Taylor's Probabilistic Method for Sequence Join Costs

[Verity S]

In section 16.5.5 Taylor describes Taylor’s probabilistic method of calculating sequence join costs.

“N-grams work on discrete entities, so frames have to be quantised as a
pre-processing step. This is done via a codebook of the type used in discrete-probability
HMMs [367], [510], which uses a bottom-up clustering to quantise the data, followed
by a mapping, which assigns any frame a number corresponding to its nearest cluster.”

Could you explain how clustering is used to quantise the data?

[Simon]

We’ll look at this in the lecture.

[Author]

Posts