Spoken Data Annotation Manual: Tundra Nenets spoken data

Author: Nikolett Mus

Date: 28.05.2025

1. Introduction

This document outlines the annotation conventions for spoken language-specific features in Tundra Nenets, including prosodic tags, discourse phenomena, and structural relations. It reflects the current annotation practice and serves as a working reference.

2. Spoken Phenomena: Tags and Relational Encoding

1. Noises
   • <n_n> = Non-linguistic noise, but speaker generated (e.g., cough, laugh, sigh)
   • <w_n> = Environmental/world noise (e.g., background chatter, traffic, mic bump)

   ID: lexical unit on its own
   FORM: tag
   LEMMA: tag
   UPOS: X
   DEPREL: noise, attached to the element it follows
   Gloss: NOISE

   
2. Speech Interruptions

   Interruptions in the speaker's normal flow, including errors, hesitations, restarts.

   1. 2.1 Disfluency: Audible disfluency
      • <a_d> = Audible disfluency (e.g., "uh", "erm")

      ID: lexical unit on its own
      FORM: tag
      LEMMA: tag
      UPOS: INTJ
      DEPREL: discourse, attached to the following element
      Gloss: DISFL

      
   2. 2.2 Disfluency: Articulatory lengthening
      • <a> = Articulatory lengthening (e.g., prolonged sound)

      ID: attached to lexeme
      FORM: The tag <a> should be attached to the relevant lexeme without a space.
      LEMMA: not relevant
      UPOS: not relevant
      DEPREL: not relevant
      Gloss: not relevant

      
   3. 2.3 Cut-offs: Unfinished lexeme
      • <u_l> = Unfinished lexeme

      ID: attached to lexeme
      FORM: The tag <u_l> should be attached to the relevant lexeme without a space.
      LEMMA: Full form of expected word
      UPOS: not relevant
      DEPREL: not relevant
      Gloss: not relevant

      
   4. 2.4 False Start
      • <f_s> = False start

      ID: attached to lexeme
      FORM: The tag <f_s> should be attached to the lexeme without a space.
      LEMMA: underlying lemma (base word) without the tag
      UPOS: POS of the base word
      DEPREL: reparandum attached to the node that replaces or repairs it
      Gloss: Gloss of base/intended word

      
   5. 2.5 Exact Repetition
      • <e_r> = Exact repetition

      ID: attached to lexeme
      FORM: The tag <e_r> attached to lexeme without a space
      LEMMA: underlying lemma (base word) without the tag
      UPOS: POS of the base word
      DEPREL: reparandum attached to the next instance (the "actual", i.e. the first use).
      Gloss: Gloss of base/intended word

      
   6. 2.6 Partial Repetition
      • <p_r> = Partial repetition

      ID: attached to lexeme
      FORM: The tag <p_r> attached to lexeme without a space
      LEMMA: underlying lemma (base word) without the tag
      UPOS: POS of the base word
      DEPREL: reparandum attached to the next instance (the "actual", i.e. the first use).
      Gloss: Gloss of base/intended word

      
3. Connected Speech: Merged lexemes
   <m_l> = Merged lexemes, when two or more lexemes are pronounced without a boundary in speech, but are distinct syntactic units.

   ID: attached to lexeme
   FORM: The lexemes must be separated and the tag <m_l> should precede the subsequent lexeme without a space.
   LEMMA: underlying lemma (base word) without the tag
   UPOS: POS of the base word
   DEPREL: As per syntax
   Gloss: Gloss of lemma

   
4. Lexical Error

   <u_w> = Unusual or incorrect word (word choice error)

   ID: attached to lexeme
   FORM: The tag <u_w> attached to word without space
   LEMMA: Lemma of incorrect word
   UPOS: POS of the word as it appears
   DEPREL: As if correct
   Gloss: Literal gloss of the incorrect word

   
5. Pauses coinciding with a clear syntactic boundary

   In Tundra Nenets it appears:

   • after a main clause
   • after an embedded clause
   • between coordinate clauses
   • between coordinate phrases
   • at the end of a sentence
   • <s_p> = Silent pause
   • <b_p> = Breathing pause

   ID: lexical unit on its own
   FORM: Tag
   LEMMA: Tag
   UPOS: PUNCT
   DEPREL: punct attached to the head of the clause/sentence it follows. Note, that the last punct is always dependent to the root
   Gloss: SIL

   
6. Pauses: disfluency-related

   In Tundra Nenets it appears:

   • between two unrelated constituents
   • inside a phrase
   • <s_p> = Silent pause
   • <b_p> = Breathing pause

   ID: lexical unit on its own
   FORM: Tag
   LEMMA: Tag
   UPOS: X
   DEPREL: discourse attached to the constituent it follows.
   Gloss: PAUSE

   

3. Attribution and Reuse

This document is part of the [Project Name] and is freely available under a CC BY 4.0 license.