SAIL is a tool which gives users the convenience of doing sentiment analysis using pre-trained models. The tool also supports incremental learning of the existing models by adding new labeled data.
The model accuracy can be improved by using domain specific lexicons and query terms.
Added basic command line arguments, so SAIL can be automated! This is how it works:
0 arguments: SAIL runs normally
1 argument : the folder is both the input and output folder.
NOTE: for input files, column headers must match the drop-down labels for SAIL to assign that column to that field. e.g. if "tweet_text" is a header in an input csv file, it'll be assigned to the tweet_text label.
2 arguments: the second argument is the output folder.
3 arguments: the third argument is the relative or absolute (recommended) model file path.
+ arguments: any arguments after 3 are ignored. there's a warning.
Separated the GUI from the functionality a little in the GUIController class to make the automation possible.
Removed all // TODO auto-gen.+ comments so property task tracking can be utilized.
Changed STDOUT and STDERR to go to the console (default) instead of files.
Fixed the model to custom and disabled the drop-down so it always uses the model in the config file. Before, it was hardcoded to the default in sentinets.Prediction.
0: no errors! success! NOTE: if the input folder is empty, success!
1: unhandled errors...
2: process input failed. this is usually because a csv format issue. SAIL will change "" to ", and also uses \ has an escape character. If a cell's text ends with a , SAIL will fail.
3: issue with your parameters. see the logs for details.
- Shubhanshu Mishra [email protected]
- Liang Tao [email protected]
- Chieh-Li Chin [email protected]
- (a little) Elias Wood [email protected]
Advisor: Jana Diesner [email protected]
The SAIL Application Executable Files (i.e. SAIL.dmg, SAIL-1.x-x64.exe, SAIL-1.x-x86.exe, SAIL.jar, and SAIL.zip) are licensed under GNU General Public License version 3.0 or later license.
The executable files include the following:
- The application code, packaged into a set of JAR files, plus any other application resources (data files, native libraries)
- A private copy of the Java and JavaFX Runtimes, to be used by this application only
- A native launcher for the application
- Metadata, such as icons
Copyright (c) 2015 University of Illinois Board of Trustees, All rights reserved. Other copyright statements provided below.
Developed at GSLIS/ the iSchool, by Dr. Jana Diesner, Shubhanshu Mishra, Liang Tao, Chieh-Li Chin.
The following files are released under GNU General Public License version 2.0 or later license:
- All files in directory "build"
- All files in directory "logo"
- All files in directory "nbproject"
- All files in directory "src"
- .classpath, .project, build.fxbuild, build.xml, mainfest.mf, and train_model.sh
Copyright (c) 2015 University of Illinois Board of Trustees, All rights reserved.
Developed at GSLIS/ the iSchool, by Dr. Jana Diesner, Shubhanshu Mishra, Liang Tao, Chieh-Li Chin.
The following dependencies are required for the application, and should be used under their licenses.
- WEKA: GNU General Public License v3.0 http://www.cs.waikato.ac.nz/ml/weka/
- Ark-Tweet-NLP: GNU General Public License v2.0 or later https://github.com/brendano/ark-tweet-nlp/blob/master/LICENSE.txt
- MPQA Subjectivity Lexicon: GNU General Public License v3.0 http://mpqa.cs.pitt.edu/
- Twitter-Text: Apache 2.0 License https://github.com/twitter/twitter-text
- Twitter4j: Apache 2.0 License http://twitter4j.org/en/index.html
- OpenCSV: Apache 2.0 License http://opencsv.sourceforge.net/license.html
- Apache Commons Lang3: Apache 2.0 License https://commons.apache.org/proper/commons-lang/project-summary.html
- SentimentSocialNets: Apache 2.0 License https://github.com/napsternxg/SentimentSocialNets
For training the word model, which is part of SAIL, we have used the SEMEVAL 2013 Task 2 part B - Twitter sentiment analysis data which contains tweet level sentiment labels for more than 20,000 tweets. We have trained our model on TRAIN+DEV+TEST data. We have only trained the model on the tweets which were labelled as positive or negative in the dataset. The data is released under a Creative Commons Attribution 3.0 Unported License (http://creativecommons.org/licenses/by/3.0/).
SemEval'2013: SemEval-2013 Task 2: Sentiment Analysis in Twitter. Preslav Nakov, Sara Rosenthal, Zornitsa Kozareva, Veselin Stoyanov, Alan Ritter, Theresa Wilson http://www.aclweb.org/anthology/S/S13/S13-2052.pdf
While not a condition of use, the developers would appreciate if you acknowledge its use with a citation:
Mishra, Shubhanshu, Jana Diesner, Jason Byrne, and Elizabeth Surbeck. "Sentiment Analysis with Incremental Human-in-the-Loop Learning and Lexical Resource Customization." In Proceedings of the 26th ACM Conference on Hypertext & Social Media, pp. 323-325. ACM, 2015.
Diesner, Jana., Mishra, Shubhanshu., Tao, Liang., Chin, Chieh-Li. (2015). SAIL: Sentiment Analysis and Incremental Learning [Software]. Available from http://people.lis.illinois.edu/~jdiesner/sail.html