Humanist Discussion Group, Vol. 39, No. 353.
Department of Digital Humanities, University of Cologne
Hosted by DH-Cologne
www.dhhumanist.org
Submit to: humanist@dhhumanist.org
[1] From: clarisse bardiot <clarisse.bardiot@univ-rennes2.fr>
Subject: [ADDA seminar] - Doug Reside - March 9th (72)
[2] From: Nathalie Aussenac-Gilles <nathalie.aussenac-gilles@irit.fr>
Subject: [CFP] First International Workshop on Extraction from Triplet Text-Table-Knowledge Graph and associated Challenge @ ESWC 2026 - deadline extended (222)
--[1]------------------------------------------------------------------------
Date: 2026-03-04 10:49:50+00:00
From: clarisse bardiot <clarisse.bardiot@univ-rennes2.fr>
Subject: [ADDA seminar] - Doug Reside - March 9th
Dear colleagues,
Next Monday (26/03/09), the STAGE seminar will feature a presentation by Doug
Reside.
Doug Reside - New York Public Library
Theatre Emulators: Preservation and Access for Digital Archives of Theatrical
Production
2026-03-09, 4 pm (Paris time) - T233 (hybrid)
Theatre makers have been using computers to write and stage
productions since the mid-1970s. However, much of the software and
hardware used in the first three decades of digitally-assisted
productions is now obsolete. Unlike other art forms (such as
computer games, photography, and written text), simple software
emulation is often insufficient for interpreting the data created
for lighting cues, sound effects, and projection designs. These
designs often depended on bespoke hardware to execute the production
files. In this presentation, New York Public Library theatre curator
Doug Reside will discuss his efforts to create an emulator for
theatrical productions. This sample recreated the lighting
instruments, physical spaces, and scenic designs in virtual reality,
allowing it to emulate the actual digital production files preserved
in theatre archives. Reside will also outline the still unsolved
problem of ensuring this emulator does not become obsolete itself.
Doug Reside is the Curator of the Billy Rose Theatre Division at the
New York Public Library for the Performing Arts, where he oversees
archival acquisitions, the Theatre on Film and Tape video
documentation program, exhibitions, programs, and researcher
support. He also founded a musical theatre writing program in 2013
for emerging theatre makers. Reside holds a PhD from the University
of Kentucky (2006). His book, Fixing the Musical: How Technologies
Shaped the Broadway Repertory, was published by Oxford University
Press in 2023.
Participation Details:
Time: Mondays from 4 PM to 6 PM Paris Time
Online, via ZOOM (registration required through this form):
https://docs.google.com/forms/d/e/1FAIpQLSc25z47cGfwqX8dgMvEdJ_OeLpaLG6DacvZYMMSr8vsjq2Ujw/viewform
If you have already registered, there is no need to fill out the form again.
The sessions will be recorded and made available on the From Stage to Data
website and YouTube channel: [https://stage-to-data.huma-num.fr](https://stage-
to-data.huma-num.fr/)
About the Seminar:
The STAGE seminar (From Stage to Data, the Digital Turn of Contemporary
Performing Arts Historiography) is led by Clarisse Bardiot as part of a European
research project. It offers an in-depth introduction to the challenges,
concepts, methods, and tools for the digital study of texts, images, and
cultural data.
The seminar explores the challenges and opportunities of using digital data in
artistic studies, particularly in the fields of performing arts, digital art
history and culture analytics, and demonstrates how digital humanities open new
research perspectives in the humanities and social sciences.
Sessions are held in English or French.
Best regards
Clarisse Bardiot
Clarisse Bardiot
Professeur des universités en études théâtrales
Département des Arts du spectacle
Université Rennes 2
http://www.clarissebardiot.info/
--[2]------------------------------------------------------------------------
Date: 2026-03-04 21:06:24+00:00
From: Nathalie Aussenac-Gilles <nathalie.aussenac-gilles@irit.fr>
Subject: [CFP] First International Workshop on Extraction from Triplet Text-Table-Knowledge Graph and associated Challenge @ ESWC 2026 - deadline extended
Call for Papers
First International Workshop on Extraction from Triplet
Text-Table-Knowledge Graph and associated Challenge
https://ecladatta.github.io/triplet2026/
in conjunction with the 23rd European Semantic Web Conference
(ESWC 2026)
https://2026.eswc-conferences.org/,
Dubrovnik, Croatia
Important dates:
- Submission deadline (extended): 13 March, 2026 (11:59pm,
AoE)
- Notifications: 31 March, 2026
- Challenge registration deadline: 15 March, 2026
- Challenge results submission: 10 April, 2026
- Camera-ready deadline: 15 April, 2026 (11:59pm, AoE)
- Workshop: Sunday 10 May OR Monday 11 May 2026
Motivation:
Understanding information spread across text and table is
essential for tasks such as question answering and fact checking.
