yannisroy / efrei-pfe2017-influence

Influence

Try it out!

python3 src/datasource-lefigaro.py
python3 src/datasource-cnews_matin.py

efrei-pfe2017-influence's Issues

10 sample articles for meeting

• Corinna: Le Monde, Le Figaro
• Tom: carenews.com
• Michael: ulule.com
• Julian: paris.fr
• Björn: Facebook

Fetch Posts for specific NGOs from Facebook

DataSource for paris.fr

Evaluate LSTM for Text Classification

Investigate Text Mining Approaches

Select 10-20 revelant Asso

Find text samples

We will need:

• ~50 texts that are clearly related to non-profits
• ~10 texts that seem to be related, but are actually not
• ~10 texts that are definitely not related

Collect articles we think are relevant in a file, so they can be verified.

Auto translate a given JSON file

Finish Crawling CareNews

Find a way to crawl all relevant pages

Define relevant, develop strategy

Load Pre-Processed Files into KNIME

test

Pre-Process JSON files and save as CSV

Evaluate Unsupervised Learning Approaches for Text Classification

CNewsMatin Data Review

Make sure Facebook posts always have an url

Even if it's not a real one, for example use
facebook://<timestamp>:<some other data>

Crawl all subpages of a given URL

• e.g. Le Monde

Evaluation Data Between Groups

Implement find_all for CareNews

Find articles not related to ngos / associations

Implement waiting between requests to prevent overload

Implement search-based crawling for CareNews

Start writing Report

Concept for a prototype in Python

Implement timeframe-based crawling for CNews Matin

Filter out Facebook Results in English

Start creating presentation

Detect Language of Posts (Twitter/Facebook)

Investigate French support KNIME

Evaluate Word2Vec for Text Classification

Concept for a prototype in KNIME for Text Mining

Classify more text samples

Extract content from webpage

• publication date
• text
• (author)
• ...

Find articles related to the NGOs provided by client

Intro Document

Investigate AFP as DataSource

https://www.afp.com/fr