Parameter Optimization for Iterative Confusion Network Decoding in Weather-Domain Speech Recognition

In this paper, we use a set of approaches to, efficiently, rescore the output of the automatic speech recognizer over weather-domain data. Since, the in-domain data is insufficient for training an accurate language model, we utilize an automatic selection method to extract domain-related sentences from an out-of-domain text resource. Then, an N-gram language model is trained on this set. We exploit this LM, along with a pre-trained acoustic model for recognition of the development and test instances. The recognizer generates a confusion network (CN) for each instance. Afterwards, we make use of the recently-introduced recurrent neural network language model (RNNLM), trained on the subtitles, in order to re-score the CNs. Re-scoring the CNs, in this way, requires estimating the weights of the RNNLM, N-gramLM and acoustic model scores. Weights optimization is the critical part of this work, whereby, we use the minimum error rate training (MERT) algorithm which has shown a great performance in machine translation. The experiments are done over weather forecast domain data that has been provided in the framework of EU_BRIDGE project.