NiosTo – Opinion Word Extraction & Sentiment Classification

Description

NiosTo is a tool for extracting opinion words, along with polarity orientation (sentiment classification), from stored opinionated data. The application uses a combination of supervised and unsupervised machine learning methods developed for sentiment analysis tasks. DEiXTo contributed the datasets required.

Details of the NiosTo algorithm were published in the following paper:

Pantelis Agathangelou, Ioannis Katakis, Fotios Kokkoras, Konstantinos Ntonas “Mining Domain-Specific Dictionaries of Opinion Words“. In Proc. of the  15th International Conference on Web Information System Engineering (WISE 2014), Thessaloniki, Greece, 12-14 October, 2014.  B. Benatallah, A. Bestavros, Y. Manolopoulos, A. Vakali, Y. Zhang (Eds), LNCS 8786, Springer, pp 47-62, 2014. DOI: http://dx.doi.org/10.1007/978-3-319-11749-2_4

Get the slides of the WISE2014 presentation of NiosTo.

The advantages of the algorithm implemented are:

• Introduces a novel resource-efficient approach for building domain-specific opinion dictionaries.
• Provides an experimental study where all steps of the proposed algorithm
  are evaluated through a sentiment analysis task.
• Provides a dataset of opinions in the Greek language containing user reviews.
  This dataset can be exploited in various opinion mining tasks.

NiosTo after data loading and pre-processing.

NiosTo takes as input opinions in csv format and presents the discovered words in each level. In addition, retrieval results as well as sentiment classification are visualized. All parameters can be changed and tuned through the various options available by means of an easy-to-use GUI.

Pattern-based evaluation.

Algorithm evaluation.

Assets

• NiosTo V.4.1.0 – (4.3MB) – includes the application, user’s guide and demo datasets. You must have the Java Runtime Environment (JRE) installed in order to run.

Version History (after v.3)

• v.4.1.0 – 2015/06/10: Operational improvements!
• v.4.0.5 – 2015/04/12: Please, check the history.txt file for changes in v.4.
• v.3.0.4 – 2014/10/01: A couple of Linux/MacOS related issues were ironed out. The User Guide and the WISE2014 dataset were included in the main distribution file.
• v.3.0.2 – 2014/09/25: Bugs affecting Linux and MacOS users were fixed.