The use of electronic media in the bedroom is an addictive behaviour of adults in modern times that compromise quality sleep and causes insomnia and other disorders.
The concept of cyborg anthropology and panic architecture stated by Amber Case (2009) motivated us in this research to propose an innovative way for creating indicators of poor quality sleep, considering the time intervals between nighttime posts in social media and their personality insights analysis, including text mining indicators of sentiment, emotion, life values and needs.
Interactive Dashboards are available.
The panic architecture is based on the fact that we consult our devices at all times. These new indicators composed by time spans and personality analysis could be used for Public Health Policy and also as input for mHealth warnings and advices for qualified users.
Language Use, as studied in Psycholinguistics, is an area where psychologists and psychiatrists nowadays could use tools that employ Machine Learning and Text Mining techniques.
A person's language usage can reveal information about their character, personality, moral values, sentiments, and emotions. In fact, in psycholinguistic analysis of texts, a verbal utterance is comprehended as an expression of personality, especially when it is closely linked with a native cultural environment.
The methodology of our research is composed by Obtaining Author Profiling Dataset; Text Mining messages; Calculating IBM Watson Scores; Optimizing Scores with Deep Learning; Evaluating Deep Learning Predictions; and Visualizing with Business Intelligence tool.
Based on the evolution of co-authorship networks in specific domains, we intend to visualize the future state of research networks, predicting links that will appear and disappear, considering their topologies.
In the graphs the co-authorship relations are extracted from Web of Science (Thomson Reuters Scientific). The graphs type are directed, showing the links from authors to co-authors. The node colors represent eccentricity. The node sizes are proportional to the degree.
Currently, we are working on the following domains:
- Aerospace Engineering
- Architecture of Information
- Construction Building Technology (Concrete)
- Natural Language Processing
- Papers on Pedophilia grouped in Clusters
- Text Analysis for Health Studies (SocialMed Project)
- Donations for the Brazilian Governors Campaigns in 2014
If you want to participate on the project, please don't hesitate to contact me.
Catching Telltale Signs and Data Discrepancies on the Corrutpion Scandal in Brazil: A Textual Analysis with Deep Learning Approach
This paper is being submitted to KCST 2018
Our main goal is to analyze telltale texts of public domain concerning the recent facts of corruption investigation in Brazil. We will try to identify personality signs and data discrepancies when comparing multiple sources of information.
Image Tags Prediction
We apply visual recognition algorithms to predict tags on images and/or videos. The system can use a general model or be tuned to build a particular model related to image collections in order to be trained, tested and evaluated using a machine learning process.