to obtain fused features, and the fused features are stored in a storage medium; the method comprises the following steps: carrying out operation on a fusion feature through a multi-layer multi-head attention mechanism to obtain an attention distribution feature, processing the attention distribution feature through an activation function to obtain a microscopic attention weight, optimizing the fusion feature based on the attention distribution feature and the microscopic attention weight, and further obtaining a sequence feature based on the sequence feature and the optimized fusion feature. According to the method, the relation between the entities in the to-be-processed text segment is determined, and due to the fact that the optimized fusion feature can better express the semantic feature of the to-be-processed text, the effect of extracting the relation between the entities from the complex to-be-processed text segment can be improved.

The invention provides a session recommendation method based on multi-intention perception and dual gating. The method comprises the following steps: firstly, capturing the position information of sessions to distinguish the sequence, and representing the importance of the sessions by frequency information; subsequently, session embedding, position embedding and frequency embedding are input into a double-gating mechanism layer, user embedding and user behavior embedding are combined through a double-gating mechanism, and importance of user behaviors is highlighted. Then, a bidirectional self-attention layer is used for self-attention generation of a plurality of contextual representations of session information related to each item, user intention embedding is emphasized through a high-speed network, the context is prevented from being excessively similar, and different intentions of users in sessions are extracted to a certain extent. However, not all session items are necessary, and some session items may be noisy, we calculate the importance of a plurality of representations, and extract important user intentions. And finally, performing maximum pooling and average pooling aggregation project distribution to obtain final output. According to the method, the problem that noise data generated by random behaviors of the user misguides final prediction of the model can be effectively solved, and various intentions of the user can be extracted from the session.

The application discloses a kind of visual-based photovoltaic module statistics method and system, it is related to computer vision technical field.In the method, the distance image set photographed by unmanned aerial vehicle is acquired;Distance image set includes n distance images collected at different distances, wherein n is greater than or equal to 2 positive integer;Distance image set is input to the preset image recognition algorithm, and multiple recognition results are obtained;The number of photovoltaic modules identified in the distance image collected at different distances is the recognition result;Difference value calculation is carried out on multiple recognition results two by two, and multiple difference value calculation results are obtained;When each difference value calculation result is less than or equal to preset threshold value, weighted average calculation is carried out on multiple recognition results, and final statistical result is obtained.The technical scheme provided in the application can effectively improve the efficiency and accuracy of statistics when the number of photovoltaic modules is counted.

The present invention relates to the field of biological data analysis technology, and in particular to a method, system and device for predicting the pathogenicity of amino acid variations. Sequence information is input into a conservative information encoding model, conservative information of the amino acid to be tested is extracted, and multiple sequence comparison information is obtained. The sequence features of the encoded sequence information and the multiple sequence comparison features of the multiple sequence comparison information are extracted through a bilateral cross-linked recurrent neural network to obtain the initial variation information of the amino acid to be tested, and the property-related index of the amino acid to be tested before the variation and the property-related index of the amino acid to be tested after the variation are obtained using the AAIndex database, respectively, and mapped to a vector space as variation information, and the variation features of the variation information are extracted. The multiple sequence comparison features, sequence features and variation features are used to obtain the amino acid variation pathogenicity prediction results, thereby improving the quality of multiple biological information and making the prediction of amino acid variation pathogenicity more accurate.

The present invention discloses an automatic calibration equipment operation status supervision system and method based on data analysis, which belongs to the technical field of equipment information management. The present invention collects data information of the product to be calibrated by entering the data information of the standard product; calculates the error value of the product to be calibrated according to the data information; considers multiple important factors in the process of collecting data information, and distributes weights to each branch of production data information and morphological data information, so as to more effectively highlight the importance of individual factors, and can more intuitively calculate the error value of the product to be calibrated; judges abnormal situations according to the error value calculation results, and alarms and prompts relevant management personnel according to the specific abnormal situations; relevant technical personnel can quickly understand the corresponding data information of the product to be calibrated and the specific problems that occur, thereby improving the efficiency of automatic calibration equipment operation status supervision and the work efficiency of relevant technical personnel.

The invention relates to the field of traffic potential hazard identification, and discloses a method for evaluating a driver's potential hazard perception ability, comprising the following steps: S1, collecting eye movement data of a driver in different potential risk scenarios through a driving simulation experiment, and converting the eye movement data into specific evaluation parameter values; S2, screening out evaluation parameters with significant differences as key evaluation parameters according to the characteristics of each risk scenario and the differences in evaluation parameters among different groups of people in the risk scenario; S3, using an entropy weight method to perform weight calculation on the key evaluation parameters in each risk scenario; S4, substituting the values ​​of the driver's key evaluation parameters into a TOPSIS model, and calculating the degree of proximity between the evaluation results of the driver in each risk scenario and positive and negative ideal solutions, so as to obtain the potential hazard perception ability score of the driver in each risk scenario, so as to solve the technical problem of complex quantization models when evaluating the potential hazard perception ability in the prior art.

This invention relates to the field of conductive materials, and more specifically, to an MXene-Ti3C2T x The preparation method, composite material, and application of the conductive agent include the following steps: Ti3AlC2 MAX phase ceramic powder is added to an etching solution and reacted at 0–60°C for 5–48 hours. The mixture is washed with water until neutral, centrifuged, and a slurry is obtained. A polar organic solvent is added to the slurry, and the mixture is ultrasonically treated to obtain a suspension. The suspension is freeze-dried to obtain the conductive agent. This conductive agent utilizes metallic conductive two-dimensional MXene to directly construct a conductive framework, combining the functions of conductive support and interface regulation. The MXene sheets not only provide electron transport but also improve the interfacial compatibility between the silicon anode and the solid electrolyte through their flexibility and surface functional groups, suppressing interfacial cracking and side reactions.