We present a framework for analyzing what the state in RNNs remembers from its input embeddings. We compute the gradients of the states with respect to the input embeddings and decompose the gradient matrix with Singular Value Decomposition to analyze which directions in the embedding space are best transferred to the hidden state space, characterized by the largest singular values. We apply our approach to LSTM language models and investigate to what extent and for how long certain classes of words are remembered on average for a certain corpus. Additionally, the extent to which a specific property or relationship is remembered by the RNN can be tracked by comparing a vector characterizing that property with the direction(s) in embedding space that are best preserved in hidden state space.

We present a Character-Word Long Short-Term Memory Language Model which both reduces the perplexity with respect to a baseline word-level language model and reduces the number of parameters of the model. Character information can reveal structural (dis)similarities between words and can even be used when a word is out-of-vocabulary, thus improving the modeling of infrequent and unknown words. By concatenating word and character embeddings, we achieve up to 2.77% relative improvement on English compared to a baseline model with a similar amount of parameters and 4.57% on Dutch. Moreover, we also outperform baseline word-level models with a larger number of parameters.

In this paper we present SCALE, a new Python toolkit that contains two extensions to n-gram language models. The first extension is a novel technique to model compound words called Semantic Head Mapping (SHM). The second extension, Bag-of-Words Language Modeling (BagLM), bundles popular models such as Latent Semantic Analysis and Continuous Skip-grams. Both extensions scale to large data and allow the integration into first-pass ASR decoding. The toolkit is open source, includes working examples and can be found on http://github.com/jorispelemans/scale.