Week 2 Task

Uploaded by

Riya Singh

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as DOCX, PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

12 views4 pages

Week 2 Task

Uploaded by

Riya Singh

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as DOCX, PDF, TXT or read online on Scribd

You are on page 1/ 4

Week 2 Task Description

- Task Title: Emotion Recognition in Text - Understanding Human

Sentiment

- Overview: In this task, you will develop a machine learning model capable
of classifying text into six distinct emotional categories: joy, anger,
sadness, fear, surprise, and neutral. Emotion detection in text is a key area
in Natural Language Processing (NLP), allowing machines to understand
and respond to human emotions, making it a powerful tool for sentiment
analysis, chatbots, and social media monitoring.

- Relevance to the Field of AI: While this task may seem like a basic
emotion classification project, it is much more. The foundational questions
included will create a deep-rooted understanding on the critical role that
emotion and sentiment plays in any AI technology, and how you can
master these techniques. By now we have all heard of the classic human
vs AI debate as next-generation AGI and superintelligence innovations are
on the horizon – with new human-like technologies being developed at a
rapid pace, it has become more important than ever to maintain the
“human touch” within these novel innovations. Regardless of what AI you
develop in the future, if your models and innovations are to succeed, they
will always have to include very deep and intelligent emotion reasoning
and understanding capabilities to match or supersede human abilities. This
initial task will set the groundwork for you to perfect the skill of not only
understanding how to include emotion detection capabilities (which is
relatively basic) but also, why each decision should be made and how it
will affect further development - for example, developing CERs known as
complex emotion recognition systems, such as multi-class models [voice
tone + facial + eye tracking models], cognitive processing models [EEG or
ECG brain waves] and more advanced emotion classification systems that
all stem from this foundational expertise.

As you complete this task, think from the mindset of an expert technologist
creating the future of emotion recognition systems:
- What complex steps may be required to advance basic emotion
detection models into the above CERs examples mentioned?
- What benefits can these foundational techniques have in the future?
Dataset

- You can pick a dataset of your own choice.

Notebook Sections

- Introduction in your own words

- Dataset loading and EDA
- Dataset Preprocessing
- Dataset Splitting
- Feature Engineering
- Model Trainings
- Model Evaluation
- Visualization Results

Questions

- Why did you choose this particular dataset for the task? What specific
features or attributes made it suitable for emotion detection?
- How did you ensure that the dataset was representative of all emotion
classes? If there were any imbalances, how did you handle them?
- Did you encounter missing, irrelevant, or noisy data in the dataset? How
did you deal with these issues?
- What key patterns or insights did you observe during the exploratory data
analysis (EDA)? How did these inform your approach?

- What preprocessing steps did you take to prepare the text data for
training? Why were these steps necessary for this task?
- How did you handle special cases such as emojis, abbreviations, or slang
in the text data? Why did you choose that approach?
- Did you perform any specific transformations on the labels (emotion
categories) before training the model? Why?
- If your preprocessing pipeline included removing elements (e.g., stopwords
or punctuation), how did you decide what to remove and what to keep?

- How did you convert the text into a format that the model could process?
Why did you choose that method?
- Did you face challenges in representing text data numerically? If so, how
did you overcome them?
- What trade-offs did you consider when selecting the features or
representations for the task?

- How did you determine the split between training and testing data? Why
was this split ratio appropriate for this task?
- What steps did you take to evaluate and compare different models or
approaches for the task? How did you make the final choice?
- What challenges did you encounter during training, such as performance
issues or overfitting? How did you address them?
- How did you decide on the hyperparameters or settings for training? Did
you experiment with different configurations?

- What evaluation metrics did you use to measure the performance of your
approach? Why were these metrics suitable for this task?
- How did you interpret and act on cases where the model misclassified the
emotion of a text? What steps did you take to reduce such errors?
- How did you ensure your model generalized well to unseen data? What
indicators helped you assess this?
- Did you notice any specific patterns in the errors made by your model?
How did these insights inform your next steps?

Submission Criteria

- A drive link containing three files which are a dataset, notebook and a doc
containing the answers of all the above questions.
- Notebook should be structured according to the specified sections above
- The last date of submission is 21 December.
NOTE: Queries should only be directed to the respective group. Avoid any
personal message (DM) to any team leads. Do not send any queries regarding
the tasks directly to our CEO, Jasmine Sharma.