Abstract: A method includes receiving a training set of utterances for training a machine-learning model to identify one or more intents for one or more utterances, and augmenting the training set of utterances with out-of-domain (OOD) examples. The augmenting includes: generating a data set of OOD examples, filtering out OOD examples from the data set of OOD examples, determining a difficulty value for each OOD example remaining within the filtered data set of the OOD examples, and generating augmented batches of utterances including utterances from the training set of utterances and utterances from the filtered data set of the OOD based on the difficulty value for each OOD. Thereafter, the machine-learning model is trained using the augmented batches of utterances in accordance with a curriculum training protocol.

Abstract: A multi-channel visible light communication system includes a carrier board and a plurality of light-emitting device stacks. The light-emitting device stacks are arranged over the carrier board. Each of the light-emitting device stacks includes a first light-emitting device, a second light-emitting device and a first adhesive layer. The second light-emitting device is disposed over the first light-emitting device. The first adhesive layer is disposed between the first light-emitting device and the second light-emitting device. The first adhesive layer includes a first metasurface. The first metasurface is configured to pass a first color light emitted from the first light-emitting device and reflect a second color light emitted from the second light-emitting device.

Abstract: Techniques for using logit values for classifying utterances and messages input to chatbot systems in natural language processing. A method can include a chatbot system receiving an utterance generated by a user interacting with the chatbot system. The chatbot system can input the utterance into a machine-learning model including a set of binary classifiers. Each binary classifier of the set of binary classifiers can be associated with a modified logit function. The method can also include the machine-learning model using the modified logit function to generate a set of distance-based logit values for the utterance. The method can also include the machine-learning model applying an enhanced activation function to the set of distance-based logit values to generate a predicted output. The method can also include the chatbot system classifying, based on the predicted output, the utterance as being associated with the particular class.

Abstract: A light-emitting device array includes a first light-emitting device, a second light-emitting device, and a third light-emitting device. A first beam shaping structure of the first light-emitting device is configured to convert light emitted by a first light-emitting structure of first light-emitting device into first structured light. A second beam shaping structure of the second light-emitting device is configured to convert light emitted by a second light-emitting structure of second light-emitting device into second structured light. Speckle patterns and spatial distributions of the first structured light and the second structured light on a projection plane are the same. A third beam shaping structure of the third light-emitting device is configured to convert light emitted by a third light-emitting structure of third light-emitting device into third structured light.

Abstract: Techniques are disclosed herein for improving the performance of an end-to-end (E2E) Automatic Speech Recognition (ASR) model in a target domain. A set of test examples are generated. The set of test examples comprise multiple subsets of test examples and each subset of test examples corresponds to a particular test category. A machine language model is then used to convert audio samples of the subset of test examples to text transcripts. A word error rate is determined for the subset of test examples. A test category is then selected based on the word error rates and a set of training examples is generated for training the ASR model in a particular target domain from a selected subset of test examples The training examples are used to fine-tune the model in the target domain. The trained model is then deployed in a cloud infrastructure of a cloud service provider.

Abstract: The present disclosure relates to chatbot systems, and more particularly, to batching techniques for handling unbalanced training data when training a model such that bias is removed from the trained machine learning model when performing inference. In an embodiment, a plurality of raw utterances is obtained. A bias eliminating distribution is determined and a subset of the plurality of raw utterances is batched according to the bias-reducing distribution. The resulting unbiased training data may be input into a prediction model for training the prediction model. The trained prediction model may be obtained and utilized to predict unbiased results from new inputs received by the trained prediction model.

Abstract: Techniques for using logit values for classifying utterances and messages input to chatbot systems in natural language processing. A method can include a chatbot system receiving an utterance generated by a user interacting with the chatbot system. The chatbot system can input the utterance into a machine-learning model including a set of binary classifiers. Each binary classifier of the set of binary classifiers can be associated with a modified logit function. The method can also include the machine-learning model using the modified logit function to generate a set of distance-based logit values for the utterance. The method can also include the machine-learning model applying an enhanced activation function to the set of distance-based logit values to generate a predicted output. The method can also include the chatbot system classifying, based on the predicted output, the utterance as being associated with the particular class.

