Research Software Sentiment Analyser
Raquel Alegre
Carl Wilson
Sinan Shi
David Perez-Suarez
Gary MacIndoe
Olivier Philippe
Problem we are trying to address:
- Measure research software credit, both negative and positive.
Audience:
- Potential users of research software
- Authors of research software
Benefits:
- To make qualitative choices based on software credit
- To find out information about their user community
Twitter Client
Sentiment Analysis
Data Analysis and Visualization
Tweets.csv
author;date;text;geo;id
raquelalegre;2015-03-20;I really like this sof
oliph;2015-03-23;Downloaded this library th
shisinan;2015-03-25;Check out the latest fro
author;date;text;geo;id;sentiment_data
raquelalegre;[...], 103
oliph;[...], 45
shisinan;[...], 97
Demo!
Twitter Client
Authors:
- Raquel Alegre
- Olivier Philippe
- Given a term, searches for global tweets about it, limited to 1000.
- Elaborates results into a dictionary of tweets.
- Saves results to a CSV file for the Sentiment Analyser.
- Test Driven Development
Technologies/Methodologies:
- Python
- Tweepy
- JSON
Sentiment Analysis
Authors:
- Gary Macindoe
- Carl Wilson
Technologies:
- Java 8
- Stanford NLP
- OpenCSV
- Maven
- Travis-CI
- Read input CSV file.
- Remove @usernames, #hashtags and URLs from tweet content using twitter-text library.
- Score tweet content using Stanford NLP library.
- Write output CSV file with score appended to each row.
Website
Authors:
- David Perez-Suarez
- Sinan Shi
Technologies:
- Python
- Flask
- Pandas
- Bokeh
- Grab user input from website.
- Call Twitter Client to search for tweets about user input.
- Call Sentiment Analyser to classify tweets.
- Make and visualize plot and GitHub badge.
-
Data mining from other sources
- getpapers, StackOverflow, ...
- Comparisons between research software tools
- Live metrics
- Geolocation
-
Improvement of the Sentiment Analysis:
- Add languages other than English
- Testing
- Bias
- tweets about one's software tools
- retweets - exponential impact