Text input method and device

Abstract

The invention relates to a text input method and device, and belongs to the technical field of input methods. The text input method comprises the steps of obtaining a pinyin sequence; inputting the pinyin sequence into a trained PS2CS model, and predicting a Chinese character sequence corresponding to the pinyin sequence through the trained PS2CS model; the training process of the PS2CS model comprises the following steps: performing vector conversion on each letter of a pinyin sequence sample in training set data according to a lookup table to generate a vector corresponding to each letter; wherein the lookup table comprises 28 characters and vector representations of the corresponding characters; wherein the 28 characters comprise 26 phonetic letters and 2 placeholder symbols; outputtinga prediction result through a prediction layer according to the pinyin sequence sample vector matrix; and comparing the prediction result with a corresponding standard Chinese character sequence sample in the training set data, and solving a loss function. Vectorization is carried out through the original pinyin character sequence, and then the corresponding Chinese character sequence is predicted, so that the memory space is greatly saved, and prediction is more accurate.

Description

technical field [0001] The invention relates to a text input method and device, belonging to the technical field of input methods. Background technique [0002] The n-gram language model n-gram is generally used in the Chinese Pinyin input method, and the n-gram model is used to determine which specific word the nth is under the condition that n-1 words are known in the given phrase (or sentence). word probabilities. In order to take into account the computational efficiency, binary or ternary (the value of n is 2 or 3) is used to model language sequences in practical applications. Therefore, in order to model longer-distance word histories, some recent works propose to use continuous-space language models instead of n-gram models for Pinyin input methods. [0003] In fact, most modern Pinyin input methods follow a serial process: 1) Pinyin word segmentation, which splits the input Pinyin sequence into legal Pinyin syllables; 2) generates candidate words for each Pinyin sy...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a text input method to solve the problems of low accuracy and large memory occupation of the existing input methods; meanwhile, it also provides a text input device to solve the problems of low accuracy and large memory occupation of the existing input devices. big memory problem

Images