Bibliographic record
Abstract
In this document, I have summarized and put into perspective three research orientations I have been developing in the past fifteen years (2006-2022).The first one concerns the way spoken language is adapted in literary works, a technique used by a number of famous writers (Dickens, Molière, and many others). I was interested in the specific strategy of the French-Canadian author Michel Tremblay, who used phonetic, syntactic and lexical features of spoken French-Canadian language not only as a means of mirroring social differences within the French Quebec community but also as a source of genuine literary effects involving in particular metarepresentational ones. I studied five plays by Tremblay over a thirty year period (1968-1998), extracting the idiolectal statistical contrasts between socially symbolic characters and discussing them in the context of textual structure.My second field of research concerns discourse relations. In order to show that they are more than just a long list of terms for labeling relations between discourse segments, I chose three illustrative examples. (I) So-called paratactic constructions, where there is no lexical marking of discourse articulation, exhibit in some cases (pseudo-declaratives and pseudo-imperatives) interesting properties, such as non-compositionality or special prosodic patterns. (ii) I examine critically a particular symbolic/statistical treatment of argumentative relations in NLP, aiming at measuring the (non-)alignment of discourse and argumentative relations between discourse segments in short normalized texts. I point out that the results delivered by the algorithms are difficult to interpret and misrepresent the real geometry of argumentation in the texts. (iii) Finally I present some aspects of the evolution of comparative discourse markers over a period of six centuries, tracing back their path of change from comparative to causal and concessive values.My last area of interest concerns the semantic status of discourse markers. My general goal was to bring closer the large literature on such markers and the more general literature on information layering, most notably the approaches on presuppositions, implicatures and side issues in the sense of Gutzmann and Turgay (2019). I show that the broad category of discourse markers (DM) gathers those indexicals which are relevant to discourse interpretation, either with respect to its organization (structuring DM) or to the monitoring of the speaker’s “stance”, i.e. her emotional, attentional or belief state, her perception of events in the discourse situation and her interaction with other participants. As a result, I divided DM into two subcategories: connectives and hic and nunc particles (HNP). Connectives convey discourse relations between the semantic objects they refer, such as illocutionary acts, states of affairs, belief states, etc. They correspond to well-known cases like but, because or therefore. Certain connectives behave as presupposition triggers. HNP refer to external or internal (psychological) events, to other participants or to the speaker's speech itself (hesitations, corrections). They are partly analogous to expressives, in the sense of Potts (2005, 2007), but include non-expressive terms and are anchored to the utterance situation even more strongly than expressives. I also consider the combinations of DM, reporting on recent work using association measures and work in progress addressing jointly their semantic, prosodic and statistical properties. This kind of approach is developped in the CODIM project framework (Compositionality and Discourse Markers). This project, which I coordinate, is funded by the ANR (https://www.codim-project.org/).
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.004 | 0.003 |
| Scholarly communication | 0.007 | 0.004 |
| Open science | 0.001 | 0.003 |
| Research integrity | 0.002 | 0.006 |
| Insufficient payload (model declined to judge) | 0.067 | 0.014 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".