😂🥰😎😥😡😖😪
The project i worked on a Tri model concept- Basically to achieve the accuracy, I have divided emotions into three different Models 1 - [Positive, Negative, Neutral] 2 - [Happy, Joy, Surprise] 3 - [Anger, Sad, Fear]
🧐 It was an English Dataset having 20k lines with 6 emotions, i have added more 10k lines with only neutral sentiment make it total 7 sentiments
🔁 With the use of AWS Translate, 30k English sentences where translated into Gujarati Language. Now most important part of any Data Science project comes
🛀 Data was cleaned, which consist of awful language, untranslated word, numbers and sentences making no sense. Intially some translated data was gibris which was removed and some was cleaned. Reducing the data by around 2%.
🔎 For the purpose I was confused between ML and DL algorithms, Obviously DL algorithm made much more sense in the preformance and accuracy