=Paper=
{{Paper
|id=Vol-2870/paper52
|storemode=property
|title=Marking up Dramatic Text: a Case Study of “7 Stories” by Morris Panych
|pdfUrl=https://ceur-ws.org/Vol-2870/paper52.pdf
|volume=Vol-2870
|authors=Ivan Bekhta,Nataliia Hrytsiv,Anastasiia Matviychuk
|dblpUrl=https://dblp.org/rec/conf/colins/BekhtaHM21
}}
==Marking up Dramatic Text: a Case Study of “7 Stories” by Morris Panych==
https://ceur-ws.org/Vol-2870/paper52.pdf
Marking up Dramatic Text: a Case Study of “7 stories” by Morris
Panych
Ivan Bekhtaa, Nataliia Hrytsivb and Anastasiia Matviychukb
a
Lviv Franko National University, Universytetska Street, 1, Lviv, 79000, Ukraine
b
Lviv Polytechnic National University, Stepana Bandery Street, 12, Lviv, 79000, Ukraine
Abstract
The paper elucidates the process, challenges and results of using computational linguistics
tools (NLP) and pre-computer technique (TEI for personage utterance tagging) in processing
dramatic text. As the material for analysis we have chosen the modern play ―7 stories‖ of the
Canadian playwright Morris Panych, researched from the viewpoint of statistical indicator’s
and textual coefficients. Special attention is paid to statistical parameters of main personages
in the play. Results obtained show numeric characteristics of such data: number of meanings
(N); maximal meaning (max); minimal meaning (min); range (R); mode (Mo); median (Md);
mean (Ẋ); standard deviation (Ϭ); coefficient of variation (ν); standard error (Sẋ);
measurement error (ε).
Keywords 1
Translation, NLP, quantitative analysis, text mark-up, applied linguistics, drama text,
tagging.
1. Introduction
Modern practices and techniques of using markup and NLP tools have proved its efficiency in
relation to systematic processing and analysis of texts, which results in generating novel systems and
well-elaborated tools for language processing. It is a powerful and promising technology to
objectively reconstruct arguments in order to amply exemplify its findings and formulate well-
grounded hypotheses.
This study is part of a larger research project on the creativity of Morris Panych and the reception
of his writings via translation(s).
Presented in the paper is the preliminary algorithm for developing analytical findings concerning
the reasons behind deviation within the aspect of statistical parameters of a source and target texts.
Further elaborations are forthcoming.
In focus is the drama ―7 stories‖ of Morris Panych [1] and its Ukrainian translation [2]. We briefly
discuss key characteristics of marking-up dramatic text and illustrate the results obtained; we also
demonstrate its primary advances.
Based on the aquired evidence, certain considerations are made regarding the usage of quantitative
comparable analysis for further comparison of ST and TT statistics and ratio findings.
Thus, linguistic research determines an effective approach to the study of text, using mathematical
methods and tools in combination with computer technology, which open new horizons for the
linguistic analysis of new broad perspectives.
A complete and comprehensive description of language and speech requires a diligent insight into
both qualitative and quantitative features of linguistic objects.
COLINS-2021: 5th International Conference on Computational Linguistics and Intelligent Systems, April 22–23, 2021, Kharkiv, Ukraine
EMAIL: ivan.bekhta@gmail.com (I. Bekhta); nataliia.m.hrytsiv@lpnu.ua (N. Hrytsiv); anastasija838@gmail.com (A. Matviychuk)
ORCID: 0000-0002-9848-1505 (I. Bekhta); 0000-0001-6660-7161 (N. Hrytsiv); 0000-0001-6056-1799 (A. Matviychuk)
2021 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
In addition, an approach towards and detailed study dramatic texts, as a unique literary genre, is a
separate challenge in current studies, which has special requirements within NLP tools application
and text mark-up. Therefore, the study of Morris Panych's playwork "7 Stories" is relevant.
The idea is that modern Canadian drama is the aspect, little studied from numerous viewpoints, i.e.
philological, translatological, rhethorical; however, least studied from the angle of mathematical
linguistics and statistics.
In order to understand the specifics of dramatic works, the concept of author's style, postmodern
literature, to which the work under study belongs, the life path of the author and translator were
additionally considered.
The play "7 Stories" by Morris Panych and translated by Ivan Krychfalushiy is an example of
postmodern literature that has become a challenge and opposition to the laws of modernism.
2. Method and preparation characteristics
Considering the vast quantities of ST and TT data available today for analysis, as discussed in [3,
4, 5, 6], Natural Language Processing is among most interesting and promising aspects of data science
[7, 8, 9, 10, 11, 12, 13].
By default, text data of the original text is difficult to process [14, 15, 16, 17, 18] given the
challenge of comparing/contrasting it to the translated drama text [19, 20, 21, 22, 23, 24], the task can
be complicated [25, 26, 27, 28], though, incredibly appealing [29, 30, 31, 32, 33].
Within this study project, we opted for exploring the way NLP techniques, especially mark-up
possibilities, can advance processing performing/drama text for statistical profiling of ST and TT.
The project outlined in the current paper explores the ddistribution of the number of words in a
sentence as well as other numeric characteristics being analyzed collectively and for all the characters
of drama under analysis in their contrast with the Ukrainian translation.
2.1. Stages of working with the text document “7 stories” by Morris Panych
A number of actions were performed for statistical analysis. Therefore, the analysis took place in
the following stages:
The books of the original text and the translation were pre-scanned for further
manipulations using ABBYY Fine Reader software;
Afterwards, it was converted from pdf to .docx to make it possible to work with text in
terms of mark-up;
The correct formatting of text was checked and discrepancies between scanned pdf file and
text documents were detected; it was normalized in the MS Word editor;
Next, the focus was on:
Selection of text marking up system according to its features;
Implementation of proper tags for the original work
Implementation of proper tags for the translated version;
Calculated texts results were processed using the Python programming language;
Afterwards, the results of the statistical parameters, such as N, max, min, R, Mo, Md, Ẋ, Ϭ,
ν, Sẋ, ε were analyzed and described.
The original text and its translation was marked up using the same marking rules.
To recall, the use was made of the XML (eXtensible Markup Language) – a text markup language.
It was used to conduct research and implement on the structural level.
The XML language was preffered since it fully determines the logical structure of a document.
The task of the XML language is to ensure certain data: images, texts, and other parts of a Web
document; it can be defined and structured regardless of the platform used to recreate them.
