<html xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:w="urn:schemas-microsoft-com:office:word" xmlns:m="http://schemas.microsoft.com/office/2004/12/omml" xmlns="http://www.w3.org/TR/REC-html40">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-2">
<meta name="Generator" content="Microsoft Word 15 (filtered medium)">
<style><!--
/* Font Definitions */
@font-face
{font-family:"Cambria Math";
panose-1:2 4 5 3 5 4 6 3 2 4;}
@font-face
{font-family:Calibri;
panose-1:2 15 5 2 2 2 4 3 2 4;}
@font-face
{font-family:Aptos;}
@font-face
{font-family:Consolas;
panose-1:2 11 6 9 2 2 4 3 2 4;}
@font-face
{font-family:Times;
panose-1:2 2 6 3 5 4 5 2 3 4;}
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
{margin:0cm;
font-size:12.0pt;
font-family:"Aptos",sans-serif;}
a:link, span.MsoHyperlink
{mso-style-priority:99;
color:#0563C1;
text-decoration:underline;}
pre
{mso-style-priority:99;
mso-style-link:"HTML Vorformatiert Zchn";
margin:0cm;
font-size:10.0pt;
font-family:"Courier New";}
p.MsoListParagraph, li.MsoListParagraph, div.MsoListParagraph
{mso-style-priority:34;
margin-top:0cm;
margin-right:0cm;
margin-bottom:0cm;
margin-left:36.0pt;
font-size:12.0pt;
font-family:"Aptos",sans-serif;}
span.HTMLVorformatiertZchn
{mso-style-name:"HTML Vorformatiert Zchn";
mso-style-priority:99;
mso-style-link:"HTML Vorformatiert";
font-family:Consolas;}
p.bibliography, li.bibliography, div.bibliography
{mso-style-name:bibliography;
mso-margin-top-alt:auto;
margin-right:0cm;
mso-margin-bottom-alt:auto;
margin-left:0cm;
font-size:12.0pt;
font-family:"Aptos",sans-serif;}
span.E-MailFormatvorlage23
{mso-style-type:personal-reply;
color:black;}
.MsoChpDefault
{mso-style-type:export-only;
font-size:10.0pt;
mso-ligatures:none;}
@page WordSection1
{size:612.0pt 792.0pt;
margin:70.85pt 70.85pt 2.0cm 70.85pt;}
div.WordSection1
{page:WordSection1;}
/* List Definitions */
@list l0
{mso-list-id:674766390;
mso-list-template-ids:1297881746;}
@list l0:level1
{mso-level-tab-stop:36.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l0:level2
{mso-level-tab-stop:72.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l0:level3
{mso-level-tab-stop:108.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l0:level4
{mso-level-tab-stop:144.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l0:level5
{mso-level-tab-stop:180.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l0:level6
{mso-level-tab-stop:216.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l0:level7
{mso-level-tab-stop:252.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l0:level8
{mso-level-tab-stop:288.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l0:level9
{mso-level-tab-stop:324.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l1
{mso-list-id:1609315210;
mso-list-template-ids:-382459850;}
@list l1:level1
{mso-level-tab-stop:36.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l1:level2
{mso-level-tab-stop:72.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l1:level3
{mso-level-tab-stop:108.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l1:level4
{mso-level-tab-stop:144.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l1:level5
{mso-level-tab-stop:180.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l1:level6
{mso-level-tab-stop:216.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l1:level7
{mso-level-tab-stop:252.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l1:level8
{mso-level-tab-stop:288.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
@list l1:level9
{mso-level-tab-stop:324.0pt;
mso-level-number-position:left;
text-indent:-18.0pt;}
ol
{margin-bottom:0cm;}
ul
{margin-bottom:0cm;}
--></style>
</head>
<body lang="DE" link="#0563C1" vlink="purple" style="word-wrap:break-word">
<div class="WordSection1">
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US">Dear All,<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US">a week ago, I posted a mail with issues concerning annotations, in particular semantic ones. I thank those few colleagues who have contributed to that discussion, and I have
been waiting over the week if there might come more. As this hasn’t been the case, I’d like to summarize some points, but also to become more precise as for the more particular background. I’m sorry that, maybe, I hadn’t been clear enough, and apologize for
this long mail.<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US">From the few reactions I got (from four colleagues) I dare infer that the topic is either considered irrelevant or it hasn’t been realized yet. The issue I caught up with
was a discussion from October 2023, in which in particular Volker Gast explained that semantic annotation yields astonishingly divergent results among annotators even for distinctions which seem to be intuitively quite clear (and, as I suppose, the researchers
had defined them well before they presented material to informants/annotators). In this line, my request primarily concerned semantic annotation.<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US"> In summarizing the responses this week, one result is that there were none concerning semantic annotation. There was consensus that, in principle, the burden
is on those writing the guidelines (or a codebook). I agree with this, in principle. The problem is, however, that some annotations are done for the purpose of exploring things that are too subtle or too much off the radar to be considered in mainstream annotation
guidelines, or designs. I come back to this below.<o:p></o:p></span></p>
<p class="MsoNormal" style="text-indent:35.4pt"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US">Apart from the linguist’s task to provide clear guidelines of annotation, Christian Lehmann raised another requirement, namely that there be “</span><span lang="EN-US">a
complete linguistic description of the language” in question. One might assume that this is possible for morphosyntax and phonology (provided one breaks down all concepts to “atomic” oppositions; see below), at least for languages we have enough data of and
