Information Quality Management: Top-Down System Design

This paper outlines several directions for research and NLP system design for detecting incorrect information or intentional deception. Besides new machine learning approaches that rely on deep learning, #CNN, and #LSTM technologies, there is significant research in other areas. For example, computational linguistics, psychology, and sociology add many valuable ideas to improve techniques based purely on machine learning. The most prominent direction appears to be based on multi-channel analysis that includes linguistic knowledge, non-verbal indicators, and broad contextual features. For information quality studies, there is also a question of the analytical granularity. An incorrect piece of information might be included in a large corpus of correct text. The authors of this paper provide an excellent discussion of individual and composite measures of linguistic indicators in such groups as lexical, syntax, verbal, paralanguage, and impression. In other papers, the authors provide a grouping of constituent indicators into quantity, specificity, diversity, complexity, expressivity, uncertainty, non-immediacy, personalism, and affect are excellent contributions.  

https://journals.sagepub.com/doi/abs/10.1177/0261927X18784119