Abstract: A computer-implemented method for fusing an end-to-end speech recognition model with an external language model (ExternalLM) is provided. The method includes obtaining an output of the end-to-end speech recognition model. The output is a probability distribution. The method further includes transforming, by a hardware processor, the probability distribution into a transformed probability distribution to relax a sharpness of the probability distribution. The method also includes fusing the transformed probability distribution and a probability distribution of the ExternalLM for decoding speech.

Abstract: A piezoelectric device includes a connection section including a first coupling portion, a second coupling portion, and a bridging portion. The first coupling portion extends along a slit and is connected to one of a pair of beam sections. The second coupling portion extends along the slit and is connected to another of the pair of beam sections. The bridging portion is located between the slit and an opening and is connected to both of the first coupling portion and the second coupling portion. The beam sections are connected to each other in a circumferential direction of a base having an annular shape via the connecting section while each of the beam sections is interposed between the slits extending in intersecting directions.

Abstract: Fusion of neural networks is performed by obtaining a first neural network and a second neural network. The first and the second neural networks are the result of a parent neural network subjected to different training. A similarity score is calculated of a first component of the first neural network and a corresponding second component of the second neural network. An interpolation weight is determined for the first and the second components by using the similarity score. A neural network parameter of the first component is updated based on the interpolation weight and a corresponding neural network parameter of the second component to obtain a fused neural network.

Abstract: A computer-implemented method of detecting a portion of audio data to be removed is provided. The method includes obtaining a recognition result of audio data. The recognition result includes recognized text data and time stamps. The method also includes extracting one or more candidate phrases from the recognition result using n-gram counts. The method further includes, for each candidate phrase, making pairs of same phrases with different time stamps and clustering the pairs of the same phrase by using differences in time stamps. The method includes further determining a portion of the audio data to be removed using results of the clustering.

Abstract: An approach for improving speech recognition is provided. A processor receives a new word to add to a prefix tree. A processor determines a bonus score for a first transition from a first node to a second node in a prefix tree on condition that the first transition is included in a path of at least one transition representing the new word. A processor determines a hypothesis score for a hypothesis that corresponds to a speech sequence based on the prefix tree, where the hypothesis score adds the bonus score to an initial hypothesis score to determine the hypothesis score. In response to a determination that the hypothesis score exceeds a threshold value, a processor generates an output text sequence for the speech sequence based on the hypothesis.

Abstract: An ultrasonic transducer includes first and second acoustic transducers and a housing with a bottomed tubular shape. The second acoustic transducer includes an annular section supporting a second membrane section and contacting an entire periphery of the second membrane section, and an acoustic matching plate facing the second membrane section and spaced apart from the second membrane section. The acoustic matching plate is connected to a surrounding wall portion defining a sealed space with the housing. An ultrasonic transmission path sandwiched between the first membrane section and the second membrane section is provided in the sealed space. A maximum inner width of the ultrasonic transmission path is smaller than a maximum inner width of the surrounding wall portion.

Abstract: A one-component type polyurethane prepolymer composition comprises a reaction product formed through a reaction between reactants comprising (a) at least one polyisocyanate, and (b) a polyol blend comprising at least one bifunctional polyether polyol, wherein the bifunctional polyether polyol is a homopolymer of propylene oxide, homopolymer of butylene oxide, or copolymer of alkylene oxide, and has a number average molecular weight from 3000 g/mol to 9000 g/mol, and at least one trifunctional polyether polyol, wherein the trifunctional polyether polyol is a copolymer of alkylene oxide and end-capped with 10 wt % to 28 wt %, by the total weight of the trifunctional polyether polyol, of ethylene oxide, and has a number average molecular weight from 5000 g/mol to 8000 g/mol, wherein the bifunctional polyether polyol and the trifunctional polyether polyol are present in a parts by weight ratio from 4:1 to 2.

Abstract: A method, computer system, and a computer program product for audio data augmentation are provided. Sets of audio data from different sources may be obtained. A respective normalization factor for at least two sources of the different sources may be calculated. The normalization factors from the at least two sources may be mixed to determine a mixed normalization factor. A first set of the sets may be normalized by using the mixed normalization factor and to obtain training data for training an acoustic model.

Abstract: A system and method for expanding a question and answer (Q&A) database. The method includes obtaining a set of Q&A documents and speech recognition results, each Q&A document in the set having an identifier, and each speech recognition result having an identifier common with the identifier of a relevant Q&A document, and adding one or more repetition parts extracted from the speech recognition results to a corresponding Q&A document in the set to generate an expanded set of Q&A documents for increasing Q&A document extraction accuracy.

Abstract: An electronic component unit includes: an electronic module in which a rear substrate is electrically connected via an electric cable to an electronic element; an external connection terminal that is electrically connected to an external circuit; a relay substrate including a terminal connection electrode to which the external connection terminal is electrically connected either directly or via a connection conductor; and a relay connector on the relay substrate. The electronic element is any one of: an imaging element; a laser element; and a sensor element.

