Low Rank Language Models for Small Training Sets

Several language model smoothing techniques are available that are effective for a variety of tasks; however, training with small data sets is still difficult. This letter introduces the low rank language model, which uses a low rank tensor representation of joint probability distributions for parameter-tying and optimizes likelihood under a rank constraint. It obtains lower perplexity than standard smoothing techniques when the training set is small and also leads to perplexity reduction when used in domain adaptation via interpolation with a general, out-of-domain model.