a body of experts who have already worked on them (and probably neglecting much of sociolinguistic variation and abstracting from diachronic change, e.g. by comparing even 19<sup>th</sup>-century English and PDE). I understand this postulate (exhaustive description)
as being based on descriptive concepts (for each particular language, or rather: language stage). Thus, how can this be made compatible with comparative concepts, which Martin draws attention to? From all the discussion I am able to recall for this topic I
learned that descriptive and comparative concepts each have their own justification (depending on one’s goals), but they are usually incommensurable. Moreover, as far as I understand (and remember from numerous examples), comparative concepts are, as a rule,
related to structural notions relevant for phonology and morphosyntax; they thus rather concern notions like “root, morpheme, affix, indexing vs flagging” etc. – some of which may not be comparative concepts, after all, but they are meant to make linguists
capable of speaking about the “same” things in linguistic comparison in terms of structure.<o:p></o:p></span></p>
<p class="MsoNormal" style="text-indent:35.4pt"><span lang="EN-US">But what about notions like, e.g., those specifying different kinds of illocution (= the speaker’s communicative intention by saying/uttering U), or information source? It has often been emphasized
that, probably, nobody would have started looking for evidential (or: information source) marking in (west) European languages if there hadn’t been linguists (with a European/western background) like Boas, Sapir and some people working on the Balkans (since
the 19<sup>th</sup> c.) and, later, on Amazonian languages who noticed that for the description of these languages it is highly relevant to take account of either bound morphology or of usage types of verbal constructions (e.g., anterior/perfect grams) because
these marked distinctions could not be adequately described by notions like mood, modality, or tense (though many tried to squeeze them into those categorial or grammatical distinctions). That is, we get those “new distinctions” that belong to a different
notional dimension than those paradigmatic oppositions you know from “classical languages”, and we get them because they are part and parcel of the morphosyntactic “outfit” of those “more exotic” languages or, at least, of their function range (e.g. as extensions
of the perfect). And once we got them we get new notional (semantic-pragmatic) distinctions which, on the level of
<u>functions</u> of utterances, you assume different kinds of information source to be discernible also in languages that don’t have such clear-cut means of marking them in their morphosyntax.<o:p></o:p></span></p>
<p class="MsoNormal" style="text-indent:35.4pt"><span lang="EN-US">The tertium comparations now concerns only functions, irrespective of the way they can be expressed. So, you can also investigate, say, sentence adverbs (or whatever one wants to call by the
vague term “particle”) from the point of view of information source marking. Sentence adverbs are not morphosyntax, they are operators on propositions and illocutions (and might be subclassified accordingly). If you want to annotate their functions in samples
of authentic speech, you normally adopt some commonly accepted classification of functions, but actually this might turn out only a bone-fide application of what linguists (i.e. experts of certain language and/or of certain linguistic phenomena) think is a
valued approximation to the linguistic reality of speakers of the respective languages. How do we know that this is exhaustive, or at least sufficient? Concomitantly, what will we do if we get rather divergent results by different annotators? Here we have
a parallel to the problem discussed by Volker Gast in 2023. We may keep the burden with the researchers who write the codebook, but how should they know whether (and which) problems arise because of a bad codebook or because the annotators didn’t understand
the distinctions made?<o:p></o:p></span></p>
<p class="MsoNormal" style="text-indent:35.4pt"><span lang="EN-US">Research in AI (and in relevant tools) has started dealing with ‘human label variation’, that is with annotation disagreements: “Human label variation arises when annotators assign different
labels to the same item for valid reasons, while annotation errors occur when labels are assigned for invalid reasons.” (Weber-Genzel et al. 2024: 2256) Here, variation is distinguished from errors (this includes misunderstanding an instruction). Simultaneously,
the amount of observed variation between annotators is taken as a signal that something is in the data which needs to be explained as their objective property, and not in the sense that one group of annotators is “right”, another one is “wrong”. Almost all
examples I have read about concern clause linkage and the semantic relation between interrelated clauses (with and without connectives). It thus has mainly been observed with distinctions relevant for semantic annotation.<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">Furthermore, Martin remarks that we need “</span><span lang="EN-US">clear and simple definitions of annotation categories”. I totally agree. I also agree that notions like ‘mood’ and ‘subordination’ are highly complex
(and usually fraught with traditions that are often rather an obstacle than helpful). In my view, the consequence must be that such complex notions are broken down into simpler oppositions. Thus, in order to get an empirically valid picture of different strategies
and degrees of subordination in</span><span lang="EN-US"> a language (or rather: its corpus examples) I would start with coding parameters like those pointed out in Verstraete 2007. I would complement them with indicators of those differences that have been
