1
00:00:03,130 --> 00:00:08,770
[Auto-generated transcript. Edits may have been applied for clarity.]
Thank you very much, all of you, for coming. I'm very, very happy, very pleased to, uh, to be introducing, uh,

2
00:00:09,070 --> 00:00:13,690
our speaker tonight for the second annual, uh, Voltaire Foundation Digital Enlightenment Studies lecture.

3
00:00:14,050 --> 00:00:17,290
Uh, this is something we started last year. I can't remember who the speaker was.

4
00:00:17,410 --> 00:00:20,530
Um, but, uh, I think it went well. Yeah.

5
00:00:21,520 --> 00:00:28,260
And that we hope to, uh, to keep going in years to come. And then we've we've coupled this with, uh, with a workshop on digital enlightenment studies.

6
00:00:28,270 --> 00:00:30,399
It's been really rewarding over the past few days.

7
00:00:30,400 --> 00:00:36,700
So it's really there's some momentum going and and our tonight's speaker is uh, is a big reason for that.

8
00:00:36,720 --> 00:00:45,700
And so I'm happy and very pleased to, uh, to, uh, to, uh, announce Michael Dolan and professor of digital humanities at the University of Helsinki.

9
00:00:46,060 --> 00:00:48,940
Uh, Mikko is, uh, is a trained intellectual historian.

10
00:00:48,940 --> 00:00:57,069
And I've always been very, uh, very impressed by the way that he juggles the sort of academic rigour, uh, of, uh, of what is in many fields,

11
00:00:57,070 --> 00:01:04,809
in many places, a very analogue, uh, field which is, uh, intellectual history, uh, with very serious computational explorations.

12
00:01:04,810 --> 00:01:11,500
Uh, and so I think, uh, he runs, uh, the computational history, uh, group there and the University of Helsinki who does amazing experiments.

13
00:01:11,500 --> 00:01:15,430
And I think he's going to talk about a few of them, if not more, uh, tonight.

14
00:01:15,430 --> 00:01:20,200
So we're very pleased to have him. He's going to speak to us about, uh, books as objects, data and meaning.

15
00:01:20,210 --> 00:01:31,200
Uh, thank you. Mikko. Uh, please give him a warm welcome. All right.

16
00:01:31,200 --> 00:01:36,929
Um, thank you very much, Glenn. Uh, and thank you, Nicholas and everybody at the Voltaire Foundation.

17
00:01:36,930 --> 00:01:40,259
This is really a great honour for me, me to do this.

18
00:01:40,260 --> 00:01:49,710
And I really, uh, love to be part of what we kind of started or you started, and we try to support that as best as we can.

19
00:01:50,340 --> 00:01:53,579
Um, let me say a few words about the title.

20
00:01:53,580 --> 00:02:03,180
So books is objects date on meaning? Uh, it aims to capture the kind of idea that that when we think about books, uh, and in the digital realm,

21
00:02:03,540 --> 00:02:09,659
uh, we want to see them to function on several different kinds of levels, which sometimes is forgotten.

22
00:02:09,660 --> 00:02:15,780
And, uh, first, for us, uh, the way that we approach this is that they are physical objects, so,

23
00:02:16,170 --> 00:02:21,900
so tangible artefacts that are shaped by production and circulation in many different ways.

24
00:02:22,440 --> 00:02:28,440
And at the same time, uh, they are data, which is what the analogue part, uh, wasn't always thinking.

25
00:02:28,440 --> 00:02:34,889
Uh, so the, the carriers or there's metadata that we can, um, mould and modify and uh,

26
00:02:34,890 --> 00:02:39,030
there's textual information that can be analysed computationally.

27
00:02:39,720 --> 00:02:46,320
And they are, of course, uh, vehicles of meaning, uh, which is the one that intellectual history always has been interested in.

28
00:02:46,650 --> 00:02:51,000
But the approach, uh, has been everything but, uh, computational.

29
00:02:51,510 --> 00:03:01,110
So, uh, cultural, uh, and intellectual expressions, uh, readers interpret and then reinterpret, uh, across different contexts.

30
00:03:01,230 --> 00:03:07,230
Uh, and that's what the thing that we are interested in is, is really about.

31
00:03:07,770 --> 00:03:11,370
And then let me emphasise this a computational approach.

32
00:03:11,370 --> 00:03:16,050
So, so we don't want to have let's try everything.

33
00:03:16,140 --> 00:03:26,850
Uh, I'll be talking how we, we sort of, uh, have had a strategy and a vision for, for some time now, already before I became a thing.

34
00:03:27,240 --> 00:03:33,600
Uh, but so, so that that's something, uh, hopefully something that you get out of this talk.

35
00:03:34,410 --> 00:03:39,840
So what I want to do is to kind of present this holistic approach, that what we mean,

36
00:03:39,870 --> 00:03:44,730
uh, by studying books as physical objects, data and vehicles of meaning.

37
00:03:45,240 --> 00:03:53,070
Uh, then I tried to offer some concrete, uh, data, uh, driven insights into enlightenment, print culture.

38
00:03:53,220 --> 00:03:59,430
Uh, so if also for for those who are maybe, uh, coming from the sort of traditional side,

39
00:03:59,460 --> 00:04:03,930
uh, I'm really that's the most important part that we are talking.

40
00:04:04,050 --> 00:04:10,800
We are not trying to do something different. We are expanding the context, maybe, and the ways of interpretation.

41
00:04:10,800 --> 00:04:16,440
But the continuum needs to be in the tradition of enlightenment studies and intellectual history.

42
00:04:16,440 --> 00:04:25,110
So all of that who are not doing computational, uh, just as important as those who are, uh, experimenting with different kinds of ways.

43
00:04:25,770 --> 00:04:31,110
Uh, but the point is that, uh, the methods they need to advance the scholarship also,

44
00:04:31,110 --> 00:04:39,390
so that we go in some ways beyond the traditional, uh, otherwise, there's no point, uh, for, for us to do anything in this room.

45
00:04:40,050 --> 00:04:45,780
And then I'll try to, uh, sketch a little bit, uh, where computational history might be heading and, uh,

46
00:04:45,780 --> 00:04:52,920
how I will shape, uh, what we call the digital enlightenment studies so that those are the three objectives.

47
00:04:54,210 --> 00:05:02,220
Now, uh, in the idea of, of working, uh, and giving a talk lecture, uh,

48
00:05:02,250 --> 00:05:08,069
in an intellectual history, uh, and enlightenment studies, usually you would have a handout,

49
00:05:08,070 --> 00:05:13,950
right, printed handout where you get a set of quotes, which kind of underlines that,

50
00:05:13,980 --> 00:05:20,640
uh, the work that we do, uh, in traditional way is somewhat anecdotal.

51
00:05:20,760 --> 00:05:23,970
So we are interpreting certain passages that are important.

52
00:05:24,780 --> 00:05:30,180
What I want to give now, uh, through this QR code and this is now an experiment a little bit.

53
00:05:30,540 --> 00:05:38,880
So, so that if you go there, uh, you will find a structure, you will find, uh, articles that embed my talk.

54
00:05:39,570 --> 00:05:48,990
So there is this, uh, read number or code and you can look at which article I'm referring to every now and then.

55
00:05:49,470 --> 00:05:54,330
You don't have to follow this, but it gives you kind of like a sense of where are we going.

56
00:05:54,330 --> 00:05:58,860
So don't feel that you're missing out necessarily if you don't want to do that.

57
00:05:59,130 --> 00:06:02,760
This gives you an opportunity for those who will want to be multitasking.

58
00:06:03,120 --> 00:06:09,150
Uh, so so I get I don't get angry when if you're looking at your phone here, uh, so so so so you can do that.

59
00:06:09,840 --> 00:06:15,780
But, but and and then another one is that if you're interested in what we do, you can go back,

60
00:06:15,810 --> 00:06:21,210
uh, after this lecture, look at the structure, uh, and then go through those articles.

61
00:06:21,930 --> 00:06:25,049
So, so most of them are available openly.

62
00:06:25,050 --> 00:06:30,330
So, so that's also the point. And you can go you don't have to use the QR code you can go to.

63
00:06:30,390 --> 00:06:34,230
To tiny. You are welcome. Uh, D.s.

64
00:06:34,590 --> 00:06:38,120
And hyphen Dolan in, uh. And you you can go there.

65
00:06:38,190 --> 00:06:43,259
So so that's that's one of the points. Uh, is there anything unclear about this?

66
00:06:43,260 --> 00:06:46,830
Because this is, uh, so so if someone is still wondering that.

67
00:06:47,010 --> 00:06:50,040
What what's going on? Uh, but but anyway. All right.

68
00:06:51,770 --> 00:06:59,210
Now, uh, and this is the structure. The this is how the also those, uh, articles are structured, uh, that I will follow today.

69
00:06:59,720 --> 00:07:03,980
So we will talk about what I call what we call bibliographic data science.

70
00:07:04,990 --> 00:07:09,319
Then we'll talk. I have two different kinds of. Examples.

71
00:07:09,320 --> 00:07:14,210
There will be more snapshots because I one hour is not enough for me to open up everything.

72
00:07:14,510 --> 00:07:20,270
And that's why you have the articles also. But I want to talk about canon formation in early modern Britain,

73
00:07:20,660 --> 00:07:26,330
a little bit about vernacular ization and what we have called democratisation of reading.

74
00:07:26,390 --> 00:07:33,350
Um, and then I will talk about reception history, uh, in different kinds of ways, uh, for enlightenment studies.

75
00:07:33,770 --> 00:07:43,640
And then I turn to this, uh, new type of place where we are now after enlightenment studies and a lot of the articles that,

76
00:07:43,670 --> 00:07:49,760
uh, we are working on that I'll be talking about through, for example, we do something called translation mining.

77
00:07:50,060 --> 00:07:57,050
Uh, that's something that hasn't yet been published. Uh, but I it's it's so good that I dare there speak about it.

78
00:07:59,430 --> 00:08:06,050
Well, let's go back. Uh, this is also a little bit of a personal journey and, uh, journey, but our, uh, research group.

79
00:08:06,060 --> 00:08:10,680
So let's go back to 2001. Here I am, uh, as a student of intellectual history.

80
00:08:11,040 --> 00:08:19,140
Uh, asking what is intellectual history? If you remember, back in 2001, that was the golden age of of this contextual reading.

81
00:08:19,170 --> 00:08:25,350
Uh, so Skinner is very, very, uh, he's still big, but even bigger in Helsinki especially, uh,

82
00:08:25,350 --> 00:08:31,620
and, uh, so, so a lot of the how to think about intellectual histories is about context.

83
00:08:32,310 --> 00:08:39,120
And what I was started to think already then this is me not being context, uh, computational at all at this stage.

84
00:08:39,540 --> 00:08:47,800
Uh, but what I was thinking, is that, okay, if we really want to do this contextual, don't we want to then really expand the day?

85
00:08:47,820 --> 00:08:54,149
I mean, what we are working with. We had to come here to Oxford Bodleian or uh, just to look at the books.

86
00:08:54,150 --> 00:08:57,430
If we're working on, uh, 18th century, this is the back in the day.

87
00:08:57,450 --> 00:09:01,710
I mean, uh, that's how it was. Another thing that was important for me.

88
00:09:01,740 --> 00:09:03,630
Uh, you see Mandeville and Hume there.

89
00:09:03,690 --> 00:09:12,270
Uh, so, so I'm a simple person coming from near Lapland, where the enlightenment, uh, people thought that, uh, not so bright people live.

90
00:09:12,420 --> 00:09:18,989
So, so I, I, I had a very, uh, when I was reading Mandeville and when I was reading Hume, I got very I mean,

91
00:09:18,990 --> 00:09:25,470
like that people at the time were reading Hume in a way that that is everything but Monday religion.

92
00:09:25,590 --> 00:09:35,010
Uh, and so for that it was just didn't I mean, for me, this kind of reception type of question that how how is, uh, ideas changing?

93
00:09:35,040 --> 00:09:40,620
It really stuck with me for, for a long time, and I haven't got rid of it anyway yet either.

94
00:09:41,940 --> 00:09:45,900
Another thing was that where do you get really the data into early 2000?

95
00:09:46,020 --> 00:09:54,000
And, uh, and this was the time we got one in a, in a trial in Helsinki for two weeks, and I wasn't sure.

96
00:09:54,240 --> 00:10:04,200
Is it going to go away? Uh, so for the two weeks, I sat in the library and I was able to get CD-ROMs and then download things.

97
00:10:04,680 --> 00:10:09,870
There was a limit 50 pages at the time. So I sat through the whole opening time.

98
00:10:09,870 --> 00:10:13,410
I should have brought the CD is I still have a big bunch of them.

99
00:10:13,740 --> 00:10:20,160
Uh, so so that that I just thinking that that that I had this so I can be able to look at the PDFs.

100
00:10:20,670 --> 00:10:23,579
Uh, so so that was the idea. The storage.

101
00:10:23,580 --> 00:10:32,550
I calculated it would have taken about 25 years to to download them at the at the time with that, that modem uh, time, uh, but didn't work.

102
00:10:33,120 --> 00:10:37,560
And of course, I mean, the I mean, if we think about the interfaces, uh,

103
00:10:37,740 --> 00:10:43,490
18th century book, as we all know, uh, survives and if you have it in good conditions,

104
00:10:43,500 --> 00:10:49,500
the CD-ROMs they break and we don't know how the cloud computing and, uh,

105
00:10:49,800 --> 00:10:55,320
supercomputers and so forth will survive in the end, but but, uh, that that's that's where we are.

106
00:10:56,430 --> 00:11:03,629
So then my steps, uh, from being a traditional intellectual historian and then being involved with, uh,

107
00:11:03,630 --> 00:11:08,760
at the great bunch of people from many different backgrounds, uh, our computational history group.

108
00:11:09,270 --> 00:11:18,170
So I really, I mean, that that. When I was doing my PhD, I was getting these ideas that, why don't we do this computationally?

109
00:11:18,590 --> 00:11:25,790
And and different kinds of ideas. Then when I finished my PhD, I started looking for a twin collaboration.

110
00:11:25,790 --> 00:11:31,490
So I called one of my friends who was a professor in medicine, asked that, do you have anybody in your group who would like to work with me?

111
00:11:31,850 --> 00:11:39,110
I have these ideas. That's what we started doing. Then we got, uh, it was two of us at the beginning, and then we had some funded projects.

112
00:11:40,040 --> 00:11:47,030
Some people came in, and then from 2014 onwards, uh, came this phase where intentionally,

113
00:11:47,330 --> 00:11:52,070
uh, took the agency to build a research group, which in the humanities, it's not.