Since in the current paper we deal with a dramatic work, text mark up and tag patterns were
selected and adjusted for the appropriate analysis of this type of work. Thus, let us now turn our sights
to text mark-up system, peculiar to drama text.
2.2. Mark-up pattern
2.2.1. Pattern
Thus, the following text markings were chosen according to the features of the dramatic work:
... — paired marking, which is used to indicate a solid whole part of the text related
to a particular character;
... — paired marking, which is used to indicate the name of the character with a
colon;
... —— paired marking, which is used to denote a sentence in the speech of the character;
... — paired marking, which is used to mark all author's remarks throughout the text
2.2.2. Example
The action of the play takes place outside an apartment building-on the ledge, outside
various windows of the seventh storey. As the play progresses, the lights emphasize the time elapsed
between early evening and late night. As the play opens, we hear a party in progress from one of the
windows, MAN stands on the ledge, in a state of perplexity, contemplating the depths below. He
seems disturbed, confused. Then he comes to what seems to be a resolution. He prepares to jump.
When he is about to leap, the window next to him flies open. CHARLOTTE appears. She holds a
MAN wAllet, which she attempts to throw out the window, RODNEY,charging up from behind,
grabs her hand. A window-ledge struggle ensues.
CHARLOTTE
Let GO of me!!! Let GO!!
RODNEY
(threatening) So-help-me-GOD, CHARLOTTE.
CHARLOTTE
(daring him) What?? WHAT??!!
RODNEY
Give me back my wallet!
She tries to throw it again. They struggle.
RODNEY
What’s WRONG with you? Are you CRAZY?!
CHARLOTTE
YES! YES, I AM!!!
RODNEY
MY GOLD CARD is in there!!
3. Results
This section of the study presents statistics taken from the calculation of data based on the number
of words in a sentence. That is, the unit of measurement in this statistical calculation is the word. The
findings illustrate the contrast of ST and TT results of statistical parameters, i.e. N, max, min, R, Mo,
Md, Ẋ, Ϭ, ν, Sẋ, ε. The schematic representation follows the data of each drama character one by one.
3.1. Analysis of the part of the text that belongs to the drama character of
"Charlotte"
Having analysed the distribution of the number of words in a sentence by absolute and relevant
frequency, we have obtained such numeric characteristics:
Charlotte: the whole ST data: 1 — 58 (90,62%); 2 — 4 (6,25%); 3 — 1 (1,56%); 4 — 1 (1,56%);.
The data for «Charlotte» presupposes that the absolute frequency of sentence lengths with word
number 1 equals to 58; consequently, with word number of 2 equals to 4; with word number 3 equals
to 1; with word number of 4 equals to 1.
Talking about translation, the most frequent are sentences with the number of words that equals to
1.
Charlotte: the whole TT data: 1 — 35 (30,97%); 4 — 17 (15,04%); 5 — 13 (11,50%); 2 — 12
(10,62%); 6 — 12 (10,62%); 3 — 10 (8,85%); 7 — 5 (4,42%); 11 — 3 (2,65%); 9 — 2 (1,77%); 10
— 2 (1,77%); 8 — 1 (0,88%); 12 — 1 (0,88%). The last two are the least frequent.
On the basis of the data above the following calculations are made of number of meanings,
maximal meaning, minimal meaning, range, mode, median, mean, standard deviation, coefficient of
variation, standard error, measurement error.
Results are presented in Table 1.
Table 1
CHARLOTTE Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 64 113
max 4 12
min 1 1
R 3 11
Mo 1 1
Md 2,5 6,5
Ẋ 1,14 3,72
Ϭ 0,25 2,69
ν 0,4347 0,7243
Sẋ 0,062 0,2532
ε 0,1065 0,1335
Table 1 proves the following results:
ST data numeric characteristic: Number of meanings (N) — 64; maximal meaning (max) — 4;
minimal meaning (min) — 1; range (R) — 3; mode (Mo) — 1; median (Md) — 2,5; mean (Ẋ) —
1,14; standard deviation (Ϭ) — 0,50; coefficient of variation (ν) — 0,4347; standard error (Sẋ) —
0,0620; measurement error (ε) — 0,1065.
TT data numeric characteristic: Number of meanings (N) — 113; maximal meaning (max) — 12;
minimal meaning (min) — 1; range (R) — 11; mode (Mo) — 1; median (Md) — 6,5; mean (Ẋ) —
3,72; standard deviation (Ϭ) — 2,69; coefficient of variation (ν) — 0,7243; standard error (Sẋ) —
0,2532; measurement error (ε) — 0,1335.
3.2. Analysis of the part of the text that belongs to the drama character of
"Rodney"
Having analysed the distribution of the number of words in a sentence by absolute and relevant
frequency, we have obtained such numeric characteristics:
Rodney: the whole ST data: 1 — 37 (90,24%); 2 — 3 (7,32%); 3 — 1 (2,44%).
The data for «Rodney» presupposes that the absolute frequency of sentence lengths with word
number 1 equals to 37; consequently, with word number of 2 equals to 3; with word number 3 equals
to 1.
Rodney: the whole TT data: 1 — 16 (21,05%); 2 — 13 (17,11%); 3 — 11 (14,47%); 4 — 9
(11,84%); 5 — 9 (11,84%); 6 — 7 (9,21%); 7 — 5 (6,58%); 9 — 3 (3,95%); 8 — 2 (2,63%); 10 — 1
(1,32%).
Based on the data above, the following calculations are made and presented in Table 2.
Table 2
RODNEY Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 41 76
max 3 10
min 1 1
R 2 9
Mo 1 1
Md 2 5,5
Ẋ 1,12 3,76
Ϭ 0,39 2,37
ν 0,3519 0,6304
Sẋ 0,0617 0,2721
ε 0,1077 0,1417
Table 2 shows the following results:
ST data numeric characteristic: Number of meanings (N) — 41; maximal meaning (max) — 3;
minimal meaning (min) — 1; range (R) — 2; mode (Mo) — 1; median (Md) — 2,0; mean (Ẋ) —
1,12; standard deviation (Ϭ) — 0,39; coefficient of variation (ν) — 0,3519; standard error (Sẋ) —
0,0617; measurement error (ε) — 0,1077.