pointed out w.r.t. a gradient between quoted/direct and indirect speech [Verstraete deliberately didn’t deal with them]. What we then get is a grid of annotational variables that is complex for its number of distinctions; moreover, it needs really good training
(especially if you want to “hire” annotators), first of all because many of these simpler distinctions are not immediately obvious (or intuitive) and they anyway require the annotator to interpret each example in a sufficiently rich context. (This is similar
to what you need to do if you annotate differences of referential status of or (in)definiteness – often you cannot determine that without a broader context.) In addition, even some rich context might not suffice to assign exactly one label from your value
set of the given criterion (all of which are specified in your codebook).<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"> The same could be said about ‘mood’, on condition that linguists sufficiently specify what they mean by it (which often is not the case). In fact, we can do without this notion, because it can probably
be shown that what linguists usually call ‘mood’ (especially “analytic mood”) is just a way to refer to the function of diverse function words (some of them are good examples of clitics, others are not) to indicate illocutionary force, and simultaneously they
often restrict marking of tense (and “synthetic mood”) on the verb (or the VP) from the inventory of TMA marking admissible in the respective language. This is where the function of such “mood” markers meets with the notion of ‘complementizer’, i.e. another
complex notion (to get a representative picture of “intersections” between what people call ‘mood’ and ‘complementizer’ cf. Wiemer 2023a, 2023b). Even if you seem to have a good definition of what a (canonical) complementizer is [it is certainly not a comparative
concept, as isn’t ‘mood’!], you need to break it down into more “atomic” distinctions, and among these we find illocutionary functions and various kinds of stance (epistemic, volitional, etc.), which have to be compatible with whatever is their complement-taking
predicate. However, do we have an exhaustive list of illocutionary functions? Moreover, to check them (in order to assign a value from your annotation grid) you again often need to rely on rich enough context. Here a simple example:<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">Polish<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">(1) <i>Niech Pan si±dzie!</i><o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"> ‘May you sit down!’<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">The function word <i>niech</i> (= uninflected trunk of obsolete *<i>nexati</i> ‘let’) requires a finite verb in the non-past (here:
<i>si±dzie</i>.3SG<i> </i>‘sit down’), this restriction applies for <i>niech</i> if used in the volitional domain (vs its use as concessive conjunction, when
<i>niech</i> allows for past tense). In isolation, this utterance could be assigned different illocutionary functions: it may simply be directive (command), it may also be permissive (if this is a reaction to a request for action), but it may also be non-curative
(‘I don’t care’, likewise reactive, but not necessarily to a request), from where we easily get into concessive use. Whether it is directive, permissive or non-curative depends on simpler distinctions like [+/- reactive], [+/- relates to request], also [+/-
interlocutors on equal social level], etc. Now, a codebook could contain the labels ‘directive’, ‘permissive’, etc., and respective explanations concerning these more atomic distinctions; but it could also just list these simpler distinctions as binary choices,
and the illocutionary function could be computed from their combination. That is, we could not use the labels of illocutionary functions; they might be better “known”, but will they be understood and be simpler for the annotators?<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"> Does anybody have experience with this kind of question (and the application in practice with annotators and interrater-reliability testing)? This question arises when you compare the Polish construction
with equivalent (often cognate) constructions in other Slavic languages, and if you want to grasp whether they have changed functionally over time (for a first impression cf. Wiemer 2023c). But there are many other examples. Which “full description” of any
of these languages would help me choose the right solution?<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">Last not least, all remarks and the information given by Sasha Berdicevskis from the perspective of computer linguistics are related and helpful to what I have been after (thanks, Sasha!). However, I’m not sure whether
the methodology he describes for investigating lexical change can be applied one-by-one to research in grammatical change (in the broad sense, as with my Polish example above), as Sasha remarks himself, and not only because it might be too expensive (with
probably depressing outcome) to be done on treebanks. Would your „</span><span lang="EN-US">very optimistic computational linguist</span><span lang="EN-US">“ use something like BERT for this purpose? If yes, the methodology seems to be inverse to composing
elaborate codebooks with a large number of criteria/variables (broken down into possibly atomic contrasts), doesn’t it? For you would assess the validity of BERT’s results ex post from human interpreters (probably language experts?), who need to understand
which contextual features BERT’s vector analysis was based on (which may remain a black box)?<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">After all, all contributors to this discussion converge in saying that being a native speaker of the language or not doesn’t matter, at least not in the first place. This was my impression as well before I posted my mail
a week ago. But the reason I asked for that, was that obviously not all peers reviewers of project applications or of journal articles share this assumption. And then it gets difficult to argue with them.
