Abstract: Methods, systems, apparatus, and tangible non-transitory carrier media encoded with one or more computer programs for classifying an input text block into a sequence of one or more classes in a multi-level hierarchical classification taxonomy. In accordance with particular embodiments, a source sequence of inputs corresponding to the input text block is processed, one at a time per time step, with an encoder recurrent neural network (RNN) to generate a respective encoder hidden state for each input, and the respective encoder hidden states are processed, one at a time per time step, with a decoder RNN to produce a sequence of outputs representing a directed classification path in a multi-level hierarchical classification taxonomy for the input text block.

Abstract: Methods, systems, apparatus, and tangible non-transitory carrier media encoded with one or more computer programs for classifying an item. In accordance with particular embodiments, a labeling task is issued to workers participating in a crowdsourcing system. The labeling task includes evaluating an inferred classification that includes one or more of the class labels in a hierarchical classification taxonomy based at least in part on a description of the item and the class labels in the classification. Evaluation decisions are received from the crowdsourcing system. The classification is validated based on the evaluation decisions to obtain a validation result. The validating includes applying at least one consensus criterion to an aggregation of the received evaluation decisions. Data corresponding to one or more of the class labels in the classification is routed to respective destinations based on the validation result.

Abstract: A unified system with a machine learning feature data pipeline that can be shared among various product areas or teams of an electronic platform is described. A set of features can be fetched from multiple feature sources. The set of features can be combined with browsing event data to generate combined data. The combined data can be sampled to generate sampled data. The sampled data can be presented in a format having a structure that is agnostic to a feature source from which the set of features was fetched. The sampled data can be joined with old features by a backfilling process to generate training data designed to train one or more machine learning models. Related methods, apparatuses, articles of manufacture, and computer program products are also described.

Abstract: A first neural network model of a user device processes audio data to extract audio embeddings that represent vocal characteristics of a user of an utterance represented in the audio data. The audio embeddings may then be hashed to remove characteristics specific to the user while still maintaining a unique set of characteristics. The hashed embeddings may be sent to a remote system, which may use them to identify the user.

Abstract: A first neural network model of a user device processes audio data to extract audio embeddings that represent vocal characteristics of a user of an utterance represented in the audio data. The audio embeddings may then be hashed to remove characteristics specific to the user while still maintaining a unique set of characteristics. The hashed embeddings may be sent to a remote system, which may use them to identify the user.

Abstract: Methods, systems, apparatus, and tangible non-transitory carrier media encoded with one or more computer programs for classifying an item. In accordance with particular embodiments, a labeling task is issued to workers participating in a crowdsourcing system. The labeling task includes evaluating an inferred classification that includes one or more of the class labels in a hierarchical classification taxonomy based at least in part on a description of the item and the class labels in the classification. Evaluation decisions are received from the crowdsourcing system. The classification is validated based on the evaluation decisions to obtain a validation result. The validating includes applying at least one consensus criterion to an aggregation of the received evaluation decisions. Data corresponding to one or more of the class labels in the classification is routed to respective destinations based on the validation result.

Abstract: Methods, systems, apparatus, and tangible non-transitory carrier media encoded with one or more computer programs for classifying an input text block into a sequence of one or more classes in a multi-level hierarchical classification taxonomy. In accordance with particular embodiments, a source sequence of inputs corresponding to the input text block is processed, one at a time per time step, with an encoder recurrent neural network (RNN) to generate a respective encoder hidden state for each input, and the respective encoder hidden states are processed, one at a time per time step, with a decoder RNN to produce a sequence of outputs representing a directed classification path in a multi-level hierarchical classification taxonomy for the input text block.

Abstract: In various example embodiments, systems and methods to identify query recommendations are presented. The system receives a query from a device and associates the query with query recommendations. Next, the system generates a probability distribution for each of the query recommendations based on experimental information associated with each of the query recommendations. Then the system samples each of the probability distributions to identify a probability for each the corresponding query recommendations. The probabilities indicate the likelihood of the corresponding query recommendation being selected. Next, the system identifies displayable query recommendations based on the probabilities for each of the corresponding query recommendations. The system then communicates an interface to the device that includes the query and the displayable query recommendations.

Abstract: The invention described a thermal printer. The thermal printer has a supply reel, a take-up and a pair of motors. Each motor connects with the supply reel and the take-up reel, respectively. A detecting apparatus detects ribbon and the reel parameters while initializing the thermal printer. A memory stores a transforming table which can obtain the driving pulse width of each motor from a radius of the supply reel, a radius of the take-up reel and the ribbon tension. A control apparatus drive each motor with different pulse width according to parameters detected by the detecting apparatus and the transforming table. A thermal print head performs a printing process.