Welcome to the BioCreative VIII challenge and workshop: Curation and evaluation in the era of ChatGPT
When: Sunday, November 12, 2023. 8am-12:30pm CST.
Where: AMIA Annual Symposium, Hilton New Orleans Riverside, New Orleans, LA
Scientific Program
| Time (CST) | Session |
|---|
| 8:00 - 8:05 am | Opening remarks |
| 8:05 - 8:30 am |
Keynote: Clinical information extraction in the era of large language models (LLMs).
Hua Xu, PhD, FACMI.
Robert T. McCluskey Professor and Vice Chair for Research and Development, Section of Biomedical Informatics and Data Science.
Assistant Dean for Biomedical Informatics, Yale School of Medicine. Yale University
|
| 8:35 - 9:30 am | Track 1: BioRED (Biomedical Relation Extraction Dataset) Track |
| 9:30 - 10:00 am | Break & BCVIII Poster session
|
| 10:00 - 10:55 am |
Track 2: SYMPTEMIST (Symptom TExt Mining Shared Task) track
|
| 11:00 - 11:55 am |
Track 3: Phenotype normalization (genetic conditions in pediatric patients)
|
| 12:00 - 12:25 pm |
PMC-Patients - a large-scale publicly available patient dataset: A call for flexible tools for complex annotation
|
| 12:25 - 12:30 pm |
Closing remarks
|
Keynote
Clinical information extraction in the era of large language models (LLMs)
Electronic Health Records (EHRs) contain abundant free text data that are valuable for research and operation in the medical domain. Natural language processing (NLP) technologies have shown great promise in unlocking information in clinical texts; however, many challenges still exist when developing and implementing NLP technologies for biomedical applications. The talk will focus on method development in clinical and biomedical NLP, including recent advancements in large language models (LLMs), along with lessons learned from building NLP-based applications.
Speaker:
Hua Xu, PhD, FACMI
Robert T. McCluskey Professor and Vice Chair for Research and Development, Section of Biomedical Informatics and Data Science
Assistant Dean for Biomedical Informatics, Yale School of Medicine
Yale University
| 
|
Bio: Dr. Hua Xu is Robert T. McCluskey Professor and Vice Chair for Research and Development, Section of Biomedical Informatics and Data Science at Yale School of Medicine (YSM), as well as Assistant Dean for Biomedical Informatics at YSM. He received his Ph.D. in Biomedical Informatics from Columbia University. His primary research interests include biomedical natural language processing (NLP) and data mining, as well as their applications in secondary use of electronic health records data for clinical and translational research. His research is funded by multiple agencies (i.e., NLM, NCI, NIGMS, NIA, AHA, and CPRIT), and methods/tools developed in his lab have been top ranked in a number of biomedical NLP shared tasks and widely used to support diverse biomedical applications. He served as the Chair of American Medical Informatics Association (AMIA) NLP Working Group and now leads the Observational Health Data Sciences and Informatics (OHDSI) NLP Working Group. Dr. Xu is a fellow of the American College of Medical Informatics (ACMI) and the International Academy of Health Sciences Informatics (IAHSI).
Back to top
Poster session
To be announced.
Back to top
Proceedings
Proceedings are available here
Back to top
Registration
To register go to AMIA symposium registration page
Note for Biocreative participants: For registration to a track please use the Google form.
Do not use the team "Team page" tab as it is non functional. For more information go to Registration section below
The BioCreative VIII workshop is scheduled to run with AMIA 2023 on November 12, 2023.
BioCreative
Critical Assessment of
Information Extraction in Biology is a community-wide effort for evaluating text mining and information extraction systems applied to the biological
domain. BioCreative has been an invaluable source for advancing state-of-the-art text mining methods since 2004, by providing reference datasets and a
collegial environment to develop and evaluate these methods in both shared and interactive models. The BioCreative VIII workshop aims to provide a
forum for clinical informatics community members, and traditional biomedical natural language processing researchers, bioinformatics researchers, and
data curators to present and discuss advances in text mining for health applications, following on the success of the previous seven BioCreative
workshops.
