Fren
HACKTHEMOUNTAINS 2020 · 4THJOURNAL · private
🔒 encrypted to the clientTHERAPIST · analysis
Honest journaling needs total privacy — but a therapist still needs the emotional signal behind the writing.
Fren lets clients write diary entries freely, but therapists never see the raw notes. An NLP layer — sentiment analysis, TF-IDF word analysis, and a network graph of context-related terms — surfaces the emotional picture instead. Built at HackTheMountains 2020, where it placed Fourth.
A MERN stack logs timestamped entries through a Node/Express API on MongoDB, while a Flask microservice runs the Python ML pipeline (scikit-learn, spaCy, NLTK, TextBlob) — a multi-layer perceptron for sentiment, TF-IDF for keywords — so only the analysis ever crosses to the therapist.
WRITE
client journals privately
React · MongoDB
ANALYZE
sentiment + TF-IDF + term graph
Flask · scikit-learn
SURFACE
the emotional picture, not the text
spaCy · NLTK
REVIEW
therapist reads the analysis only
React
I built the full React/TypeScript frontend — both the client journaling interface and the therapist analytics panel, where raw entries are deliberately never shown:
private journaling
free-form entries, encrypted to the client
sentiment gauge
MLP classifier surfaces mood over time
TF-IDF keywords
the themes behind the writing
therapist panel
analysis only — never the raw notes