METHOD AND SYSTEM TO FILTER OUT HARASSMENT FROM INCOMING SOCIAL MEDIA DATA
Budget: $15 – $25 USD
Emakia is a 501(c) 3 nonprofit startup that researches and develops the filtering out of harassment on social media platforms. The purpose of Emakia is to protect the receiver of social media content by developing a system that filters trolls, harassment, and fake news. Emakia’s first app, Enaëlle will provide users access to their social media feeds that filter out unwanted content. Using machine learning (ML) classifiers, the app filters text, image, audio, and video data.
Emakia is designed not only to filter the unwanted content on Enaëlle, it will also
automate the reporting of harassment via emails. The reports will be sent to individuals or affinity organizations chosen by the user. In addition, Emakia will coordinate with these organizations that will provide emotional and psychological assistance to the user. At the same time, the relevant social media platforms will also receive an email alerting them to remove the harassing content.
Emakia’s board is all-women-led.
Emakia's goal is to provide a system that filters out unwanted content at the receiving end for social media users. In addition, an automated report is generated when harassment is detected and supports are enabled where a user can elect to get professional intervention.
Emakia aims to provide social media users with an application, Enaëlle, that filters out unwanted content free of charge. To accomplish this goal we plan to utilize the latest AI technologies (Core ML and Auto ML) and to collaborate with academic researchers with whom Emakia will share its data sets and code.
The Emakia system applies classifier models ML on incoming social media data, (text, image, video, and audio). The classifier models from Apple Core ML 3 or Google Auto ML determine if the incoming data qualifies for harassment. The classifier models separate the data into two sets: the harassment data set and the neutral data set. Only the neutral data are displayed on the receiver's main screen; the harassing content is filtered out. The classification process is similar to the ML classifiers used to filter out spam from emails.
The proposed solution to filter out harassment involves many components such as:
● Performing text classification using Core ML and Auto ML classifiers to train machine learning models with English and Italian labeled data
● Fine-tuning of the training models with a bag-of-words and word embedding as an adaptive filter of the ML models and additional Natural Language Processing (NLP) tools to train the model and clean the data
● Transferring big set of data using Webhooks and cloud server
● Implementing a test application to evaluate the trained models and fine-tune the training using real-time data
● Implementing the Enaëlle application
● Using Core ML and Auto ML Classifiers to filter out harassment on images, videos, and audio.
In the future updated version, classifiers will be trained with fake news data to detect incoming fake news. The future version of the Emakia system would include the following
● Training models in multiple languages
● Detecting Fake News and Fake Videos
● Report harassment to external entities such as
○ the social media platform where the content comes from for removal of the content
o social support group for emotional support.
Emakia’s board is all-women-led; Corinne David founder, CEO, and CTO; Sadhana Seelam co-founder and CFO; Mary Riley Open Source Manager; Shawna Lee Grant Application Manager.
To define our success, we evaluate the efficiency and security of the Enaëlle application to remove harassment.
Emakia plans to support system by reaching out to the following organizations: the International Women's Media Foundation, Amnesty International, National Organization for Women (NOW), Planned Parenthood, Association for Women’s Rights, LGBT National Help Center; Human Rights Campaign (HRC), National Association for Advancement for Colored People (NAACP).
Emakia is designed not only to filter the unwanted content on Enaëlle, it will also
automate the reporting of harassment via emails. The reports will be sent to individuals or affinity organizations chosen by the user. In addition, Emakia will coordinate with these organizations that will provide emotional and psychological assistance to the user. At the same time, the relevant social media platforms will also receive an email alerting them to remove the harassing content.
Emakia’s board is all-women-led.
Emakia's goal is to provide a system that filters out unwanted content at the receiving end for social media users. In addition, an automated report is generated when harassment is detected and supports are enabled where a user can elect to get professional intervention.
Emakia aims to provide social media users with an application, Enaëlle, that filters out unwanted content free of charge. To accomplish this goal we plan to utilize the latest AI technologies (Core ML and Auto ML) and to collaborate with academic researchers with whom Emakia will share its data sets and code.
The Emakia system applies classifier models ML on incoming social media data, (text, image, video, and audio). The classifier models from Apple Core ML 3 or Google Auto ML determine if the incoming data qualifies for harassment. The classifier models separate the data into two sets: the harassment data set and the neutral data set. Only the neutral data are displayed on the receiver's main screen; the harassing content is filtered out. The classification process is similar to the ML classifiers used to filter out spam from emails.
The proposed solution to filter out harassment involves many components such as:
● Performing text classification using Core ML and Auto ML classifiers to train machine learning models with English and Italian labeled data
● Fine-tuning of the training models with a bag-of-words and word embedding as an adaptive filter of the ML models and additional Natural Language Processing (NLP) tools to train the model and clean the data
● Transferring big set of data using Webhooks and cloud server
● Implementing a test application to evaluate the trained models and fine-tune the training using real-time data
● Implementing the Enaëlle application
● Using Core ML and Auto ML Classifiers to filter out harassment on images, videos, and audio.
In the future updated version, classifiers will be trained with fake news data to detect incoming fake news. The future version of the Emakia system would include the following
● Training models in multiple languages
● Detecting Fake News and Fake Videos
● Report harassment to external entities such as
○ the social media platform where the content comes from for removal of the content
o social support group for emotional support.
Emakia’s board is all-women-led; Corinne David founder, CEO, and CTO; Sadhana Seelam co-founder and CFO; Mary Riley Open Source Manager; Shawna Lee Grant Application Manager.
To define our success, we evaluate the efficiency and security of the Enaëlle application to remove harassment.
Emakia plans to support system by reaching out to the following organizations: the International Women's Media Foundation, Amnesty International, National Organization for Women (NOW), Planned Parenthood, Association for Women’s Rights, LGBT National Help Center; Human Rights Campaign (HRC), National Association for Advancement for Colored People (NAACP).