Deep Learning and Explainability for Sentiment Analysis

Dear Colleagues,

People use online social platforms to express opinions about products and/or services in a wide range of domains, influencing the point of view and behavior of their peers. Understanding individuals’ satisfaction is a key element for businesses, policy makers, organizations, and social institutions to make decisions. This has led to a growing amount of interest within the scientific community, and, as a result, to a host of new challenges that need to be solved. Sentiment analysis methodologies have been investigated and employed by researchers in the past to provide methodologies and resources to stakeholders. In the field of machine learning, deep learning models which combine several neural networks have emerged and have become the state-of-the-art technologies in various domains for a variety of natural language processing tasks. The most prominent deep learning solutions are combined with word embeddings. However, how to include sentiment information in word-embedding representations to boost the performances of deep learning models, as well as explain what deep learning models (often employed as a black-box) learn are questions that still remain open and need further research and development.

The investigation of these key points will answer to why and how design choices for creating embedding representations and designing deep learning should be made. This goes toward the direction of Explainable Deep Learning (XDL), whose aim is to address how deep learning systems make decisions. This Special Issue aims to foster discussions about the design, development, and use of deep learning models and embedding representations which can help to improve state-of-the-art results, and at the same time enable interpreting and explaining the effectiveness of the use of deep learning for sentiment analysis. We invite theoretical works, implementations, and practical use cases that show benefits in the use of deep learning with a high focus on explainability for various domains.

The Special Issue is focused but not limited to these topics:

• Deep learning topics
  • Aspect-based DL and XDL models; 
  • Bias detection within DL and XDL for sentiment analysis; 
  • DL and XDL for toxicity and hate speech detection; 
  • Multilingual DL and XDL for sentiment analysis; 
  • DL and XDL for emotions detection; 
  • Weak-supervised DL and XDL for sentiment analysis; 
  • XDL design methodologies for sentiment analysis; 
  • Analysis of DL models for sentiment analysis.
• Data representations topics
  • Word embeddings for sentiment analysis; 
  • Knowledge graph and knowledge graph embeddings for sentiment analysis; 
  • Use of external knowledge (e.g., knowledge graphs) to feed DL for sentiment analysis; 
  • Combination of existing sentiment analysis resources (e.g., SenticNet) with embedding representations; 
  • Analysis of the performance of data representations for sentiment analysis tasks 
  • Lexicon-based explainability for sentiment analysis.
• Case studies
  • Educational environments; 
  • Healthcare systems; 
  • Scholarly discussions (e.g., peer review process discussions, mailing lists, etc.); 
  • News platforms; 
  • Mental health systems; 
  • Social networks.

Prof. Dr. Diego Reforgiato Recupero
Prof. Dr. Harald Sack
Dr. Danilo Dessi'
Guest Editors

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Electronics is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2400 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.