<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"> The same applies to reviewers who reject the existence of phenomena (e.g., of data that defy unambiguous annotation) because they are not discussed or noted in UD ressources… This is another anecdote,
but a really annoying and sad one.<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">However, let’s be optimistic, though, and be confident that research in diachronic change can be done without native speakers, and the parameters indicating change even be quantified (to some extent), even if data is
not that abundant and no full descriptions of earlier languages stages are available (and testable). And that peer reviewers will be constructive in this regard.<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">Best,<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">Björn.<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal">References<o:p></o:p></p>
<p class="MsoNormal">Verstraete, Jean-Christophe. 2007. <i><span lang="EN-US">Rethinking the Coordinate-Subordinate Dichotomy Interpersonal Grammar and the Analysis of Adverbial Clauses in English</span></i><span lang="EN-US">. Berlin, New York: Mouton de Gruyter.<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">Weber-Genzel, Leon, Siyao Peng, Marie-Catherine de Marneffez & Barbara Plank. 2024. VariErr NLI: Separating Annotation Error from Human Label Variation.
<i>Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics</i> (Volume 1: Long Papers), 2256-2269.
<o:p></o:p></span></p>
<p class="MsoNormal">Wiemer, Björn. 2023a. <span lang="EN-US">Between analytical mood and clause-initial particles – on the diagnostics of subordination for (emergent) complementizers.
<i>Zeitschrift für Slawistik</i> 68-2, 187-260.</span><o:p></o:p></p>
<p class="MsoNormal">Wiemer, Björn. 2023b. <a name="_Hlk216096944"><span lang="EN-US">Clause-initial connectives, bound and unbound: Indicators of mood, of subordination, or of something more fundamental?
<i>Slavia Meridionalis</i> 23 (Special issue: <i>Comparative and typological approaches to Slavic languages</i>.
</span></a><span style="mso-bookmark:_Hlk216096944"><span lang="PL">Ed. by</span></span><span style="mso-bookmark:_Hlk216096944"><span lang="PL"> Jakub Banasiak, Julia Mazurkiewicz-Su³kowska, Bo¿ena Rozwadowska, Dorota Klimek-Jankowska). DOI: 10.11649/sm.3194</span></span><span style="mso-bookmark:_Hlk216096944"></span><o:p></o:p></p>
<p class="MsoNormal">Wiemer, Björn. 2023c. <span lang="EN-US">Directive-optative markers in Slavic: observations on their persistence and change.
<i>Linguistica Brunensia</i> 71-1, 5-45.</span><span style="color:black;mso-fareast-language:EN-US"><o:p></o:p></span></p>
<p class="MsoNormal"><span style="color:black;mso-fareast-language:EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span style="color:black;mso-fareast-language:EN-US">
<o:p></o:p></span></p>
<p class="MsoNormal"><span style="color:black;mso-fareast-language:EN-US"><o:p> </o:p></span></p>
<div>
<div style="border:none;border-top:solid #E1E1E1 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal"><b><span style="font-size:11.0pt;font-family:"Calibri",sans-serif">Von:</span></b><span style="font-size:11.0pt;font-family:"Calibri",sans-serif"> Lingtyp <lingtyp-bounces@listserv.linguistlist.org>
<b>Im Auftrag von </b>Martin Haspelmath via Lingtyp<br>
<b>Gesendet:</b> Montag, 5. </span><span lang="EN-US" style="font-size:11.0pt;font-family:"Calibri",sans-serif">Januar 2026 12:51<br>
<b>An:</b> lingtyp@listserv.linguistlist.org<br>
<b>Betreff:</b> Re: [Lingtyp] complex annotations and inter-rater reliability<o:p></o:p></span></p>
</div>
</div>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p><span lang="EN-US">Dear Björn,<o:p></o:p></span></p>
<p><span lang="EN-US">Since you mentioned works on cross-linguistic inter-coder reliability as well (e.g. Himmelmann et al. 2018 on the universality of intonational phrases):<o:p></o:p></span></p>
<p><span lang="EN-US">I think it's important to have clear and simple definitions of annotation categories, so if you are interested, for example, in "</span><span lang="EN-US" style="color:black;mso-fareast-language:EN-US">the coding of clause-initial “particles”
(are they just particles, operators of “analytical mood”, or complementizers?)", you need to have clear and simple definitions of
<i>particle</i>, <i>mood</i>, and <i>complementizer</i> as comparative concepts. ("</span><span lang="EN-US">The burden is on those who formulate the guidelines", as Christian Lehmann said.)<o:p></o:p></span></p>
<p><span lang="EN-US" style="color:black;mso-fareast-language:EN-US">I think one can define
<i>particle</i> as "a bound morph that is neither a root nor an affix nor a person form nor a linker", but this definition of course presupposes that one has a definition of "root", of "affix", and so on. These terms are not understood uniformly either, and
<i>mood</i> is perhaps the worst of all traditional terms (even worse than "subordination", I think).</span><span lang="EN-US"><o:p></o:p></span></p>
<p><span lang="EN-US">Matters are quite different with materials from little-studied languages, i.e. with "transcribing and annotating recordings", as described by Jürgen Bohnemeyer. Language-particular descriptive categories are much easier to identify across
