config-docker.yml

version: "0.8.1"

corpusPath: "./resources/dataset/dataseer/corpus"
templatePath: "./resources/dataset/dataseer/crfpp-templates/dataseer.template"
grobidHome: "/opt/grobid/grobid-home"
tmpPath: "/opt/grobid/grobid-home/tmp/"

# path to Pub2TEI repository as available at https://github.com/kermitt2/Pub2TEI
pub2teiPath: "/opt/Pub2TEI/"

gluttonHost:
gluttonPort: 

# entity-fishing server information for performing entity disambiguation
# for https, indicate 443 as port
entityFishingHost: cloud.science-miner.com/nerd
entityFishingPort: 443
#entityFishingHost: localhost
#entityFishingPort: 8090

# if true we use binary classifiers for the contexts, otherwise use a single multi-label classifier
# binary classifiers perform better, but heavier to use
useBinaryContextClassifiers: false

# sequence labeling model (identify data-related sections)
models:

  # model for zones
  - name: "dataseer"
    engine: "wapiti"
    #engine: "delft"
    wapiti:
      # wapiti training parameters, they will be used at training time only
      epsilon: 0.00001
      window: 20
      nbMaxIterations: 2000

  # classifier model, dataset binary (dataset or not dataset in the current sentence)
  - name: "dataseer-binary"
    engine: "delft"
    delft:
      # deep learning parameters
      #architecture: "gru"
      architecture: "bert"
      #embeddings_name: "word2vec"
      transformer: "allenai/scibert_scivocab_cased"

  # identification of the data type (first level hierarchy) 
  - name: "dataseer-first"
    engine: "delft"
    delft:
      # deep learning parameters
      #architecture: "gru"
      architecture: "bert"
      #embeddings_name: "word2vec"
      transformer: "allenai/scibert_scivocab_cased"

  # mention context classification (reuse binary for the moment)
  - name: "dataseer-reuse"
    engine: "delft"
    delft:
      # deep learning parameters
      #architecture: "gru"
      architecture: "bert"
      #embeddings_name: "word2vec"
      transformer: "allenai/scibert_scivocab_cased"

  # model for dataset mention recognition
  - name: "datasets"
    #engine: "wapiti"
    engine: "delft"
    wapiti:
      # wapiti training parameters, they will be used at training time only
      epsilon: 0.00001
      window: 20
      nbMaxIterations: 2000
    delft:
      # deep learning parameters
      #architecture: "BidLSTM_CRF"
      architecture: "BERT_CRF"
      #transformer: "allenai/scibert_scivocab_cased"
      transformer: "michiyasunaga/LinkBERT-basecased"
      #useELMo: true
      #embeddings_name: "glove-840B"
      runtime:
        # parameters used at runtime/prediction
        max_sequence_length: 200
        #max_sequence_length: 300
        batch_size: 20

  - name: "context"
    engine: "delft"
    delft:
      #architecture: "gru"
      #embeddings_name: "glove-840B"
      architecture: "bert"
      transformer: "michiyasunaga/LinkBERT-basecased"

  - name: "context_used"
    engine: "delft"
    delft:
      #architecture: "gru"
      #embeddings_name: "glove-840B"
      architecture: "bert"
      transformer: "michiyasunaga/LinkBERT-basecased"

  - name: "context_creation"
    engine: "delft"
    delft:
      #architecture: "gru"
      #embeddings_name: "glove-840B"
      architecture: "bert"
      transformer: "michiyasunaga/LinkBERT-basecased"

  - name: "context_shared"
    engine: "delft"
    delft:
      #architecture: "gru"
      #embeddings_name: "glove-840B"
      architecture: "bert"
      transformer: "michiyasunaga/LinkBERT-basecased"

# Limit the maximum number of requests (0, no limit)
maxParallelRequests: 0

# CORS configuration for the web API service
corsAllowedOrigins: "*"
corsAllowedMethods: "OPTIONS,GET,PUT,POST,DELETE,HEAD"
corsAllowedHeaders: "X-Requested-With,Content-Type,Accept,Origin"

server:
  type: custom
  applicationConnectors:
    - type: http
      port: 8060
      idleTimeout: 120 seconds
      acceptQueueSize: 2048
  adminConnectors:
    - type: http
      port: 8061
  registerDefaultExceptionMappers: false
  maxThreads: 2048
  maxQueuedRequests: 2048

  requestLog:
    appenders: []

# these logging settings apply to the service usage mode
logging:
  level: INFO
  loggers:
    org.apache.pdfbox.pdmodel.font.PDSimpleFont: "OFF"
    org.glassfish.jersey.internal: "OFF"
    com.squarespace.jersey2.guice.JerseyGuiceUtils: "OFF"
  appenders:
    - type: console
      threshold: INFO
      timeZone: UTC
      # uncomment to have the logs in json format
      #layout:
      #  type: json
#    - type: file
#      currentLogFilename: logs/datastet-service.log
#      threshold: INFO
#      archive: true
#      archivedLogFilenamePattern: logs/datastet-service-%d.log
#      archivedFileCount: 7
#      timeZone: UTC
      # uncomment to have the logs in json format
      #layout:
      #  type: json