DOING@ 3rd MaDICS Symposium (on line)

Webinar Action DOING

Twitter: https://twitter.com/NetworkDoing
BBB : https://visio.icube.unistra.fr/b/gan-by7-7eg

5 July 2021

13h30min-17h30min (CET)

INSCRIPTIONS at MADICS web page (free)

PROGRAM

  • 13h30min. Introduction
  • 13h45min. KEYNOTE: Natural language processing for epidemiology & public health Aurélie Névéol, CNRS, LISN, France
  • 14h45min. KEYNOTE: Evaluating navigational queries over graphs, Domagoj Vrogoc, Pontificia Universidad Católica de Chile, Chili
  • 15h45min. Coffee Break
  • 16h. TUTORIAL : Say the word, and you’ll be free: Methods and Techniques for Natural Language Database Interfaces, Altigran Soares da Silva, Universidade Federal do Amazonas, Brésil
  • Aurélie
  • Domagoj
  • Altigran

ABSTRACTS & BIO:

Natural language processing for epidemiology and public health


Abstract: Much medical knowledge and information about patients is contained in free text, such as scientific articles, electronic patient records or social media. Natural language processing can help with the extraction of health information from free texts and processing of this information in an aggregated form that can be used for clinical practice and public health. This presentation will outline the challenges of clinical text analysis and discuss their practical impact on patient care.

Aurélie Névéol is a researcher at the Centre National pour la Recherche Scientifique (CNRS). She leads research on clinical natural language processing for languages other than English. Her research work includes using NLP to create representations of clinical information to support information extraction from unstructured clinical narrative text in the electronic health record, which can then be used for high throughput phenotyping. She earned an MSc in Linguistics in 2002 and a PhD in Computer Science in 2005. She has then contributed to research projects at the National Library of Medicine to improve the retrieval and analysis of biomedical text from the litterature. She has also contributed to the evaluation of research methods and workflows through her participation in the H2020 MIROR project and international evaluation campaigns such as CLEF eHealth and the biomedical task at WMT.

Evaluating navigational queries over graphs: from theory to practice

Abstract: In this talk we will focus on graph queries which allow traversing paths of arbitrary length, as opposed to fixed size patterns. We will discuss some theoretical results regarding this class of queries, and see how they have been supported in existing graph database engines. We will also try to determine what algorithms one could potentially use to execute such queries efficiently, and show that tried and tested solutions sometimes reign supreme to new proposals.

Domagoj Vrgoč is an assistant professor at the Institute for Mathematical and Computational Engineering at Pontificia Universidad Católica de Chile, and an associate researcher at the Institute for Foundational Research on Data (IMFD Chile). He did his PhD with Leonid Libkin at the University of Edinburgh on the topic of graph query languages. His research interests include databases, Semantic Web, and theory of computation.

Say the word, and you’ll be free: Methods and Techniques for Natural Language Database Interfaces

Abstract Since the dawn of database technology in the 1970s, Natural Language Database Interfaces (NLIDBs) have been almost a utopian aspiration, both in academia and industry. In fact, despite the wide adoption and popularity of databases in the last decades, effective methods for the development of NLIDBs have only recently emerged in the literature, allowing users without technical knowledge to effectively explore the data maintained by Database Management Systems (DBMS). This renewed interest in NLIDBs is mainly due to the current stage of technological maturity in areas such as Machine Learning, Natural Language Processing, and Information Retrieval, whose recent advances allow extracting semantics from user-written text with great precision and efficiency. Two main approaches have been studied in this regard. The first is the use of keyword searches, similar to what happens in search engines. The second is the use of sentences written in natural language to express queries. While keyword-based systems offer a more straightforward and more intuitive way to express queries, natural language-based systems allow one to express more complex queries involving, for example, aggregations. In this tutorial, we will present an overview of current methods and techniques that have improved in many ways the algorithms and models used to build NLIDBs. 

Companion hands-on notebook: https://github.com/pr3martins/nalir-sbbd

Altigran da Silva is a professor at the Instituto de Computação in the Universidade Federal do Amazonas (IComp/UFAM), where he works as a researcher, lecturer, and advisor in undergraduate, master’s, and doctorate degrees. He completed his doctorate in Computer Science at UFMG in 2002. His research interests involve Data Management, Information Retrieval, and Data Mining, emphasizing the World-Wide Web and Social Media environment. He has coordinated and participated in dozens of research projects on these topics, resulting in more than 140 scientific publications in good-quality journals and conference proceedings. In addition, he has coordinated several conference program committees in Brazil and abroad, having also participated as a technical program committee member in around 50 international conferences and workshops. He served as the Dean of Research and Graduate Studies at UFAM (2007/2009), as the Deputy Coordinator of the Computing area