The VIIIth BioCreative workshop seeks to attract researchers
interested in automatic methods of extracting medically relevant information from clinical data and aims to bring together the medical NLP community
and the health professionals community. Proposed tracks include SYMPTEMIST (Symptom TExt Mining Shared
Task) and Phenotype extraction (genetic conditions in pediatric patients), addressing symptom
and phenotype extraction from clinical records (in English and Spanish); in addition, the new BioRED (Biomedical Relation
Extraction Dataset) Track will continue to address information extraction from biomedical literature; finally BioCreative
VIII proposes a new Annotation Tool track focused on developing annotation tools to facilitate the job of domain experts, offering seamless
integration with relevant ontologies and other features to improve user experience and efficiency. Please see below for more on the
tracks:
Track 1: BioRED (Biomedical Relation Extraction Dataset) Track (Rezarta Islamaj and Zhiyong
Lu)
This track aims to foster the development of systems that automatically extract biomedical relations in
journal articles, and the final resource -- freely available to the community -- will consist of 1000 MEDLINE articles fully annotated with biological
and medically relevant entities, biomedical relations between them, and the novelty of the relation (whether the relation is a key point of the
article versus background knowledge that can be found elsewhere). The participants will use the training data (600 articles) to design and develop
their NLP systems to extract asserted relationships from free text and are encouraged to classify relations that are novel findings. In the
BioCreative setting we will enrich the BioRED training dataset with 400 recently published MEDLINE articles fully annotated, bringing this valuable
resource to 1000 articles. This track serves as a continuation of previous BioCreative Workshops that addressed the individual extraction of bio
entities and/or specific relations such as disease-gene, protein-protein, or chemical-chemical, in biomedical articles. In contrast from previous
challenges, this track calls for the extraction of all semantic relations expressed in the article and their novelty factor.
Track 2: SYMPTEMIST (Symptom TExt Mining Shared Task) (Martin Krallinger)
A considerable
effort has been made to automatically extract from clinical texts relevant variables and concepts using advanced entity recognition approaches.
Despite the importance of clinical signs and symptoms for diagnosis, prognosis and healthcare data analytics strategies, this kind of clinical entity
has received far less attention when compared to other entity classes such as medications or diseases. To understand and characterize relationships
between different symptoms, their onset, or associations of symptoms to diseases is a central question for medical research. Due to the complexity
underlying the annotation process and normalization or mapping of symptom mentions to controlled vocabularies, very few datasets or corpora have been
generated to train and evaluate advanced clinical named entity recognition systems. To foster the development, research and evaluation of semantic
annotation strategies that can be useful for systematically extracting and harmonizing symptoms from clinical documents we propose the
SYMPTEMIST track. We will invite researchers, health-tech professionals, NLP, and ontology experts to develop tools capable of
detecting automatically mentions of clinical symptoms from clinical texts in Spanish and normalizing or mapping them to a widely used multilingual
clinical vocabulary, namely SNOMED CT. For this task we will release a large collection of manually annotated symptoms mentions, together with
detailed annotation guidelines, consistency analysis and additional resources. For this track we plan also to release a multilingual version of the
corpus (English, Italian, Romanian, Catalan, Portuguese, French, Dutch, Swedish and Czech). This is a new challenge.
Track 3: Phenotype normalization (genetic conditions in pediatric patients) (Graciela Gonzalez, Ian Campbell, Davy Weissenbacher)
The dysmorphology physical examination is a critical component of the diagnostic evaluation in clinical genetics. This
process catalogs often minor morphological differences of the patient's facial structure or body, but it may also identify more general medical
signs such as neurologic dysfunction. The findings enable the correlation of the patient with known rare genetic diseases. Although the medical
findings are key information, they are nearly always captured within the electronic health record (EHR) as unstructured free text, making them
unavailable for downstream computational analysis. Advanced Natural Language Processing methods are therefore required to retrieve the information
from the records. This is a new challenge.
Track 4: Annotation Tool track (Rezarta Islamaj, Cecilia Arighi,
Lynette Hirschman, Martin Krallinger, Graciela Gonzalez)
Recognizing the need for freely available,
time-saving tools that help build quality gold-standard resources, the goal of BioCreative 2023 Annotation Tool Track is to foster development of such
biocuration annotation systems. This track calls for text mining developers to submit systems that are: 1) both publicly available, and offer local
setup options to allow for data with privacy concerns, such as clinical records, 2) able to support team annotation, and collaboration between
annotators to ensure data annotation quality, 3) able to annotate documents for triage, entities, and/or relations, and 4) able to integrate the
selected ontology, and provide search capabilities/browsing, as well as suggestions to the curator for the selected ontology. A select number of
systems will be showcased at the workshop.
Publication
The BioCreative VIII Proceedings will host all the submissions from participating
teams, and it will be freely available by the time of the workshop.
In addition, we are working with a
journal to host the BioCreative VIII special issue for work that passes their peer-review process. Invitation to submit will be sent after the
workshop.
Team Registration
Teams can participate in one or more of these tracks. Team registration will continue until final commitment is requested
by the individual tracks.
To register a team go to the Registration form. If you have restrictions accessing Google forms please send e-mail to
BiocreativeChallenge@gmail.com.
BioCreative Organizing Committee
- Dr.
Rezarta Islamaj, National Library of Medicine
- Dr. Cecilia Arighi, University of
Delaware
- Dr. Ian M. Campbell, Children Hospital of Philadelphia
- Dr. Graciela Gonzalez-Hernandez, Cedars-Sinai Medical Center
- Dr. Lynette Hirschman, MITRE
- Dr. Martin Krallinger, Barcelona Supercomputing
Center
- Dr. Davy Weissenbacher, Cedars-Sinai Medical Center
- Dr. Zhiyong Lu, National Library of Medicine
