The real-world big data are largely dynamic, interconnected and unstructured text. It is highly desirable to transform such massive unstructured data into structured knowledge. Many researchers rely on labor-intensive labeling and curation to extract knowledge from such data. Such approaches, however, are not scalable. We vision that massive text data itself may disclose a large body of hidden structures and knowledge. Equipped with pretrained language models and text embedding methods, it is promising to transform unstructured data into structured knowledge. In this talk, we introduce a set of methods developed recently in our group for such an exploration, including joint spherical text embedding, discriminative topic mining, taxonomy construction, text classification, and taxonomy-guided text analysis. We show that data-driven approach could be promising at transforming massive text data into structured knowledge.

Never before in history, network and social media data have been generated and collected at such high volumes as it is today. As the volumes of data available to companies, scientists, and the public increase, their effective use becomes more challenging. Keeping up to date with the flood of data, using standard tools for data analysis and exploration, is fraught with difficulty. The field of visual analytics seeks to provide people with better and more effective ways to understand and analyze large datasets, while also enabling them to immediately act upon their findings. Visual analytics integrates the analytic capabilities of the computer and the abilities of the human analyst, allowing novel discoveries and empowering individuals to take control of the analytical process. The talk presents the challenges of visual analytics and exemplifies them with several application examples from the areas of network and social media analysis, illustrating the exiting potential of visual analysis techniques but also their limitations.

With the increasing legalization of medical and recreational use of substances, more research is needed to understand the association between mental health and user behavior related to drug consumption. Specifically, drug overdose and substance use- related mental health issues have become two major topics that have been widely discussed on social media platforms. Big social media data has the potential to provide deeper insights about these associations to public health analysts for making policy decisions. Multiple national population surveys have found that about half of those who experience a mental health illness during their lives will also experience a substance use disorder and vice versa. The communications related to addiction and mental health are complex to process and understand given their language and contextual characteristics. Surface-level data analysis alone is not sufficient to understand the complex nature of relationships among the addiction and mental health context. Moreover, dark web vendors have been using social media as a new marketplace for drugs. Social media users also discuss the novel drugs emerging in dark web marketplaces and associated side effects/health conditions. These communications get complex when researchers try to annotate them or link them to a specific mental health entity. Considering the significant sensitivity of such communications and to protect user privacy on social media, a potential solution requires reliable algorithms for modeling such communications. We demonstrate the value of incorporating domain-specific knowledge in natural language understanding to identify the relationship between mental health and drug addiction. We discuss end-to-end knowledge-infused deep learning frameworks that leverage the pre-trained language representation model and domain-specific declarative knowledge source to extract entities and their relationships jointly. Our model is further tailored to focus on the entities mentioned in the sentence where ontology is used to locate the target entity, position. We also demonstrate the capabilities of inclusion of the knowledge-aware representation in association with language models that can extract the Drug-Mental health condition associations. Acknowledgements: Usha Lokala, Raminta Daniulaityte, Francois Lamy, Manas Gaur, Jyotishman Pathak, and collaborators on NIDA/NIH and NSF funded projects on Addiction and Mental Health.