In my project I have a set of sequences (elements are letters from English alphabet) and some of the sequences have missing elements. I need to fill them with the most probable elements.
I've been reading about Hidden Markov Models. But I haven't exactly understood the capabilities of an HMM. If I build an HMM, is it possible to use it to fill the gaps in a given sequence?
Best Answer
Yes, HMM is generative model and can predict next character given input sequence. It basically emits next symbol based on its current state and then changes state according to transition probabilities (thats informal but reasonably good description). What you want to achive is called HMM character-level language model. You can also read more about language models, for example neural-network based language models were shown to be particularly powerful. Also, you may want to consider building model that takes into account next know character.