114
00:11:52,340 --> 00:11:56,390
There are people who do it, but it's still a little bit different kind of undertaking,

115
00:11:56,810 --> 00:12:01,460
uh, than working just in the funded projects that we were, we were doing also.

116
00:12:02,060 --> 00:12:05,900
So, so that that's kind of kind of where how things were going.

117
00:12:07,310 --> 00:12:15,110
Now for me, Voltaire Foundation. So my book on Mandeville and Hume, Anatomy of Civil Society, that was published in 2013.

118
00:12:15,650 --> 00:12:23,540
And that's the kind of from there ten years to this point where digital enlightenment studies and we had an article there also.

119
00:12:23,780 --> 00:12:28,669
So, so that this period of time here was very formidable in many ways.

120
00:12:28,670 --> 00:12:38,090
And I feel like that there's a little bit of a kind of kind of overlap with, uh, how, uh, Voltaire Foundation is also, uh, transferring and so forth.

121
00:12:38,420 --> 00:12:42,709
This period when I had finished, uh, this book and was very serious.

122
00:12:42,710 --> 00:12:50,300
Okay. Now I'm going to do I'm going to start this computation. Seriously, people, uh, advising me, my supervisors, that you have a decent career.

123
00:12:50,360 --> 00:12:53,360
Uh, don't do it. Stop. Stop now.

124
00:12:53,370 --> 00:12:54,440
Uh, so.

125
00:12:54,560 --> 00:13:02,560
And, uh, but and then they said that, uh, well, if you are successful, that then that means that we also need to start doing this computationally.

126
00:13:02,570 --> 00:13:09,440
We don't want to do it. Uh, so so that but but anyway, here we are in Helsinki, our group.

127
00:13:09,530 --> 00:13:14,990
Uh, so it's formed so that we have data scientists, uh, and what is and,

128
00:13:15,010 --> 00:13:20,660
and linguists who look at the world and concepts in a different way than intellectual historians.

129
00:13:21,290 --> 00:13:27,019
And then we have a lot of elements of book history, uh, people interested in semantic change,

130
00:13:27,020 --> 00:13:30,260
uh, different kinds of historical processes and so forth.

131
00:13:30,920 --> 00:13:35,600
The what is not always so easy is that to understand that, for example,

132
00:13:35,600 --> 00:13:39,620
if you have data scientists, these are not people who come to fix your computer.

133
00:13:40,310 --> 00:13:43,370
You know, they they don't they are serious scientists, uh,

134
00:13:43,370 --> 00:13:53,140
who need to be thought of in that way so that we are doing work where the identity for everyone is so that you can retain it.

135
00:13:53,150 --> 00:13:56,719
And, uh, we are not like bioinformatics quite yet, so.

136
00:13:56,720 --> 00:14:02,990
So that to think about this, uh, and not, uh, kind of, uh, take advantage of anyone.

137
00:14:03,170 --> 00:14:09,350
So that's, that's the important part. But at the same time, we need to advance together.

138
00:14:09,380 --> 00:14:15,200
Uh, and, and also think that where we're going to get our next funding and and so forth and so forth.

139
00:14:15,200 --> 00:14:23,270
So, so there's a lot of, uh, different kinds of elements and this, this sort of, uh, collaboration that need to be taken, taken really seriously.

140
00:14:24,890 --> 00:14:29,480
Now our aims. This was done. This slide is from year 2017.

141
00:14:29,540 --> 00:14:37,550
Uh, so that what we are trying to do and the reason why I want to show this is that that we've had quite, uh, consistent.

142
00:14:37,940 --> 00:14:46,850
So the strategy that we started already then so understanding public communication and then also looking at the early modern Europe more broadly,

143
00:14:47,270 --> 00:14:53,770
movement of ideas, so clearly saying that metadata work, uh, looking at uh,

144
00:14:53,780 --> 00:14:59,340
genres, looking at intellectual traditions, uh, ancient texts and how they, uh.

145
00:15:00,360 --> 00:15:07,470
Preserved, or how their reception and so forth goes in the early modern period.

146
00:15:07,500 --> 00:15:14,130
Uh, and then looking at uh, also texture use and this coupled with conceptual change.

147
00:15:14,340 --> 00:15:20,130
Very much so the intellectual history underpinnings there. We haven't done nearly enough.

148
00:15:20,280 --> 00:15:24,089
Theoretical work in this line as we would like to uh,

149
00:15:24,090 --> 00:15:29,520
and it's something that also should be encouraged more and more in the enlightenment studies in general.

150
00:15:30,060 --> 00:15:37,020
Uh, but then doing these data releases, uh, doing tools, building tools for others.

151
00:15:37,050 --> 00:15:42,510
Uh, so, so all of this kind of going together was, was what we aimed at.

152
00:15:44,440 --> 00:15:48,010
All right, let me start now. Uh, that from this background.

153
00:15:48,400 --> 00:15:55,180
Go to this, uh, conceptual, uh, and practical level of talking about bibliographic data science.

154
00:15:55,180 --> 00:15:59,950
And this is now the first part where you can look at some of the articles, uh, as well, if you like.

155
00:16:01,150 --> 00:16:12,100
So 2015, uh, it was very clear to us, uh, and we were very enthusiastic that we can map ideas in different kinds of ways,

156
00:16:12,100 --> 00:16:20,170
looking at coming up with this, this network that that is great and bibliographic data and treated properly.

157
00:16:20,200 --> 00:16:27,100
Uh, is something that where we can take these steps and, uh, really look how how, for example,

158
00:16:27,490 --> 00:16:35,889
natural law tradition in some, some sense is, is uh, being, uh, published, uh, across Europe and um,

159
00:16:35,890 --> 00:16:41,950
and take that and combine different elements and move forward and also thinking

160
00:16:41,950 --> 00:16:48,249
about what is the reception of of Mandeville or Hume and this type of elements,

161
00:16:48,250 --> 00:16:52,540
where it was really clear that that this is a good way of, of thinking about it.

162
00:16:54,210 --> 00:17:03,690
And conceptually, uh, we started or I started reading quite a bit about what in bibliography, uh, and in book history, people have been doing.

163
00:17:04,080 --> 00:17:11,050
And there is this debate about potential use of library catalogues, uh, and decades of debate about that.

164
00:17:11,070 --> 00:17:14,910
Of course, the debate about bibliography and science that, uh,

165
00:17:14,910 --> 00:17:23,520
Tencel and Mackenzie and others were engaging it with was a little bit different than what we are now thinking when we are talking about data science.

166
00:17:23,880 --> 00:17:30,690
And then the questions that, in what way are we sure that you can actually use library catalogues to receive?

167
00:17:30,720 --> 00:17:39,660
I mean, to get, uh, decent interpretations of, of, of, uh, how ideas are formed and how they change and,

168
00:17:39,930 --> 00:17:46,290
and what can you do also with respect to book history, only relying on library catalogues?

169
00:17:46,680 --> 00:17:53,700
There's a lot of book historians who think that, uh, the archival, uh, evidence is the only way to, to actually move forward.

170
00:17:54,030 --> 00:18:02,939
We took the other way and, uh, so that you can still debate, uh, and these haven't been resolved in, in any sense.

171
00:18:02,940 --> 00:18:06,179
And it's obvious that, that you need different kinds of approaches.

172
00:18:06,180 --> 00:18:14,880
Of course. Uh, but our argument was that if we take this what we called bibliographic data science approach,

173
00:18:15,210 --> 00:18:19,980
you can actually use in a scientific way, uh, these catalogues.

