Prediction of Meaning of Bisyllabic Chinese Words Using Back Propagation Neural Network

A three layer back propagation neural net is set up to determine the dependence between the semantic class of a bisyllabic Chinese word and its two characters. Simulations were performed using a three–layer back–propagation neural net (BPNN) with various combinations of inputs. The inputs are (1) semantic classes of the characters, (2) Entropy of the characters and (3) semantic strengths [1] of the characters. Our results show that we can obtain the bias of meaning class of a bisyllabic word with an accuracy of 81% by inputting the semantic classes and strengths of the characters. The BPNN can also be trained to predict the meaning of a word to a high precision of 83%. With the results obtained from this system, more research on the formation of word and meaning can be performed.