Abstract: A pose calculating apparatus and method are provided. The pose calculating apparatus receives a plurality of real-time images and a plurality of inertial measurement parameters corresponding to at least one inertial sensor worn by a user. The pose calculating apparatus determines a pose calculating mode corresponding to each of a plurality of body regions of the user based on the real-time images and the inertial measurement parameters, wherein the pose calculating mode corresponds to a static mode or a motion mode. The pose calculating apparatus calculates a pose corresponding to each of the body regions based on the pose calculating mode corresponding to each of the body regions.

Abstract: A light-emitting device array includes a first light-emitting device, a second light-emitting device, and a third light-emitting device. A first beam shaping structure of the first light-emitting device is configured to convert light emitted by a first light-emitting structure of first light-emitting device into first structured light. A second beam shaping structure of the second light-emitting device is configured to convert light emitted by a second light-emitting structure of second light-emitting device into second structured light. Speckle patterns and spatial distributions of the first structured light and the second structured light on a projection plane are the same. A third beam shaping structure of the third light-emitting device is configured to convert light emitted by a third light-emitting structure of third light-emitting device into third structured light.

Abstract: An optical detection system includes a first optical transmitter module, a first optical receiver module, a second optical receiver module, and a second optical transmitter module. The first optical transmitter module is configured to emit detection light. A first light transmitter of the first optical transmitter module includes a photonic crystal surface emitting laser. A first metasurface of the first optical transmitter module is located above the first light transmitter. The second optical receiver module is configured to receive first communication light. The second communication light has information of signal light and information in first communication light. A second light transmitter of the second optical transmitter module includes a photonic crystal surface emitting laser. A second metasurface of the second optical transmitter module is located above the second light transmitter.

Abstract: A method includes receiving a training set of utterances for training a machine-learning model to identify one or more intents for one or more utterances, and augmenting the training set of utterances with out-of-domain (OOD) examples. The augmenting includes: generating a data set of OOD examples, filtering out OOD examples from the data set of OOD examples, determining a difficulty value for each OOD example remaining within the filtered data set of the OOD examples, and generating augmented batches of utterances including utterances from the training set of utterances and utterances from the filtered data set of the OOD based on the difficulty value for each OOD. Thereafter, the machine-learning model is trained using the augmented batches of utterances in accordance with a curriculum training protocol.

Abstract: Techniques for using enhanced logit values for classifying utterances and messages input to chatbot systems in natural language processing. A method can include a chatbot system receiving an utterance generated by a user interacting with the chatbot system and inputting the utterance into a machine-learning model including a series of network layers. A final network layer of the series of network layers can include a logit function. The machine-learning model can map a first probability for a resolvable class to a first logit value using the logit function. The machine-learning model can map a second probability for a unresolvable class to an enhanced logit value. The method can also include the chatbot system classifying the utterance as the resolvable class or the unresolvable class based on the first logit value and the enhanced logit value.

Abstract: Techniques for out-of-domain data augmentation for training chatbot systems in natural language processing. In one particular aspect, a method is provided that includes receiving a training set of utterances for training a machine-learning model to identify one or more intents for one or more utterances, and augmenting the training set of utterances with out-of-domain (OOD) examples. The augmenting includes: generating a data set of OOD examples, filtering out OOD examples from the data set of OOD examples, determining a difficulty value for each OOD example remaining within the filtered data set of the OOD examples, and generating augmented batches of utterances comprising utterances from the training set of utterances and utterances from the filtered data set of the OOD based on the difficulty value for each OOD. Thereafter, the machine-learning model is trained using the augmented batches of utterances in accordance with a curriculum training protocol.

