What Is Subtitle Segmentation?

advertisement
Optimizing Subtitle Segmentation on
Smartphones
Olivia Gerber –Morón
Universitat Autònoma de Barcelona
The Consortium
2 Academic
Institutions
4 Public
Broadcasters
2 Research
Institutes
4 SMEs
3
Work Package Structure
Enhancement of the quality of subtitles
Readability
“quality that makes possible the recognition of the
information content of material when it is represented by
alphanumeric characters in meaningful groupings, such as
words, sentences, or continuous text” (Sanders and
McCormick 1993)
5
Readability of subtitles
Textual constraints:
Legibility of the characters
Synchronization of the text
Line-breaks
Formal constraints:
Space
Time factors
(Gottlieb 1992:164)
6
What Is Subtitle Segmentation?
“To segment means to divide something into separate
parts or sections. In subtitling, segmentation is the
division of the subtitled dialogue, narration etc. into
sections or segments – subtitles – that the viewers can
understand at a glance.”
(Díaz Cintas & Remael 2007: 172)
7
Categories of segmentation
Line-breaks (intrasegmentation)
“a sentence may have to be distributed over the two
available lines of a subtitle”
Intersegmentation
“[…] or it may run on, into two or more subtitles”
8
Recommendations
Readability for subtitle segmentation
“[…] lines should be divided in such a way that words
intimately connected by logic, semantics or grammar are
written on the same line whenever possible” (Ivarsson 1998:
77).
“[…] make the breaks coincide with the beginnings and ends
of phrases” (Ivarsson 1998: 90).
9
Recommendations
“Subtitle text should appear segmented at the highest
syntactic nodes possible. This means that each subtitle flash
should ideally contain one complete sentence. In cases where
the sentence cannot fit in a single-line subtitle and has to
continue over a second line or even over a new subtitle flash, the
segmentation on each of the lines should be arranged to
coincide with the highest syntactic node possible.”
(Karamitroglou, 1998)
10
Recommendations
(Karamitroglou, 1998)
11
Recommendations
The destruction of the city was inevitable.
(44 characters)
Segmentation at N5:
The destruction of the
city was inevitable.
Segmentation at N2:
The destruction of the city
was inevitable.
12
Recommendations
(Ivarsson 1998: 91)
13
Past research on subtitle segmentation
The Cognitive Effectiveness of Subtitle Processing (Perego et
al., 2010).
“[…] subtitle segmentation quality did not have a significant
impact in our study. Participants appeared to process wellsegmented and ill-segmented subtitles basically in the same way
and with the same outcomes.”
14
Past research on subtitle segmentation
Effects of text chunking on subtitling: A quantitative and
qualitative examination (Rajendran et al., 2011).
“[…] word-for-word subtitles was deemed unfavorable
compared to the other subtitling methods, notably chunked by
phrase[…]. This allowed for a steadier, more natural viewing
experience.”
15
Past research on the effects of
screen size on the reading speed
“It is easier to read big characters at a convenient
distance than very small characters at a short distance,
even though the visual angle remains the same”
(Perego, 2005)
16
Past research on screen size
The effect of mobile phone screen size on video based
learning (Maniar et al., 2008).
“[…]people tend to pay more attention when viewing
a larger screen display.”
“[…] the effectiveness of the learning experience may
be inhibited.”
17
Past research on screen size
The Effects of Screen-Size and Communication Modality
on Psychology of Mobile Device Users (Joon Kim et al.,
2011).
“Screen-size, however, had no effects on the
participants’ perception of the news story nor on the
perceived ease of use of the device.”
“[…] larger screen-size was the key to greater
enjoyment.”
18
Past research on subtitles and screen size
Reading Across Devices
The smartphone got the most negative results
19
Testing different subtitle
segmentation layouts
Geometrical Segmentation vs. Syntactical Segmentation
“The upper line and the lower line of a two-line subtitle should
be proportionally as equal in length as possible, since the
viewers’ eye is more accustomed to reading text in a rectangular
rather than a triangular format. This happens because the
conventional text format of printed material is rectangular (in
columns or pages). […] this means that the segmentation of
subtitled text should be a compromise between syntax and
geometry. However, if we had to sacrifice the one for the sake of
the other, we should prefer to sacrifice geometry.”
(Karamitroglou, 1998)
20
Geometrical Segmentation vs. Syntactical Segmentation
“[…] the segmentation of the text should follow syntactic and
grammatical considerations rather than aesthetic rules, e.g.
having lines with a symmetrical layout.”
(Díaz Cintas & Remael 2007: 172)
“Some subtitlers feel that a degree of equilibrium in line
length is more pleasing aesthetically[…].”
(Díaz Cintas & Remael 2007: 176)
21
Geometrical Segmentation vs. Syntactical Segmentation
My whole life, I’ve been
followed by loneliness
vs.
My whole life,
I’ve been followed by loneliness
(Karamitroglou, 1998)
22
23
Design
Thirty Italian native participants from Ticino
(Switzerland):
Deaf or hard of hearing
Control group
No knowledge of the original language of the film fragments
Two fragments of 10 minutes each from the first
episode of an American TV series dubbed in
Hungarian with Italian subtitles
24
Design
Presentation of two different styles of subtitling segmentation
quality of the same video
10 min clip using geometrical segmentation
vs.
10 min clip using syntactical segmentation
After each fragment, questionnaires were distributed:
Multiple-choice questionnaires on comprehension and
performance on word recognition memory (lexical memory)
Questionnaire about the preference of subtitle segmentation
25
Aims
Check if there are differences depending on the user
profile
Analyze if there are differences of comprehesion and
lexical memory according to the segmentation layout
New recommendations for broadcasters, subtitlers,
various organizations; live-subtitles and automatic
subtitles softwares
26
Results
GENERAL COMPREHENSION
27
RECALL QUESTIONNAIRE (lexical memory)
28
Preferences
Deaf and hard of hearing
29
Control group
30
Conclusions
Suboptimal line segmentation will not significantly
disrupt processing
Geometrical style is not a burden on cognitive
processing
Individuals with different profiles are able to adopt
different strategies to process segmentation layouts
Screen size does not seem to be a determining factor
for subtitle segmentation
The general preferences is for syntactical layout,
although viewers do not dislike geometrical 31layout
Next steps…
Measuring text processing by making use of eyetracking data
Previous studies tended to define areas of interests
(i.e. the entire subtitle area)
Measure of the visual processing of subtitles over
extended text to calculate the degree to which each
individual subtitle was read (Kruger, 2013)
Data on subtitle reading vs. data on visual attention
to subtitles
32
Subtitle segmentation
on dynamic subtitle placement
33
34
35
Dziękuję 
DD/MM/2014
www.hbb4all.eu
36
Download