Abstract: A method includes receiving training data that includes a set of transcribed speech utterances where each respective transcribed speech utterance is paired with a corresponding transcription. For each respective transcribed speech utterance, the method includes generating an encoded audio representation and an encoded textual representation, generating a higher order audio feature representation for a corresponding encoded audio representation, generating a higher order textual feature representation for a corresponding encoded textual representation, and determining a loss for the respective transcribed speech utterance based on the higher order audio feature representation and the higher order textual feature representation. The method also includes training a speech encoder and a text encoder of a correction model based on the loss determined for each transcribed speech utterance of the set of transcribed speech utterances.

Abstract: The present invention discloses a blue luminescent host composition, an organic electroluminescent device comprising the blue luminescent host composition, and a display device. The blue luminescent host composition comprises at least a first host compound, a second host compound, and a third host compound, wherein the structure of the first host compound is as represented by formula I, the structure of the second host compound is as represented by formula II, and the structure of the third host compound is as represented by formula I or II. The blue luminescent host composition provided by the present invention has a melting point lower than that of each individual compound. The blue luminescent host compounds provided by the present invention are evaporated to form a host composition thin film under continuous working conditions and at different evaporation rates.

Abstract: A method includes obtaining a sequence of audio embeddings derived from speech features characterizing a spoken prompt. The method also includes, for each candidate biasing phrase in a candidate phrase corpus: obtaining a phrase embedding; obtaining a sequence of wordpiece embeddings, and generating, using a scoring function, a ranking score that indicates a relevance of the phrase embedding to the sequence of audio embeddings. Based on the ranking scores generated for the candidate biasing phrases in the candidate phrase corpus, the method includes identifying the top-K biasing phrases from the candidate phrase corpus and processing, using a biaser module, the sequence of audio embeddings and the sequences of wordpiece embeddings obtained for the top-K biasing phrases to generate a context vector. The method also includes generating, using a speech recognizer, a transcription of the spoken prompt based on the context vector and the speech features characterizing the spoken prompt.

Abstract: A method includes receiving training data that includes a set of transcribed speech utterances where each respective transcribed speech utterance is paired with a corresponding transcription. For each respective transcribed speech utterance, the method includes generating an encoded audio representation and an encoded textual representation, generating a higher order audio feature representation for a corresponding encoded audio representation, generating a higher order textual feature representation for a corresponding encoded textual representation, and determining a loss for the respective transcribed speech utterance based on the higher order audio feature representation and the higher order textual feature representation. The method also includes training a speech encoder and a text encoder of a correction model based on the loss determined for each transcribed speech utterance of the set of transcribed speech utterances.

Abstract: A method includes receiving, as input to a speech recognition model, audio data corresponding to a spoken utterance. The method also includes performing, using the speech recognition model, speech recognition on the audio data by, at each of a plurality of time steps, encoding, using an audio encoder, the audio data corresponding to the spoken utterance into a corresponding audio encoding, and decoding, using a speech recognition joint network, the corresponding audio encoding into a probability distribution over possible output labels. At each of the plurality of time steps, the method also includes determining, using an intended query (IQ) joint network configured to receive a label history representation associated with a sequence of non-blank symbols output by a final softmax layer, an intended query decision indicating whether or not the spoken utterance includes a query intended for a digital assistant.

Abstract: A method includes receiving, as input to a speech recognition model, audio data corresponding to a spoken utterance. The method also includes performing, using the speech recognition model, speech recognition on the audio data by, at each of a plurality of time steps, encoding, using an audio encoder, the audio data corresponding to the spoken utterance into a corresponding audio encoding, and decoding, using a speech recognition joint network, the corresponding audio encoding into a probability distribution over possible output labels. At each of the plurality of time steps, the method also includes determining, using an intended query (IQ) joint network configured to receive a label history representation associated with a sequence of non-blank symbols output by a final softmax layer, an intended query decision indicating whether or not the spoken utterance includes a query intended for a digital assistant.

Abstract: A method includes receiving training data that includes a set of transcribed speech utterances where each respective transcribed speech utterance is paired with a corresponding transcription. For each respective transcribed speech utterance, the method includes generating an encoded audio representation and an encoded textual representation, generating a higher order audio feature representation for a corresponding encoded audio representation, generating a higher order textual feature representation for a corresponding encoded textual representation, and determining a loss for the respective transcribed speech utterance based on the higher order audio feature representation and the higher order textual feature representation. The method also includes training a speech encoder and a text encoder of a correction model based on the loss determined for each transcribed speech utterance of the set of transcribed speech utterances.

Abstract: Example implementations relate to an inspection method for training a measurement machine to accurately measure side joint lengths and detecting a defect among a plurality of solder joints. The method includes receiving a first data representing the side joint lengths of the plurality of solder joints measured by a first measurement machine and a second data representing the side joint lengths measured by a second measurement machine. Further, the method includes determining a correlation value based on a statistical analysis of a relationship between the first data and the second data. The method further includes updating an algorithm used by the first measurement machine to measure the side joint lengths, based on the correlation value to reduce deviation between the first data and the second data. Later, the updated algorithm is used as a dimensional metrology in the first measurement machine for detecting the defect in the solder joints.