Abstract: A piezoelectric device includes a connecting section connecting a pair of beam sections adjacent to each other. The connecting section is connected to one of the pair of beam sections at a first end portion. The connecting section is connected to another of the pair of beam sections at a second end portion. The second end portion faces the first end portion in a direction in which the pair of beam sections are aligned. A second coupling portion is located along a first coupling portion. The connecting section includes only one first end portion. The connecting section includes only one second end portion. Each of the first end portion and the second end portion is closer to a tip end portion than to a fixed end portion of each of the pair of beam sections.

Abstract: Methods and systems for information retrieval include analyzing audio data to produce word hypotheses. Displaying the word hypotheses in motion at different respective speeds at once across a graphical display. Information is retrieved in accordance with one or more selected terms from the displayed word hypotheses.

Abstract: In the multi-layer coil component, the hole portion is provided in the vicinity of the side surface of the element body, and the distance from the tip position of the external electrode to the hole portion is shorter than the distance from the tip position of the external electrode to the coil. Therefore, when a crack is generated starting from the tip position of the external electrode on the mounting surface, the crack progress toward the hole portion located at a distance closer than the coil. Therefore, in the multi-layer coil component, splitting of the coil due to the crack can be prevented.

Abstract: A transducer includes a base, beams, and a coupler. The beams each include a piezoelectric layer, a first electrode layer, and a second electrode layer. The coupler is fitted in slits between adjacent beams to define a connection between the beams. The coupler extends from an upper portion of the base into each of the slits without a break. A Young's modulus of the material of the coupler is lower than a Young's modulus of the material of the piezoelectric layer. A maximum thickness of the coupler in the upper portion of the base in the direction of the central axis of the base is greater than a thickness of each of the beams.

Abstract: A computer-implemented method includes obtaining, using a hardware processor, training data including utterances of speakers and performing tasks to train a machine learning model that converts an utterance into a feature vector, each task using one subset of multiple subsets of training data. The subsets of training data include a first subset of training data including utterances of a first number of speakers and at least one second subset of training data. Each second subset of training data includes utterances of a number of speakers that is less than the first number of speakers.

Abstract: A vehicle control device including a communication section, a first microcomputer that can communicate with an exterior via the communication section and a second microcomputer that cannot communicate directly with the exterior and can communicate with the first microcomputer. The communication section receives a control signal from the exterior, based on the control signal received by the communication section, the first microcomputer controls operation of an object of control, and outputs the control signal to the second microcomputer. In a case in which a state of the object of control that corresponds to the control signal and a state exhibited by the object of control differ, the second microcomputer effects control so as to stop operation of the communication section.

Abstract: A piezoelectric element includes a second electrode layer on a second surface of a single-crystal piezoelectric layer. A hole continuous with a through-hole is provided in the second electrode layer. The second electrode layer is made of Pt, Ti, Al, Cu, Au, Ag, Mg, or an alloy including at least one of the metals as a main ingredient. A third electrode layer is on one side of the second electrode layer opposite to the single-crystal piezoelectric layer. The third electrode layer includes at least a portion outside of an edge of the hole with a distance maintained relative to the edge of the hole when viewed in a direction perpendicular or substantially perpendicular to the second surface. The third electrode layer is made of Ni or an alloy including Ni as a main ingredient.

Abstract: A computer-implemented method is provided for generating a language model for an application. The method includes estimating interpolation weights of each of a plurality of language models according to an Expectation Maximization (EM) algorithm based on a first metric. The method further includes classifying the plurality of language models into two or more sets based on characteristics of the two or more sets. The method also includes estimating a hyper interpolation weight for the two or more sets based on a second metric specific to the application. The method additionally includes interpolating the plurality of language models using the interpolation weights and the hyper interpolation weight to generate a final language model.

Abstract: A piezoelectric device includes a beam portion with a fixed end portion and a free end portion opposite to the fixed end portion. The beam portion extends from the fixed end portion towards the free end portion. The beam portion includes a multilayer body including a piezoelectric layer and first and second electrode layers. A base is connected with the fixed end portion of the beam portion. As viewed from a layering direction of the multilayer body, the base surrounds the beam portion at an interval from the beam portion except for the fixed end portion. As viewed from the layering direction, an average of a dimension of a width of the beam portion in an orthogonal or substantially orthogonal direction to an extension direction of the beam portion is greater than a maximal dimension of a length thereof in the extension direction.

Abstract: A sprayable water blown low density polyurethane foam-forming formulation including: (A) an isocyanate component; (B) a polyol component, wherein the polyol component includes a surfactant composition; wherein the surfactant composition includes at least one phenol-free secondary alkyl alcohol ethoxylate, and wherein the at least one phenol-free secondary alkyl alcohol ethoxylate is at least one linear alkyl chain phenol-free secondary alcohol ethoxylate; and (C) water; a process for making the above sprayable water blown low density polyurethane foam-forming formulation; and a polyurethane foam made from the above sprayable water blown low density polyurethane foam-forming formulation.