1
00:00:11,592 --> 00:00:13,416
Live Text Access.

2
00:00:13,490 --> 00:00:16,760
Training
for real time intralingual subtitlers.

3
00:00:36,410 --> 00:00:41,672
The introduction to this Unit
has already explained what respeaking is.

4
00:00:41,704 --> 00:00:45,576
The introduction
to this element has already explained

5
00:00:45,608 --> 00:00:49,480
the Psycho-cognitive skills
a respeaker must possess.

6
00:00:49,944 --> 00:00:55,192
This video lecture is aimed
to let you know about the Effort Model,

7
00:00:55,224 --> 00:01:00,088
which is at the basis of the skill
of listening and speaking simultaneously.

8
00:01:00,740 --> 00:01:05,768
And at letting you learn the skill
of listening and speaking simultaneously

9
00:01:05,784 --> 00:01:07,560
in two specific contexts.

10
00:01:08,120 --> 00:01:10,600
Where verbatim accuracy is required,

11
00:01:10,632 --> 00:01:14,296
you will learn
to listen and repeat at the same time.

12
00:01:14,760 --> 00:01:17,960
Where sensatim accuracy is required,

13
00:01:17,992 --> 00:01:22,712
you will learn to listen and reformulate
at the same time.

14
00:01:25,512 --> 00:01:28,472
This is the agenda
of this presentation.

15
00:01:28,940 --> 00:01:31,930
I will first recall
the skills of the respeaker.

16
00:01:32,180 --> 00:01:35,350
Then, I will concentrate
on the so-called effort model,

17
00:01:35,384 --> 00:01:38,020
at the basis
of simultaneous interpreting,

18
00:01:38,040 --> 00:01:41,980
a mental activity,
very close to real time subtitling.

19
00:01:42,740 --> 00:01:46,430
 I will finally focus on shadowing,
a manifold exercise,

20
00:01:46,800 --> 00:01:50,200
you will be requested to do
along the course

21
00:01:50,232 --> 00:01:55,080
to develop, reinforce,
and finally maintain this skill

22
00:01:55,336 --> 00:02:01,280
in two different contexts:
verbatim and sensatim subtitling.

23
00:02:08,480 --> 00:02:09,970
As you have seen,

24
00:02:10,008 --> 00:02:14,030
a respeaker has to do
many things at the same time.

25
00:02:15,336 --> 00:02:18,792
In this Element, we will deal
with the Psychocognitive skill

26
00:02:18,824 --> 00:02:21,400
of listening and speaking
at the same time,

27
00:02:21,432 --> 00:02:23,880
the bulk of the respeaking skills.

28
00:02:23,912 --> 00:02:27,850
In particular, for a respeaker
to become a professional,

29
00:02:27,870 --> 00:02:33,040
he or she has to first develop
the skill of doing two things at a time.

30
00:02:33,060 --> 00:02:38,150
Listening and understanding the speaker
while talking to the computer,

31
00:02:38,170 --> 00:02:41,864
so that real-time subtitles are produced.

32
00:02:42,040 --> 00:02:46,930
You can develop this skill by the time
thanks to ad hoc exercises.

33
00:02:46,950 --> 00:02:50,760
But you may be interested in understanding
the rationale behind it,

34
00:02:50,780 --> 00:02:55,410
the effort model,
and then shadowing as a way to develop it.

35
00:03:04,020 --> 00:03:06,660
So, what is the Effort Model?

36
00:03:06,990 --> 00:03:12,860
In 1985, Daniel Gile proposes
the so-called <i>Modèle d'efforts</i>,

37
00:03:12,880 --> 00:03:14,840
or Effort Model,

38
00:03:14,860 --> 00:03:20,550
based on observing the mistakes
of trainees in simultaneous interpreting.

39
00:03:21,440 --> 00:03:26,470
Gile realises that trainees
do mistakes that are so banal

40
00:03:26,490 --> 00:03:30,810
the reason for these mistakes
cannot be a poor command of the language.

41
00:03:31,240 --> 00:03:34,640
He understands
that this is due to other reasons,

42
00:03:34,640 --> 00:03:38,770
mainly that our brain is to be trained

43
00:03:38,770 --> 00:03:43,390
if we want it do more than one
non-automatic thing at the same time.

44
00:03:44,540 --> 00:03:50,850
indeed, our brain can process
only one non-automatic process at a time.

45
00:03:51,930 --> 00:03:53,160
To understand this,

46
00:03:53,180 --> 00:03:58,770
it is important to know
that mental activities can be of 3 types.

47
00:03:59,190 --> 00:04:03,980
They can be automatic,
like that at basis of breathing.

