A Chinese word segmentation method and apparatus based on deep learning. The method comprises: converting training corpus data into character-level data; converting the character-level data into sequence data; segmenting the sequence data according to pre-set symbols to obtain a plurality of pieces of sub-sequence data, and grouping the plurality of pieces of sub-sequence data according to the lengths of the sub-sequence data to obtain K data sets; according to the K data sets, obtaining K trained time sequence convolutional neural network-conditional random field models; and inputting data obtained after the processing of target corpus data into at least one of the K trained time sequence convolutional neural network-conditional random field models to obtain a word segmentation result for the target corpus data. Therefore, the method can solve the problem of the low accuracy of Chinese word segmentation in the prior art.
展开▼
