Smart Resume Matcher using NLTK is an NLP-based Python tool that helps job seekers compare their resumes against job descriptions. It highlights missing keywords, suggests improvements, and provides a match score — all using fundamental NLP techniques powered by NLTK.
This project was built to:
Practice core NLP tasks like tokenization, lemmatization, and similarity scoring.
Solve a real-world problem that many job seekers face: aligning resumes with job requirements.
Uses pure NLTK — no external ML/AI APIs
Solves a practical problem — helps improve resume-job alignment
Shows cosine similarity between resume and job description
Identifies missing or underused keywords in the resume
Fully contained in one clean script and easy to run
Text preprocessing is the foundational step in NLP. In this project, we applied several preprocessing techniques to clean and normalize the input resume and job description text:
Lowercasing: Converts all text to lowercase to ensure consistency (e.g., "Python" and "python" are treated the same).
Stopword Removal: Common words like “the,” “is,” “and,” etc., don’t contribute to meaning in NLP tasks, so we remove them using NLTK’s stopword list.
Lemmatization: Reduces words to their base/root form (e.g., "running" → "run"). We used NLTK’s WordNet Lemmatizer for this.
This ensures we’re working with meaningful, standardized tokens before any analysis.
Once the text is cleaned, we tokenize it — meaning we break it into individual words (tokens) using nltk.word_tokenize().
After tokenizing:
We use collections.Counter to count how many times each word appears.
These word frequency dictionaries are then used to build feature vectors for each text.
Frequency analysis lets us numerically compare text content in a meaningful way.
Instead of using a pre-built ML library, we implemented cosine similarity ourselves using Python’s math module.
Cosine similarity calculates the angle between two frequency vectors.
A smaller angle (i.e., higher cosine value) means the resume and job description are more similar.
Formula used: Cos(tita)=A.B/(||A|| . ||B||)
Where A and B are frequency vectors of the resume and job description.
This shows an understanding of both linear algebra and vector space models in NLP.
We used Python’s built-in file handling methods to:
Load the resume and job description from .txt files.
Support flexible input for real-world usage (can swap in any resume or JD).
Though this project is in a single file, it’s structured into modular functions:
preprocess_text() – handles all text cleaning
cosine_similarity() – performs the vector math
load_and_process() – handles file reading + preprocessing
main() – coordinates everything cleanly
This design makes the code readable, testable, and scalable for future improvements.
In this project, I documented each major NLP step and added comments throughout the code to explain why each step is important.
Additionally:
The project README provides a clear walkthrough of the logic.
Output is printed in an intuitive way (match % and suggestions).
Clear communication of NLP pipelines is crucial when working in teams or applying for roles that require explainability.