48
00:04:04,510 --> 00:04:08,350
They can be semi-automatic,
like that at the basis of walking,

49
00:04:08,370 --> 00:04:12,630
They can be non-automatic, like that
at the basis of active listening

50
00:04:12,710 --> 00:04:17,440
or – guess! – repeating or reformulating
what somebody has just said.

51
00:04:18,040 --> 00:04:22,620
And while you can breathe,
walk and speak at the same time,

52
00:04:22,664 --> 00:04:25,310
you cannot listen and speak
at the same time,

53
00:04:25,350 --> 00:04:29,550
unless you are trained to do so,
in courses like this one.

54
00:04:34,000 --> 00:04:35,880
As said in the introduction,

55
00:04:35,920 --> 00:04:39,350
the effort model is thought
for Simultaneous interpreting,

56
00:04:39,400 --> 00:04:43,900
but it can be easily adapted
to real-time intralingual subtitling.

57
00:04:44,860 --> 00:04:48,880
Adapting Gile’s notions
o real-time subtitling,

58
00:04:48,880 --> 00:04:53,410
while respeaking,
we mainly do 4 non-automatic actions
at the same time.

59
00:04:54,060 --> 00:04:56,570
Listening to the speaker
and understanding,

60
00:04:56,600 --> 00:05:02,350
which implies we analyze the waves
she produces and turn them into meanings.

61
00:05:03,370 --> 00:05:04,890
Then there is memory,

62
00:05:04,950 --> 00:05:10,660
which is the capacity to temporarily store
meaning before we do something with it.

63
00:05:11,820 --> 00:05:15,150
Production is the fact
of producing sentences

64
00:05:15,150 --> 00:05:20,200
that need to both adequately render
the meanings uttered by the speaker

65
00:05:20,200 --> 00:05:24,520
and be understandable
by the Automatic Speech Recognition tool.

66
00:05:25,120 --> 00:05:29,930
Coordination is the fact
of monitoring one’s performance,

67
00:05:29,930 --> 00:05:33,860
so that a balance is found
between these three efforts.

68
00:05:36,312 --> 00:05:42,488
Now, these efforts require
a lot of processing capacity.

69
00:05:43,180 --> 00:05:44,180
As said,

70
00:05:44,248 --> 00:05:49,970
we need to train our brain to find
a balance between these four efforts.

71
00:05:50,930 --> 00:05:54,920
As long as our processing capacity
is enough

72
00:05:54,980 --> 00:05:58,260
to carry on these mental activities
simultaneously,

73
00:05:58,310 --> 00:05:59,930
we then do a good job.

74
00:06:00,400 --> 00:06:04,000
If it is not,
because of fatigue or other reasons,

75
00:06:04,000 --> 00:06:08,750
we may experience breakdowns
in one of the activities,

76
00:06:08,750 --> 00:06:12,640
with negative consequences
on the quality of the subtitles.

77
00:06:16,310 --> 00:06:21,590
In particular, a good performance
is possible under these conditions.

78
00:06:22,380 --> 00:06:25,720
First of all,
the speaker should speak properly,

79
00:06:25,720 --> 00:06:29,540
not too fast, with a clear voice
and not too technically.

80
00:06:30,470 --> 00:06:36,280
Then it is important the respeaker
knows the topic or has studied it.

81
00:06:37,370 --> 00:06:41,260
Moreover, the respeaker must work
in a well-equipped environment

82
00:06:41,260 --> 00:06:45,680
and with working turns
which are not too stressful.

83
00:06:46,080 --> 00:06:51,272
And finally, the interaction
with the machine should be at its utmost.

84
00:06:52,480 --> 00:06:56,360
In case one
or more of these conditions are not met,

85
00:06:56,872 --> 00:06:59,704
more processing capacity is required

86
00:06:59,800 --> 00:07:03,336
in one of the four efforts
to bridge the gap.

87
00:07:04,072 --> 00:07:08,290
This may cause a breakdown
in one of the other activities

88
00:07:08,390 --> 00:07:12,770
which, in the end, go to the detriment
of the quality of the subtitles.

89
00:07:13,400 --> 00:07:16,470
In particular,
if the speaker is challenging,

90
00:07:16,470 --> 00:07:20,070
it will be harder for the respeaker
to listen and understand.

91
00:07:20,584 --> 00:07:23,760
If the topic is not well mastered
or totally unknown,

92
00:07:23,864 --> 00:07:27,480
it will be harder to remember
what the speaker says.

93
00:07:27,830 --> 00:07:32,872
If the respeaker is not in good mental
and or physical conditions,

94
00:07:32,952 --> 00:07:36,050
it will be harder to produce
coherent subtitles.

