Viterbi is a Python 3.5 application developed on Jupyter Notebook, mainly to correct the misspelled characters. The project details and the data were obtained from COS 402: Artificial Intelligence Homework #5 - Correcting typos without a dictionary.
An Hidden Markov Model was developed to predict the following character.
The Initial State probabilities of each character are given in the table below.
| Character | Probability |
|---|---|
| a | 0.1059087895 |
| b | 0.0500994622 |
| c | 0.0418846239 |
| d | 0.0285493259 |
| e | 0.0299123259 |
| f | 0.0350327857 |
| g | 0.0162086495 |
| h | 0.0358063803 |
| i | 0.0808590584 |
| j | 0.0017682163 |
| k | 0.0028733515 |
| l | 0.0241656229 |
| m | 0.0411847049 |
| n | 0.0251234068 |
| o | 0.0810800855 |
| p | 0.051757165 |
| q | 0.0010314595 |
| r | 0.0265600825 |
| s | 0.0745597878 |
| t | 0.1736904148 |
| u | 0.0101672438 |
| v | 0.0047889192 |
| w | 0.05474103 |
| x | 0 |
| y | 0.0022471082 |
| z | 0 |
The success rate of the algorithm is given in the following two tables.
| Words | Correct(#) | Wrong(#) | Correct(%) | Wrong(%) |
|---|---|---|---|---|
| Before | 2121 | 1290 | 62.18% | 37.82% |
| Viterbi | 2363 | 1048 | 69.28% | 30.72% |
| Characters | Correct(#) | Wrong(#) | Correct(%) | Wrong(%) |
|---|---|---|---|---|
| Before | 14934 | 1654 | 90.03% | 9.97% |
| Viterbi | 15258 | 1330 | 91.98% | 8.02% |
- numpy library is required to perform matrix operations.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent