AplikaceAplikace
Nastavení

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
en:cnk:orator [2025/05/28 13:28] – [Corpus composition and data acquisition] martinawaclawicovaen:cnk:orator [2026/06/30 12:39] (current) – michalkren
Line 3: Line 3:
 The ORATOR corpus contains monologues by native Czech speakers. The typical situations include a lecture, instruction, guided tour, welcome address, sermon etc. The speech is usually prepared and the speaker has to fit within the given time frame. To our knowledge, there is no corpus with this kind of data available for Czech. The ORATOR corpus contains monologues by native Czech speakers. The typical situations include a lecture, instruction, guided tour, welcome address, sermon etc. The speech is usually prepared and the speaker has to fit within the given time frame. To our knowledge, there is no corpus with this kind of data available for Czech.
  
-Transcription rules, linking to the corresponding audio track and most metadata follow the [[en:cnk:ortofon|ORTOFON]] and [[en:cnk:oral|ORAL]] corpora, structural attributes used in ORATOR are described [[pojmy:atributy_strukturni|here]] (Czech only). The corpus is [[en:cnk:lemtag_mluv|lemmatized and morphologically tagged]] in the same way as the ORAL and ORTOFON corpora. The corpus is not balanced in any way.+Transcription rules, linking to the corresponding audio track (using the [[https://archive.mpi.nl/tla/elan|ELAN]] tool, developed in the Max Planck Institute for Psycholinguistics, Nijmegen((ELAN (Version 7.1) [Computer software]. (2026). Nijmegen: Max Planck Institute for Psycholinguistics. Retrieved from https://archive.mpi.nl/tla/elan 
 +)))  and most metadata follow the [[en:cnk:ortofon|ORTOFON]] and [[en:cnk:oral|ORAL]] corpora, structural attributes used in ORATOR are described [[pojmy:atributy_strukturni|here]] (Czech only). The corpus is [[en:cnk:lemtag_mluv|lemmatized and morphologically tagged]] in the same way as the ORAL and ORTOFON corpora. The corpus is not balanced in any way.
  
 <WRAP 45%> <WRAP 45%>
Line 24: Line 25:
 The original criteria for including a recording in the corpus were that it must not be a read speech and that it must take place in the presence of an audience. This means that it must not be a speech prepared for the web, as it is impossible to know whether it has been repeatedly recorded and additionally edited. It is therefore not possible to guarantee an authentic capture of a monologue under normal conditions, where the speaker is exposed to certain expectations from the audience, is influenced by their presence and by the form of the event in question. However, during the data collection we repeatedly came across the fact that on certain occasions the (partially) read form is a usual part of monologues, for example, because they are ceremonies where the form has to be observed (e.g. graduation) or even legally binding (wedding ceremony), quotations are usually part of the lecture, or the speech is interpreted. We have therefore decided to include a small number (18 recordings) of read or partially read speeches to complete the picture of monologues more comprehensively. For the same reasons, we have also selected a few (9 in total) recordings without an audience, made available to the general public via the internet; these are lectures or New Year's speeches. Each of these types accounts for about 3% of the scope of the whole corpus. The original criteria for including a recording in the corpus were that it must not be a read speech and that it must take place in the presence of an audience. This means that it must not be a speech prepared for the web, as it is impossible to know whether it has been repeatedly recorded and additionally edited. It is therefore not possible to guarantee an authentic capture of a monologue under normal conditions, where the speaker is exposed to certain expectations from the audience, is influenced by their presence and by the form of the event in question. However, during the data collection we repeatedly came across the fact that on certain occasions the (partially) read form is a usual part of monologues, for example, because they are ceremonies where the form has to be observed (e.g. graduation) or even legally binding (wedding ceremony), quotations are usually part of the lecture, or the speech is interpreted. We have therefore decided to include a small number (18 recordings) of read or partially read speeches to complete the picture of monologues more comprehensively. For the same reasons, we have also selected a few (9 in total) recordings without an audience, made available to the general public via the internet; these are lectures or New Year's speeches. Each of these types accounts for about 3% of the scope of the whole corpus.
  
-The recordings were made at various locations in the Czech Republic or were downloaded from the internet with the consent of the spokesperson. Except for the 9 cases mentioned above, the recordings always capture the communication situation in the presence of the audience and in an authentic environment. The corpus is also not balanced by the gender of the speakers, with a predominance of men.+The recordings were made at various locations in the Czech Republic or were downloaded from the internet with the consent of the speaker. Except for the 9 cases mentioned above, the recordings always capture the communication situation in the presence of the audience and in an authentic environment. The corpus is also not balanced by the gender of the speakers, with a predominance of men. 
 + 
 +===== Morphological tagging of the ORATOR corpus ===== 
 + 
 +The ORATOR v3 corpus is automatically [[en:pojmy:tag|annotated]] with [[en:cnk:syn2020#morphological_tagging|a new morphological tag]] according to the [[en:cnk:anotacni_standard_cnk|unified CNC annotation scheme]]. It recognizes [[en:cnk:syn2020#multiple_lemmatization_and_tagging_aggregate|aggregates]] (e.g., //vidělas//, //zač//), uses [[en:cnk:syn2020|double-level lemmatization]], and has a verb tag ([[en:cnk:syn2020#verb_tagging_verbtag|verbtag]]).  
 + 
 +Substandard variants and forms typical of dialects and spontaneous speech are also tagged in the corpus (according to the ORTOFON corpus, see [[en:cnk:ortofon#morphological_tagging_of_the_ortofon_corpus|Morphological tagging of the ORTOFON corpus]]).  
 + 
 +The following specific tags are used in the first tag position (word type): 
 + 
 +^  Tag  ^  Meaning  ^ 
 +|  E | fragments (incomplete words) | 
 +|  H    | nonverbal sounds (e.g. hezitation) | 
 +|  M | comments by transcribers (in round brackets) | 
 +|  W | anonymised sections (mainly names) |  
 + 
 +Note: The anonymised sections are specified on a basic level ''%%word%%'': NP – surname, NJ – first name, NN – nickname, NM – place name, NO – other proper names, NT – last two digits of the telephone number. 
 + 
 +The ORATOR v2 corpus is tagged with the prior morphological tagset used until 2020. Detailed information on the annotation of these previously published corpora can be found on a [[en:cnk:lemtag_mluv|separate page]].
  
 ====== ORATOR v1 (2019) ====== ====== ORATOR v1 (2019) ======
Line 36: Line 55:
 ====== ORATOR v3 (2025) ====== ====== ORATOR v3 (2025) ======
  
-The ORATOR corpus in its third version contains the same recordings and transcripts as the second version (i.e. over 1.5 million tokens), but they are newly annotated according to the SYN2020 standard. The genphone attribute is also newly included in the corpus, indicating the automatically generated phonetic form of a word. In addition, a number of transcription corrections have been made.+The ORATOR corpus in its third version contains the same recordings and transcripts as the second version (i.e. over 1.5 million tokens) but annotated according to the new Unified CNC Annotation Scheme using a language model trained also on spoken data. The ''genphone'' attribute is also newly included in the corpus, indicating the automatically generated phonetic form of a word. In addition, several transcription corrections have been made.
 ===== How to cite ===== ===== How to cite =====
  
 <WRAP round tip 70%> <WRAP round tip 70%>
-Kopřivová, M. – Laubeová, Z. – Lukeš, D. – Poukarová, P. – Horký, V. – Jelínek, T. – Křivan, J.: //ORATOR v3: Corpus of monologues, version 3, 28. 5. 2025//. Ústav Českého národního korpusu FF UK, Praha 2025. Retrieved from [[https://www.korpus.cz]].+Kopřivová, M. – Laubeová, Z. – Lukeš, D. – Poukarová, P. – Horký, V. – Jelínek, T. – Křivan, J.: //ORATOR: Corpus of monologues, version 3, 28. 5. 2025//. Ústav lingvistiky FF UK, Praha 2025. Retrieved from [[https://www.korpus.cz]].
  
-Kopřivová, M. – Laubeová, Z. – Lukeš, D. – Poukarová, P.: //ORATOR v2: Korpus monologů//. Ústav Českého národního korpusu FF UK, Praha 2020. Retrieved from [[https://www.korpus.cz]].+Kopřivová, M. – Laubeová, Z. – Lukeš, D. – Poukarová, P.: //ORATOR: Corpus of monologues, version 2, 18. 12. 2020//. Ústav Českého národního korpusu FF UK, Praha 2020. Retrieved from [[https://www.korpus.cz]].
  
-Kopřivová, M. – Laubeová, Z. – Lukeš, D. – Poukarová, P.: //ORATOR v1: Korpus monologů//. Ústav Českého národního korpusu FF UK, Praha 2019. Retrieved from [[https://www.korpus.cz]].+Kopřivová, M. – Laubeová, Z. – Lukeš, D. – Poukarová, P.: //ORATOR: Corpus of monologues, version 1, 19. 12. 2019//. Ústav Českého národního korpusu FF UK, Praha 2019. Retrieved from [[https://www.korpus.cz]].
 </WRAP> </WRAP>
  

Trace: • segmentace • syn2006pub • orator • verze14 • czesl-sgt-basic • frekvence • ulozit • morfio • prehled_negativnich_lemmat • syn