texts than comparatively defined categories are to identify across languages.<o:p></o:p></span></p>
<p><span lang="EN-US">Best wishes for the New Year,<o:p></o:p></span></p>
<p><span lang="EN-US">Martin<o:p></o:p></span></p>
<div>
<p class="MsoNormal"><span lang="EN-US">On 03.01.26 12:54, Wiemer, Bjoern via Lingtyp wrote:<o:p></o:p></span></p>
</div>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US">Dear All,</span><span lang="EN-US"><o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US">since this seems to be the first post on this list this year, I wish everybody a successful, more peaceful and decent year than the previous one.</span><span lang="EN-US"><o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US"> </span><span lang="EN-US"><o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US">I want to raise an issue which gets back to a discussion from October 2023 on this list (see the thread below, in inverse chronological order). I’m interested to know whether
anybody has a satisfying answer to the question how to deal with semantic annotation, or the annotation of more complex (and less obvious) relations, in particular with the annotation of interclausal relations, both in terms of syntax and in semantic terms.
Problems arise already with the coordination-subordination gradient, which ultimately is an outcome of a complex bunch of semantic criteria (like independence of illocutionary force, perspective from which referential expressions like tense or person deixis
are interpreted; see also the factors that were analyzed meticulously, e.g., by Verstraete 2007). Other questions concern the coding of clause-initial “particles”: are they just particles, operators of “analytical mood”, or complementizers? (Notably, these
things do not exclude one another, but they heavily depend on one’s theory, in particular one’s stance toward complementation and mood.) Another case in point is the annotation of the functions and properties of constructions in TAME-domains, especially if
the annotation grid is more fine-grained than mainstream categorizing.</span><span lang="EN-US"><o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US"> The problems which I have encountered (in pilot studies) are very similar to those discussed in October 2023 for seemingly even “simpler”, or more coarse-grained
annotations. And they aggravate a lot when we turn to data from diachronic corpora: even if being an informed native speaker is usually an asset, with diachronic data this asset is often useless, and native knowledge may be even a hindrance since it leads
the analyst to project one’s habits and norms of contemporary usage to earlier stages of the “same” language. (Similar points apply for closely related languages.) I entirely agree that annotators have to be trained, and grids of annotation to be tested, first
of all because you have to exclude the (very likely) possibility that raters disagree just because some of the criteria are not clear to at least one of them (with the consequence that you cannot know whether disagreement or low Kappa doesn’t result from misunderstandings,
instead of reflecting properties of your object of study). I also agree that each criterion of a grid has to be sufficiently defined, and the annotation grid (or even its “history”) as such be documented in order to save objective criteria for replicability
and comparability (for cross-linguistic research, but also for diachronic studies based on a series of “synchronic cuts” of the given language).</span><span lang="EN-US"><o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US"> </span><span lang="EN-US"><o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US">On this background, I’d like to formulate the following questions:</span><span lang="EN-US"><o:p></o:p></span></p>
<ol style="margin-top:0cm" start="1" type="1">
<li class="MsoListParagraph" style="color:black;margin-left:0cm;mso-list:l1 level1 lfo3">
<span lang="EN-US" style="mso-fareast-language:EN-US">Which arguments are there that (informed) native speakers are better annotators than linguistically well-trained students/linguists who are not native speakers of the respective language(s), but can be considered
experts?</span><span lang="EN-US"><o:p></o:p></span></li><li class="MsoListParagraph" style="color:black;margin-left:0cm;mso-list:l1 level1 lfo3">
<span lang="EN-US" style="mso-fareast-language:EN-US">Conversely, which arguments are there that non-native speaker experts might be even better suited as annotators (for this or that kind of issue)?</span><span lang="EN-US"><o:p></o:p></span></li><li class="MsoListParagraph" style="color:black;margin-left:0cm;mso-list:l1 level1 lfo3">
<span lang="EN-US" style="mso-fareast-language:EN-US">Have assumptions about pluses and minuses of both kinds of annotators been tested in practice? That is, do we have empirical evidence for any such assumptions (or do we just rely on some sort of common sense,
or on the personal experience of those who have done more complicated annotation work)?</span><span lang="EN-US"><o:p></o:p></span></li><li class="MsoListParagraph" style="color:black;margin-left:0cm;mso-list:l1 level1 lfo3">
