<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body>
<p>Dear All, <br>
</p>
<p>the next JenLIng talk is coming up this <b>Wednesday, January
18th, 18:00</b></p>
<p>Where? either in person at Fürstengraben 30, Seminarraum 1.
OG., or via zoom with the following link: <span class="HALYaf
XQINac R21Rlc KKjvXb" role="tabpanel" id="tabEventDetails"><a
class="moz-txt-link-freetext"
href="https://uni-jena-de.zoom.us/j/92382391570">https://uni-jena-de.zoom.us/j/92382391570</a></span><span
class="HALYaf XQINac R21Rlc KKjvXb" role="tabpanel"
id="tabEventDetails"></span><span class="HALYaf XQINac R21Rlc
KKjvXb" role="tabpanel" id="tabEventDetails"> , Passcode:
JenLing . We hope to go for drinks and more linguistic
discussions afterwards, so do come in person!</span><span
class="HALYaf XQINac R21Rlc KKjvXb" role="tabpanel"
id="tabEventDetails"><br>
</span></p>
<p><b>Felix Bildhauer (IDS Mannheim/HU Berlin) and Roland Schäfer
(FSU Jena) </b>will give a talk with the title<b><i>:<br>
</i></b></p>
<p><b><i>Discovering Written Registers</i></b><b><i><br>
</i></b><br>
In this talk, we present a novel approach in corpus-based
data-driven register research (developed in project A04 "Situated
Syntax" of the CRC 1412 "Register" at HU Berlin). In existing
corpus-heavy approaches to register (prominently Douglas Biber’s
MDA), it is always assumed that the set of registers represented
in the corpus is known beforehand (except in very recent work by
Biber et al.). For situations in which this assumption does not
hold because of the size and the composition of the corpus (e.g.,
for large web corpora such as the DECOW web corpus of German), a
principled approach to discovering the registers instantiated in
the corpus has not yet been proposed (to the best of our
knowledge). This is most likely because registers are obviously
hard to detect latent dimensions of texts. Based on a fully
probabilistic model of registers as sets of (probabilities of)
linguistic signs (lexical or grammatical) used in certain types of
situations and thus occurring in texts written in such situations,
we present a discovery procedure which starts with the fully
automatic identification of register candidates based on the
distributions of linguistic signs across texts. The potential
registers discovered in this procedure are then manually annotated
for situational-functional parameters (such as formality or
interactivity) in order to find texts which (i) show a
characteristic distribution of linguistic signs and (ii) have
characteristic situational-functional properties. Only the
distributions of signs and the situational-functional properties
together constitute true register candidates in our view.<br>
<span class="HALYaf XQINac R21Rlc KKjvXb" role="tabpanel"
id="tabEventDetails"></span><br>
<span class="HALYaf XQINac R21Rlc KKjvXb" role="tabpanel"
id="tabEventDetails"></span><span class="HALYaf XQINac R21Rlc
KKjvXb" role="tabpanel" id="tabEventDetails"></span></p>
<p><span class="HALYaf XQINac R21Rlc KKjvXb" role="tabpanel"
id="tabEventDetails">Looking forward to seeing you there! <br>
</span></p>
<p><span class="HALYaf XQINac R21Rlc KKjvXb" role="tabpanel"
id="tabEventDetails">Ruprecht</span></p>
Prof. Dr. Ruprecht von Waldenfels<br>
Institutsdirektor<br>
Institut für Slawistik und Kaukasusstudien<br>
Friedrich-Schiller-Universität Jena<br>
Ernst-Abbe-Platz 8<br>
D-07743 Jena<br>
<br>
Tel. +49 3641-9 44720<br>
Mob +49 163 230 34 23<br>
skype: rvwaldenfels<br>
<br>
</body>
</html>