Abstract: This document relates to natural language processing using a framework such as a neural network. One example method involves obtaining a first document and a second document and propagating attention from the first document to the second document. The example method also involves producing contextualized semantic representations of individual words in the second document based at least on the propagating. The contextualized semantic representations can provide a basis for performing one or more natural language processing operations.

Abstract: A multi-phase voltage converter can include: a control chip configured to generate N pulse distribution signals, where N is a positive integer greater than 1; a power conversion module comprising N first-level conversion modules; where the first-level conversion module comprises at least one second-level conversion module, and the second-level conversion module comprises at least one third-level conversion module; where when the second-level conversion module comprises multiple third-level conversion modules, the multiple third-level conversion modules are coupled in parallel with each other; and where the first-level conversion module receives a corresponding one of the N pulse distribution signals, the second-level conversion module receives a first phase distribution signal generated based on the corresponding pulse distribution signal, and the third-level conversion module receives a second phase distribution signal generated based on the first phase distribution signal.

Abstract: The present invention provides a display panel and a mobile terminal, including a panel layer including, but not limited to, a first portion, a second portion, and a curved third portion connected therebetween; a protective layer on the third portion; a first support layer on a side of the panel layer away from the protective layer and disposed opposite to the first portion; and a cover layer on the first portion and including an edge as an ink layer; the protective layer has a minimum thickness on a side close to the first portion and between a boundary of the first portion and an inner boundary of the ink layer.

Abstract: Presented herein are techniques for identifying and/or validating sequence variants in genomic sequence data. The techniques include generating an error rate reflective of sequence errors present in the genomic sequence data. The error rate may be used to validate potential sequence variants. The error rate may be based on errors identified during consensus sequence confirmation for sequence reads associated with individual unique molecular identifiers.

Abstract: A wearable device includes a ground element, a first radiation element, a second radiation element, a third radiation element, a fourth radiation element, a fifth radiation element, and a sixth radiation element. The first radiation element has a feeding point. The first radiation element is coupled to the ground element. The third radiation element is coupled through the second radiation element to the first radiation element. The fourth radiation element is coupled through the second radiation element to the first radiation element. The fifth radiation element is coupled to the first radiation element. The sixth radiation element is coupled to the ground element. The sixth radiation element is adjacent to the fourth radiation element. An antenna structure is formed by the first radiation element, the second radiation element, the third radiation element, the fourth radiation element, the fifth radiation element, and the sixth radiation element.

Abstract: Provided are a power supply conversion device and a charging control method. The device includes: a transformer; a first rectifier circuit connected to a primary winding of the transformer, for converting a received alternating current into a first direct current, a voltage value of the first direct current being a first direct-current voltage, and the transformer is used for converting the first direct-current voltage into a second direct-current voltage; a voltage converter connected to a secondary winding of the transformer, for converting the second direct-current voltage to output a constant direct-current voltage or a pulsating direct-current voltage; and a controller connected to the first rectifier circuit and the voltage converter, for controlling the voltage converter to selectably output the constant direct-current voltage or the pulsating direct-current voltage according to a desired charging mode of a device to be charged.

Abstract: Methods and systems are provided for determining a variant of interest by analyzing sizes and sequences of cfDNA fragments obtained from a test sample. The methods and systems provided herein implement processes that synergistically combine size and sequence information, thereby improving specificity and sensitivity of assays over conventional methods.

Abstract: A driving circuit and a driving method are provided. The driving circuit includes an energy-storage capacitor, a charging circuit and a discharging circuit. The energy-storage capacitor is coupled between a piezoelectric load and the charging circuit. Operation states of the charging circuit and the discharging circuit are controlled, so that the charging circuit charges the energy-storage capacitor during a first operation interval of an operation period, to adjust a power supply voltage signal provided to the piezoelectric load to change with a reference voltage in a first interval, and at least the piezoelectric load discharges electricity through the discharging circuit during a second operation interval of the operation period, to adjust the power supply voltage signal to change with the reference voltage in a second interval. The driving circuit according to the present disclosure requires a few switches, thereby facilitating circuit integration.

Abstract: A method includes: generating fault alarm information, and sending the fault alarm information through a serializer/deserializer (SerDes) bus. The fault alarm information indicates a fault that occurs in a first apparatus and/or a target communication link, and the target communication link is a link for communication between the first apparatus and a second apparatus.

Abstract: The disclosed embodiments concern methods, systems and computer program products for determining sequences of interest using unique molecular indexes (UMIs) that are uniquely associable with individual polynucleotide fragments, including sequences with low allele frequencies or long sequence length. In some implementations, the UMIs include variable-length nonrandom UMIs (vNRUMIs). Methods and systems for making and using sequencing adapters comprising vNRUMIs are also provided.

Abstract: A power converter can include: a plurality of circuit modules coupled in parallel between a first port and a second port, where each of the plurality of circuit modules includes a switching power stage circuit having a first magnetic element coupled between a switch node of the switching power stage circuit and a first terminal of the second port, at least one switch group having first and second transistors and being coupled between a first terminal of the first port and a first terminal of the switching power stage circuit, and at least one first energy storage capacitor for providing energy to a load of the power converter; and a plurality of second energy storage capacitors configured to periodically store energy and release energy to corresponding first energy storage capacitors.

