Bryan Hua
Production
«
Development of a Pipeline for Text Mining Steam Reviews with
Natural Language Processing and Topic Modeling: A Case Study of
Dota 2
The personal computer gaming industry the Steam platform. I employed Natural
industry. Developing a successful game techniques to improve the efficiency and
has become a multi-billion dollar
is challenging due to the large scale,
and developers need to understand user
feedback to improve the quality of their
products. Most online game publishing
platforms allow users to leave text-
accuracy of analysis using the pipeline
and tested the results using reviews of
Dota 2, one of the most downloaded and
reviewed games on Steam.
based reviews for a game they bought. While I worked on this project, I dived into
rich source of potential feedback. As a on experience building natural language
Such reviews provide developers with a
result, there is a need to develop efficient
methods for analyzing large volumes of
text reviews.
In this nine-month project, I explored
a general pipeline of text data mining
and iterated on them to establish a
dedicated pipeline to analyze reviews on
68
Language Processing and topic modeling
PRODUCTION
the area of data analysis and got hands-
processing and topic modeling tools with
R language.
The result of this work could be beneficial
to all developers, especially community
managers, to discover insights from large
volumes of player feedback and improve
the quality of their products.