Existing benchmarks primarily deal with semantic table
interpretation or reasoning over tables for question answering,
leaving a gap in evaluating models that integrate tabular and
textual information, perform joint information extraction across
modalities, or can automatically detect inconsistencies between
modalities.
This workshop aims to provide a forum for exchanging ideas between
the NLP community working on open information extraction and the
vibrant Semantic Web community working on the core challenge of
matching tabular data to Knowledge Graphs, on populating knowledge
graphs using texts and on reasoning across text, tabular data and
knowledge graphs. The workshop also targets researchers focusing
on the intersection of learning over structured data and
information retrieval, for example, in retrieval augmented
generation (RAG) and question answering (QA) systems. Hence, the
goal of the workshop is to connect researchers and trigger
collaboration opportunities by bringing together views from the
Semantic Web, NLP, database, and IR disciplines.
Scope:
The topics of interest include but are not limited to:
- Semantic Table Interpretation
- Automated Tabular Data Understanding
- Using Large Language Models (LLMs) for Information Extraction
- Generative Models and LLMs for Structured Data
- Knowledge Graph Construction and Completion with Tabular Data
and Texts
- Analysis of Tabular Data on the Web (Web Tables)
- Benchmarking and Evaluation Frameworks for Joint Text-Table Data
Analysis
- Applications (e.g. data search, fact-checking,
Question-Answering, KG alignment)
Submission Guidelines:
We invite two types of submissions:
1. Full research papers (12-15 pages) including references and
appendices
2. Challenge papers (6-8 pages) including references and
appendices
All submissions should be formatted in the CEUR layout format,
https://www.overleaf.com/latex/templates/template-for-submissions-to-ceur-
workshop-proceedings-ceur-ws-dot-org/wqyfdgftmcfw
This workshop is double-blind and non-archival. Submissions are
managed through EasyChair at
https://easychair.org/conferences/?conf=triplet2026.
All accepted papers will be presented as posters or as oral talks.
TRIPLET Challenge:
In recent years, the research community has shown increasing
interest in the joint understanding of text and tabular data,
often, for performing tasks such as question answering or fact
checking where evidences can be found in texts and tables. Hence,
various benchmarks have been developed for jointly querying
tabular data and textual documents in domains such as finance,
scientific publications, and open domain. While benchmarks for
triple extraction from text for Knowledge Graph construction and
semantic annotation of tabular data exist in the community, there
remains a gap in benchmarks and tasks that specifically address
the joint extraction of triples from text and tables by leveraging
complementary clues across these different modalities.
The TRIPLET 2026 challenge is proposing three sub-tasks and
benchmarks for understanding the complementarity between tables,
texts, and knowledge graphs, and in particular to propose a joint
knowledge extraction and reconciliation process.
#Sub-Task 1: Assessing the Relatedness Between Tables and Textual
Passages
The goal of this task is to assess the relatedness between tables
and textual passages (within documents and across documents). For
this purpose, we have constructed LATTE (Linking Across Table and
Text for Relatedness Evaluation), a human annotated dataset
comprising table–text pairs with relatedness labels. LATTE
consists of 7,674 unique tables and 41,880 unique textual
paragraphs originating from 3,826 distinct Wikipedia pages. Each
text paragraph is drawn from the same or contextually linked pages
as the corresponding table, rather than being artificially
generated. LATTE provides a challenging benchmark for cross-modal
reasoning by requiring classification of related and unrelated
table–text pairs. Unlike prior resources centered on table-to-text
generation or text retrieval, LATTE emphasizes fine-grained
semantic relatedness between structured and unstructured data.
The Figure below provides an example, using a web-annotation tool
we developed, of how we identify the relatedness between the
sentence containing the entity AirPort Extreme 802.11n
(highlighted in Orange) and the data table providing information
about output power and frequency for this entity. Participants are
provided with tables and textual passages that would need to be
ranked. The evaluation will use metrics such as P@k, R@k and F1@k.