174
00:18:21,620 --> 00:18:26,630
So for us, uh, the early on, uh, little update from University of Turkey.

175
00:18:26,870 --> 00:18:30,560
It was him. Him and me, we. I worked plenty with him early on.

176
00:18:30,620 --> 00:18:34,340
Uh, and this was kind of like a white paper for us.

177
00:18:35,000 --> 00:18:39,980
Bibliographic data science and the history of the book that was published eventually in 2018.

178
00:18:40,730 --> 00:18:48,500
Um, so if you want to read that, what what where the really the ideas that we had for for this kind of approach,

179
00:18:48,530 --> 00:18:56,960
uh, that is the article, uh, to look at, uh, much of it has to do with the aspects of harmonisation.

180
00:18:57,440 --> 00:19:07,340
So, uh, you if you just take any kind of metadata and think that now I'm going to, uh, get some results that are relevant.

181
00:19:07,700 --> 00:19:13,480
What happens often is that you end up playing with the data. So you are maybe just, um.

182
00:19:13,790 --> 00:19:23,359
Well, explorations can be good. Uh, but our argument was, and still is that that you need to harmonise, uh, data.

183
00:19:23,360 --> 00:19:31,460
There's many different ways of doing that, uh, in order to, uh, get the kind of results that are robust and, and what,

184
00:19:31,550 --> 00:19:40,490
uh, so the thesis, also the analogue traditional, uh, scholars, it mainly it's about the research questions, of course.

185
00:19:40,850 --> 00:19:48,320
So, so what you are asking, uh, is the relevant part, uh, and then what data can you use in order to, to answer those?

186
00:19:49,780 --> 00:20:01,300
In 2018. Uh, it was a little bit of a shift, uh, which was towards, uh, full data, uh, meaning that we have images and we have, uh, the text.

187
00:20:01,330 --> 00:20:07,479
So the echo for us, we came here at the centre, and the idea was very,

188
00:20:07,480 --> 00:20:12,730
very seriously that we are building this kind of knowledge infrastructure for enlightenment studies.

189
00:20:13,180 --> 00:20:18,280
And, and this is also the reason we are able to do, uh, different kinds of things.

190
00:20:19,190 --> 00:20:32,330
These days. Uh, because the combination of the SDC metadata, uh, that includes, for example, uh, having work level and additional level,

191
00:20:32,360 --> 00:20:39,410
uh, sorted, which when you want to do some kind of a text mining becomes a very, very crucial issue.

192
00:20:39,560 --> 00:20:48,320
So the combination between SDC and Agco and then combine of course, with other data sets like VCF, uh,

193
00:20:48,740 --> 00:20:57,860
if we want to look at the influence between French and, uh, English, uh, or British, uh, we need BNF, uh, we need garlic.

194
00:20:57,860 --> 00:21:01,490
Uh, we need information about the printers and publishers.

195
00:21:01,940 --> 00:21:06,560
Uh, and also what we really want is also the newspapers.

196
00:21:07,070 --> 00:21:15,800
So the idea that that if we want to study the public, uh, we need to also, why would we only look at books?

197
00:21:16,160 --> 00:21:24,620
Why don't we also look at the overlap between, uh, newspapers when we are talking about public discourse in the 18th century?

198
00:21:24,620 --> 00:21:31,370
And this was something that not that many people are actually yet even doing that cross catalogue thing.

199
00:21:31,400 --> 00:21:37,760
So, so for us to build from here onwards that we started really seriously in 2018,

200
00:21:38,120 --> 00:21:43,160
moving from the library catalogues to this kind of a knowledge infrastructure.

201
00:21:43,190 --> 00:21:50,750
Uh, that's also something that, uh, has been very, very beneficial for us, uh, throughout and will continue to be so.

202
00:21:52,110 --> 00:22:02,159
Now, one thing that, uh, bibliographic data science combined to this type of work means is in this one article, uh, a two, uh,

203
00:22:02,160 --> 00:22:13,830
anatomy of 18th century collections online, uh, where the idea was that we analyse, uh, what is in echo just based on the STC, which is the best.

204
00:22:14,310 --> 00:22:21,000
Uh, it's not it isn't perfect, but it's the best, uh, information that we have available.

205
00:22:21,360 --> 00:22:27,180
A lot of people want to use the use echo, but, uh, if you don't really know what is in there,

206
00:22:27,660 --> 00:22:31,140
uh, might be a little bit difficult to justify your results.

207
00:22:31,740 --> 00:22:43,680
So just one example. If you want to study pamphlets in an echo and there is this kind of drop, uh, of coverage compared to SDC.

208
00:22:44,040 --> 00:22:52,480
So going from 50, 60% to 30%, uh, I would at least want to be aware of this before I start that kind of work.

209
00:22:52,500 --> 00:22:56,430
And and this is what that, uh, combination enables.

210
00:22:56,850 --> 00:23:00,749
Also, the, uh, like these big gaps, some of them are known.

211
00:23:00,750 --> 00:23:07,380
So people know that there are gaps, but I wouldn't trust what what, for example, the glass company, because they haven't done this.

212
00:23:07,740 --> 00:23:14,969
Uh, so they don't know. So and, and and I think that this kind of it's like preliminary work for,

213
00:23:14,970 --> 00:23:19,680
for any type of, uh, research you want to do using, using this kind of material.

214
00:23:19,980 --> 00:23:25,950
So for us, this is the bibliographic data science, what we want to mean, uh,

215
00:23:26,460 --> 00:23:32,040
when before we start engaging with, with certain, uh, analysis, uh, that we are doing.

216
00:23:33,640 --> 00:23:41,770
Now, another part of the bibliographic data science is the relationship between unstructured and structured data.

217
00:23:42,550 --> 00:23:51,850
So the structured data meaning, for example, metadata from the catalogues, this kind of classic bibliographic data approach.

218
00:23:52,120 --> 00:23:57,009
It's great because it's rather systematic. And then you have unstructured data.

219
00:23:57,010 --> 00:24:04,480
Could be text image whatever, audio collections, anything uh, that can also be analysed in a systematic way.

220
00:24:05,260 --> 00:24:09,900
But it's not I mean, and I don't understand why people don't think about it, that okay,

221
00:24:09,910 --> 00:24:15,100
it might be good to do some analysis, but why would you not think about it at the same time?

222
00:24:15,100 --> 00:24:26,380
Creation of new structured data. So there's a lot of, uh, mark fields, for example, that are missing, uh, from the catalogue information.

223
00:24:26,830 --> 00:24:36,760
But if we can use the unstructured data, for example, to create, uh, genre information or, you know, uh, subject topics, whatever you want to call it,

224
00:24:37,240 --> 00:24:47,410
uh, that kind of a way of working together, uh, in a shared infrastructure, if possible, this then will, uh, make us go much further.

225
00:24:48,670 --> 00:24:58,480
Now, example of this is that, uh, we've started working more and more with the physical aspect and, uh, images, for example, from the echo.

226
00:24:59,320 --> 00:25:04,330
So already I was wondering one part that this comes I was doing it.

227
00:25:05,340 --> 00:25:11,520
By hand. Uh, so I was printing and cutting, uh, head pieces when I was doing my PhD.

228
00:25:12,080 --> 00:25:16,910
And then gluing them to the to the title.

229
00:25:16,920 --> 00:25:24,530
Uh, so sort of, uh, kind of looking at four variants, uh, of, of the different ornaments and.

