Gabriel_Jai - Machine Learning Pathway

Lessons Learned
Technical:

  • Embeddings and advanced NLP pre-processing tecniques
  • Latest NLP deep learning model such as BERT

Tools:

  • Selenium
  • BERT
  • Google Colab

Soft Skills:

  • Working with a team remotely

Achievement Highlights

  • Wrote automated web scraping scripts to collect more than 8000 posts from Amazon discourse forum
  • Implemented XGBoost and LSTM to classify topics on Amazon forum into their corresponding categories

List of Meetings/Training Attended

  • All team meetings

Goals for the Coming Week

  • Finalize the BERT model and determine the best one for the classification task

Detailed Statement of Tasks Done

  • Among all basic models, the logistic regression model was the best one to classify the topics because of its high accuracy and short running time. Although I tried to optimize my LSTM classifier, the amount of time and effort invested was not worth it.

Request Change of Role
Although I am comfortable with my participant role, I am interested in taking up more responsibility

Final self-assessment of June session

Technical area:

  1. Improved my familiarity of Git
  2. Expended my knowledge in advanced NLP models such as and BERT and XLNet.
  3. Learned different data augmentation techniques

Tools:

  1. Selenium + BeautifulSoup
  2. Git
  3. Google Colab and its GPU/TPU functionalities

Soft skills:

  1. Improved my communication with teammates through Google meet.
  2. Improved my skills in dealing with different schools of thought and convincing my teams with quantitative evidence

Achievement highlights

  1. incorporating selenium with BeautifulSoup to automate web scraping
  2. Improved my knowledge in BERT and proposed XLNet as a better choice than BERT
  3. Contributed myself in preparing the final presentation

List of meetings/training attended including social team events

  • all team meetings