Abstract: Techniques for iterative query-based analysis of text are described. According to various implementations, a neural network architecture is implemented receives a query for information about text content, and iteratively analyzes the content using the query. During the analysis a state of the query evolves until it reaches a termination state, at which point the state of the query is output as an answer to the initial query.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for selecting actions to be performed by an agent interacting with an environment. In one aspect, a method comprises: obtaining a graph of nodes and edges that represents an interaction history of the agent with the environment; generating an encoded representation of the graph representing the interaction history of the agent with the environment; processing an input based on the encoded representation of the graph using an action selection neural network, in accordance with current values of action selection neural network parameters, to generate an action selection output; and selecting an action from a plurality of possible actions to be performed by the agent using the action selection output generated by the action selection neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for selecting actions to be performed by an agent interacting with an environment. In one aspect, a method comprises: obtaining a graph of nodes and edges that represents an interaction history of the agent with the environment; generating an encoded representation of the graph representing the interaction history of the agent with the environment; processing an input based on the encoded representation of the graph using an action selection neural network, in accordance with current values of action selection neural network parameters, to generate an action selection output; and selecting an action from a plurality of possible actions to be performed by the agent using the action selection output generated by the action selection neural network.

Abstract: In neural-network-based approaches to sequence modeling, an output sequence may be modeled via segmentations, the probability of the output sequence being constructed as a sum of products of output-segment probabilities, taken over all valid output-sequence segmentations. A set of artificial neural networks may model the distribution of the output-sequence probability with a recurrent neural network modeling the distributions of the individual output-segment probabilities, optionally in conjunction with a second recurrent neural network modeling concatenations of output segments. In various embodiments, this approach is applied to neural phrase-based machine translation.

Abstract: A deep structured semantic module (DSSM) is described herein which uses a model that is discriminatively trained based on click-through data, e.g., such that a conditional likelihood of clicked documents, given respective queries, is maximized, and a condition likelihood of non-clicked documents, given the queries, is reduced. In operation, after training is complete, the DSSM maps an input item into an output item expressed in a semantic space, using the trained model. To facilitate training and runtime operation, a dimensionality-reduction module (DRM) can reduce the dimensionality of the input item that is fed to the DSSM. A search engine may use the above-summarized functionality to convert a query and a plurality of documents into the common semantic space, and then determine the similarity between the query and documents in the semantic space. The search engine may then rank the documents based, at least in part, on the similarity measures.

Abstract: Techniques for iterative query-based analysis of text are described. According to various implementations, a neural network architecture is implemented receives a query for information about text content, and iteratively analyzes the content using the query. During the analysis a state of the query evolves until it reaches a termination state, at which point the state of the query is output as an answer to the initial query.

Abstract: A speaker device with an acoustic tuning assembly can adjust sound in variety. The acoustic tuning assembly includes a main body and a housing receiving the main body. The main body receives an acoustic unit, and includes a first sound chamber and an upper peripheral wall. The housing defines a plurality of through holes. The upper peripheral wall defines a plurality of match holes. The housing can be rotated relatively to the main body to make the through holes of the housing overlapped partially or completely with the match holes to form a plurality of overlap holes, or make the through holes of the housing non-overlapping with the match holes. Therefore, sound from the acoustic unit can be transmitted through the first sound chamber and selectively through the overlap holes.

Abstract: An speaker device with an acoustic tuning assembly can adjust sound in variety. The acoustic tuning assembly includes a main body and a housing receiving the main body. The main body receives an acoustic unit, and includes a first sound chamber and an upper peripheral wall. The housing defines a plurality of through holes. The upper peripheral wall defines a plurality of match holes. The housing can be rotated relatively to the main body to make the through holes of the housing are overlapped partially or completely with the match holes to formed a plurality of overlap holes, or make the through holes of the housing are non-overlapping with the match holes. Therefore, sound from the acoustic unit can be transmitted through the first sound chamber and selectively through the overlap holes.

Abstract: A deep structured semantic module (DSSM) is described herein which uses a model that is discriminatively trained based on click-through data, e.g., such that a conditional likelihood of clicked documents, given respective queries, is maximized, and a condition likelihood of non-clicked documents, given the queries, is reduced. In operation, after training is complete, the DSSM maps an input item into an output item expressed in a semantic space, using the trained model. To facilitate training and runtime operation, a dimensionality-reduction module (DRM) can reduce the dimensionality of the input item that is fed to the DSSM. A search engine may use the above-summarized functionality to convert a query and a plurality of documents into the common semantic space, and then determine the similarity between the query and documents in the semantic space. The search engine may then rank the documents based, at least in part, on the similarity measures.

Abstract: An earphone cord managing device of the invention has a resilient bushing, which can clip and be fixed onto an earphone cord, and a fixing cover for fixing the resilient bushing. One of resilient members of the resilient bushing is integrally formed with a loop body. A gap for passage of the earphone cord is formed on an annular body of the loop body. When necessary, the earphone cord may be firstly wound a number of turns, and then the wound earphone cord is buckled into the loop body of the resilient bushing to achieve the object of rapidly and really storing and managing the earphone cord.

