The objective of this project is dual. First, you will learn how to deal with a specific case of sentiment analysis (emotions) considering a conversational context. Second, you will learn to manipulate and design a model considering the final task.
We have multiple conversations.
In each conversation, there are utterances (textual content).
Each utterance's text is made of multiple tokens (word/punctuation/etc.).
This means our data will have the following structure:
- Dataset
- Conversation
- Utterance
- Tokens
- Utterance
- Conversation
We want to predict the emotion of each utterance. Which means only utterances have a label:
- Dataset
- Conversation
- Utterance [ one of these labels: 🙂 😞 😠 😯 😐 🤮 😱 ]
- Tokens
- Utterance [ one of these labels: 🙂 😞 😠 😯 😐 🤮 😱 ]
- Conversation