95
00:07:36,580 --> 00:07:40,952
Finally, if the interaction
with the machine is not good,

96
00:07:41,208 --> 00:07:43,816
coordination breakdowns may occur.

97
00:07:46,180 --> 00:07:49,608
In the next video lectures,

98
00:07:49,700 --> 00:07:53,464
we will see
how to cope with each of these situations

99
00:07:53,560 --> 00:07:55,976
that may cause bad performance.

100
00:07:56,340 --> 00:07:57,760
In this video lecture,

101
00:07:57,860 --> 00:08:03,500
we will try and make sure that, in good
conditions, you are going to perform well.

102
00:08:04,360 --> 00:08:08,440
Moreover, we will try to see
how it is possible to semi-automate

103
00:08:08,540 --> 00:08:13,848
some of the simultaneous, non-automatic
actions which are required in respeaking

104
00:08:14,160 --> 00:08:17,960
to reduce the processing capacity
needed for each of them.

105
00:08:18,552 --> 00:08:24,630
One way to go for that is shadowing,
which we will see in the next section.

106
00:08:32,200 --> 00:08:33,950
In this second section,

107
00:08:34,050 --> 00:08:38,210
we will deal with the notion of shadowing
as a tool to develop

108
00:08:38,360 --> 00:08:43,950
and semi-automate the psycho-cognitive skill
of listening and speaking at the same time.

109
00:08:45,224 --> 00:08:51,208
Shadowing means you listen to a speech
and you repeat it.

110
00:08:52,520 --> 00:08:56,070
There is no need to involve
the software for the moment.

111
00:08:56,232 --> 00:08:59,960
However, you should know
what it means to repeat.

112
00:09:00,936 --> 00:09:02,630
What should you repeat.

113
00:09:03,384 --> 00:09:06,820
To our understanding,
there are four types of shadowing.

114
00:09:07,400 --> 00:09:13,384
Litteratim shadowing, verbatim shadowing,
sensatim shadowing and signatim shadowing.

115
00:09:14,040 --> 00:09:19,660
The difference lies
in the degree to which you should repeat
the source text,

116
00:09:19,704 --> 00:09:22,168
that is the speech
you are listening to.

117
00:09:24,750 --> 00:09:30,760
Litteratim shadowing means you repeat
every sound you hear of the speaker.

118
00:09:31,520 --> 00:09:34,584
Every speaker,
even the most trained one,

119
00:09:34,744 --> 00:09:40,030
makes use of given and recurrent
features of orality when speaking.

120
00:09:40,890 --> 00:09:45,440
This means that you should repeat
not just words,

121
00:09:45,540 --> 00:09:47,420
but any of them,

122
00:09:47,620 --> 00:09:52,210
even words
that you don’t understand the meaning of,

123
00:09:52,310 --> 00:09:55,480
as well as recurrent examples of orality.

124
00:09:56,230 --> 00:09:58,952
These may be fillers

125
00:09:59,112 --> 00:10:02,900
like "well", "you know",
"kind of", "you see", etc.

126
00:10:03,560 --> 00:10:08,870
Extra sounds like "uhm", "er", "mh", etc.

127
00:10:09,470 --> 00:10:14,020
Interjections
like "Uh!", "Ah!", "Ow!", "Ack!" etc.

128
00:10:17,190 --> 00:10:19,400
Barbarisms, or words

129
00:10:19,544 --> 00:10:24,940
that each of us pronounces
his or her own way,

130
00:10:25,224 --> 00:10:30,152
like "heyyy", "naaa", "wassuup", etc.

131
00:10:33,140 --> 00:10:35,048
Verbatim shadowing means

132
00:10:35,144 --> 00:10:40,140
that you should repeat every distinct,
meaningful element of the speech you hear

133
00:10:40,140 --> 00:10:44,040
as uttered by the speaker,
mistakes included.

134
00:10:44,456 --> 00:10:48,744
This means that you should repeat
every single lexical item,

135
00:10:48,920 --> 00:10:51,880
like "house", "government", "Spain", etc.

136
00:10:52,020 --> 00:10:57,192
Grammar items like "the", "when", "for", etc.

137
00:10:57,500 --> 00:11:02,824
Acronyms like "GDP", "NCRA", "OWL", etc.

138
00:11:02,910 --> 00:11:05,420
Foreign words and expressions,

139
00:11:05,520 --> 00:11:09,180
like "ad nauseam",
"bon voyage", "gnocchi", etc.

140
00:11:12,430 --> 00:11:16,690
Sensatim shadowing is a bit more complex
and less precise as a notion.

