Franchesca Ivane Manuntag, Mari Jo Pabilonia, Julia Krista Villadiego and Charina Ortiz of Team OMQ were initially tasked to create a Headline Analyzer for a Content Marketing Agency to capture the attention of their niche market. But is headline enough? There is a shift in the viewership and the client needs to identify what exactly do their target readers want. Using advanced data science techniques such as LDA (Latent Dirichlet Allocation) Topic Modeling and XGBoost Regression, the team took on the challenge of creating ART-E (ARTicle Evaluator), a tool that can describe what exactly makes an article popular and predict viewership even before an article gets published.
8. Digging deeper into the biz
problem:
WHAT IS MY TARGET MARKET
UP TO? (DESCRIPTIVE)
What did they like in my website
before? What have they been reading?
What are their topic preferences?
9. Digging deeper into the biz
problem:
HOW DO I INCREASE MY
READERSHIP? (PREDICTIVE)
What elements or features in the articles can
I change to adapt to the preferences of my
target market?
10. How can Data Science help?
Our solution enables the
editorial team to make
strategic decisions,
maximizing the impact of
their content.
Descriptive
The app provides context
to writers telling them
what makes a fleek.ph
article popular.
Predictive
The app helps predicts
the popularity (in page
views) of a fleek.ph
article before
publishing.
Enabling+ =
11. Our Solution Adds Seamless Value
Check effectivity of
article title using online
Headline Analyzer (Co-
scheduler)
FLEEK.PH CURRENT PRACTICES
12. Our Solution Adds Seamless Value
Check for good topics using an online social listening tool
FLEEK.PH CURRENT PRACTICES
13. Our Solution Adds Seamless Value
Come up with article topics based on writer’s intuition/
team’s brainstorming
FLEEK.PH CURRENT PRACTICES
14. 2-in-1 web app: analyze BOTH headline
and content, gives page view prediction
Tool is localized to PH market and 18-24
audience.
Know topics 18-24 y/o will want
tomorrow based on preferences
yesterday.
Our Solution Adds Seamless Value
WHAT THEY CAN DO W/ OUR TOOL
15. Here’s ART-e!
We built fleek.ph a web-based
Article Evaluator with
a machine learning backend,
using Latent Dirichlet
Allocation (LDA) Topic
Modeling technique and
Regression.
17. #DataScienceXContentMarketing
You type in: article title and
article content
App gives you: predicted
number of article views +
automated tagging of topic AND
if it’s good for 25-34 years old
39. Recommendations
…aside from using the Article Analyzer ;)
Optimized Data Bank
= better web app
performance
= better descriptive and
predictive power