The role of the Typo Correction module is to fix many potential typos that the user can accidentally type while chatting with the bot.
Not all languages have this feature
To know if you can use this feature in a specific language, please check the Language Support page.
Below is a list of the most common mistakes managed by Smartly.AI spell checker.
jjoobbb => job
weke => wake
nonster => monster
Reduction and vowel swaps
CUNsperrICY => conspiracy
nmnster => monster
One of the interesting things with our implementation of typo correction is that it takes in account your training dataset in addition to a generic dictionary.
In other words, if you have some very specific jargon in your bot that you are also expecting from your users, those words wont be affected by the typo correction.
All you have to do is to make sure all your jargon is used in intents and entities used and trained in your bot.
Train your bot first
Because our typo correction is contextual, you wont get any result from the spell checker if you don't train your bot beforehand.
Once you have created and trained your bot, you can start checking the results of the typo correction in the API Ouput tab of the simulator as shown below.
You can also check a sample test we did here.
Once we have the raw input + a corrected candidate, we send the two options to the intents classifier wich will keep the best result for the rest of the remaining NLU tasks of the pipeline.
Updated about a year ago