TT data numeric characteristic: Number of meanings (N) — 76; maximal meaning (max) — 10;
minimal meaning (min) — 1; range (R) — 9; mode (Mo) — 1; median (Md) — 5,5; mean (Ẋ) —
3,76; standard deviation (Ϭ) — 2,37; coefficient of variation (ν) — 0,6304; standard error (Sẋ) —
0,2721; measurement error (ε) — 0,1417.
3.3. Analysis of the part of the text that belongs to the drama character of
"Man"
By analogue to the previous characters (Charlotte and Rondey) we obtain the results for other
characters; here – Man.
Man: the whole ST data: 1 — 228 (87,36%); 2 — 27 (10,34%); 3 — 6 (2,30%).
Thus, the data for «Man» states that the absolute frequency of sentence lengths with word number
1 equals to 228; consequently, with word number of 2 equals to 27; with word number 3 equals to 6.
Man: the whole TT data: 1 — 99 (18,50%); 3 — 90 (16,82%); 4 — 78 (14,58%); 2 — 61
(11,40%); 5 — 48 (8,97%); 6 — 47 (8,79%); 7 — 37 (6,92%); 8 — 18 (3,36%); 9 — 16 (2,99%); 10
— 9 (1,68%); 11 — 8 (1,50%); 12 — 7 (1,31%); 15 — 4 (0,75%); 13 — 3 (0,56%); 16 — 3 (0,56%);
18 — 2 (0,37%); 14 — 1 (0,19%); 17 — 1 (0,19%); 19 — 1 (0,19%); 23 — 1 (0,19%); 27 — 1
(0,19%).
Next, we have calculated number of meanings, maximal meaning, minimal meaning, range, mode,
median, mean, standard deviation, coefficient of variation, standard error, measurement error. The
results are demonstrated in Table 3.
Table 3
MAN Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 261 535
max 3 27
min 1 1
R 2 26
Mo 1 1
Md 2 11
Ẋ 1,15 4,52
Ϭ 0,42 3,47
ν 0,3619 0,7678
Sẋ 0,0258 0,15
ε 0,0439 0,0651
ST data numeric characteristic: Number of meanings (N) — 261; maximal meaning (max) — 3;
minimal meaning (min) — 1; range (R) — 2; mode (Mo) — 1; median (Md) — 2,0; mean (Ẋ) —
1,15; standard deviation (Ϭ) — 0,42; coefficient of variation (ν) — 0,3619; standard error (Sẋ) —
0,0258; measurement error (ε) — 0,0439.
TT data numeric characteristic: Number of meanings (N) — 535; maximal meaning (max) — 27;
minimal meaning (min) — 1; range (R) — 26; mode (Mo) — 1; median (Md) — 11,0; mean (Ẋ) —
4,52; standard deviation (Ϭ) — 3,47; coefficient of variation (ν) — 0,7678; standard error (Sẋ) —
0,1500; measurement error (ε) — 0,0651.
3.4. Analysis of the part of the text that belongs to the drama character of
"Leonard"
By analogue to the previous characters we obtain the results for the character – Leonard.
Leonard: the whole ST data: 1 — 92 (86,79%); 2 — 12 (11,32%); 3 — 2 (1,89%).
Leonard: the whole TT data: 1 — 30 (14,49%); 5 — 28 (13,53%); 2 — 27 (13,04%); 3 — 27
(13,04%); 4 — 24 (11,59%); 6 — 23 (11,11%); 8 — 15 (7,25%); 7 — 8 (3,86%); 9 — 5 (2,42%); 10
— 5 (2,42%); 12 — 4 (1,93%); 14 — 3 (1,45%); 13 — 2 (0,97%); 16 — 2 (0,97%); 17 — 2 (0,97%);
11 — 1 (0,48%); 19 — 1 (0,48%).
Table 4
LEONARD Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 106 207
max 3 19
min 1 1
R 2 18
Mo 1 1
Md 2 9
Ẋ 1,15 4,94
Ϭ 0,41 3,53
ν 0,3539 0,7148
Sẋ 0,0396 0,2453
ε 0,0674 0,0974
ST data numeric characteristic:
Number of meanings (N) — 106; maximal meaning (max) — 3; minimal meaning (min) — 1;
range (R) — 2; mode (Mo) — 1; median (Md) — 2,0; mean (Ẋ) — 1,15; standard deviation (Ϭ) —
0,41; coefficient of variation (ν) — 0,3539; standard error (Sẋ) — 0,0396; measurement error (ε) —
0,0674.
TT data numeric characteristic:
Number of meanings (N) — 207; maximal meaning (max) — 19; minimal meaning (min) — 1;
range (R) — 18; mode (Mo) — 1; median (Md) — 9,0; mean (Ẋ) — 4,94; standard deviation (Ϭ) —
3,53; coefficient of variation (ν) — 0,7148; standard error (Sẋ) — 0,2453; measurement error (ε) —
0,0974.
3.5. Analysis of the part of the text that belongs to the drama character of
"Jennifer"
Jennifer: the whole ST data: 1 — 21 (84,00%); 2 — 3 (12,00%); 6 — 1 (4,00%);.
Jennifer: the whole TT data: 6 — 5 (19,23%); 4 — 4 (15,38%); 2 — 3 (11,54%); 3 — 3
(11,54%); 5 — 2 (7,69%); 9 — 2 (7,69%); 1 — 1 (3,85%); 7 — 1 (3,85%); 8 — 1 (3,85%); 10 — 1
(3,85%); 11 — 1 (3,85%); 14 — 1 (3,85%); 15 — 1 (3,85%).
Table 5
JENNIFER Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 25 26
max 6 15
min 1 1
R 5 14
Mo 1 6
Md 2 7
Ẋ 1,32 5,96
Ϭ 1,01 3,55
ν 0,7642 0,5948
Sẋ 0,2018 0,6955
ε 0,2996 0,2287
ST data numeric characteristic: Number of meanings (N) — 25; maximal meaning (max) — 6;
minimal meaning (min) — 1; range (R) — 5; mode (Mo) — 1; median (Md) — 2,0; mean (Ẋ) —
1,32; standard deviation (Ϭ) — 1,01; coefficient of variation (ν) — 0,7642; standard error (Sẋ) —
0,2018; measurement error (ε) — 0,2996.
TT data numeric characteristic: Number of meanings (N) — 26; maximal meaning (max) — 15;
minimal meaning (min) — 1; range (R) — 14; mode (Mo) — 6; median (Md) — 7,0; mean (Ẋ) —
5,96; standard deviation (Ϭ) — 3,55; coefficient of variation (ν) — 0,5948; standard error (Sẋ) —
0,6955; measurement error (ε) — 0,2287.
3.6. Analysis of the part of the text that belongs to the drama character of
"Marshall"
Marshall: the whole ST data: 1 — 94 (85,45%); 2 — 15 (13,64%); 4 — 1 (0,91%).
Marshall: the whole TT data: 2 — 31 (15,74%); 4 — 27 (13,71%); 3 — 26 (13,20%); 5 — 25
(12,69%); 6 — 21 (10,66%); 8 — 16 (8,12%); 7 — 11 (5,58%); 9 — 11 (5,58%); 1 — 9 (4,57%); 10
— 7 (3,55%); 11 — 6 (3,05%); 12 — 2 (1,02%); 16 — 2 (1,02%); 17 — 2 (1,02%); 23 — 1 (0,51%).
Table 6
MARSHALL Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 110 197
max 4 23
min 1 1
R 3 22
Mo 1 2
Md 2 8
Ẋ 1,16 5,39
Ϭ 0,44 3,38
ν 0,376 0,6279
Sẋ 0,0417 0,2409
ε 0,0703 0,0877
ST data numeric characteristic: Number of meanings (N) — 110; maximal meaning (max) — 4;
minimal meaning (min) — 1; range (R) — 3; mode (Mo) — 1; median (Md) — 2,0; mean (Ẋ) —
1,16; standard deviation (Ϭ) — 0,44; coefficient of variation (ν) — 0,3760; standard error (Sẋ) —
0,0417; measurement error (ε) — 0,0703.
TT data numeric characteristic: Number of meanings (N) — 197; maximal meaning (max) — 23;
minimal meaning (min) — 1; range (R) — 22; mode (Mo) — 2; median (Md) — 8,0; mean (Ẋ) —
5,39; standard deviation (Ϭ) — 3,38; coefficient of variation (ν) — 0,6279; standard error (Sẋ) —
0,2409; measurement error (ε) — 0,0877.
3.7. Analysis of the part of the text that belongs to the drama character of
"Joan"
Joan: the whole ST data: 1 — 43 (84,31%); 2 — 7 (13,73%); 3 — 1 (1,96%);.
Joan: the whole TT data: 3 — 16 (16,49%); 4 — 16 (16,49%); 1 — 13 (13,40%); 5 — 12
(12,37%); 2 — 10 (10,31%); 7 — 10 (10,31%); 6 — 6 (6,19%); 9 — 4 (4,12%); 8 — 3 (3,09%); 12
— 3 (3,09%); 11 — 1 (1,03%); 14 — 1 (1,03%); 17 — 1 (1,03%); 18 — 1 (1,03%).
Table 7
JOAN Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 51 97
max 3 18
min 1 1
R 2 17
Mo 1 3
Md 2 7,5
Ẋ 1,18 4,81
Ϭ 0,43 3,35
ν 0,3651 0,6958
Sẋ 0,0602 0,3402
ε 0,1002 0,1385
ST data numeric characteristic: Number of meanings (N) — 51; maximal meaning (max) — 3;
minimal meaning (min) — 1; range (R) — 2; mode (Mo) — 1; median (Md) — 2,0; mean (Ẋ) —
1,18; standard deviation (Ϭ) — 0,43; coefficient of variation (ν) — 0,3651; standard error (Sẋ) —
0,0602; measurement error (ε) — 0,1002.
TT data numeric characteristic: Number of meanings (N) — 97; maximal meaning (max) — 18;
minimal meaning (min) — 1; range (R) — 17; mode (Mo) — 3; median (Md) — 7,5; mean (Ẋ) —
4,81; standard deviation (Ϭ) — 3,35; coefficient of variation (ν) — 0,6958; standard error (Sẋ) —
0,3402; measurement error (ε) — 0,1385.
3.8. Analysis of the part of the text that belongs to the drama character of
"Michael"
Michael: the whole ST data: 1 — 34 (91,89%); 2 — 3 (8,11%);.
Michael: the whole TT data: 4 — 11 (20,00%); 5 — 10 (18,18%); 3 — 7 (12,73%); 7 — 7
(12,73%); 6 — 6 (10,91%); 2 — 5 (9,09%); 12 — 3 (5,45%); 1 — 2 (3,64%); 8 — 2 (3,64%); 10 — 2
(3,64%).
Table 8
MICHAEL Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 37 55
max 2 12
min 1 1
R 1 11
Mo 1 4
Md 1,5 5,5
Ẋ 1,08 5,16
Ϭ 0,27 2,57
ν 0,2525 0,4979
Sẋ 0,0449 0,3467
ε 0,0814 0,1316
ST data numeric characteristic: Number of meanings (N) — 37; maximal meaning (max) — 2;
minimal meaning (min) — 1; range (R) — 1; mode (Mo) — 1; median (Md) — 1,5; mean (Ẋ) —
1,08; standard deviation (Ϭ) — 0,27; coefficient of variation (ν) — 0,2525; standard error (Sẋ) —
0,0449; measurement error (ε) — 0,0814.
TT data numeric characteristic: Number of meanings (N) — 55; maximal meaning (max) — 12;
minimal meaning (min) — 1; range (R) — 11; mode (Mo) — 4; median (Md) — 5,5; mean (Ẋ) —
5,16; standard deviation (Ϭ) — 2,57; coefficient of variation (ν) — 0,4979; standard error (Sẋ) —
0,3467; measurement error (ε) — 0,1316.
3.9. Analysis of the part of the text that belongs to the drama character of
"Rachel"
Rachel: the whole ST data: 1 — 53 (91,38%); 2 — 5 (8,62%).
Rachel: the whole TT data: 4 — 18 (15,00%); 5 — 14 (11,67%); 7 — 14 (11,67%); 3 — 12
(10,00%); 2 — 11 (9,17%); 6 — 11 (9,17%); 1 — 10 (8,33%); 8 — 5 (4,17%); 9 — 4 (3,33%); 11 —
4 (3,33%); 10 — 3 (2,50%); 12 — 3 (2,50%); 13 — 3 (2,50%); 14 — 3 (2,50%); 16 — 3 (2,50%); 15
— 1 (0,83%); 20 — 1 (0,83%).