<span lang="EN-US" style="mso-fareast-language:EN-US">How can pluses and minuses of both kinds of annotators be counterbalanced in a not too time (and money) consuming way?</span><span lang="EN-US"><o:p></o:p></span></li><li class="MsoListParagraph" style="color:black;margin-left:0cm;mso-list:l1 level1 lfo3">
<span lang="EN-US" style="mso-fareast-language:EN-US">What can we do with data from diachronic corpora if we have to admit that (informed) native speakers are of no use, and non-native experts are not acknowledged, either? Are we just deemed to refrain from
any reliable and valid in-depth research based on annotations (and statistics) for diachronically earlier stages and for diachronic change?</span><span lang="EN-US"><o:p></o:p></span></li><li class="MsoListParagraph" style="color:black;margin-left:0cm;mso-list:l1 level1 lfo3">
<span lang="EN-US" style="mso-fareast-language:EN-US">In connection with this, has any cross-linguistic research that is interested in diachrony tried to implement insights from such fields like historical semantics and pragmatics into annotations? In typology,
linguistic change has increasingly become more prominent during the last 10-15 years (not only from a macro-perspective). I thus wonder whether typologists have tried to “borrow” methodology from fields that have possibly been better in interpreting diachronic
data, and even quantify them (to some extent).</span><span lang="EN-US"><o:p></o:p></span></li></ol>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US"> </span><span lang="EN-US"><o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US">I don’t want to be too pessimistic, but if we have no good answers as for who should be doing annotations – informed native speakers or non-native experts (or only those
who are both native and experts)? – and how we might be able to test the validity of annotation grids (for comparisons across time and/or languages), there won’t be convincing arguments how to deal with diachronic data (or data of lesser studied languages
for which there might be no native speakers available) in empirical studies that are to disclose more fine-grained distinctions and changes, also in order to quantify them. In particular, reviewers of project applications may always ask for a convincing methodology,
and if no such research is funded we’ll remain ignorant of quite many reasons and backgrounds of language change.
</span><span lang="EN-US"><o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US"> </span><span lang="EN-US"><o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US">I’d appreciate advice, in particular if it provides answers to any of the questions under 1-6 above.</span><span lang="EN-US"><o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="color:black;mso-fareast-language:EN-US"> </span><span lang="EN-US"><o:p></o:p></span></p>
<p class="MsoNormal"><span style="color:black;mso-fareast-language:EN-US">Best,</span><o:p></o:p></p>
<p class="MsoNormal"><span style="color:black;mso-fareast-language:EN-US">Björn (Wiemer).</span><o:p></o:p></p>
<p class="MsoNormal"><span style="color:black;mso-fareast-language:EN-US"> </span><o:p></o:p></p>
<p class="MsoNormal"><span style="color:black;mso-fareast-language:EN-US"> </span><o:p></o:p></p>
<div>
<div style="border:none;border-top:solid #E1E1E1 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal"><b><span style="font-size:11.0pt;font-family:"Calibri",sans-serif">Von:</span></b><span style="font-size:11.0pt;font-family:"Calibri",sans-serif"> Lingtyp
</span><a href="mailto:lingtyp-bounces@listserv.linguistlist.org"><span style="font-size:11.0pt;font-family:"Calibri",sans-serif"><lingtyp-bounces@listserv.linguistlist.org></span></a><span style="font-size:11.0pt;font-family:"Calibri",sans-serif">
<b>Im Auftrag von </b>William Croft<br>
<b>Gesendet:</b> Montag, 16. </span><span lang="EN-US" style="font-size:11.0pt;font-family:"Calibri",sans-serif">Oktober 2023 15:52<br>
<b>An:</b> Volker Gast </span><a href="mailto:volker.gast@uni-jena.de"><span lang="EN-US" style="font-size:11.0pt;font-family:"Calibri",sans-serif"><volker.gast@uni-jena.de></span></a><span lang="EN-US" style="font-size:11.0pt;font-family:"Calibri",sans-serif"><br>
<b>Cc:</b> </span><a href="mailto:LINGTYP@LISTSERV.LINGUISTLIST.ORG"><span lang="EN-US" style="font-size:11.0pt;font-family:"Calibri",sans-serif">LINGTYP@LISTSERV.LINGUISTLIST.ORG</span></a><span lang="EN-US" style="font-size:11.0pt;font-family:"Calibri",sans-serif"><br>
<b>Betreff:</b> Re: [Lingtyp] typology projects that use inter-rater reliability?</span><span lang="EN-US"><o:p></o:p></span></p>
</div>
</div>
<p class="MsoNormal"><span lang="EN-US"> <o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">An early cross-linguistic study with multiple annotators is this one:<o:p></o:p></span></p>
<div>
<p class="MsoNormal"><span lang="EN-US"> <o:p></o:p></span></p>
</div>
<div>
<p class="bibliography" style="mso-margin-top-alt:0cm;margin-right:0cm;margin-bottom:0cm;margin-left:18.0pt;text-align:justify;text-indent:-18.0pt">
<span style="font-size:13.5pt;font-family:"Times",serif">Gundel, Jeannette K., Nancy Hedberg & Ron Zacharski.