Abstract: A deep structured semantic module (DSSM) is described herein which uses a model that is discriminatively trained based on click-through data, e.g., such that a conditional likelihood of clicked documents, given respective queries, is maximized, and a condition likelihood of non-clicked documents, given the queries, is reduced. In operation, after training is complete, the DSSM maps an input item into an output item expressed in a semantic space, using the trained model. To facilitate training and runtime operation, a dimensionality-reduction module (DRM) can reduce the dimensionality of the input item that is fed to the DSSM. A search engine may use the above-summarized functionality to convert a query and a plurality of documents into the common semantic space, and then determine the similarity between the query and documents in the semantic space. The search engine may then rank the documents based, at least in part, on the similarity measures.

Abstract: A head strap includes a head band and a pair of holders. The pair of holders are connected to two ends of the head band. Each of the pair of holders includes a receiving portion. The receiving portion defines a receiving cavity with a cutout. The disclosure also provides an earphone assembly.

Abstract: A protection member for a transmission cable includes a resilient coat and a fixing member. The resilient coat includes two resilient members connected to each other so as to receive the flexible wire of the transmission cable. A first connection portion is formed on the facing side of each of the resilient members. The two resilient members are connected to each other by the two respective first connection portions. The fixing member includes two hard members connected to each other so as to receive the resilient coat therein. A second connection portion is formed on the facing side of each of the hard members. The two hard members are connected to each other by the two respective second connection portions. The fixing member protects the flexible wire to be overly bent.

Abstract: An earphone device includes a housing, a speaker diaphragm, and a sealing cup. The housing includes a cavity and a plurality of vent holes acoustically coupled to the cavity. The speaker diaphragm is received in the cavity and configured to produce sound. The sealing cup is hermetically closed with the housing, and has at least one opening defined therethrough. The sealing cup and the housing cooperate to define an air chamber acoustically coupled to both the vent holes and the opening. The vent holes are configured to transfer the sound produced by the speaker diaphragm into the air chamber. The air chamber serves as an acoustic resonator to resonate the sound. The opening is configured to allow the sound transfer from the air chamber to outside of the air chamber.

Abstract: A water bottle includes a body having an open top, and two first snapping members are connected to two outsides of the body. A top part is mounted to the open top of the body with a first seal ring located between the top part and the open top. An outlet is defined in the top of the top part and communicates with the interior of the body. Two second snapping members extend from two outsides of the top part so as to be snapped with the first snapping members to securely connect the top part to the body. A cap is pivotably connected to the top part and covers the outlet. A second seal ring is located between the outlet and the cap. The interior of the body is easily cleaned when removing the top part from the body.

Abstract: A head strap includes a head band and a pair of holders. The pair of holders are connected to two ends of the head band. Each of the pair of holders includes a receiving portion. The receiving portion defines a receiving cavity with a cutout. The disclosure also provides an earphone assembly.

Abstract: A soap holder includes a holder having a first wall on the periphery thereof, and a support face which is enclosed by the first wall. A first rib and a second rib respectively extend from the top of the support face. The first rib is parallel to the second rib, and they have different heights so as to keep the soap on the first and second ribs at an angle relative to the holder. The first and second ribs provide better ventilation between the soap and the support face.

Abstract: An earphone device includes a housing, a speaker diaphragm and a tuner. The housing has an exterior surface and defines an interior cavity. The speaker diaphragm is positioned within the interior cavity and divides the interior cavity into a first chamber and a second chamber. The housing further includes a plurality of openings defined to connect the first chamber to the exterior of the housing, and a plurality of vents defined to connect the second chamber to the exterior surface. The speaker diaphragm is configured to selectively produce sound and sound produced by the speaker diaphragm is acoustically transferred from the first chamber through the openings and away from the housing. The tuner is configured to selectively open and close one or more of the plurality of vents.

Abstract: An earphone device includes a housing, a speaker diaphragm, and a sealing cup. The housing includes a cavity and a plurality of vent holes acoustically coupled to the cavity. The speaker diaphragm is received in the cavity and configured to produce sound. The sealing cup is hermetically closed with the housing, and has at least one opening defined therethrough. The sealing cup and the housing cooperate to define an air chamber acoustically coupled to both the vent holes and the opening. The vent holes are configured to transfer the sound produced by the speaker diaphragm into the air chamber. The air chamber serves as an acoustic resonator to resonate the sound. The opening is configured to allow the sound transfer from the air chamber to outside of the air chamber.

Abstract: A deep structured semantic module (DSSM) is described herein which uses a model that is discriminatively trained based on click-through data, e.g., such that a conditional likelihood of clicked documents, given respective queries, is maximized, and a condition likelihood of non-clicked documents, given the queries, is reduced. In operation, after training is complete, the DSSM maps an input item into an output item expressed in a semantic space, using the trained model. To facilitate training and runtime operation, a dimensionality-reduction module (DRM) can reduce the dimensionality of the input item that is fed to the DSSM. A search engine may use the above-summarized functionality to convert a query and a plurality of documents into the common semantic space, and then determine the similarity between the query and documents in the semantic space. The search engine may then rank the documents based, at least in part, on the similarity measures.