| W12-6216 | . Figure 2 shows a WFST for a | backoff bigram | model . It is also quite straight-forward |
| W00-1309 | the N-1 previous tags . Here , | backoff bigram | ( N = 2 ) model is used . The |
| W00-0737 | the N-1 previous tags . Here , | backoff bigram | ( N = 2 ) model is used . The |
| H93-1016 | purpose we used a conventional | backoff bigram | for that pur - pose . Secondly |
| P00-1073 | prior SLM research , including | backoff bigram | LM , perplexity , and related |
| P10-1028 | model ( the second factor ) is a | backoff bigram | model trained on the tokenized |
| H93-1022 | system is about 1200 words , and a | backoff bigram | language model was trained using |
| P02-1024 | probabilities are estimated by | backoff bigram | probabilities divided by the |
| H93-1022 | of the Switchboard database . A | backoff bigram | language model was trained as |
| P00-1012 | this dataset and constructed a | backoff bigram | model from the remaining 90 % |
| W98-0206 | SDR corpus , including standard | backoff bigram | ( 6.1 million ) and trigam ( |
| P00-1073 | our conclusions in section 5 . 2 | Backoff Bigram | and Cutoff One of the most successful |
