LLODREAM Presentation

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 18

Interlinking Lexicographic

Data in the MORDigital Project


Anas Fahad Khan, CNR-ILC, Ana Salgado, Centro de Linguística da
Universidade Nova de Lisboa (CLUNL) & Academia das Ciências de Lisboa,
Margarida Ramos, CLUNL, Sara Carvalho, CLUNL & Centro de Línguas,
Literaturas e Culturas, Laurent Romary, Automatic Language Modelling and
ANAlysis & Computational Humanities Inria de Paris (ALMAnaCH), Bruno
Almeida, CLUNL, Mohamed Khemakhem, ArcaScience, Raquel Silva,
CLUNL, Toma Tasovac, Belgrade Center for Digital Humanities (BCDH), Rute
Costa, CLUNL.
Introduction

● In this talk we will introduce the Portuguese national project MORDigital and present
some updates on its current progress.
● MORDigital aims to make a historic Portuguese-language dictionary, Diccionario da
Lingua Portugueza aka Morais, available as a digital resource.
● The project brings together some of the latest innovations in computational
lexicography, the digital humanities and linguistic linked data including work on
modelling lexicographic resources using standards such as OntoLex, TEI Lex-0 and LMF.
● It also innovative in its creation and use of pipelines for converting retrodigitised
dictionaries into computational lexical resources.
The Diccionario da Lingua Portugueza

● As the first Portuguese monolingual dictionary Morais was instrumental


in normalising the language and become the model for subsequent
Portuguese language dictionaries.
● It was influenced by other modern language dictionaries published in
Europe in the 16th and 17th centuries during the age of Enlightenment.
● Authorship of the 1st edition is attributed to Rafael Bluteau, a Portuguese
priest and lexicographer, whose Portuguese-Latin Vocabulary (10 vols.,
1712-1728) was the basis for Morais.
● Morais directly oversaw the 2nd (1813) and 3rd (1823) editions, which
greatly overhauled the dictionary.
The Diccionario da Lingua Portugueza
Frontispiece of Morais dictionaries (1789, 1813, 1823)

Two volumes Two volumes Two volumes


A to K, 752 pp. A to E, 889 pp. A to K, 952 pp.
L to Z, 541 pp. F to Z, 886 pp. L to Z, 872 pp.
Who was Morais?

● The dictionary in question is usually referred to by the name of its compiler, the renowned
Brazilian lexicographer ANTÓNIO DE MORAIS SILVA (1757?–1824)
● Born in Rio de Janeiro, Morais graduated in Civil and Canon Law from the University of
Coimbra in Portugal.
● After being accused of heresy by the Inquisition, he fled to England and devoted himself to
the study of languages; it was here that he planned the structure of his future dictionary
● Morais moved to Brazil in 1794 where he entered the judiciary and held the position of judge
in the Bahia Court of Appeal (a position from which he soon resigned). He subsequently
moved to Muribeca, in Pernambuco, where he lived until his death on April 11, 1824
● We can’t find any images of Morais himself!
MORDigital

● MORDigital – Digitalização do Diccionario da Lingua Portugueza de António de


Morais Silva [PTDC/LLT-LIN/6841/2020] is a project financed by the Portuguese
National Funding agency through the FCT – Fundação para a Ciência e Tecnologia
● Although it is a Portuguese national project, it also includes collaborators from all
over Europe.
● The main aim of the project is to convert three editions (1789; 1813; 1823) of Morais
into a structured lexical resource in both TEI-XML (based on the ISO LMF standard)
and in RDF (based on the OntoLex-Lemon model and its recent extensions).
● These editions will also be made available via an online interface on the website (at
the moment only PDFs are available).
MORDigital - The website
Workflow
Optical Character Recognition (OCR) Corrections

● We already had access to OCR’ed


versions of the dictionary editions at the
beginning of the project.
● These needed to be post-corrected. For
this we decided to use ABBYY
FineReader.
● An established tool with user-friendly
interface.
● It creates editable, searchable files.
Produces several output formats (RTF,
DOCX, PDF, HTML, XML, etc.).
● The tool preserves typographical
features.
Noise produced by the OCRization Description Example

Archaic characters The long /s/ is recognised as a lower- f instead of ſ


cased /f/ , or /j/ if the original is typed in
italic
j instead of ∫

Case-sensitivity Lower- and upper-case characters can I instead of i


be mixed up

Wrong characters inserted Characters are misrecognised as wrong rn instead of m


characters

Segmentation errors Different spacings between lines, words. temtres instead of tem tres
Misrecognition of white-spaces

Ligature A combination of two or three fi


characters set as a single glyph

Ink spots Text has ink spots, which prevents both


human and machine from reading

Costa et al. MDTT 2022


Structuring of Digital Editions

● For the structuring of the digital editions we are using GROBID-Dictionaries


● This is a machine learning library for structuring digitised lexical resources
and entry-based documents with encyclopedic or bibliographic content.
● In particular, it allows the automatic parsing, extraction and structuring of
lexical information from PDFs.
● GROBID-Dictionaries takes as input lexical resources digitised in PDF format
and generates a TEI-encoded hierarchy of the different text structures which
it has recognised.
TEI and OntoLex

● The TEI-XML sources will subsequently be converted to OntoLex (both the


original model and its follow-up modules) using an XSLT stylesheet.
● Having an RDF version available allows us to make the dictionary available
using both a dedicated platform and via a SPARQL endpoint.
● In addition, the RDF versions of each edition of the dictionary will be published
on the LLOD cloud. This will be an important contribution to adding
Portuguese language resources to the cloud.
● At the end of the project, based on our experiences, we will draft technical
guidelines to help lexicographers and digital humanists carry out these tasks.
An Example Entry

Almeida et al. COMHUM 2022


Organising knowledge

Costa et al. MDTT 2022


Organising knowledge
Organising knowledge

OntoDomLab-Med: https://github.com/sarampcarvalho/OntoDomLab-Med
Ongoing Work

● Start OCR corrections on the two other editions of the dictionary


● Run the GROBID tool iteratively on the corrected output of the OCR of the first edition to
ensure a correct TEI-XML (LMF) encoding of the different components of single entries (e.g.,
authoritative examples, collocations, )
● Start testing the XSLT transformation to OntoLex and its extensions
● In addition throughout the project our approach foresees a rigorous linguistic treatment of the
source texts. This will make it possible to organise and structure the lexicographic
components, and to elicit lexical relationships between various elements.
● We also propose combining semasiological and onomasiological approaches in our treatment
of the different editions of Morais. For this we foresee the inclusion of ontologies (e.g.
diasystematic marking, namely domain labels, registers and part of speech categories).
Thank You! Obrigado! Ačiū!
Anas Fahad Khan: [email protected]

Ana Salgado: [email protected]

Rute Costa: [email protected]

Sara Carvalho: [email protected]

Laurent Romary: [email protected]

Bruno Almeida: [email protected]

Margarida Ramos: [email protected]

Mohamed Khemakhem: [email protected]

Toma Tasovac: [email protected]

Raquel Silva: [email protected]

Toma Tasovac: [email protected]

You might also like