Go to https://www.codabench.org/competitions/12776/
and enroll to participate in this Task.
#Sub-Task 2: Joint Relation Extraction Between Texts and Tables
The goal of this task is to automatically extract knowledge
jointly from tables and related texts. For this purpose, we
created ReTaT, a dataset that can be used to train and evaluate
systems for extracting such relations. This dataset is composed of
(table, surrounding text) pairs extracted from Wikipedia pages and
has been manually annotated with relation triples. ReTaT is
organized in three subsets with distinct characteristics: domain
(business, telecommunication and female celebrities), size (from
50 to 255 pairs), language (English vs French), type of relations
(data vs object properties), close vs open list of relation, size
of the surrounding text (paragraph vs full page). We then assessed
its quality and suitability for the joint table-text relation
extraction task using Large Language Models (LLMs).
Given a Wikipedia page containing texts and tables and a list of
predicates defined in Wikidata, a participant system should
extract triples composed of mentions located partly in the text
and partly in the table and disambiguated with entities and
predicates identified in the Wikidata reference knowledge graph.
For example, in the Figure below, an annotation triple
<Q13567390, P2109, 24.57> is associated with mentions
highlighted in orange (subject), blue (predicate) and green
(object) to annotate the document available at
https://en.wikipedia.org/wiki/AirPort_Extreme.
Similar to the Text2KGBench evaluation
(https://link.springer.com/chapter/10.1007/978-3-031-47243-5_14),
and because the set of triples are not exhaustive for a given
sentence, to avoid false negatives, we follow a locally closed
approach by only considering the relations that are part of the
ground truth. The evaluation then uses metrics such as P, R and
F1.
Go to https://www.codabench.org/competitions/12936/
and enroll to participate in this Task.
# Sub-Task 3: Detecting Inconsistencies Between Texts, Tables and
Knowledge Graphs
The goal of this task is to check the consistency of knowledge
extracted from tables and texts with existing triples in the
Wikidata knowledge graph. Different kind of inconsistencies will
be considered in this task. Participants to this task will be able
to report on their findings in their system paper.
See the Figure at https://ecladatta.github.io/images/triplet_annotation_tool.png
# Data & Evaluation:
For the first 2 sub-tasks, we have released a training dataset
with ground-truth annotations, enabling participant teams to
develop machine learning-based systems, and in particular for
training purposes and for hyperparameter optimizations and
internal validations.
A separate blind test dataset will remain private and be used for
ranking the submissions.
Participants should register on Codabench and then enroll for each
sub-task separately (Task 1: https://www.codabench.org/competitions/12776/
and Task 2: https://www.codabench.org/competitions/12936/).
Each team are allowed a limited number of daily submissions, and
the highest achieved accuracy will be reported as the team's final
result. We encourage participants to develop open-source
solutions, to utilise and fine-tune pre-trained language models
and to experiment with LLMs of various size in zero-shot or
few-shot settings.
# Challenge Important Dates:
- Release of training set: 13 February 2026
- Deadline for registering to the challenge: 15 March 2026
- Release of test set: 24 March 2026
- Submission of results: 10 April 2026
- System Results & Notification of Acceptance: 17 April 2026
- Submission of System Papers: 28 April 2026
- Presentations @ TRIPLET Workshop: May 2026
Workshop Organizers
- Raphael Troncy (EURECOM, France)
- Yoan Chabot (Orange, France)
- Véronique Moriceau (IRIT, France)
- Nathalie Aussenac-Gilles (IRIT, France)
- Mouna Kamel(IRIT, France)
Contact:
For discussions, please use our Google Group,
https://groups.google.com/g/triplet-challenge
The workshop is supported by the ECLADATTA project funded by the
French National Funding Agency ANR under the grant
ANR-22-CE23-0020.
--
Raphaël Troncy
EURECOM, Campus SophiaTech
Data Science Department
450 route des Chappes, 06410 Biot, France.
e-mail: raphael.troncy@eurecom.fr & raphael.troncy@gmail.com
Tel: +33 (0)4 - 9300 8242
Fax: +33 (0)4 - 9000 8200
Web: http://www.eurecom.fr/~troncy/
_______________________________________________
Unsubscribe at: http://dhhumanist.org/Restricted
List posts to: humanist@dhhumanist.org
List info and archives at at: http://dhhumanist.org
Listmember interface at: http://dhhumanist.org/Restricted/
Subscribe at: http://dhhumanist.org/membership_form.php