141
00:11:17,576 --> 00:11:23,288
It means you should repeat every single
meaning uttered by the speaker,

142
00:11:23,450 --> 00:11:28,210
a meaning being
any concept expressed as a word,

143
00:11:28,310 --> 00:11:30,380
a clause, or a sentence.

144
00:11:30,950 --> 00:11:34,776
You can either repeat these words
as in verbatim shadowing,

145
00:11:34,970 --> 00:11:39,920
or re-express them using other words,
normally shorter words.

146
00:11:40,420 --> 00:11:43,650
For example,
you can make use of synonyms,

147
00:11:43,750 --> 00:11:47,940
both horizontal synonyms,
like "aim" instead of "objective",

148
00:11:48,040 --> 00:11:52,070
and vertical synonyms,
like "flower" instead of "anemone".

149
00:11:52,890 --> 00:11:55,500
You can also reformulate the utterance.

150
00:11:55,760 --> 00:12:00,392
For example, you can say "yes"
instead of "I tend to agree",

151
00:12:01,010 --> 00:12:04,180
or "we" instead of "you and I".

152
00:12:06,120 --> 00:12:10,580
Finally, you can strategically omit
some elements in a sentence

153
00:12:10,680 --> 00:12:14,376
that not necessarily add something
to the overall meaning.

154
00:12:14,488 --> 00:12:16,290
For example, instead of saying:

155
00:12:16,360 --> 00:12:20,680
"Carlo talks a lot
so as to try and explain this notion",

156
00:12:21,110 --> 00:12:24,600
you can simply say:
"Carlo talks a lot to explain this".

157
00:12:27,260 --> 00:12:30,630
Finally,
signatim shadowing is the form of shadowing

158
00:12:30,744 --> 00:12:34,552
that is aimed at repeating
every single sign,

159
00:12:35,280 --> 00:12:38,792
meaning any meaningful occurrence
in a speech event

160
00:12:39,040 --> 00:12:42,568
that adds information to the ideal reader

161
00:12:42,770 --> 00:12:47,330
of the potential transcription
of signatim shadowing.

162
00:12:48,150 --> 00:12:52,936
This means not just the single sentences
that one hears like "Good morning",

163
00:12:53,400 --> 00:12:59,256
but also any other relevant element
such as punctuation marks,

164
00:12:59,880 --> 00:13:04,950
like comma, period, question mark etc.,
as in "good morning period".

165
00:13:06,270 --> 00:13:09,470
Signatim also means paraverbal events,

166
00:13:09,470 --> 00:13:12,970
like: the tone of voice,
intonation, volume etc.,

167
00:13:13,144 --> 00:13:17,820
as in "good morning period [he shouts]".

168
00:13:19,800 --> 00:13:23,100
Signatim finally means
non-verbal events,

169
00:13:23,100 --> 00:13:26,760
like applauses,
cheering, a bell ringing,

170
00:13:26,760 --> 00:13:31,350
as in "good morning period [applauses]".

171
00:13:36,970 --> 00:13:41,680
In this video lecture we have introduced
the main competence of the respeaker,

172
00:13:41,784 --> 00:13:47,370
the psycho-cognitive skill of listening
to the ST and simultaneously repeating it.

173
00:13:48,240 --> 00:13:51,784
In particular,
we have dealt with two important aspects:

174
00:13:51,940 --> 00:13:54,840
a theoretical one,
that is the Effort Model,

175
00:13:54,984 --> 00:13:57,560
and a more practical one, shadowing.

176
00:13:58,120 --> 00:14:02,056
The effort model is at the basis
of this cognitive process

177
00:14:02,150 --> 00:14:05,816
that involves doing more things
at the same time.

178
00:14:05,976 --> 00:14:10,856
We have seen that balancing the efforts
these simultaneous actions require

179
00:14:11,032 --> 00:14:13,910
is always needed to avoid breakdowns.

180
00:14:14,968 --> 00:14:21,432
One way to go for that is trying
to semi-automate these actions.

181
00:14:22,024 --> 00:14:26,232
To do so,
shadowing is the exercise to start with.

182
00:14:26,520 --> 00:14:30,456
It consists in listening
a speech and repeating it,

183
00:14:30,792 --> 00:14:34,184
either litteratim or sound for sound,

184
00:14:34,280 --> 00:14:37,160
verbatim or word for word,

185
00:14:37,288 --> 00:14:39,576
sensatim or meaning for meaning

186
00:14:39,670 --> 00:14:42,280
and signatim or sign for sign.

187
00:14:42,584 --> 00:14:47,544
In the homework session,
we will see how to do this in practice.