Meeting minutes
Open Issues/Actions on W3C tracker
see issue/tracker https://
Presentation by Rana
On: Building a Corpus of Physical Health Data Disclosure on Twitter during the Covid-19 Pandemic
rana: users share personal data in social media networks (OSN) which has implications regarding privacy threats
rana: NLP can help with detecting such threats, more specifically: discrimination in job searchers, harassment, bullying, identity theft, misuse of health information
rana: detection of persona ldata or PII is a (current and topical) challenge
rana: PHDD: A Corpus of Physical Health Data Disclosure on Twitter during COVID-19
rana: tweets wre collected using keywords, hastags, regex, and were tagged based on criteria regarding health information or subject
rana: we published corpus in RDF and JSON; for RDF we created lightweight ontology "privacy tags for health information"
rana: DPV provides broad categories for personal data and health categories (physical, mental health)
rana: used HL7 concepts regarding confidentiality and sensitivity
… Health Level Seven International (HL7) is a standard for health data https://
rana: descriptive blog with example at https://
rana: future work includes use of supervised ML techniques for detection of health sensitive information, to notify users if shared content are sensitive, implement fine-grained access control mechanism
… Q&A / discussion
beatriz: why did you not use the HL7 concepts to tag the tweets?
rana: for this work, we focused on the data subject; whereas the concepts from HL7 regarding sensitivity or confidentiality are relevant for information providers
beatriz: how do you measure confidentiality?
rana: based on contents or related concept in sensitive data categories e.g. special categories in GDPR
paul: why did you break up the text into nouns, pronouns, etc.?
rana: this information is relevant and useful for NLP tasks
harsh: regarding HL7, is this the same as FHIR, or are they tw o separate things?
rana: we used HL7 v4 privacy and security ontology
harsh: do you think we should take up some concepts from HL7 related to privacy/sensitivity?
rana: HL7 is more focused on healthcare, whereas DPV is more generic
terms proposed by Beatriz and Rana
rana: See https://
beatriz: Rana and myself have been analysing fitness apps policies, and we extracted terms missing in DPV
… There are a lot of fine-grained concepts which might not be suitable to DPV
… Date of birth is relevant to age
… Some others are not immediately relevant e.g. water intake as they can be broken down into more related concepts e.g. number of glasses
… We seem to have a gap in concepts regarding Health, Medical Health, physical health, etc. and the concepts proposed
… We need to figure out the proper structure for this in terms of hierarchy
next meeting
beatriz: We will move to having a meeting next week 24-FEB 13:00 WET / 14:00 CET
… Tentative topic for the agenda is continued discussion on proposed topics
RDF Vocab tool by Nishad
nishad: zazuko/
nishad: I have added DPV to the tool, so all terms from DPV show up within the term/vocab finder
nishad: so when we update vocabulary, the urls must stay same for the updates