Abstract: A dual-stack air-cooled fuel cell includes two stack groups, a fan assembly, an air supply channel, a spoiler and a temperature control device. The fan assembly is centrally located in the air-cooled fuel cell, and is configured to generate an air flow field in the air-cooled fuel cell. The two stack groups are located at an air inlet side and an air outlet side of the fan assembly, respectively. The spoiler is arranged between the fan assembly and the two stack groups. The above components are connected through the air supply channel. The temperature control device is configured to detect surface temperature of the stack groups to adjust the rotation direction of the fan assembly, so as to alternately perform suction and blowing operation for the two stack groups to reach the heat equilibrium of the fuel cell.

Abstract: Presented herein are techniques for identifying and/or validating sequence variants in genomic sequence data. The techniques include generating an error rate reflective of sequence errors present in the genomic sequence data. The error rate may be used to validate potential sequence variants. The error rate may be based on errors identified during consensus sequence confirmation for sequence reads associated with individual unique molecular identifiers.

Abstract: A water-pollution environmental-protection verification method based on power grid and tax data fusion includes: acquiring power data and tax data corresponding to a plurality of enterprise identities in a target region; acquiring a pollutant output and a sewage purification consumable consumption corresponding to each enterprise identity based on the tax data; determining a first candidate enterprise identity set based on the pollutant output; determining a second candidate enterprise identity set based on the sewage purification consumable consumption and a sewage purification consumable consumption of an actual unit yield of each enterprise identity; acquiring production information of the plurality of enterprise identities based on the power data, and determining a third candidate enterprise identity set; and processing based on the first candidate enterprise identity set, the second candidate enterprise identity set and the third candidate enterprise identity set to determine a target enterprise identity

Abstract: An enterprise screening method and apparatus, an electronic device and a storage medium are provided. The method includes: acquiring original enterprise data belonging to a preset industry category, and extracting enterprise business behavior information corresponding to each enterprise from a business scope of each enterprise in the original enterprise data; extracting first target business behavior information from the business scope of the confirmed target enterprise data belonging to the preset industry category, the enterprise business behavior information and the first target business behavior information both including business mode information and business object information; and matching the enterprise business behavior information corresponding to each enterprise with the first target business behavior information respectively, and when the matching is successful, determining that the enterprise is a first target enterprise. An accuracy of enterprise screening can be improved.

Abstract: A method and apparatus for screening enterprises in Yangtze River Basin, an electronic device and a storage medium are provided. The method includes: acquiring original enterprise data belonging to a preset industry category, and comparing the original enterprise data with screened local enterprise data to obtain common enterprise data of the original enterprise data and the local enterprise data; extracting a first text feature from a business scope of the common enterprise data, and extracting a second text feature from a business scope of each enterprise in the original enterprise data; and performing feature matching on the second text feature corresponding to each enterprise and the first text feature respectively, and when a matching result meets a preset condition, determining that the enterprise is a first target enterprise. An accuracy of enterprise screening can be improved.

Abstract: An enterprise activation degree determining method and apparatus, an electronic device and a storage medium are provided. The method includes: acquiring original activation degree index data in P dimensions corresponding to N enterprises respectively, and performing dimensionless processing on the original activation degree index data to obtain target activation degree index data in P dimensions; calculating a correlation coefficient of target activation degree index data in every two dimensions to obtain a correlation coefficient matrix, and determining feature values and feature vectors of the correlation coefficient matrix; determining M principal components and accumulated contribution rates respectively corresponding to the M principal components based on the feature values and the feature vectors; calculating weights respectively corresponding to the target activation degree index data in the P dimensions; and determining the activation degree of each enterprise.

Abstract: A pollution emission determination method and apparatus based on a digital watershed space-time model is provided. In the method, a to-be-detected river watershed includes a plurality of sections, each section is provided with a monitoring station, a water quality sensor is mounted on each monitoring station, and the water quality sensor is used for acquiring water quality data. The method includes: acquiring currently monitored water quality data of a current monitoring station; acquiring actual water quality data of a previous monitoring station, calculating the actual water quality data through a calculated one-dimensional steady-state river model, and acquiring theoretical water quality data of the current monitoring station; and determining whether pollution emission occurs between the current monitoring station and the previous monitoring station according to a comparison result of the currently monitored water quality data and the theoretical water quality data.

Abstract: The disclosure provides a method for estimating the nondirectional wave spectrum from the sea echoes of multiple HF radar frequencies. The method includes: dividing the radar detection area into a plurality of fan-shaped units at an equal range interval and angle interval according to the distance resolution and the angular resolution of an HF radar; obtaining the Doppler spectrum from the sea echo of a single radar frequency at a fan-shaped unit by performing the first fast Fourier transform (FFT) in distance dimension, the second FFT in Doppler frequency dimension and the digital beamforming; extracting the positive first-order peak and the negative first-order peak from the aforementioned Doppler spectrum by the peak-searching method; and selecting the stronger first-order peak ?R(1)(?); dividing the second-order spectrum on the stronger first-order peak side into an inner second-order spectrum and an outer second-order spectrum.