230
00:25:25,410 --> 00:25:31,500
Okay. Why wouldn't you do this in an automated way and, uh, well, because you didn't have the data.

231
00:25:31,500 --> 00:25:39,600
You couldn't. So, so that this is something, uh, so, for example, my when I was looking for Mandeville, being obsessed, as I always was, that's.

232
00:25:39,600 --> 00:25:48,360
How can people say that Thomson wouldn't be his publisher when all of the everything you look at in the book points exactly that direction.

233
00:25:48,360 --> 00:25:55,170
So and so to me that the idea that that what can we actually do with these, uh, is clear.

234
00:25:55,260 --> 00:26:02,190
There were has, of course, been people who have been looking at printers ornaments, uh, for a long time, usually doing by hand.

235
00:26:02,640 --> 00:26:10,650
But there the idea has been that the, that the woodblock is owned by the, by the printer, and there might be some lending going on.

236
00:26:11,190 --> 00:26:18,930
Uh, but the interest for collecting the information was always just to look for that one tied to Boya or somebody,

237
00:26:19,380 --> 00:26:24,240
uh, famous, famous, uh, printer and, uh, not looking at the variants.

238
00:26:24,510 --> 00:26:29,280
So what we've done, uh, there's also, of course, other people who work with ornaments.

239
00:26:29,560 --> 00:26:38,520
Not saying that, but but there's been discussion whether the echo quality, which is quite poor, uh, is good enough for doing this and well,

240
00:26:38,520 --> 00:26:48,780
we been able to take the 700,000 head pieces from echo and then cluster them in this kind of super and subclasses where we get these variants and,

241
00:26:48,780 --> 00:26:57,209
and some of these are very interesting. So what is the intentionality of producing, uh, these where the eagle, uh,

242
00:26:57,210 --> 00:27:05,400
is looking inwards and some that are looking outward and they are printed in Ireland, the other ones and the other ones in London.

243
00:27:05,730 --> 00:27:13,020
And, and so there's all kinds of it becomes like a big puzzle piece that, that we can then start mapping in different types of ways.

244
00:27:13,680 --> 00:27:20,130
And uh, that combined to better printer, uh, metadata information, publisher information.

245
00:27:20,400 --> 00:27:27,270
There's also, of course, a lot of thought that you can do looking at the type, uh, looking at the fonts, uh, and, and so forth.

246
00:27:27,690 --> 00:27:33,509
But a great way of going here so that we are combining the unstructured and structured.

247
00:27:33,510 --> 00:27:39,510
So, so with the intention that, that we, we get, uh, better data sets and that the project doesn't end here.

248
00:27:39,990 --> 00:27:44,879
It somebody will when we are able to offer it to others, the the work continues.

249
00:27:44,880 --> 00:27:48,980
Uh, and, and we move forward. Also.

250
00:27:48,980 --> 00:27:54,500
Uh, so. So then you can really I mean, that, uh, if you look at these two, two books here,

251
00:27:54,890 --> 00:27:59,870
uh, one should be said set to be published by Roberts and the other one by Thomson.

252
00:28:00,230 --> 00:28:11,510
And that's the one of his, uh, later fable. Uh, and you look at the details, uh, you may start thinking also about the, the author author attribution.

253
00:28:11,530 --> 00:28:17,120
And in this one. Oh, sorry. Uh, the publisher, uh, attribution in this kind of cases.

254
00:28:17,930 --> 00:28:22,399
And, uh, so just an observation. Uh, so they are quite intriguing patterns.

255
00:28:22,400 --> 00:28:26,840
So, so some people say that. Okay, Thomson was a publisher, uh, the younger.

256
00:28:26,840 --> 00:28:32,000
So he didn't have in-house printing. And it was this what, who did all the or the printing?

257
00:28:32,210 --> 00:28:37,550
Uh, but when he dies, uh, some of the uses of the head pieces drops.

258
00:28:37,910 --> 00:28:42,500
Uh, they're not used anymore. So if there were watches, I don't know why that that happened.

259
00:28:42,500 --> 00:28:46,580
And so. So there's a lot of this kind of kind of quite an intriguing.

260
00:28:47,680 --> 00:28:54,940
Questions. We are now working. Uh. I'm not there's no, uh, final, uh, thing that we have said about Thompson,

261
00:28:54,940 --> 00:28:59,830
but but that's something that we are working towards with this head piece of information.

262
00:29:01,330 --> 00:29:10,390
So with bibliographic, uh, data science, what I want to say is that this virtuous cycle of better data, which is a never ending process.

263
00:29:10,840 --> 00:29:18,520
We're talking about better data. So you combine the harmonised bibliographic data, text and image sources, and you come out with, uh,

264
00:29:18,520 --> 00:29:25,480
something that supplements, uh, and helps you use the images then to enrich the bibliographic data.

265
00:29:25,960 --> 00:29:33,340
Uh, there's a loop that move forward. Uh, and then you can also use the text information in many, many different kinds of ways.

266
00:29:33,790 --> 00:29:39,490
And what is really important is that you develop this kind of wit together with the use cases.

267
00:29:39,610 --> 00:29:46,560
So thinking about what kind of research question where in what sense is the data good enough?

268
00:29:46,570 --> 00:29:48,310
Because it never will never be perfect.

269
00:29:48,730 --> 00:29:54,879
We would never be doing an intellectual history if we are just there to let's get this dataset to a perfect condition,

270
00:29:54,880 --> 00:29:58,900
and then you end up annotating everything by hand and so forth.

271
00:29:59,290 --> 00:30:05,360
So good enough and usable and but then you need to be also able to evaluate that.

272
00:30:05,470 --> 00:30:08,530
What does that mean in in different kinds of context.

273
00:30:09,370 --> 00:30:16,000
But this is very much the kind of core of why I want to talk about bibliographic data science.

274
00:30:18,280 --> 00:30:19,899
And okay, I'll make this.

275
00:30:19,900 --> 00:30:28,630
So I publicly say that we are going to finish the cleaning of the our version of the STC and make that available as open data.

276
00:30:29,480 --> 00:30:35,200
We corrected it from 25 to 26. So I give our give us a little break here here.

277
00:30:35,200 --> 00:30:40,149
But but but anyway so so we think that this is uh, a bit of a milestone.

278
00:30:40,150 --> 00:30:46,900
So then after that anyone can can work with it. It will include for example, addition work relationship.

279
00:30:46,930 --> 00:30:52,870
Uh, that will help you a lot if you're working, for example, with the echo, echo data and so forth.

280
00:30:53,260 --> 00:31:00,070
Uh, so hopefully uh, that will also, uh, help, uh, collaborative projects when people work with,

281
00:31:00,520 --> 00:31:06,250
uh, text mining and for example, uh, lamps and, and and so forth.

282
00:31:07,100 --> 00:31:10,880
If we are not done by 26, it will be no fault.

283
00:31:12,140 --> 00:31:22,430
Okay. Uh, then a couple of snapshots. Uh, about what can you do when, when when we are in a situation that that is your data is kind of in a,

284
00:31:22,430 --> 00:31:25,490
in a sense, uh, good enough that we can start working with it.

285
00:31:26,000 --> 00:31:29,870
Can information is something something that we've been very, very interested in,

286
00:31:30,380 --> 00:31:36,110
uh, in, in many different, uh, subprojects or however you want to call them.

287
00:31:36,770 --> 00:31:43,010
Uh, this is one, one project, uh, that we worked with, uh, Mark Hill, who's here?