Abstract: A method and an apparatus for controlling a mobile robot, and a control system are provided. The method includes: in response to receiving a start request sent by a mobile robot, obtaining at least one of the position information or condition information of the mobile robot; determining a speed limit value of the mobile robot according to the obtained information; and sending a start instruction including the speed limit value to the mobile robot, so as to control the mobile robot to move at a speed smaller than or equal to the speed limit value. The implementation improves the motion safety of a mobile robot.

Abstract: Embodiments of the present disclosure disclose a method and an apparatus for pushing information, a method and an apparatus for controlling presentation of content. The method for pushing information includes: determining vehicle information of a communicatively connected autonomous vehicle, the vehicle information including autonomous vehicle driving route information; selecting to-be-played content matching the vehicle information from a pre-stored set of to-be-played contents according to the vehicle information, and storing the elected to-be-played content to the sub-set of to-be-played contents; and transmitting the sub-set of to-be-played contents to the autonomous vehicle so that the autonomous vehicle plays the to-be-played content in the sub-set of to-be-played contents. According to the embodiment, targeted to-be-played content can be sent to the autonomous vehicle.

Abstract: A method includes receiving, as input to a speech recognition model, audio data corresponding to a spoken utterance. The method also includes performing, using the speech recognition model, speech recognition on the audio data by, at each of a plurality of time steps, encoding, using an audio encoder, the audio data corresponding to the spoken utterance into a corresponding audio encoding, and decoding, using a speech recognition joint network, the corresponding audio encoding into a probability distribution over possible output labels. At each of the plurality of time steps, the method also includes determining, using an intended query (IQ) joint network configured to receive a label history representation associated with a sequence of non-blank symbols output by a final softmax layer, an intended query decision indicating whether or not the spoken utterance includes a query intended for a digital assistant.

Abstract: Example implementations relate to an inspection method for training a measurement machine to accurately measure side joint lengths and detecting a defect among a plurality of solder joints. The method includes receiving a first data representing the side joint lengths of the plurality of solder joints measured by a first measurement machine and a second data representing the side joint lengths measured by a second measurement machine. Further, the method includes determining a correlation value based on a statistical analysis of a relationship between the first data and the second data. The method further includes updating an algorithm used by the first measurement machine to measure the side joint lengths, based on the correlation value to reduce deviation between the first data and the second data. Later, the updated algorithm is used as a dimensional metrology in the first measurement machine for detecting the defect in the solder joints.

Abstract: A method includes receiving audio data encoding an utterance, processing, using a speech recognition model, the audio data to generate speech recognition scores for speech elements, and determining context scores for the speech elements based on context data indicating a context for the utterance. The method also includes executing, using the speech recognition scores and the context scores, a beam search decoding process to determine one or more candidate transcriptions for the utterance. The method also includes selecting a transcription for the utterance from the one or more candidate transcriptions.

Abstract: A method and an apparatus for controlling an autonomous driving vehicle are provided. The method includes: receiving environment information sent by an autonomous driving vehicle, the environment information including vehicle exterior environment information; determining whether the autonomous driving vehicle is in an abnormal operation status, based on the vehicle exterior environment information and operation information of an operation executed by the autonomous driving vehicle; and sending a braking control instruction and a data acquisition instruction to the autonomous driving vehicle, in response to determining that the autonomous driving vehicle is in the abnormal operation status, the braking control instruction being used for controlling braking of the autonomous driving vehicle, and the data acquisition instruction being used for acquiring data of a driving recorder in the autonomous driving vehicle.

Abstract: Disclosed are a pulse magnet device based on magnetic flux compression, and a high-flux measurement method. The device includes a diamagnetic block, reinforcing plates, screw rods and a magnet coil. The diamagnetic block and the magnet coil are concentrically arranged in the axial direction; the reinforcing plates are arranged at ends of the magnet coil and the diamagnetic block and are connected by means of the screw rods. The diamagnetic block is used for inducing the induction current opposite the coil current during the discharge of the magnet coil, and for compressing the magnetic field to the area between the diamagnetic block and the magnet coil. The intensity and uniformity of the magnetic field around the magnet coil are improved by means of increasing the magnetic flux density.

Abstract: Electromagnetic manufacturing method and forming device of mesoscale plate are provided. The manufacturing method includes: oppositely and parallelly disposing a first workpiece to be formed on top of a mold, side-press restraining two ends of the first workpiece, and disposing a deceleration block on two sides of the mold; controlling the first workpiece to tend toward the mold and to be deformed under the drive of uniform electromagnetic force; and colliding a middle area of the first workpiece firstly with the mold under the drive of uniform electromagnetic force, and driving the speed of the middle area of the first workpiece to decelerate to zero. When an area close to the two ends collides with the deceleration block and until the speed of all areas of first workpiece decelerates to zero, forming is completed. Shaping is tending further toward the mold through electromagnetic force until completely fitted to the mold.

Abstract: A method and apparatus for controlling an unmanned vehicle to perform route verification are disclosed. A method may include: receiving to-be-verified route information sent by a cloud server in communication connection, the to-be-verified route information including road information and expected traveling state information corresponding to a road section on a road characterized by the road information, and the expected traveling state information being used to characterize an expected traveling state of the unmanned vehicle on the corresponding road section; and performing following verification according to a set period during driving of the unmanned vehicle along the road: detecting a traveling state of the unmanned vehicle; and determining a verification result of the to-be-verified route information, based on a matching relationship between the detected traveling state and an expected traveling state corresponding to a road section of a current location in the to-be-verified route information.