</span><span lang="EN-US" style="font-size:13.5pt;font-family:"Times",serif">1993. Cognitive status and the form of referring expressions in discourse. <i>Language</i> 69.274-307.</span><span lang="EN-US"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span lang="EN-US"> <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span lang="EN-US">It doesn’t have all the documentation that Volker suggests; our standards for providing documentation has risen.<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span lang="EN-US"> <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span lang="EN-US">I have been involved in annotation projects in natural language processing, where the aim is to annotate corpora so that automated methods can “learn” the annotation categories from the “gold standard” (i.e. “expert”)
annotation -- this is supervised learning in NLP. Recent efforts are aiming at developing a single annotation scheme for use across languages, such as Universal Dependencies (for syntactic annotation), Uniform Meaning Representation (for semantic annotation),
and Unimorph (for morphological annotation). My experience is somewhat similar to Volker’s: even when the annotation scheme is very coarse-grained (from a theoretical linguist’s point of view), getting good enough interannotator agreement is hard, even when
the annotators are the ones who designed the scheme, or are native speakers or have done fieldwork on the language. I would add to Volker’s comments that one has to be trained for annotation; but that training can introduce (mostly implicit) bases, at least
in the eyes of proponents of a different theoretical approach -- something that is more apparent in a field such as linguistics where there are large differences in theoretical approaches.<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span lang="EN-US"> <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span lang="EN-US">Bill<o:p></o:p></span></p>
<div>
<p class="MsoNormal"><span lang="EN-US"><br>
<br>
<br>
<o:p></o:p></span></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<p class="MsoNormal"><span lang="EN-US">On Oct 16, 2023, at 6:02 AM, Volker Gast <</span><a href="mailto:volker.gast@uni-jena.de"><span lang="EN-US">volker.gast@uni-jena.de</span></a><span lang="EN-US">> wrote:<o:p></o:p></span></p>
</div>
<p class="MsoNormal"><span lang="EN-US"> <o:p></o:p></span></p>
<div>
<div>
<p class="MsoNormal"><span lang="EN-US"><br>
Hey Adam (and others),<br>
<br>
I think you could phrase the question differently: What typological studies have been carried out with multiple annotators and careful documentation of the annotation process, including precise annotation guidelines, the training of the annotators, publication
of all the (individual) annotations, calculation of inter-annotator agreement etc.?<br>
<br>
I think there are very few. The reason is that the process is very time-consuming, and "risky". I was a member of a project co-directed with Vahram Atayan (Heidelberg) where we carried out very careful annotations dealing with what we call 'adverbials of immediate
posteriority' (see the references below). Even though we only dealt with a few well-known European languages, it took us quite some time to develop annotation guidelines and train annotators. The inter-rater agreement was surprisingly low even for categories
that appeared straightforward to us, e.g. agentivity of a predicate; and we were dealing with well-known languages (English, German, French, Spanish, Italian). So the outcomes of this process were very moderate in comparison with the work that went into the
annotations. (Note that the project was primarily situated in the field of contrastive linguistics and translation studies, not linguistic typology, but the challenges are the same).<br>
<br>
It's a dilemma: as a field, we often fail to meet even the most basic methodological requirements that are standardly made in other fields (most notably psychology). I know of at least two typological projects where inter-rater agreement tests were run, but
the results were so poor that a decision was made to not pursue this any further (meaning, the projects were continued, but without inter-annotator agreement tests; that's what makes annotation projects "risky": what do you do if you never reach a satisfactory
level of inter-annotator agreement?). Most annotation projects, including some of my own earlier work, are based on what we euphemistically call 'expert annotation', with 'expert' referring to ourselves, the authors. Today I would minimally expect the annotations
to be done by someone who is not an author, and I try to implement that requirement in my role as a journal editor (Linguistics), but it's hard. We do want to see more empirical work published, and if the methodological standards are too high, we will end
publishing nothing at all.<br>
<br>
I'd be very happy if there were community standards for this, and I'd like to hear about any iniatives implementing more rigorous methodological standards in lingusitic typology. Honestly, I wouldn't know what to require. But it seems clear to me that we cannot
simply go on like this, annotating our own data, which we subsequently analyze, as it is well known that annotation decisions are influenced by (mostly implicit) biases.<br>
<br>
Best,<br>
Volker<br>
<br>
Gast, Volker & Vahram Atayan (2019). 'Adverbials of immediate posteriority in French and German: A contrastive corpus study of tout de suite, immédiatement, gleich and sofort'. In Emonds, J., M. Janebová & L. Veselovská (eds.): Language Use and Linguistic Structure.
Proceedings of the Olomouc Linguistics Colloquium 2018, 403-430. Olomouc Modern Lanuage Series.
