user side spam email detection using CNNs
Akash Chaudhari
Mumbai, Maharashtra
- 0 Collaborators
while most email spam detectors work by classifying the emails you receive , lot of times your own legitimate emails get classified as spam and the person we are sending it to misses it , my projects aims at giving the user an idea of whether his/her email will be classified as spam , this is done while the user is typing the email ...learn more
Project status: Under Development
Overview / Usage
the project is essentially text classification , but using conv_net
the net classifies the email as spam / not spam
this project being done though for academic research purposes , does has have application outside of academia, this can be implemented in any email service provider in the text area and be used to classify the text the user is typing and hence give him a idea whether there is a chance his/her email might end up in spam
Methodology / Approach
i am following a research paper by kim yoon " text classification suing CNNs" essentially i am creating a CNN with multiple layers to classify the text
the training dataset is first converted to word vectors (each word is given a vector so that the cnn gets a idea of its symatical meaning ) , these word vectors are fed to a embedding layer which acts a s the input and then passes it on to multiple conv_layers with diffrent kernel sizes the outputs of these layer are maxpooled and then concatenated into a single tensor , which is again processed upon to get the output
Technologies Used
python(numpy ,os, pandas,csv)
anaconda
keras
gensim
intel tech - i5 8300 processor
nvidea 1050ti gpu