A method for training a non-autoregressive translation (NAT) model includes: acquiring a source language text, a target language text corresponding to the source language text and a target length of the target language text; generating a target language prediction text and a prediction length by inputting the source language text into the NAT model, in which initialization parameters of the NAT model are determined based on parameters of a pre-trained translation model; and obtaining a target NAT model by training the NAT model based on the target language text, the target language prediction text, the target length and the prediction length.

Techniques are described herein for determining a long-form of an abbreviation using a machine learning based approach that takes into consideration both sequential context and structural context, where the long-form corresponds to a meaning of the abbreviation as used in a sequence of words that form a sentence. In some embodiments, word representations are generated for different words in the sequence of words, and a combined representation is generated for the abbreviation based on a word representation corresponding to the abbreviation, a sequential context representation, and a structural context representation. The sequential context representation can be generated based on word representations for words positioned near the abbreviation. The structural context representation can be generated based on word representations for words that are syntactically related to the abbreviation. The combined representation can be input to a classification neural network trained to output a label representing the long-form of the abbreviation.

Embodiments of the present disclosure include a method for token-position handling comprising: processing a first sequence of tokens to produce a second sequence of tokens, wherein the second sequence of tokens has a smaller number of tokens than the first sequence of tokens; masking at least some tokens in the second sequence to produce masked tokens; moving the masked tokens to the beginning of the second sequence to produce a third sequence; encoding tokens in the third sequence into a set of numeric vectors in a first array; and processing the first array in a transformer neural network to determine correlations among the third sequence, the processing the first array producing a second array.

A method includes obtaining first training data having multiple first linguistic samples. The method also includes generating second training data using the first training data and multiple symmetries. The symmetries identify how to modify the first linguistic samples while maintaining structural invariants within the first linguistic samples, and the second training data has multiple second linguistic samples. The method further includes training a machine learning model using at least the second training data. At least some of the second linguistic samples in the second training data are selected during the training based on a likelihood of being misclassified by the machine learning model.

A method of identifying text (e.g., a sentence or sentence portion) in a word processing text editor; automatically identifying a domain-specific deep-learning neural network that corresponds to an identified context, from among one or more domain-specific deep-learning neural networks; automatically identifying at least one suggested replacement word using the identified domain specific deep-learning neural network that corresponds to the identified context; and automatically controlling a display to display a user interface that includes functionality that presents prompt information that includes the at least one suggested replacement word. Changes for errors that are common in academic papers written by non-native speakers may be suggested.

A syntax analyzing device includes: a syntax analyzing unit that analyzes syntax of a sentence received by a receiving unit, thereby acquiring a first analysis result, which is an analysis result having one or more elements constituting the sentence and parts of speech of the respective one or more elements and has one or more binary trees each having the parts of speech or the elements as nodes; a category acquiring unit that acquires categories of the respective one or more elements constituting the sentence; a category inserting unit that acquires a second analysis result in which the categories of the elements are respectively inserted between the elements and the parts of speech of the elements, which respectively correspond to the one or more categories, and constituting the first analysis result; and a learning unit that outputs the second analysis result acquired by the category inserting unit.

Provided is an automated interpretation method, apparatus, and system. The automated interpretation method includes encoding a voice signal in a first language to generate a first feature vector, decoding the first feature vector to generate a first language sentence in the first language, encoding the first language sentence to generate a second feature vector with respect to a second language, decoding the second feature vector to generate a second language sentence in the second language, controlling a generating of a candidate sentence list based on any one or any combination of the first feature vector, the first language sentence, the second feature vector, and the second language sentence, and selecting, from the candidate sentence list, a final second language sentence as a translation of the voice signal.

A symbol prediction method includes storing a statistic for each of a set of symbols w in at least one context, each context including a string of k preceding symbols and a string of l subsequent symbols, the statistic being based on observations of a string kwl in training data. For an input sequence of symbols, a prediction is computed for at least one symbol in the input sequence, based on the stored statistics. The computing includes, where the symbol is in a context in the sequence not having a stored statistic, computing the prediction for the symbol in that context based on a stored statistic for the symbol in a more general context.

Embodiments of this application include an information conversion method for translating source information. The source information is encoded to obtain a first code. A preset conversion condition is obtained. The preset conversion condition indicates a mapping relationship between the source information and a conversion result. The first code is decoded according to the source information, the preset conversion condition, and translated information to obtain target information. The target information and the source information are in different languages. Further, the translated information includes a word obtained through conversion of the source information into a language of the target information.

A system includes a narrative repository which stores a plurality of narratives and, for each narrative, a corresponding outcome. A narrative evaluator receives the plurality of narratives and the outcome for each narrative. For each received narrative, a subset of the narrative is determined to retain based on rules. For each determined subset, a entropy matrix is determined which includes, for each word in the subset, a measure associated with whether the word is expected to appear in a sentence with another word in the subset. For each entropy matrix, a distance matrix is determined which includes, for each word in the subset, a numerical representation of a difference in meaning of the word and another word. Using one or more distance matrix(es), a first threshold distance is determined for a first word of the subset. The first word and first threshold are stored as a first word-threshold pair associated with the first outcome.