The present disclosure provides a holistic model for feature specification and dependency representation for requirement specification documents where the conventional models fail to provide. The present disclosure receives a plurality of requirement specification documents and a related data. A product feature model is generated based on the plurality of requirement specification documents and the related data using a feature model generation technique. The product feature model includes a plurality of product feature elements. The plurality of product feature elements includes a feature area, a major feature and a plurality of features. A specification model is generated further for each of the plurality of features using a specification extraction technique. Post generating the specification model, a plurality of dependency associations are generated for each of a plurality of specification elements of the specification model using a dependency extraction technique. Finally, the plurality of dependency associations are updated in the specification model.

The present disclosure provides a holistic model for feature specification and dependency representation for requirement specification documents where the conventional models fail to provide. The present disclosure receives a plurality of requirement specification documents and a related data. A product feature model is generated based on the plurality of requirement specification documents and the related data using a feature model generation technique. The product feature model includes a plurality of product feature elements. The plurality of product feature elements includes a feature area, a major feature and a plurality of features. A specification model is generated further for each of the plurality of features using a specification extraction technique. Post generating the specification model, a plurality of dependency associations are generated for each of a plurality of specification elements of the specification model using a dependency extraction technique. Finally, the plurality of dependency associations are updated in the specification model.

This application relates to a text classification method. The method includes obtaining, by a computer device, a to-be-classified text, and calculating an original text vector corresponding to the text; determining, by the computer device according to the original text vector, an input text vector corresponding to each channel of a trained text classification model; inputting, by the computer device, the input text vector corresponding to each channel into a convolution layer of the corresponding channel of the trained text classification model, the trained text classification model comprising a plurality of channels, each channel being corresponding to a sub-text classification model, and the trained text classification model being used for determining a classification result according to a sub-classification parameter outputted by each sub-text classification model; and obtaining, by the computer device, a classification result outputted by the trained text classification model, and classifying the text according to the classification result.

This application relates to a text classification method. The method includes obtaining, by a computer device, a to-be-classified text, and calculating an original text vector corresponding to the text; determining, by the computer device according to the original text vector, an input text vector corresponding to each channel of a trained text classification model; inputting, by the computer device, the input text vector corresponding to each channel into a convolution layer of the corresponding channel of the trained text classification model, the trained text classification model comprising a plurality of channels, each channel being corresponding to a sub-text classification model, and the trained text classification model being used for determining a classification result according to a sub-classification parameter outputted by each sub-text classification model; and obtaining, by the computer device, a classification result outputted by the trained text classification model, and classifying the text according to the classification result.

A system retrains a natural language understanding (NLU) model by regularly analyzing electronic documents including web publications such as online newspapers, blogs, social media posts, etc. to understand how word and phrase usage is evolving. Generally, the system determines the frequency of words and phrases in the electronic documents and updates an NLU dictionary depending on whether certain words or phrases are being used more frequently or less frequently. This dictionary is then used to retrain the NLU model, which is then applied to predict the meaning of text or speech communicated by a people group. By analyzing electronic documents such as web publications, the system is able to stay up-to-date on the vocabulary of the people group and make correct predictions as the vocabulary changes (e.g., due to natural disaster). In this manner, the safety of the people is improved.

A system retrains a natural language understanding (NLU) model by regularly analyzing electronic documents including web publications such as online newspapers, blogs, social media posts, etc. to understand how word and phrase usage is evolving. Generally, the system determines the frequency of words and phrases in the electronic documents and updates an NLU dictionary depending on whether certain words or phrases are being used more frequently or less frequently. This dictionary is then used to retrain the NLU model, which is then applied to predict the meaning of text or speech communicated by a people group. By analyzing electronic documents such as web publications, the system is able to stay up-to-date on the vocabulary of the people group and make correct predictions as the vocabulary changes (e.g., due to natural disaster). In this manner, the safety of the people is improved.

Embodiments of a production-quality text-to-speech (TTS) system constructed from deep neural networks are described. System embodiments comprise five major building blocks: a segmentation model for locating phoneme boundaries, a grapheme-to-phoneme conversion model, a phoneme duration prediction model, a fundamental frequency prediction model, and an audio synthesis model. For embodiments of the segmentation model, phoneme boundary detection was performed with deep neural networks using Connectionist Temporal Classification (CTC) loss. For embodiments of the audio synthesis model, a variant of WaveNet was created that requires fewer parameters and trains faster than the original. By using a neural network for each component, system embodiments are simpler and more flexible than traditional TTS systems, where each component requires laborious feature engineering and extensive domain expertise. Inference with system embodiments may be performed faster than real time.

Embodiments of the present invention are directed to methods and resulting structures for integrated circuits having metal-insulator-metal (MIM) capacitors that serve as both decoupling capacitors and crack stops. In a non-limiting embodiment, an interconnect is formed on a first portion of a substrate in an interior region of the integrated circuit. A second portion of the substrate is exposed in an edge region of the integrated circuit. A MIM capacitor is formed over the second portion of the substrate in the edge region. The MIM capacitor includes two or more plates and one or more dielectric layers. Each dielectric layer is positioned between an adjacent pair of the two or more plates and a portion of the two or more plates extends over the interconnect in the interior region. A plate of the two or more plates is electrically coupled to a last metal wiring level of the interconnect.

Item codes for items are mapped to multidimensional space as item vectors based on transaction contexts. Similarities between item codes are based on distances between the item codes within the multidimensional space. Substitute items for out-of-stock items are automatically identified based on the item similarities and based on collected feedback from transactions. The substitute items are provided in real time to customers during transactions, item picking services during item fulfillment, and shelf management services for item shelf stocking. In an embodiment, the substitute items are further determined based on a specific transaction history for a given customer and specific feedback collected for the given customer from the specific transaction history.

Item codes for items are mapped to multidimensional space as item vectors based on transaction contexts. Similarities between item codes are based on distances between the item codes within the multidimensional space. Substitute items for out-of-stock items are automatically identified based on the item similarities and based on collected feedback from transactions. The substitute items are provided in real time to customers during transactions, item picking services during item fulfillment, and shelf management services for item shelf stocking. In an embodiment, the substitute items are further determined based on a specific transaction history for a given customer and specific feedback collected for the given customer from the specific transaction history.