I have read many papers explaining how to assign polarity to each word in the lexicon of positive and negative terms. Some of them such as SentiWordNet assigned 3 polarity scores(Positive, Negative, and Objective). In my lexicon, I want to assign degree of polarity (e.g. weakly positive, mildly positive, and strongly positive) and similarly with negative terms. I have collected multiple of labeled dataset in order to accomplish this task.
Some research proposed a naive approaches by using a simple equation (e.g. PositivePolarity=#occurrence of positive word in positive examples/ sum of #occurrence of positive word in positive and negative examples).
I hope you to help me on how to extract features from labeled dataset and train my machine learning algorithms to obtain the 6 polarity scores for every term in my lexicon.
Dear Ali,
Please read the following:
Kulisiewicz, M., Kajdanowicz, T., Kazienko, P., & Piasecki, M. (2015, June). On sentiment polarity assignment in the Wordnet using Loopy Belief Propagation. In International Conference on Hybrid Artificial Intelligence Systems (pp. 451-462). Springer, Cham.
https://medium.com/@datamonsters/sentiment-analysis-tools-overview-part-1-positive-and-negative-words-databases-ae35431a470c
Hi sir,
Sometimes there are no clear cut answers, you gotta get creative. Of course there are methods like the one you defined. You can also use real people data by making them evaluate each word in different axes like arousal, violence, positivity etc. I've seen works like that. You should/might consider axes other than positive-negative axis.
You can also use lexical maps/graphs like https://wordnet.princeton.edu/
For example you put "best" in the center but "better" gets relatively lower score etc.
You can also find similar studies in different languages and apply some of their scorings.
Dear Ali,
Please read the following:
Kulisiewicz, M., Kajdanowicz, T., Kazienko, P., & Piasecki, M. (2015, June). On sentiment polarity assignment in the Wordnet using Loopy Belief Propagation. In International Conference on Hybrid Artificial Intelligence Systems (pp. 451-462). Springer, Cham.
- Badges
- Science topic