288
00:31:43,220 --> 00:31:49,670
He was, uh, postdoc, uh, at the time. So we wanted to examine the early morning cannon with the idea that.

289
00:31:49,670 --> 00:31:55,520
What is a cannon here? We just took it so that what is published most often, most frequently.

290
00:31:55,700 --> 00:32:04,820
And for the longest period of time. So. So, uh, means that cookbooks and Shakespeare go on level, uh, as a as a starting point.

291
00:32:05,390 --> 00:32:11,480
And then, uh, we use the addition field information, the work edition, uh, that was really crucial for this.

292
00:32:11,930 --> 00:32:14,930
Uh, and then, uh, publisher and printer information.

293
00:32:15,470 --> 00:32:22,490
And the main interest was really, uh, kind of if we can see an epistemological shift in this kind of kind of information.

294
00:32:22,500 --> 00:32:31,459
Uh, there is an article, uh, before, uh, so, so if you if you want to read it, I'm just going to show a couple very simple,

295
00:32:31,460 --> 00:32:37,280
uh, slides that, that what, what you can, you can do when you are examining this kind of data.

296
00:32:37,290 --> 00:32:44,900
So, uh, here, uh, okay. On the left here, here's, uh, the full, uh, STC most published.

297
00:32:45,350 --> 00:32:51,380
Uh, then if we look at, uh, the new, I mean, first editions, uh, it looks like this.

298
00:32:51,770 --> 00:32:55,550
And then our, uh, uh, cannon.

299
00:32:55,580 --> 00:33:01,910
Cannon data is, is is here, and it kind of breaks down quite nicely in the end.

300
00:33:01,940 --> 00:33:07,820
Uh, when when they are sorted by the first year of publication, this is the number of times they were published.

301
00:33:08,210 --> 00:33:16,730
There's a little bit of way, way in going on, uh, so that later, if you, there's no, uh, data ends at 1800.

302
00:33:16,730 --> 00:33:21,500
So, so that these, these, uh, works can be also included.

303
00:33:21,500 --> 00:33:24,809
And it's not only only earlier. Once. Now.

304
00:33:24,810 --> 00:33:28,110
Uh, so then you can, of course, upset in different ways.

305
00:33:28,650 --> 00:33:35,490
You can start looking at most frequently printed at least during one decade between 1500 and 1800.

306
00:33:35,760 --> 00:33:41,069
So if you look at, uh, what you find, uh, in this way, uh, the works,

307
00:33:41,070 --> 00:33:47,310
there's a lot of things that you would expect to find, maybe some that that wouldn't be so common.

308
00:33:47,730 --> 00:33:56,639
Uh, but that's not the point here. The point is that we we can do this kind of examination, uh, and set it up, uh, when you have a certain interest.

309
00:33:56,640 --> 00:34:08,040
And ours was that that if, uh, we are taking this, this type of, uh, uh, canon, uh, which is justified also by the by the reason that, uh,

310
00:34:08,850 --> 00:34:16,440
the you would the so-called London average of print runs, uh, works and, and that's,

311
00:34:16,440 --> 00:34:21,780
that's if you want to have more books printed, you will get out with the new edition.

312
00:34:22,110 --> 00:34:27,750
So, so that's also a reason why why this this is justified to be doing.

313
00:34:29,270 --> 00:34:34,370
Also, of course. Authors. Uh, publishing after death is important.

314
00:34:34,430 --> 00:34:41,420
Uh, usually, uh, the normal thing is for the non-canonical ones that once you die, they don't print you anymore.

315
00:34:41,450 --> 00:34:47,779
Uh, for whatever reason. Uh, so here also, you find very Voltaire's there, of course.

316
00:34:47,780 --> 00:34:54,320
Uh, and, uh, these are now. Now, remember the the British SDC, uh, data, uh, used here.

317
00:34:55,310 --> 00:34:58,910
So, uh, this was when 21, uh, we were doing.

318
00:34:59,570 --> 00:35:02,690
But then our interest in canon is constant.

319
00:35:02,700 --> 00:35:07,460
Uh, so there's this many more articles. I just want to show one recent thing that we did.

320
00:35:08,300 --> 00:35:13,720
So we took, uh. Um. Well, the article's, uh, title tells you a lot.

321
00:35:13,870 --> 00:35:18,010
Quantifying the presence of ancient Greek and Latin classics in early modern Britain.

322
00:35:18,760 --> 00:35:23,140
And, uh, we won when we were exploring the data.

323
00:35:23,920 --> 00:35:27,790
And it was a lot of work to get the cannon engine cannon right.

324
00:35:27,820 --> 00:35:29,680
Uh, so so that was a lot of work.

325
00:35:30,460 --> 00:35:40,360
But, uh, what we noticed in a data driven way, and what the article is very much about is that one while there was a as a whole,

326
00:35:40,360 --> 00:35:43,900
there's a there's a growing popularity of ancient printing.

327
00:35:44,650 --> 00:35:48,250
But what happens at the same time is the diversity.

328
00:35:48,280 --> 00:35:50,110
Diversity diminishes.

329
00:35:50,380 --> 00:35:59,740
So so the ones that they want, there's many more authors that are printed in the 17th century, but in the 18th century it becomes very concentrated.

330
00:35:59,740 --> 00:36:04,390
So so the so-called martell's effect becomes very, uh, prominent.

331
00:36:04,780 --> 00:36:09,940
So the ones that are printed, most, uh, printed and the other ones, uh, kind of dropped.

332
00:36:10,240 --> 00:36:18,280
So, so we did a lot of, uh, different kinds of, uh, uh, testing, uh, whether statistical testing, whether whether that is right.

333
00:36:18,280 --> 00:36:21,750
If you're interested in that, please, please look at the, uh, article.

334
00:36:21,910 --> 00:36:31,410
Uh. Be good. So then, uh, Verna Colourisation and what we call democratisation of reading.

335
00:36:31,420 --> 00:36:40,320
Uh, these are also now just snapshots. So, uh, this work was done long ago, but I still find it quite interesting.

336
00:36:40,770 --> 00:36:45,360
So if we look at book printing in Latin vernacular in northern Europe.

337
00:36:45,370 --> 00:36:51,300
So there's now many more, uh, data base, uh, for, for this than just the British case.

338
00:36:51,900 --> 00:37:03,630
Uh, what do you notice quickly, for example, from this heritage of the printed book database, is that, uh, book size sizes, uh, shift considerably.

339
00:37:04,050 --> 00:37:09,180
So we go towards much more mobile. So October is the start of the enlightenment.

340
00:37:09,180 --> 00:37:14,400
Uh, not Voltaire. Uh, so so that that, uh, becomes quite clear.

341
00:37:14,550 --> 00:37:19,940
Uh, at the same time, the Latin share that that is dropping, but it's so,

342
00:37:19,940 --> 00:37:28,500
so one thing to note about the Latin share is that, uh, it's not a synchronised move.

343
00:37:28,560 --> 00:37:34,740
Uh, so very looking at the local local differences, different cities have different profiles.

344
00:37:35,220 --> 00:37:39,540
Uh, so we can't say that there was a uniform move, uh, or something like that.

345
00:37:40,020 --> 00:37:44,040
Yet from here, uh, you could start. I mean, not not saying.

346
00:37:44,040 --> 00:37:46,169
I mean, the the Latin, uh, part.

