PADI-web for animal disease surveillance

PADI-web pipeline involves six steps ranging from online news collection to the extraction of epidemiological features: (1) data collection, (2) data processing, (3) data classification, (4) sentence classification, (5) information extraction and (6) user notification

PADI-web collects news articles based on Google News due to its international coverage and flexible RSS feeds. To detect news dealing with dedicated diseases (e.g., avian influenza, West Nile virus, African swine fever) or unknown diseases (i.e., Disease X), the RSS feeds use specific keywords (e.g., disease names, association of terms on hosts and symptoms).

The collected news items are automatically classified as "relevant" or "irrelevant" using machine learning techniques. The relevant news corresponds to recent or current infectious animal health events. Moreover, a more fine-grained classification of documents and sentences using machine learning approaches has been implemented to highlight more accurate topics. To identify key pieces of epidemiological information in the news (i.e., location and date of outbreaks, affected hosts, their numbers and encountered clinical signs, and so forth), PADI-web integrates information extraction (IE) methods. The PADI-Web pipeline detailed in our scientific publications collected more than 490,000 news items dealing with animal health to extract the epidemiological events. Finally, our system proposes automatic notifications to end-users.

PADI-web: pipeline

In the new extensions of PADI-Web, five types of extractions are proposed to end users:

  • Events extracted in relevant articles based on locations extracted with spaCy;
  • Events extracted in relevant articles based on locations extracted with spaCy learnt with labeled data;
  • Events extracted at the beginning of the articles;
  • Events extracted in Outbreak articles (i.e., document-based classification);
  • Events extracted in Outbreak articles (i.e., document-based classification) and Current event sentences (i.e., sentence-based classification).

Examples of diseases monitored: Avian Influenza (AI), African swine fever (ASF), West Nile Virus (WN), TBE, RVF, PPR, CCHF, CHIK, Brucella, Bluetongue, Usutu, FMD, LSD, LYME, SBV, AMR, etc.

Reference (Agence pour la Protection des Programmes - APP): IDDN.FR.001.130023.000.S.C.2019.000.10000 (22/03/2019)

Software: https://padi-web.cirad.fr/en/

PADI-web: Automatic notifications to end-users