Open vocabulary language modeling for binary response typing interfaces

We contrast language modeling for binary response typing interfaces with the more standard use of language models for full sequence disambiguation in applications like speech recognition. We highlight a key issue for construction of these language models: using Huffman coding versus simpler binary coding tree topologies. We then introduce novel methods for handling of selection error within the model itself. Surprisingly, we find that the difference between optimal Huffman coding and simpler linear coding is reduced not only with improved language models (which was expected and is verified) but also with increased selection error rates.