Table 9
RACHEL Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 58 120
max 2 20
min 1 1
R 1 19
Mo 1 4
Md 1,5 9
Ẋ 1,09 6,03
Ϭ 0,28 3,94
ν 0,2584 0,6529
Sẋ 0,0369 0,3596
ε 0,0665 0,1168
ST data numeric characteristic: Number of meanings (N) — 58; maximal meaning (max) — 2;
minimal meaning (min) — 1; range (R) — 1; mode (Mo) — 1; median (Md) — 1,5; mean (Ẋ) —
1,09; standard deviation (Ϭ) — 0,28; coefficient of variation (ν) — 0,2584; standard error (Sẋ) —
0,0369; measurement error (ε) — 0,0665.
TT data numeric characteristic: Number of meanings (N) — 120; maximal meaning (max) — 20;
minimal meaning (min) — 1; range (R) — 19; mode (Mo) — 4; median (Md) — 9,0; mean (Ẋ) —
6,03; standard deviation (Ϭ) — 3,94; coefficient of variation (ν) — 0,6529; standard error (Sẋ) —
0,3596; measurement error (ε) — 0,1168.
3.10. Analysis of the part of the text that belongs to the drama character of
"Percy"
Percy: the whole ST data: 1 — 34 (80,95%); 2 — 7 (16,67%); 3 — 1 (2,38%).
Percy: the whole TT data: 6 — 12 (16,44%); 3 — 11 (15,07%); 4 — 10 (13,70%); 5 — 7 (9,59%);
1 — 6 (8,22%); 2 — 5 (6,85%); 7 — 4 (5,48%); 8 — 4 (5,48%); 9 — 3 (4,11%); 11 — 3 (4,11%); 14
— 3 (4,11%); 10 — 2 (2,74%); 12 — 1 (1,37%); 18 — 1 (1,37%); 23 — 1 (1,37%).
Table 10
PERCY Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 42 73
max 3 23
min 1 1
R 2 22
Mo 1 6
Md 2 8
Ẋ 1,21 5,9
Ϭ 0,46 4,04
ν 0,3827 0,6839
Sẋ 0,0717 0,4727
ε 0,1158 0,1569
ST data numeric characteristic: Number of meanings (N) — 42; maximal meaning (max) — 3;
minimal meaning (min) — 1; range (R) — 2; mode (Mo) — 1; median (Md) — 2,0; mean (Ẋ) —
1,21; standard deviation (Ϭ) — 0,46; coefficient of variation (ν) — 0,3827; standard error (Sẋ) —
0,0717; measurement error (ε) — 0,1158.
TT data numeric characteristic: Number of meanings (N) — 73; maximal meaning (max) — 23;
minimal meaning (min) — 1; range (R) — 22; mode (Mo) — 6; median (Md) — 8,0; mean (Ẋ) —
5,90; standard deviation (Ϭ) — 4,04; coefficient of variation (ν) — 0,6839; standard error (Sẋ) —
0,4726; measurement error (ε) — 0,1569.
3.11. Analysis of the part of the text that belongs to the drama character of
"Al"
Al: the whole ST data: 1 — 23 (74,19%); 2 — 6 (19,35%); 3 — 2 (6,45%).
Al: the whole TT data: 6 — 11 (18,97%); 3 — 8 (13,79%); 4 — 7 (12,07%); 5 — 7 (12,07%); 1
— 6 (10,34%); 2 — 5 (8,62%); 7 — 3 (5,17%); 10 — 3 (5,17%); 9 — 2 (3,45%); 12 — 2 (3,45%); 8
— 1 (1,72%); 13 — 1 (1,72%); 14 — 1 (1,72%); 16 — 1 (1,72%).
Table 11
AL Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 31 58
max 3 16
min 1 1
R 2 15
Mo 1 6
Md 2 7,5
Ẋ 1,32 5,4
Ϭ 0,59 3,41
ν 0,4457 0,6316
Sẋ 0,1059 0,4476
ε 0,1569 0,1626
ST data numeric characteristic: Number of meanings (N) — 31; maximal meaning (max) — 3;
minimal meaning (min) — 1; range (R) — 2; mode (Mo) — 1; median (Md) — 2,0; mean (Ẋ) —
1,32; standard deviation (Ϭ) — 0,59; coefficient of variation (ν) — 0,4457; standard error (Sẋ) —
0,1059; measurement error (ε) — 0,1569.
TT data numeric characteristic: Number of meanings (N) — 58; maximal meaning (max) — 16;
minimal meaning (min) — 1; range (R) — 15; mode (Mo) — 6; median (Md) — 7,5; mean (Ẋ) —
5,40; standard deviation (Ϭ) — 3,41; coefficient of variation (ν) — 0,6316; standard error (Sẋ) —
0,4476; measurement error (ε) — 0,1626.
3.12. Analysis of the part of the text that belongs to the drama character of
"Nurse Wilson"
Nurse Wilson: the whole ST data: 1 — 42 (87,50%); 2 — 5 (10,42%); 3 — 1 (2,08%);.
Nurse Wilson: the whole TT data: 3 — 10 (13,16%); 4 — 10 (13,16%); 1 — 9 (11,84%); 5 — 9
(11,84%); 2 — 7 (9,21%); 6 — 6 (7,89%); 7 — 6 (7,89%); 12 — 4 (5,26%); 8 — 3 (3,95%); 9 — 3
(3,95%); 11 — 2 (2,63%); 13 — 2 (2,63%); 18 — 2 (2,63%); 10 — 1 (1,32%); 17 — 1 (1,32%); 23
— 1 (1,32%).
Table 12
NURSE WILSON Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 48 76
max 3 23
min 1 22
R 2 3
Mo 1 3
Md 2 8,5
Ẋ 1,15 5,91
Ϭ 0,41 4,48
ν 0,3558 0,7586
Sẋ 0,0588 0,5141
ε 0,1007 0,1705
ST data numeric characteristic: Number of meanings (N) — 48; maximal meaning (max) — 3;
minimal meaning (min) — 1; range (R) — 2; mode (Mo) — 1; median (Md) — 2,0; mean (Ẋ) —
1,15; standard deviation (Ϭ) — 0,41; coefficient of variation (ν) — 0,3558; standard error (Sẋ) —
0,0588; measurement error (ε) — 0,1007.
TT data numeric characteristic: Number of meanings (N) — 76; maximal meaning (max) — 23;
minimal meaning (min) — 1; range (R) — 22; mode (Mo) — 3; median (Md) — 8,5; mean (Ẋ) —
5,91; standard deviation (Ϭ) — 4,48; coefficient of variation (ν) — 0,7586; standard error (Sẋ) —
0,5141; measurement error (ε) — 0,1705.
3.13. Analysis of the part of the text that belongs to the drama character of
"Lilian"
Lilian: the whole ST data: 1 — 68 (91,89%); 2 — 6 (8,11%);.
Lilian: the whole TT data: 2 — 23 (14,94%); 4 — 19 (12,34%); 3 — 18 (11,69%); 5 — 17
(11,04%); 1 — 14 (9,09%); 6 — 13 (8,44%); 10 — 12 (7,79%); 8 — 10 (6,49%); 7 — 9 (5,84%); 9
— 7 (4,55%); 11 — 3 (1,95%); 13 — 2 (1,30%); 16 — 2 (1,30%); 17 — 2 (1,30%); 12 — 1 (0,65%);
14 — 1 (0,65%); 18 — 1 (0,65%).
Table 13
LILIAN Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 74 154
max 2 18
min 1 1
R 1 17
Mo 1 2
Md 1,5 9
Ẋ 1,08 5,51
Ϭ 0,27 3,69
ν 0,2525 0,6704
Sẋ 0,0317 0,2975
ε 0,0575 0,1059
ST data numeric characteristic: Number of meanings (N) — 74; maximal meaning (max) — 2;
minimal meaning (min) — 1; range (R) — 1; mode (Mo) — 1; median (Md) — 1,5; mean (Ẋ) —
1,08; standard deviation (Ϭ) — 0,27; coefficient of variation (ν) — 0,2525; standard error (Sẋ) —
0,0317; measurement error (ε) — 0,0575.
TT data numeric characteristic: Number of meanings (N) — 154; maximal meaning (max) — 18;
minimal meaning (min) — 1; range (R) — 17; mode (Mo) — 2; median (Md) — 9,0; mean (Ẋ) —
5,51; standard deviation (Ϭ) — 3,69; coefficient of variation (ν) — 0,6704; standard error (Sẋ) —
0,2975; measurement error (ε) — 0,1059.
3.14. Analysis of the part of the text that belongs to the secondary drama
characters
3.14.1. Character "One"
One: the whole ST data: 1 — 1 (100,00%). One: the whole TT data: 4 — 2 (40,00%); 3 — 1
(20,00%); 5 — 1 (20,00%); 6 — 1 (20,00%).
Table 14
ONE Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 1 5
max 1 6
min 1 3
R 0 3
Mo 1 4
Md 1 4,5
Ẋ 1 4,4
Ϭ 0 1,02
ν 0 0,2318
Sẋ 0 0,4561
ε 0 0,2032
ST data numeric characteristic: Number of meanings (N) — 1; maximal meaning (max) — 1;
minimal meaning (min) — 1; range (R) — 0; mode (Mo) — 1; median (Md) — 1,0; mean (Ẋ) —
1,00; standard deviation (Ϭ) — 0,00; coefficient of variation (ν) — 0,0000; standard error (Sẋ) —
0,0000; measurement error (ε) — 0,0000.
TT data numeric characteristic: Number of meanings (N) — 5; maximal meaning (max) — 6;
minimal meaning (min) — 3; range (R) — 3; mode (Mo) — 4; median (Md) — 4,5; mean (Ẋ) —
4,40; standard deviation (Ϭ) — 1,02; coefficient of variation (ν) — 0,2318; standard error (Sẋ) —
0,4561; measurement error (ε) — 0,2032.
3.14.2. Character "Two"
Two: the whole ST data: 1 — 2 (66,67%); 2 — 1 (33,33%). Two: the whole TT data: 4 — 2
(33,33%); 8 — 2 (33,33%); 5 — 1 (16,67%); 10 — 1 (16,67%).
Table 15
TWO Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 3 6
max 2 10
min 1 4
R 1 6
Mo 1 4
Md 1,5 6,5
Ẋ 1,33 6,5
Ϭ 0,47 2,29
ν 0,3536 0,3525
Sẋ 0,2722 0,9354
ε 0,4001 0,2821
ST data numeric characteristic: Number of meanings (N) — 3; maximal meaning (max) — 2;
minimal meaning (min) — 1; range (R) — 1; mode (Mo) — 1; median (Md) — 1,5; mean (Ẋ) —
1,33; standard deviation (Ϭ) — 0,47; coefficient of variation (ν) — 0,3536; standard error (Sẋ) —
0,2722; measurement error (ε) — 0,4001.
TT data numeric characteristic: Number of meanings (N) — 6; maximal meaning (max) — 10;
minimal meaning (min) — 4; range (R) — 6; mode (Mo) — 4; median (Md) — 6,5; mean (Ẋ) —
6,50; standard deviation (Ϭ) — 2,29; coefficient of variation (ν) — 0,3525; standard error (Sẋ) —
0,9354; measurement error (ε) — 0,2821.
3.14.3. Character "Three"
Three: the whole ST data: 1 — 4 (80,00%); 2 — 1 (20,00%).
Three: the whole TT data: 4 — 2 (40,00%); 3 — 1 (20,00%); 6 — 1 (20,00%); 8 — 1 (20,00%).
Table 16
THREE Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 5 5
max 2 8
min 1 3
R 1 5
Mo 1 4
Md 1,5 5
Ẋ 1,2 5
Ϭ 0,4 1,79
ν 0,3333 0,3578
Sẋ 0,1789 0,8
ε 0,2922 0,3136
ST data numeric characteristic: Number of meanings (N) — 5; maximal meaning (max) — 2;
minimal meaning (min) — 1; range (R) — 1; mode (Mo) — 1; median (Md) — 1,5; mean (Ẋ) —
1,20; standard deviation (Ϭ) — 0,40; coefficient of variation (ν) — 0,3333; standard error (Sẋ) —
0,1789; measurement error (ε) — 0,2922.
TT data numeric characteristic: Number of meanings (N) — 5; maximal meaning (max) — 8;
minimal meaning (min) — 3; range (R) — 5; mode (Mo) — 4; median (Md) — 5,0; mean (Ẋ) —
5,00; standard deviation (Ϭ) — 1,79; coefficient of variation (ν) — 0,3578; standard error (Sẋ) —
0,8000; measurement error (ε) — 0,3136.
3.14.4. Character "Four"
Four: the whole ST data: 1 — 2 (100,00%).
Four: the whole TT data: 4 — 2 (50,00%); 1 — 1 (25,00%); 2 — 1 (25,00%).
