Advanced Paper Analyzer takes a set of research papers and extracts its metadata to obtain information. It accesses Wikidata and ROR to expand the information and also has processes that compare the similarity between the abstracts taken from the papers and that analyze the possible topics the paper is about.
You have the choice to run the application in a container (note that you need a VNC-client) here or in your computer as follows :
- Clone the repository:
git clone https://gitlab.utc.fr/royhucheradorni/ia04.git
- Python
The code runs on Python 3.10, so it must be installed in the system to be able to use Advanced Paper Analyzer.
- Dependencies
Dependencies can be installed by using Poetry. You simply must go to the root directory of the repository and run:
poetry install
Or install all dependencies with pip using requirements.txt in the root directory of the repository by running:
pip install -r requirements.txt
- Grobid
Grobid is used to extract metadata from the papers, which are then used for further analysis. For this reason you must install either the full or light version of the Grobid 0.8.0 Docker image. To run Grobid use one of this commands depending on version you have:
Full image: https://hub.docker.com/r/grobid/grobid
docker pull grobid/grobid
Light image: https://hub.docker.com/r/lfoppiano/grobid/
docker pull lfoppiano/grobid
- Apache Jena Fuseki
Jena Fuseki is used to create the triple-store and the SPARQL endpoint, so it must be installed and run as described in the section to create the dataset
docker pull stain/jena-fuseki
- run Jena-fuseki and grobid with :
docker run -p 8070:8070 lfoppiano/grobid:latest-develop
docker run -p 3035:3030 -e ADMIN_PASSWORD=pw123 -e FUSEKI_DATASET_1=KG_dataset stain/jena-fuseki
it creates the dataset at the same time
- Run the script interface.py with the parameter 0
poetry run python interface.py 0
You can now :
-
PROCESS PDF WITH GROBID : process all the pdf in the directory Corpus_pdf to reformat the data/metadata in a XML format.
-
EXTRACT DATA : Extract the data (title, date, author) from the processed pdf and do some topic modeling and compute similarity between the abstract of each pdf.
-
Enrich DATA : Add more information coming from ROR and WIKIDATA (name, authors, organizations_founder of referenced papers).
-
INSERT DATA FROM RDF : Add all this data to KG server Jena-fuseki.
-
SUBMIT QUERY : in the input box, write your SPARQL queries and submit.
Example of queries :
- select each topic of which the papers have more than 0.90 probability of belonging to that topic :
- to request all the pair of article with more than 70% of similarity :
In order to display the Graphic User Interface running in a docker container, we create a VNC-server. Therefore, you will need to have a VNC-client software (such as RealVNC Viewer).
- Go to the location of the docker-compose and run :
docker-compose build
docker-compose up -d
-
Connect to the container using your VNC-client at the adress : localhost:5901 The password is : pw123
-
Open a terminal and execute :
poetry run python interface.py 1