← Back to Case Studies
Case Study

Automated Blogging

How to save editors' time using scraping and deep learning - at scale.

Challenge

Creating high-quality SEO content for a large network of blogs requires a lot of resources and is often cost-prohibitive. If a major part of the process could be automated, it would lead to massive efficiencies and enable rapid scaling.

Solution

To automate content creation for a U.S.-based network of 800 blogs, we created a system that downloads speech transcripts from various sources (e.g. YouTube), transforms them into readable articles using deep learning and posts them automatically to WordPress API as prepared drafts according to a schedule. We provided editors with ready-made article drafts based on automatically scraped speech transcripts that they can edit instead of creating articles from scratch. The system can be easily adapted to handle any topic in any language.

Looking for something similar?

TALK TO OUR EXPERT

System Architecture

Technology

  • The system is built using Python.
  • Admin interface is based on Django framework.
  • Text transformation uses a bi-directional LSTM neural network trained on TED talk transcripts, plus various pre & post-processing rules.
  • The model is trained using Theano framework.
  • Various tasks are scheduled using Celery.
  • Articles are posted via WordPress XML-RPC API.

Raw transcript

if my strategy was to least-expect my way into true love then the variable that I had to deal with was serendipity in short I was trying to figure out what's the probability of my finding Mr. Right well at the time I was living in the city of Philadelphia and it's a big city and I figured in this entire place there are lots of possibilities so again I started doing some math population of Philadelphia it has 1.5 million people I figure about half of that are men so that takes the number down to 750,000 I'm looking for a guy between the ages of 30 and 36 which was only four percent of the population so now I'm dealing with the possibility of 30,000 men I was looking for somebody who was Jewish because I am and that was important to me that's only 2.3 percent of the population I figure I'm attracted to maybe one out of 10 of those men and there was no way I was going to deal with somebody who was an avid golfer so that basically meant there were 35 men for me that I could possibly date in the entire city of Philadelphia

if my strategy was to least-expect my way into true love then the variable that I had to deal with was serendipity in short I was trying to figure out what's the probability of my finding Mr. Right well at the time I was living in the city of Philadelphia and it's a big city and I figured in this entire place there are lots of possibilities so again I started doing some math population of Philadelphia

Source: TED talk by Amy Webb - How I Hacked Online Dating

Article Draft

If my strategy was to least-expect my way into true love, then the variable that I had to deal with was serendipity. In short, I was trying to figure out what's the probability of my finding Mr. Right.

Well, at the time I was living in the city of Philadelphia and it's a big city and I figured in this entire place. There are lots of possibilities. So again I started doing some math population of Philadelphia. It has 1.5 million people. I figure about half of that are men, so that takes the number down to 750,000, I'm looking for a guy between the ages of 30 and 36, which was only four percent of the population. So now I'm dealing with the possibility of 30,000 men.

I was looking for somebody who was Jewish because I am - and that was important to me - that's only 2.3 percent of the population. I figure I'm attracted to maybe one out of 10 of those men and there was no way I was going to deal with somebody who was an avid golfer, so that basically meant there were 35 men for me that I could possibly date in the entire city of Philadelphia.

If my strategy was to least-expect my way into true love, then the variable that I had to deal with was serendipity. In short, I was trying to figure out what's the probability of my finding Mr. Right.

Well, at the time I was living in the city of Philadelphia and it's a big city and I figured in this entire place. There are lots of possibilities. So again I started doing some math population of Philadelphia.

Results

800

blogs

serviced in the network

250

articles

posted every day

91%

editors' time

freed up by automation

Read More Case Studies