The goal of Part I of the task is to use raw textual data in language models for recommendation based application.

Sentence comparison using N$-$gram:

Let a search engine powered by language model recommend which of the below sentences are most relevant w$.$r$.$t to given training corpus. Design a probabilistic language model to compare below test sentences for recommendation using Trigram. Use all the instances in the dataset as a training corpus.

Test Sentence $1$:

This is a great, reliable phone. I also purchased this phone after my samsung A$460$ died. The menu is easily comprehendable and speed dialing is available for around $300$ numbers. Voice dialing is also a nice feature, but it takes longer than speed dialing. The only thing that bothers me is the games...Nokia seems to have taken snake $(1$ and $2)$ off their phones. There is a skydiving game, bowling, and tennis $($like pong$).$ The ringers are very nice, and a feature is available to choose a different ringer for each person calling. However, ringtones are not available online to download to this phone. You're pretty much stuck with what you have. There are vibrating ringtones and regular $($midi$)$ polyphonic tones. All they need are covers in a reasonable price range...

Test Sentence $2$:

Very bad experience with this iPhone xr phone. Back camera found defective and non functional. Amazon is very careless. I haven't got any response on confirmation for replacing. I am just hanging with a damaged piece worth of more than $91000$ rupees.