Table 17
FOUR Numeric characteristics of word distribution within the sentence of the drama
Unit ST TT
N 2 4
max 1 4
min 1 1
R 0 3
Mo 1 4
Md 1 2
Ẋ 1 2,75
Ϭ 0 1,3
ν 0 0,4724
Sẋ 0 0,6495
ε 0 0,4629
ST data numeric characteristic: Number of meanings (N) — 2; maximal meaning (max) — 1;
minimal meaning (min) — 1; range (R) — 0; mode (Mo) — 1; median (Md) — 1,0; mean (Ẋ) —
1,00; standard deviation (Ϭ) — 0,00; coefficient of variation (ν) — 0,0000; standard error (Sẋ) —
0,0000; measurement error (ε) — 0,0000.
TT data numeric characteristic: Number of meanings (N) — 4; maximal meaning (max) — 4;
minimal meaning (min) — 1; range (R) — 3; mode (Mo) — 4; median (Md) — 2,0; mean (Ẋ) —
2,75; standard deviation (Ϭ) — 1,30; coefficient of variation (ν) — 0,4724; standard error (Sẋ) —
0,6495; measurement error (ε) — 0,4629.
4. Comparative analysis of word distribution in sentences
4.1. Difference in the Number of meanings (N) in ST and TT
Given form the results above that the translated variant statistical parameters data exceeds the
original drama in the majority of cases, we now turn our sights to one parameter – Number of
meanings (N). We tend to compare the data and find the difference (if present). Our assumption 1 is
that the TT is much longer in terms of word usage within the sentence.
Character’s name ST TT Difference
Charlotte 64 113 +49
Rodney 41 76 +35
Man 261 535 +274
Leonard 106 207 +101
Jennifer 25 26 +1
Marshal 110 197 +87
Joan 51 97 +46
Michael 37 55 +18
Rachel 58 120 +62
Percy 42 73 +31
Al 31 58 +27
Nurse Wilson 48 76 +28
Lilian 74 154 +80
One 1 5 +4
Two 3 6 +3
Three 5 5 0
Four 2 4 +2
Figure 1: Comparative statistics of Number of meanings (N) in ST and TT
To recall, character "Man" is the protagonist and the main character of the play. He is a well-
dressed gentleman who is willing to jump off the seventh story.
He has a number of conversations with the residents of the building. He feels lost and compelled to
stand on the seventh story of the building. Taking into account the results of Figure 1 we hold
assumption 2 that the translator adds a considerable number of words (274), or, he rather, doubles the
ST quantity, due to a number of reasons:
to explain the original;
to compensate literary imagery losses;
to add something from the translator himself, to recreate, so to say, the original;
due to structural and lexico-gramatical allomorphic features of a language pair.
Whatever reason stands behind this translator’s decision-making, it is a prosperous ground for
further Translation Studies analysis.
4.2. Analysis of the whole text
Here we focus on statistical parameters with the defined unit of measurement – a word. The
number of words in a drama text utterunces is important due to a couple of reasons:
the length of lines of the written script;
chronometry and metrics of the whole drama act;
pithiness and iconicity of each phrase.
Below are the results on the distribution of the number of words in a TT sentence by absolute and
relevant frequency.
The most frequent are sentences in the translated text with the number of words 4 – 259 (14,2%),
1 – 255(13,98%), 3– 255(13,98%), 2 – 219(12,01%) 5 – 205(11,24%), 6 -182 (9,98), 7-121 (6,63%),
8 – 84(4,61), 9- 62 (3,4%), 10 – 49 (2,69%), 11 – 32 (1,75), 12 – 31 (1,7%), 14 – 14 (0,77%), 13 – 13
(0,71%), 16 – 13(0,71%), 17 – 9 (0,49%), 18 – 7 (0,38%), 15 – 6 (0,33%), 23 – 4 (0,22%), 19 – 2
(0,11%), 20 – 1 (0,05%), 27 – 1 (0,05%). The last two results are the least frequent.
In the following Figure 2 we can see a comparison of the number of words in the sentences of the
whole TT drama work.
The x-axis is the number of sentences, and the y-axis is the number of words in a sentence.
Figure 2: Number of words in sentences of TT
5. Conclusions
The main advances of statistical linguistics have been retrieved in the article. The original
Canadian play has been compared with the corresponding translated text in terms of statistical
parameters, which has never been done before.
The paper is of practical and applied value; however, the scientific value of the paper is seen as
such that the suggested approach and methods will eventually allow formulating and substantiating a
plausible scientific hypothesis in the realm of statistical linguistics and translation studies. At this
point it is proven that bilingual drama texts are well adoptable for NLP and reveal promising
outcomes.
We have verified absolute and relevant distribution, probability measurement, also: N, max, min,
R, Mo, Md, Ẋ, Ϭ, ν, Sẋ, ε in the sentences of both texts.
Specifically designed software, which is represented as a combination of XML markup language,
Microsoft Excel spreadsheet, and Python programming language, has been used. Results of statistical
calculations of the drama ―7 stories‖ by Morris Panych by unit of measure word are presented in the
corresponding Tables 1 – 17.
Structural recognition provides useful information about the characters of the play, original and
translation, namely the length of the sentence in word units that will help with further comparisons of
ST and TT. The quantitative characteristics of the original play and its Ukrainian translation on the
lexical level relying on the linguistic statistical analysis have been clarified: the amount of translated
text Numbers of meaning (N) exceeds considerably and demands further analysis. The discrepancy
becomes obvious with number of characters (Man, Leonard, Marshal, Lilian)
The correlation of coefficients has been presented in tables and figures to illustrate the material
under research.
The prospect of the study is to further explore the problems of translator’s meaningful choices
which resulted in the declared above data.
6. Acknowledgement
The project has been carried out within the complex academic topic ―Application of modern
technologies for optimization of information processes in natural language‖ at Lviv Polytechnic
National University. At the initial stage the project underwent the consultancy of Ihor Kulchytskyy, to
whom we express our gratitude.
7. References
[1] M. Panych, Seven Stories, Vancouver: Talonbooks, 2013.
[2] M. Panych, 7 istorii, [per. Z anhliiskoi Ivana Krychfalushiia], Brusturiv: Dyskursus, 2014.
[3] S. Laviosa (Ed.), Corpus-based Translation Studies: Theory, Findings, Applications, Rodopy,
2002.
[4] K. H. Chen, and H. H. Chen, Aligning bilingual corpora especially for language pairs from
different families. Information Sciences Applications, 1995, 42, pp. 57–81.
