Code Switching Language Model Using Monolingual Training Data

Abstract: Training a code-switching (CS) LLM using only monolingual data is still an ongoing research problem. In this paper, a CS LLM is trained using only monolingual training data. As recurrent neural network (RNN) models are best suited for predicting sequential data. In this work, an RNN LLM is trained using alternate batches from only monolingual English and Spanish data and the perplexity of the LLM is computed. From the results, it is concluded that using alternate batches of monolingual data in training reduced the perplexity of a CS LLM. The results were consistently improved using mean square error (MSE) in the output embeddings of RNN based LLM. By combining both methods, perplexity is reduced from 299.63 to 80.38. The proposed methods were comparable to the LLM fine tune with code-switch training data.