347
00:37:46,170 --> 00:37:54,510
We looked quite carefully, uh, the different profiles, uh, but from the, the document size, uh, what do you want to make out of that?

348
00:37:55,230 --> 00:37:59,219
Um, that's another other question. And how how red or thick?

349
00:37:59,220 --> 00:38:03,150
You want to see the connection? That's that's another issue.

350
00:38:03,810 --> 00:38:12,000
But, uh, especially the drop of the Latin share, uh, might suggest that the reader community is at least getting wider.

351
00:38:12,060 --> 00:38:18,690
Uh, so also, uh, those who didn't go to college, uh, are starting to read, uh, and which is,

352
00:38:18,690 --> 00:38:26,240
uh, important part and part of, uh, this movement and then combined to this, this is a,

353
00:38:26,300 --> 00:38:35,160
uh, article led by, uh, doctoral researcher Eero de Honan, uh, who's very interested about the kind of like,

354
00:38:35,160 --> 00:38:39,180
uh, looking at book history from the economic perspective as well.

355
00:38:39,840 --> 00:38:50,820
So, uh, probably everybody knows this, uh, idea that there was a monopoly in London and the book prices were kept artificially high.

356
00:38:51,420 --> 00:39:02,070
Uh, and then when a lot of people have been interested about this copyright ruling, uh, that that there is this public domain that comes,

357
00:39:02,100 --> 00:39:11,010
uh, and then that the, the, the prices would have gone down considerably when we looked at this from the perspective of price constraint.

358
00:39:11,700 --> 00:39:17,550
Uh, the obvious conclusion was that that didn't have such an effect.

359
00:39:18,060 --> 00:39:27,870
So the, the reading still was, uh, mainly for the, for the, well, I mean, well, not reading, but but book ownership was, was for the wealthy.

360
00:39:27,960 --> 00:39:38,340
Uh, and then more emphasis could and should be put on this kind of different kinds of circulation, uh, resale or lending practices and so forth.

361
00:39:38,940 --> 00:39:47,460
But, uh, a good way, in my opinion, to show how, uh, this bibliographic data, uh,

362
00:39:48,270 --> 00:39:55,810
science perspective can bring a new aspect to something that has been debated over decades.

363
00:39:55,940 --> 00:40:01,080
The copyright question is, uh, is a huge issue in, in, in book history.

364
00:40:01,500 --> 00:40:06,090
And, uh. This is about bringing a new perspective.

365
00:40:06,390 --> 00:40:11,760
There was, uh, we calculated, uh, probability for the whole SDC for the book prices.

366
00:40:11,760 --> 00:40:14,880
And there was, uh, quite a lot of work gone into that.

367
00:40:15,330 --> 00:40:20,370
But if you're interested, please look at that, too. Now reception history.

368
00:40:21,300 --> 00:40:31,680
Which is also something that, uh, in our workshop now, over the two days, uh, we've talked quite a bit about texture use, uh, and such things.

369
00:40:32,250 --> 00:40:36,690
This is also something that we in Helsinki we've been working quite a bit with.

370
00:40:37,110 --> 00:40:42,689
Uh, so we have run this blast bioinformatics, uh,

371
00:40:42,690 --> 00:40:51,750
software and all of the TCP 18th century collections online and also then the British Library newspapers, Burnie and Nichols collection.

372
00:40:52,110 --> 00:40:56,400
And we have about a half a billion, uh, text reuse pairs.

373
00:40:57,120 --> 00:41:03,779
And, uh, what do you I mean, why that bibliographic data science for doing this is so important is that

374
00:41:03,780 --> 00:41:10,019
that we also need to have the SDC frame and the possibility of getting rid of,

375
00:41:10,020 --> 00:41:18,600
uh, multiple editions, uh, in certain cases if we want to really analyse that, that what is happening here.

376
00:41:19,200 --> 00:41:23,339
But, uh, so we doing different kinds of reception studies, uh,

377
00:41:23,340 --> 00:41:30,060
patterns of borrowing impact in some sense and, and also what we might call discourses through the,

378
00:41:30,420 --> 00:41:39,930
this kind of dissemination of texts and what again, that knowledge uh, in uh graph uh,

379
00:41:39,930 --> 00:41:48,360
that we want to see uh, so that we go beyond one single digital archive, we have plenty of them.

380
00:41:48,720 --> 00:41:58,800
And then we are looking to step from some kind of simple textual overlaps to mapping ideas so that we are we are able to classify and understand,

381
00:41:59,130 --> 00:42:07,740
uh, what the texture uses, for example, are or what uh, Glenn call today, uh, intellectually relevant, uh,

382
00:42:08,160 --> 00:42:17,850
texture uses that are more interesting than just mere Bible being, uh, quoted many times.

383
00:42:19,410 --> 00:42:26,670
So, uh, for us, uh, just to again, I want to go back to, uh, back in time, 2019 was still relevant.

384
00:42:26,700 --> 00:42:34,910
Uh, the ideas that we had then, uh, these were the basic questions that we came with, uh, when looking at the tax free use.

385
00:42:34,920 --> 00:42:45,450
So if we take Mandeville, uh, what kind of patterns of borrowing, uh, did he have or what kind of habits of reusing his own works did Mandeville have?

386
00:42:45,450 --> 00:42:51,000
Often with texture use, we tend to think that let's get rid of the author's own works.

387
00:42:51,480 --> 00:42:55,020
So because they are not interesting, in some cases they are very interesting.

388
00:42:55,470 --> 00:42:58,090
And then, of course, who quoted Mandeville and why?

389
00:42:58,110 --> 00:43:05,970
So there's quite a lot of, uh, if we are a little bit clever with like the simple data, we are able to do quite interesting things.

390
00:43:06,520 --> 00:43:13,799
And it's not always running for the next, uh, biggest model or something like that, but using what we have.

391
00:43:13,800 --> 00:43:20,790
And that's where, uh, humanities people could be much more effective and, uh, thoughtful in many sense.

392
00:43:22,080 --> 00:43:27,690
This is again, uh, the, uh, image from 2017.

393
00:43:28,200 --> 00:43:32,279
But I really like it because a lot of things came out because of this.

394
00:43:32,280 --> 00:43:36,030
So what we called anatomy of the impact.

395
00:43:36,030 --> 00:43:42,480
Mandeville. Fable of the beast in 18th century. So what we have here is the beginning of the book.

396
00:43:43,080 --> 00:43:50,310
And here's the end of the book. These, uh, count of how many times different chapters, uh, reused.

397
00:43:50,610 --> 00:43:54,750
And then the colour tells you, uh, when is the reuse happening?

398
00:43:55,380 --> 00:44:02,820
So what we notice here, uh, that the beginning of the book, when there is, uh, uh, Mandeville makes a lot of noise.

399
00:44:03,330 --> 00:44:14,340
So this part where, uh, private vices, public benefits is said, uh, it's used in 20s and 30s, of course, people complaining quite a bit about it.

400
00:44:14,820 --> 00:44:19,500
Uh, but it's something that, that then, uh, vanishes a little bit after that.

401
00:44:19,860 --> 00:44:21,570
And then there's an interesting shift.

402
00:44:21,720 --> 00:44:32,730
Uh, so the part, the chapter where he discusses luxury and pride, 1760s, there are many more times that this is this part is quoted.

403
00:44:33,300 --> 00:44:37,710
And then, of course, we can see what is not quoted at all. So Mandeville was a vegetarian.

