Sentiment Analysis (SA) is a natural language processing task that aims to determine the emotional tone of a text.
In this work, a classifier was built on a pair of high-order hidden Markov models. To train the model, we used latent semantic clusters obtained using the LSA method. The classifier supports multiprocessing.
The highest accuracy is achieved on a weighted composition of two models.
| clusters | polarity, accuracy | subjectivity, accuracy | |
|---|---|---|---|
| order=1 | 50 | 0.727 | 0.856 |
| order=2 | 35 | 0.709 | 0.854 |
| ensemble | [35, 50] | 0.731 | 0.865 |
More information in my report:
@software{samarin-igor-shmm,
author = {Samarin, I.},
doi = {10.5281/zenodo.7957936},
title = {Negative Binomial Distribution Model in Categorical
Sequence Analysis},
url = {http://hdl.handle.net/11701/43006},
version = {2.0.0},
year = {2023}
}