Determining the polarity of words is an important task in sentiment analysis and its applications. The most comprehensive dictionaries could be found in English, however, many other low-resource languages lack established polarity dictionaries,the existing ones are small in size or contain only polarity identifier. In this study, we propose a new lexicon-based approach for text polarity detection using sentiment triggers that add contextual semantics during the analysis. To this end, the existing word polarity dictionary in Serbian has been extended as to contain approximately 15000 words annotated with polarity strength. Serbian sentiment framework (SRPOL), relying on the new lexicon and proposed sentiment triggers, has shown an overall accuracy score of 79% on validation datasets from different domains annotated for sentiment, which is in the range with the state-of-the-art approaches on this task.