Table of contents

• What is DBpedia Spotlight
• Run DBpedia Spotlight as a single Docker container
  • Setup Docker container
  • Stop the Docker container
• Run DBpedia Spotlight Multilingual
  • Setup Docker Compose
  • Stop Docker Compose
• Supported Docker versions
• Documentation

DBpedia Spotlight is a tool for automatically annotating mentions of DBpedia resources in text, providing a solution for linking unstructured information sources to the Linked Open Data cloud through DBpedia.

The dbpedia/spotlight-multilingual-databus is a docker image to run the DBpedia Spotlight server with the most recent language models, downloaded from the DBpedia Databus repository (as shown in Figure 1)., e.g., English (en), German (nl), Italian (it), etc.

Figure 1. DBpedia Databus and DBpedia Spotlight

The following are the instructions to run a DBpedia Spotlight service for one or more language models and to deploy a web page to annotate text with any of the running services.

1. Create a volume to persist models

    docker volume create spotlight-models
   

2. build docker image

    cd multilingual-databus
    docker build -t dbpedia/dbpedia-spotlight
   

3. Run docker image

    docker run -tid \
    --restart unless-stopped \
    --name dbpedia-spotlight.[LANG] \
    --mount source=spotlight-models,target=/opt/spotlight \
    -p 2222:80 \
    dbpedia/dbpedia-spotlight \
    spotlight.sh [LANG]      
   

Where [LANG] is a two-digit code to define the language model to be run e.g., en, de, it.

4. Send a query to DBpedia Spotlight

    curl http://localhost:2222/rest/annotate \
    --data-urlencode "text=President Obama called Wednesday on Congress to extend a tax break for students included in last year's economic stimulus package, arguing that the policy provides more generous assistance." \
    --data "confidence=0.35" \
    -H "Accept: text/turtle"
   

The "Accept: text/turtle" returns a NIF output but this option could be changed by "Accept: application/json" to returns a JSON output format.

To run more than one DBpedia Spotlight service, repeat the same step 3 changing the port number (-p 222X:80) and [LANG] e.g., http://localhost:2222 for English ([LANG]=en) language and http://localhost:2223 for German ([LANG]=de) language.

• It is possible to monitor each DBpedia Spotlight service through docker tool such as:

  • docker logs dbpedia-spotlight.[LANG] : Displays the log information for the corresponding service
  • docker stats dbpedia-spotlight.[LANG]: Shows the statistics (e.g., the amount of memory and CPU) for the corresponding service

An option to stop the DBpedia Spotlight server is by the commands:

 docker kill dbpedia-spotlight.[LANG]
 docker rm dbpedia-spotlight.[LANG]

The docker kill command will stop the running container and the docker rm command will remove the container. The dbpedia-spotlight corresponds to the name given with the --name option of the docker run command.

The DBpedia Spotlight Multilingual runs over a web application to provide a visual interface, annotating text for different languages. The interface provides features such as type selection (defined by DBpedia, Freebase, Schema or user defined, Figure 2-A), confidence value selection (from 0.0 to 1.0, Figure 2-B) and the n-best candidates (Figure 2-C).

Figure 2. Example of DBpedia Spotlight web application

The web application provides support for n-DBpedia Spotlight services, depending on the resource of the host machine (Figure 3-a). Each DBpedia Spotlight service is a docker container running in the same network (Figure 3-b), facilitating the communication between the web server and each DBpedia Spotlight service. The independency between each service helps to monitor the resources consumed by each one via docker stats or any monitor tool.

Figure 3. Example of DBpedia Spotlight Multilingual running in a Docker network over a web application

1. Create a new docker network (if it was not created before)

    docker network create spotlight-net
   

2. Create a volume to persist models (if it was not created before)

    docker volume create spotlight-models
   

3. Run Docker Compose file:

    cd multilingual-databus
    docker-compose -f spotlight-compose.yml up -d
   

   The Docker compose file (spotlight-compose.yml) defines one or more DBpedia Spotlight service through the following configuration block:

    spotlight.[LANG]:
    image: [DOCKER-IMAGE]
    container_name: dbepdia-spotlight.[LANG]
    volumes:
       - spotlight-model:/opt/spotlight
    restart: unless-stopped   
    ports:
       - "0.0.0.0:2224-2250:80"  
    networks:
       - spotlight-net          
    command: /bin/spotlight.sh [LANG]
   

Where

• [LANG]: is a two digits code to define the language model e.g., en, de, it.

• [DOCKER-IMAGE]: the name of the docker image.

  The sites.xml file defines the models available on the web application (<text> .... </text> tag is optional, it is used to provide an example text), at this moment it is updated by the user based on their necessities (the number of services run)

   <site>
           <i18n>[LANG]</i18n>
           <language>German</language>
           <url>http://[IP]:2222/[LANG]/rest</url>
           <text>Erstmals 1237 urkundlich erwähnt, war Berlin im Verlauf der Geschichte und in verschiedenen Staatsformen Hauptstadt Brandenburgs, Preußens und des Deutschen Reichs. Faktisch war der Ostteil der Stadt Hauptstadt der Deutschen Demokratischen Republik. Seit der Wiedervereinigung im Jahr 1990 ist Berlin gesamtdeutsche Hauptstadt mit Sitz des Bundespräsidenten seit 1994, des Deutschen Bundestags seit 1999 sowie des Bundesrats seit 2000.</text>
   </site>
  

Where:

• <i18n>: it is the same two digits code as [LANG] e.g., en, de, it.
• <language>: it is a label to denotes the language to be used e.g., English, German, Italian, etc.
• <url>: by default, the web application runs over the 2222 port but, this could be changed in the compose file. [LANG] defines the language model to be queried.
• [IP]: Defines the IP address of the host machine or it could be set as localhost or 0.0.0.0

The sites.xml file is copied to the web application docker image through the directive:

    ....
    volumes:
       - [PATH_TO_FILE]/sites.xml:/var/www/html/config/sites.xml 
    ...

4. Run Docker Compose file

    docker-compose -f spotlight-compose.yml up -d
   

5. In a web browser access to http://localhost:2222/

6. To query via curl

    curl http://localhost:2222/[LANG]/annotate \
    --data-urlencode "text=President Obama called Wednesday on Congress to extend a tax break for students included in last year's economic stimulus package, arguing that the policy provides more generous assistance." \
    --data "confidence=0.35" \
    -H "Accept: text/turtle"
   

An option to stop the DBpedia Spotlight server is by the commands:

docker-compose -f spotlight-compose.yml stop

The stop command will stop the running containers. The spotlight-compose.yml corresponds to the file name used to run the compose file.

This image is officially supported on Docker version 19.03.11.

Please see the Docker installation documentation for details on how to upgrade your Docker daemon.

Documentation for this image is stored in GitHub repo.