</span>Olomouc: Palacký University Olomouc.<br>
<br>
in German:<br>
<br>
Atayan, V., B. Fetzer, V. Gast, D. Möller, T. Ronalter (2019). 'Ausdrucksformen der unmittelbaren Nachzeitigkeit in Originalen und Übersetzungen: Eine Pilotstudie zu den deutschen Adverbien gleich und sofort und ihren Äquivalenten im Französischen, Italienischen,
Spanischen und Englischen'. In Ahrens, B., S. Hansen-Schirra, M. Krein-Kühle, M. Schreiber, U. Wienen (eds.): Translation -- Linguistik -- Semiotik, 11-82. Berlin: Frank & Timme.<br>
<br>
Gast, V., V. Atayan, J. Biege, B. Fetzer, S. Hettrich, A. Weber (2019). 'Unmittelbare Nachzeitigkeit im Deutschen und Französischen: Eine Studie auf Grundlage des OpenSubtitles-Korpus'.
<span lang="EN-US">In Konecny, C., C. Konzett, E. Lavric, W. Pöckl (eds.): Comparatio delectat III.
</span>Akten der VIII. Internationalen Arbeitstagung zum romanisch-deutschen und innerromanischen Sprachvergleich, 223-249.
<span lang="EN-US">Frankfurt: Lang.<br>
<br>
<br>
---<br>
Prof. V. Gast<br>
</span><a href="https://linktype.iaa.uni-jena.de/VG"><span lang="EN-US">https://linktype.iaa.uni-jena.de/VG</span></a><span lang="EN-US"><br>
<br>
On Sat, 14 Oct 2023, Adam James Ross Tallman wrote:<br>
<br>
<br>
<br>
<o:p></o:p></span></p>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<p class="MsoNormal"><span lang="EN-US">Hello all,<br>
I am gathering a list of projects / citations / papers that use or refer to inter-rater reliability. So far I have.<br>
Himmelmann et al. 2018. On the universality of intonational phrases: a cross-linguistic interrater study. Phonology 35.<br>
Gast & Koptjevskaja-Tamm. 2022. Patterns of persistence and diffusibility in the European lexicon. Linguistic Typology (not explicitly the topic of the paper, but interrater reliability metrics are used)<br>
I understand people working with Grambank have used it, but I don't know if there is a publication on that.<br>
best,<br>
Adam<br>
--<br>
Adam J.R. Tallman<br>
Post-doctoral Researcher<br>
Friedrich Schiller Universität<br>
Department of English Studies<o:p></o:p></span></p>
</blockquote>
<p class="MsoNormal"><span lang="EN-US">_______________________________________________<br>
Lingtyp mailing list<br>
</span><a href="mailto:Lingtyp@listserv.linguistlist.org"><span lang="EN-US">Lingtyp@listserv.linguistlist.org</span></a><span lang="EN-US"><br>
</span><a href="https://listserv.linguistlist.org/cgi-bin/mailman/listinfo/lingtyp"><span lang="EN-US">https://listserv.linguistlist.org/cgi-bin/mailman/listinfo/lingtyp</span></a><span lang="EN-US"><o:p></o:p></span></p>
</div>
</div>
</blockquote>
</div>
<p class="MsoNormal"><span lang="EN-US"> <o:p></o:p></span></p>
</div>
<p class="MsoNormal"><span lang="EN-US"><br>
<br>
<o:p></o:p></span></p>
<pre><span lang="EN-US">_______________________________________________<o:p></o:p></span></pre>
<pre><span lang="EN-US">Lingtyp mailing list<o:p></o:p></span></pre>
<pre><a href="mailto:Lingtyp@listserv.linguistlist.org"><span lang="EN-US">Lingtyp@listserv.linguistlist.org</span></a><span lang="EN-US"><o:p></o:p></span></pre>
<pre><a href="https://listserv.linguistlist.org/cgi-bin/mailman/listinfo/lingtyp"><span lang="EN-US">https://listserv.linguistlist.org/cgi-bin/mailman/listinfo/lingtyp</span></a><span lang="EN-US"><o:p></o:p></span></pre>
</blockquote>
<pre><span lang="EN-US">-- <o:p></o:p></span></pre>
<pre><span lang="EN-US">Martin Haspelmath<o:p></o:p></span></pre>
<pre><span lang="EN-US">Max Planck Institute for Evolutionary Anthropology<o:p></o:p></span></pre>
<pre>Deutscher Platz 6<o:p></o:p></pre>
<pre>D-04103 Leipzig<o:p></o:p></pre>
<pre><a href="https://www.eva.mpg.de/linguistic-and-cultural-evolution/staff/martin-haspelmath/">https://www.eva.mpg.de/linguistic-and-cultural-evolution/staff/martin-haspelmath/</a><o:p></o:p></pre>
</div>
</body>
</html>