Abstract: Techniques for using logit values for classifying utterances and messages input to chatbot systems in natural language processing. A method can include a chatbot system receiving an utterance generated by a user interacting with the chatbot system. The chatbot system can input the utterance into a machine-learning model including a set of binary classifiers. Each binary classifier of the set of binary classifiers can be associated with a modified logit function. The method can also include the machine-learning model using the modified logit function to generate a set of distance-based logit values for the utterance. The method can also include the machine-learning model applying an enhanced activation function to the set of distance-based logit values to generate a predicted output. The method can also include the chatbot system classifying, based on the predicted output, the utterance as being associated with the particular class.

Abstract: Techniques for using enhanced logit values for classifying utterances and messages input to chatbot systems in natural language processing. A method can include a chatbot system receiving an utterance generated by a user interacting with the chatbot system and inputting the utterance into a machine-learning model including a series of network layers. A final network layer of the series of network layers can include a logit function. The machine-learning model can map a first probability for a resolvable class to a first logit value using the logit function. The machine-learning model can map a second probability for a unresolvable class to an enhanced logit value. The method can also include the chatbot system classifying the utterance as the resolvable class or the unresolvable class based on the first logit value and the enhanced logit value.

Abstract: Techniques for noise data augmentation for training chatbot systems in natural language processing. In one particular aspect, a method is provided that includes receiving a training set of utterances for training an intent classifier to identify one or more intents for one or more utterances; augmenting the training set of utterances with noise text to generate an augmented training set of utterances; and training the intent classifier using the augmented training set of utterances. The augmenting includes: obtaining the noise text from a list of words, a text corpus, a publication, a dictionary, or any combination thereof irrelevant of original text within the utterances of the training set of utterances, and incorporating the noise text within the utterances relative to the original text in the utterances of the training set of utterances at a predefined augmentation ratio to generate augmented utterances.

Abstract: Techniques for using logit values for classifying utterances and messages input to chatbot systems in natural language processing. A method can include a chatbot system receiving an utterance generated by a user interacting with the chatbot system. The chatbot system can input the utterance into a machine-learning model including a set of binary classifiers. Each binary classifier of the set of binary classifiers can be associated with a modified logit function. The method can also include the machine-learning model using the modified logit function to generate a set of distance-based logit values for the utterance. The method can also include the machine-learning model applying an enhanced activation function to the set of distance-based logit values to generate a predicted output. The method can also include the chatbot system classifying, based on the predicted output, the utterance as being associated with the particular class.

Abstract: A semiconductor device includes a first contact layer, a second contact layer, an active layer, a photonic crystal layer, a passivation layer, a first electrode and a second electrode. The first contact layer has a first surface and a second surface opposite to each other. Microstructures are located on the second surface. The second contact layer is located below the first surface. The active layer is located between the first contact layer and the second contact layer. The photonic crystal layer is located between the active layer and the second contact layer. The passivation layer is located on the second contact layer. The first electrode is located on the passivation layer and is electrically connected the first surface of the first contact layer. The second electrode is located on the passivation layer and is electrically connected to the second contact layer.

Abstract: A display device includes a first substrate, a plurality of light-emitting diodes, a first wavelength conversion layer and a metasurface. The light-emitting diodes are arranged on the first substrate, in which the light-emitting diodes emit a first color light, and the light-emitting diodes includes a first light-emitting diode, a second light-emitting diode and a third light-emitting diode. The first wavelength conversion layer is on the first light-emitting diode, and configured to convert the first color light emitted from the first light-emitting diode into a second color light, in which the second color light is different from the first color light. The metasurface is above the first wavelength conversion layer, and configured to reflect the first color light and pass the second color light.

Abstract: A method of manufacturing a light-emitting device includes a number of operations. A light-emitting element is formed. A simulation of a metasurface is performed. The metasurface is formed based on the simulation of the metasurface. The metasurface is disposed on a light-emitting side of the light-emitting element. Performing the simulation of the metasurface includes establishing a metasurface model of the metasurface, in which the metasurface model has a plurality of unit cells, and phase compensation values of the unit cells are periodically distributed with a supercell period length in a deflection direction. The phase compensation values of the unit cell are adjusted and the light source is set to simulate the multiple transmittances of the metasurface model under different phase compensation values. The phase compensation values at a peak value of transmittance are selected as process parameters of the metasurface.