N-GRAMS
Remarks:
N-GRAM models are very important when we have to identify words in a noisy and ambiguous input. N-GRAM models are used in:
- Speech Recognition
- Hand Writing Recognition
- Spell Correction
- Machine Translation
- many other applications
You can read more about N-GRAM models in:
- Speech and Language Processing Book by Daniel Jurafsky and James H. Martin
Computing the Conditional Probability
P( glasses | reading ) = Count( reading glasses ) / Count( reading )
We count the sequences reading glasses and glasses from corpus and compute the probability.
Syntax:
- The conditional probability of the next most likely word can be obtained by using a big corpus(Managed Collection of text or speech data), it is all bout counting things(words) from the corpus. The goal is to find P(w|h), which the probability of next word in the sequence given some history h.
- The Concept of the N-GRAM model is that instead of computing the probability of a word given its entire history, it shortens the history to previous few words. When we use only a single previous word to predict the next word it is called a Bi-GRAM model. For Example, we have P(glasses|reading), the probability of the word "glasses" given the previous word "reading" is computed as:(Refer to the example)
Contributors
Topic Id: 8851
Example Ids: 27572
This site is not affiliated with any of the contributors.