Microsoft Technology Licensing, LLC

Touchtype Limited

Joseph Osborne

Juha Iso-Sipilä

Douglas Alexander Harper Orr

Marco Fiscato

Matthew James Willson

Richard Tunnicliffe

Matthew James Wilson

Marisa Clare Montaldi

Jessica Margaret Pumphrey

Hwasung Lee

Julien Baley

Abigail Bunyan

James Peter John Withers

A data input system is described for inputting text items to an electronic device. The data input system has a store holding a vocabulary of embeddings of text items, each embedding being a numerical encoding of a text item. The data input system receives user input comprising a sequence of one or more context text items and a new text item, the new text item being a text item with an embedding to be computed and added to the vocabulary or with an embedding already in the vocabulary and to be updated. A neural network predictor predicts a next text item in the sequence given the context text items and the vocabulary. An online training module updates the vocabulary by using either a direction associated with the predicted next item, or, by comparing the new text item and the predicted next text item.

Data input system with online learning

An electronic device is described which has a user interface which receives an input comprising a sequence of target indicators of data items. The data entry system has a search component which searches for candidate expanded sequences of indicators comprising the target indicators. The search component searches amongst indicators generated by a trained conditional language model, the conditional language model having been trained using pairs, each individual pair comprising a sequence of indicators and a corresponding expanded sequence of indicators.

Sequence expander for data entry/information retrieval

A data input system is described of the type which has a virtual keyboard which enables a user to type a text sequence into a computing device. The data input system has an input probability generator which is configured to compute keypress evidence. The keypress evidence comprises probabilities that user input events at the virtual keyboard correspond to characters or keyboard functions. The data input system has a trained keypress encoder, having been trained using keypress evidence and corresponding words. The trained keypress encoder encodes the keypress evidence into a numerical encoding. The data input system has a completion/correction predictor which is configured to take as input, the numerical encoding and an encoding of one or more text items of the text sequence already input to the computing device, in order to predict a text item in the text sequence.

Data input system using trained keypress encoder

A data entry system is described which has a user interface which receives a sequence of one or more context text items input by a user. The data entry system has a predictor trained to predict a next item in the sequence. The predictor comprises a plurality of learnt text item embeddings each text item embedding representing a text item in a numerical form, the text item embeddings having a plurality of different lengths. A projection component obtains text item embeddings of the context text items and projects these to be of the same length. The predictor comprises a trained neural network which is fed the projected text item embeddings and which computes a numerical output associated with the predicted next item.

Neural network data entry system

A text input system is described for inputting text to a computing device. The text input system has a memory storing first evidence comprising text selected by a user for input to the computing device in a first attempt by a user to input intended text. The memory stores second evidence comprising either information about text deleted by the user or text selected by the user in a second attempt at inputting the intended text. The text input system has an input model configured to combine at least the first and second evidence to produce combined evidence; and a text predictor configured to take the combined evidence as input and use the combined evidence to compute a plurality of predicted text items for input to the computing device.

Text input system using evidence from corrections

Systems for inputting text into an electronic device are provided. The systems are configured to receive a sequence of characters input into the device. The systems comprise a means configured to generate from the sequence of characters a sequence of word-segments. The systems also comprise a text prediction engine comprising a language model having stored sequences of word-segments. The text prediction engine is configured to receive the sequence of word-segments. In a first embodiment, the text prediction engine is configured to determine whether each word-segment of the sequence of word-segments corresponds to a stored word-segment of the language model and output the sequence of word-segments as a candidate prediction when each of the word-segments of the sequence of word-segments corresponds to a stored word-segment of the language model, regardless of whether the sequence of word-segments corresponds to a stored sequence of word-segments. Various other systems and corresponding methods are provided. A system is provided comprising a word-segment language model comprising stored sequences of word segments and a candidate filter. The candidate filter is used to filter word predictions generated by the word-segment language model.

System and method for inputting text into electronic devices

A data input system is described which has a user interface which receives one or more context text items in a sequence of text items input by a user. A processor of the data input system uses a plurality of language models to predict, from each language model, a next item in the sequence of text items. The processor uses a dynamic model which is bespoke to the user as a result of learning text items which the user has previously used, to predict a next item in the sequence of text items. The processor weights the predicted next item from the dynamic model using at least per term weights, each per term weight representing a likelihood of an associated term of the dynamic model given one of the language models.

Growing community of inventors

London, United Kingdom

Joseph Osborne