Introduction

The WoPoss project

• A world of possibilities. Modal pathways over an extra-long period of time: the diachrony of modality in the Latin language . SNF project n° 176778
• The team:
  • PI: Francesca Dell’Oro
  • Senior researcher: Helena Bermúdez Sabel
  • PhD student and project assistant: Paola Marongiu

The WoPoss project

• Research question: how have Latin modal markers evolved during a long period of time (1000 years)?
• Method: automatic linguistic annotation + manual semantic annotation of a representative corpus

WoPoss as an extrapolable use case

• Generic goals
  • to have an annotated corpus that is useful and shareable
  • to have a GUI that is useful and intuitive

WoPoss as an extrapolable use case

• Generic challenges
  • source text retrieval: copyright, philological quality, heterogeneity of formats
  • preprocessing: orthographic variants, typographical conventions, abbreviations, editorial information
  • annotation: pipeline, tool-dependency, formats
  • publication: formats

An open science workflow

Corpus preparation

• Text retrieval (online, open-source)
• Homogenization

Source example: TEI-Epidoc

Source example: HTML

From “anything” to plain text

• Analysis of each file
• Conversion of typographical conventions and/or markup to pseudo-markup
• Plain text output

Automatic linguistic analysis

• Input: plain text
• Method: StanfordNLP library for Python (Stanza)
• Output: CONLL-U

CONLL-U

Manual annotation

Output: UIMA CAS XMI

• Exportation of results of annotation process: UIMA CAS XMI
  • Unstructured Information Management Applications: standard for annotation
  • Feature structures are represented in the UIMA Common Analysis Structure (CAS)

XMI snippets

From XMI to TEI

• More widespread format than UIMA CAS in the DH community
• More suitable for editorial information

Correction and enrichment

• Validation of the annotation
• Correction of textual issues
• Transformation of pseudo-markup into TEI elements
• Addition of metadata (DHTK, Picca & Egloff 2017)

Diachronic semantic maps
of modal markers

• Initial source: synthesis of lexicographical works
• To be reviewed with the results of the corpus-based, empirical analysis

Multi-format publishing
as a re-usability strategy

Resource sharing

• Datasets
  • Diachronic modal semantic maps in JSON, SVG and PNG. Interactive version online
  • Plain text version of the source texts
  • Automatic annotation results in CONLL-U format

Resource sharing

• Datasets
  • Manual semantic annotation results in:
    • UIMA CAS XMI
    • TEI
    • RDF/XML (see LiLa: Linking Latin)
• GUI
• Web application
  • GUI-based access to the dataset and analysis functions as an eXist-DB application

Conclusions

Multi-format publishing
for different specialist groups

• Use of standardized formats in the workflow
• External software dependencies: free and open source
• Creation of customized programs tailored to the project specifications: open source
• Open science workfow and FAIR principles (both during development and results)

Thank you!