[5] J. Munday, A Computer-assisted approach to the Analysis of Translation Shifts, Meta, 1998,
XLIII, 4.
[6] F. Zanettin, Parallel corpora in translation studies: Issues in corpus design and analysis. In
Intercultural Faultlines. Research Models in Translation Studies I: Textual and Cognitive
Aspects, ed. M. Olohan, pp. 105–118. Manchester: St. Jerome, 2000.
[7] J. Allen, Natural Language Understanding. Cummings Publishing Company, Redwood City,
1995.
[8] D. Barnard, et al. ―SGML-Based Markup for Literary Texts: Two Problems and Some
Solutions.‖ Computers and the Humanities, vol. 22, no. 4, 1988, pp. 265–276. JSTOR, URL:
www.jstor.org/stable/30200136. Accessed 28 Feb. 2021.
[9] P. Blackburn, J. Bos, M. Kohlhase, & H. De Nivelle, Inference and computational semantics.
In Computing Meaning, Springer Netherlands, 2001, pp. 11–28.
[10] I. Dagan, and O. Glickman, Probabilistic textual entailment: generic applied modeling
of language variability. In Proceedings of the PACAL Workshop on Learning Methods for
Text Understanding and Mining, Grenoble, France, 2004, pp. 26–29.
[11] R. Dale, H. Moisl, H. Somers (Eds.), Handbook of natural language processing. CRC
press, 2000.
[12] M. Dilai, O. Levchenko, Discourses Surrounding Feminism in Ukraine: A Sentiment
Analysis of Twitter Data 2018 IEEE 13th International Scientific and Technical Conference
on Computer Sciences and Information Technologies, CSIT 2018 – Proceedings
2018 | conference-paper doi: 10.1109/STC-CSIT.2018.8526694
[13] A. Hogan, The Web of Data. Springer, 2020.
[14] V. Lytvyn, V. Vysotska, T. Hamon, N. Grabar, N. Sharonova, O. Cherednichenko,
O. Kanishcheva (Eds.), Computational Linguistics and Intelligent Systems. Proc. 4thInt.
Conf. COLINS 2020. Volume I:Workshop. Lviv, Ukraine, April 23-24, 2020, CEUR-
WS.org, online
[15] M. Marcus, B. Santorini, M. Marcinkiewicz, Building a Large Annotated Corpus of
English: Penn TreeBank. Computational linguistics: Special Issue on Using Large Corpora,
1993, 19(2), pp. 313–330.
[16] C. Matthews, An Introduction to Natural Language Processing Through Prolog,
Routledge: London and New York, 2014.
[17] M. Oakes, Sentence and word alignment in the CARTER project. In Using Corpora
for Language Research, ed. J. Thomas, and M. Short, London: Longman, 1996, pp. 211–233.
[18] P. Pavis, Theatre at the Crossroads of Culture, Routledge, 1992.
[19] S. Bassnett, Translating for the Theatre: The Case Against Performability. TTR :
traduction, terminologie, rédaction, 1991, 4(1), pp. 99–111. URL:
https://doi.org/10.7202/037084ar.
[20] S. Bassnett, Still Trapped in the Labyrinth: Further Reflections on Translation and
Theatre, Constructing Cultures: Essays on Literary Translation.-Multilingual Matters, 1998,
pp. 90–108.
[21] T.H. Howard-Hill, Modern Textual Theories and the Editing of Plays. The Library,
6th ser., 1989, 11, pp. 89–115.
[22] M. Issacharoff, F. Robin Jones (Eds.), Performing Texts. Philadelphia: University of
Pennsylvania Press, 1988.
[23] J. Lavagnino, E. Mylonas, The show must go on: Problems of tagging performance
texts. Comput Hum, 1995, pp. 113–121. URL: https://doi.org/10.1007/BF01830705
[24] Corpus-based Language Studies: An Advanced Resource Book, ed. T. McEnery, R.
Xiao, Y. Tono, Routledge, 2006.
[25] N. Dershowitz, E. Nissan (Eds.), Language, Culture, Computation: Computing for the
Humanities, Law and Narratives. Springer, 2014.
[26] O. Levchenko, O. Tyshchenko and M. Dilai. Associative Verbal Network of the
Conceptual Domain БІДА (MISERY) in Ukrainian. Proceedings of the 4th International
Conference on Computational Linguistics and Intelligent Systems (COLINS 2020). Volume
I: Main Conference. URL: http://ceur-ws.org/Vol-2604/Associative Verbal Network of the
Conceptual Domain БІДА (MISERY) in Ukrainian
[27] N. Shakhovska, and M. Medykovskyy (Eds), Advances in Intelligent Systems and
Computing III: Selected papers from the International Conference on Computer Science and
Information Technologies, CSIT 2018, September 11–14 Lviv, Ukraine. Springer: Springer
Nature Switzerland, 2019.
[28] C.M. Sperberg-McQueen, Text in the Electronic Age: Textual Study and Text
Encoding, with Examples from Medieval texts. Literary and Linguistic Computing, 6 (1991),
pp.34–46.
[29] C.M. Sperberg-McQueen, and B. Lou (Eds.), Guidelines for Electronic Text
Encoding and Interchange (TEI P3). Chicago and Oxford: Text Encoding Initiative, 1994.
[30] S. Shaheen, and M. Spruit. Full-Text or Abstract? Examining Topic Coherence
Scores Using Latent Dirichlet Allocation. 2017 IEEE International Conference on Data
Science and Advanced Analytics (DSAA), 2017, doi:10.1109/dsaa.2017.61.
[31] Topic Modeling in Python with Gensim. Machine Learning Plus, 16 Apr. 2020, URL:
www.machinelearningplus.com/nlp/topic-modeling-gensim-python.
[32] K. Aguilar, NLP Techniques with Shakespeare’s Plays: Cleaning and Classifying
Text with the Bard, 2020. URL: https://medium.com/analytics-vidhya/nlp-techniques-with-
shakespeares-plays-d8843ba26a4f.
[33] O. Levchenko, M. Dilai, (2019) Attitudes Toward Feminism in Ukraine: A
Sentiment Analysis of Tweets. In: Shakhovska N., Medykovskyy M. (eds) Advances in
Intelligent Systems and Computing III. CSIT 2018. Advances in Intelligent Systems and
Computing, vol 871. Springer, Cham. doi:10.1007/978-3-030-01069-0_9