What is it about?
The article comes under the domain of sentiment analysis or opinion mining. The work is all about analyzing user reviews regarding an entity which may be a product or a service. The article presents a robust model that detects the features/aspects about which the user has expressed some opinion that may be anywhere on the scale of extremely strong positive to extremely strong negative. These features may be either directly discussed in the review or may be hidden in the sense, that the feature may be implied by some clue in the text. The proposed model not only extracts the directly stated features but also brings out the hidden features. Analyzing the opinion at the individual feature level is the most fine-grained approach that aids in gaining valuable insights from the user's reviews at the grass root level.
Photo by Robert Collins on Unsplash
Why is it important?
The unique features of the proposed work are: 1. A Robust set of rules that could identify aspect-opinion pairs even though they may be connected via complex dependency relationship. 2. A domain-specific adjective-noun collocation list is developed as part of the research work. It helps in bringing out the hidden aspects. 3. A novel nine-point scale is framed to measure the sentiment orientation in a fine-tuned manner. The scale covers nine levels of polarity in contrast to the existing three-point scale (positive, negative, and neutral). It helps in finding the strength of user opinion on an accurate scale. For example, instead of simply finding that the user has expressed a positive opinion about a feature, the proposed model can go in-depth and find whether the positivity is extremely strong positive, strongly positive, or simply positive. The same approach applies to negativity too.
Read the Original
This page is a summary of: Efficacy improvement of aspect-based sentiment analysis using enhanced rule –based approach and domain-specific lexicon (ERBA-DSL), Journal of Intelligent & Fuzzy Systems, July 2022, IOS Press,
You can read the full text:
The following have contributed to this page