Pragm.A.Lab

What is Prag.Ma.Lab.

The Pragmatic Annotation Lab (Pragm.A.Lab) aims to adopt methods and theories from Pragmatics and Social Psychology in the design of annotation schemes for Natural Language Processing. This integration has a broad range of applications, such as the adoption of annotation layers based on implicatures (Levinson, 2000; George et al, 2020) to study conversations on controversial topics or the measuring of misalignment in the prediction of emotions when observed from different standpoints (Braun et al, 2022). Pragm.A.Lab is conceived as a recurring initiative, with each edition dedicated to a specific theme and methodological integrations.

The first edition: Talking about another

The first edition of Pragm.A.Lab will take place at LREC, during the SoCon-NLPS workshop. The theme of the first edition will be the annotation of self and otherness in conversation.

Creating the other is a universal property of language use (Coseriu, 1975): whenever we communicate we are always talking with (alter) or referring to (alius) someone else (Jungbluth, 2015). This process is intrinsically dynamic: people speak to be understood and make an effort to understand their interlocutors (Schrott, 2021). The same interaction happens at the level of groups and communities whose self-representations arise from the relationship established with the other (Balibar, 2005).

This phenomenon is traceable in many NLP lines of research that study communication practices from several perspectives: persuasion (Boissin et al, 2025), code-switching (Claassen et al, 2025), micro-aggressions (Zhou et al, 2023), coping strategies (Troiano et al, 2024). However, an explicitly formalization of alterity as a core principle of language use remains underdeveloped.

The lab will focus on the conversation as the dynamic context where people and groups reciprocally co-shape themselves. The lab will be data-driven: starting from datasets selected by the participants we will discuss the identification of annotation schemes that can be generalized to different domains. The annotation schemes proposed during the lab will be made publicly available and support a community annotation task. Examples of research questions that could drive the discussion are the following: is the annotation of implicatures effective in explaining human-label variation in subjective tasks? Can the writer-reader misalignment be a proxy for the analysis of bias in content moderation?

We encourage the participation of researchers from diverse domains who are interested in the study of conversation from different research perspectives and methodological approaches, with the aim of promoting the exchange of expertise, practices, and suggestions.

How to participate

Pragm.A.Lab. will take place during the SoCon-NLPSI workshop. All its participants are welcomed to join the lab. Additional ways to participate are the following:

  • Share your data. You may submit a data sample via this form. Submissions can include any type of data, provided they are relevant to the workshop topic and do not exceed 100 items.

  • Lightning talk video. If you are unable to attend in person, you can still participate by submitting a lightning talk: a one-minute video presenting your take on the workshop topic. Videos can be sent to: socon-nlpsi-workshop-organizers.nlproc@uni-bamberg.de.