Predicting Formality of Written Texts Using Machine Learning Algorithms

Overview

Objectives/Goals My project goal was to create a program that could automatically classify a given text document with >80% accuracy into three levels of Formality: Formal, Casual and Semi-Formal. Methods/Materials Python program, Natural Language Tool Kit (NLTK) Library, 8 kinds of machine learning algorithms, 2 kinds of feature extraction methods, and 130 pieces of documents as labeled data. 10 experiments were conducted. The classifiers were trained using 100 labeled documents. The model was tested using 30 documents and accuracy was measured against the correct labels. Results 5 of the 8 classifiers were able to successfully predict the levels of formality with an average accuracy of 88.6%. Bag-of-Words feature extraction method was found more accurate (92.28%) than Part-of-Speech feature extraction method (84.5%). Conclusions/Discussion In conclusion, the project achieved its goal and successfully demonstrated a working program that used machine learning algorithms to predict levels of formality of any given text with >80% accuracy.

Awards (1)

• Fourth Place

Competition history

• CSEF 2017 Computational Systems & Analysis · Entry J0814

Resources

• CSEF project summary ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Systems Software and Computer Science projects
• CSEF Systems Software and Computer Science projects
• Every CSEF 2017 project
• CSEF Fourth Place winners