Opinion Mining in Latvian Text Using Semantic Polarity Analysis and Machine Learning Approach

Gatis Špats, Ilze Birzniece

Abstract


In this paper we demonstrate approaches for opinion mining in Latvian text. Authors have applied, combined and extended results of several previous studies and public resources to perform opinion mining in Latvian text using two approaches, namely, semantic polarity analysis and machine learning. One of the most significant constraints that make application of opinion mining for written content classification in Latvian text challenging is the limited publicly available text corpora for classifier training. We have joined several sources and created a publically available extended lexicon. Our results are comparable to or outperform current achievements in opinion mining in Latvian. Experiments show that lexicon-based methods provide more accurate opinion mining than the application of Naive Bayes machine learning classifier on Latvian tweets. Methods used during this study could be further extended using human annotators, unsupervised machine learning and bootstrapping to create larger corpora of classified text.


Keywords:

Sentiment analysis; opinion mining; semantic polarity; automatic classification of Latvian text

Full Text:

PDF


DOI: 10.7250/csimq.2016-7.03

Cited-By

1. Sentiment Analysis in Latvian and Russian: A Survey
Rinalds Vīksna, Gints Jēkabsons
Applied Computer Systems  vol: 23  issue: 1  first page: 45  year: 2018  
doi: 10.2478/acss-2018-0006

Refbacks

  • There are currently no refbacks.