Skip to content

Demo of browser extension that flags atypical news articles based on their fringiness

Notifications You must be signed in to change notification settings

martinholub/HackZurich2017

Repository files navigation

HackZurich2017

Repository housing a chrome extension demo developed at HackZurich2017. The extension wraps a python script that can be used to analyze text from a news article to determine its fringiness (measure of non-meanstreamness). This is done by comparing the entites extracted from the article by PermID against entities in real-time stream of articles fetched via Thomson Reuters ® API.

Example Usage

text = <few-paragraphs-of-news-text>
res = fastrun(text)

x, y, f = fringiness(res_to_matrix(res_times)[0])
plot = embedding_plot_bokeh(x, y, f, res)

from bokeh.resources import CDN
html = file_html(plot, CDN, title = "my plot")
with open("file.html", "w") as file:
    file.write(html)

See also the Jupyter Notebook


Team Members:


Get the word vectors here https://drive.google.com/file/d/0B7XkCwpI5KDYNlNUTTlSS21pQmM/edit and change the path in document_similarity.py

You need the gensim and NLTK libraries: pip install gensim nltk

About

Demo of browser extension that flags atypical news articles based on their fringiness

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published