404
00:44:38,010 --> 00:44:42,750
So he wrote a chapter called Why Man's Craving flesh for food is unnatural.

405
00:44:42,960 --> 00:44:51,490
Nobody is quoting it in the 18th century. But, but but uh, so this kind of, I mean, like that we can do this kind of anatomy of a book.

406
00:44:51,540 --> 00:44:57,959
I mean, of course, this is, uh, like, like a visual way of looking at how, uh,

407
00:44:57,960 --> 00:45:03,450
different parts of a book are reused, but it's so useful and especially when it's scalable.

408
00:45:03,720 --> 00:45:10,530
So you can do it, uh, for basically for any, any 18th century books, because we have the, The Echo collection.

409
00:45:10,830 --> 00:45:18,870
And this kind of shift especially to me is extremely interesting that that and and also that the

410
00:45:18,870 --> 00:45:27,509
question that why quantity uh sometimes is matters and we can see some kind of shift then of course,

411
00:45:27,510 --> 00:45:36,000
I mean, if you want to go uh, we developed based on, on mainly this idea here, uh, we developed this, uh, reception reader.

412
00:45:36,870 --> 00:45:43,170
Uh, so if you go w w w reception reader accom uh, you can use this interface.

413
00:45:43,500 --> 00:45:50,340
We are very grateful to Gail, uh, that they let us, uh, publish this openly so anybody can use it.

414
00:45:50,340 --> 00:45:54,540
You don't have to have a subscription. Uh, and it works so that.

415
00:45:54,570 --> 00:45:58,110
Okay. And here is the beginning of the book, and here's the end of the book.

416
00:45:58,470 --> 00:46:04,200
And there is a timeline, all of these, uh, dots here, uh, reuse cases.

417
00:46:04,650 --> 00:46:12,480
So you hover over one, uh, here, you see, when you are hovering over it's interactive, you see which book are we talking about?

418
00:46:12,870 --> 00:46:18,290
And then you click on the on the dot and it opens up the pages.

419
00:46:18,300 --> 00:46:22,770
So here on the left is the original. And here on the right is the blue.

420
00:46:22,770 --> 00:46:27,480
It's uh, uh, book that is quoting this, uh, part of wonderful.

421
00:46:27,900 --> 00:46:34,049
And, and it's highlighted here again, I mean, very easy, intuitive, uh,

422
00:46:34,050 --> 00:46:38,370
and meant for intellectual historians who don't want to do necessarily anything statistical.

423
00:46:38,700 --> 00:46:44,760
But this is what people used to do. I mean, that you are looking I mean that, uh, do I find the same things?

424
00:46:44,820 --> 00:46:48,780
Uh, and here you can have an interface that does it, does it very easily.

425
00:46:49,380 --> 00:46:55,800
So David Rossen, uh, from our group is the one, one the front end guys, so to say, uh,

426
00:46:55,800 --> 00:47:01,020
who, who very quickly, when he came with us, uh, we were able to, to put this together.

427
00:47:02,570 --> 00:47:08,380
Now, uh, I want to show a couple of things that you could do with text reviews that you maybe wouldn't think, uh, very ugly graph.

428
00:47:08,390 --> 00:47:16,700
But what you have in here, in this matrix, uh, is, uh, Manuel's, uh, all of his own books, uh, and, and textual overlaps between them.

429
00:47:17,420 --> 00:47:24,470
And so one thing that Mandeville is very well known for is he recycles passages from his own works,

430
00:47:24,860 --> 00:47:31,940
and then he cannot be without quoting Bayle or taking something from Bayle, not quoting, but but just taking material.

431
00:47:33,020 --> 00:47:38,200
This happens in almost all of the works except one, which is public use.

432
00:47:38,720 --> 00:47:47,780
Uh, if you look at the reason why public use is attributed to Mandeville, it becomes quite questionable.

433
00:47:47,780 --> 00:47:53,630
That is he actually the author? Uh, so we there's there's an article.

434
00:47:53,660 --> 00:47:59,600
Uh, that's in the digital enlightenment studies. Uh, we we wrote about Mandeville and where we talk about it.

435
00:48:00,290 --> 00:48:10,070
Uh, we are not saying that this is now concrete evidence that we have done everything to show, but this is evidence to at least think about it.

436
00:48:10,550 --> 00:48:14,070
That that. The way that oath.

437
00:48:14,190 --> 00:48:22,230
I mean, people frequently quote, if people want to talk about prostitution and then they, they quote, say that public statues is by Mandeville.

438
00:48:22,590 --> 00:48:28,040
Uh, but the reason why is this by Mandeville is that everybody thought that it is by Mandeville.

439
00:48:28,050 --> 00:48:30,000
So, so that the evidence is super slim.

440
00:48:30,210 --> 00:48:37,530
Also, if you look at the publisher, uh, information, uh, about it, um, there are reasons that you might want to doubt it.

441
00:48:39,410 --> 00:48:50,090
Also, uh, one thing. So. So we mean, the way that we did this addition work and then collection level was based on, uh, titles from the HTC first,

442
00:48:50,870 --> 00:48:58,670
but then when we have this text reuse information, what we did, we computed a coverage metric from echo and bought TCP.

443
00:48:59,210 --> 00:49:08,060
And then it's a directed one. So which way is the I mean now we are talking about works of the same, same same book.

444
00:49:08,210 --> 00:49:11,720
Uh, or might be a collection of, of something else.

445
00:49:12,410 --> 00:49:17,120
But the question that how much overlap there is between the same, uh,

446
00:49:17,120 --> 00:49:22,970
additions we are involved in Hume's history of England, uh, for Clarendon edition.

447
00:49:23,180 --> 00:49:25,639
So we need this kind of, uh, overlap.

448
00:49:25,640 --> 00:49:33,469
Uh, question for, for many, many different reasons, but what do you get here is a possibility based on the text we use.

449
00:49:33,470 --> 00:49:47,000
So if we have a work and edition one in addition to have, there's 90% of material that is the same going this way and 75% going that way.

450
00:49:47,330 --> 00:49:54,740
Okay. You can make, uh, different kinds of, uh, assumptions from there, but they are still probably the same addition.

451
00:49:55,370 --> 00:50:06,500
But when you have a situation where 90% of the text goes to this edition here, but only 30%, uh, is is kind of covered like this,

452
00:50:06,830 --> 00:50:11,390
there's a good reason to start thinking that maybe we're talking about collection level instead of a work level,

453
00:50:12,110 --> 00:50:17,540
but for many, many different kinds of reasons. Uh, extremely important.

454
00:50:17,640 --> 00:50:22,970
And one good way of thinking that how do we want to use, uh, this, this type of, uh,

455
00:50:23,480 --> 00:50:29,240
text reuse material that might be seemed trivial to someone, but it most definitely is not.

456
00:50:31,500 --> 00:50:37,800
So. Yeah. So just to string matching, uh, the newspapers as an intellectual platform.

457
00:50:38,130 --> 00:50:41,460
Uh, that's something that I would really encourage people to think about.

458
00:50:41,850 --> 00:50:46,590
Uh, and also this kind of way of using this for critical editions.

459
00:50:46,620 --> 00:50:55,620
Uh, we are building, uh, scholarly edition tool that we're and we talked about that, that today, uh, where, where this is,

460
00:50:55,620 --> 00:51:03,510
uh, coming to use and also we want to make this available for others so that, that, uh, making, uh, kind of.

461
00:51:05,210 --> 00:51:09,410
Enlightenment studies, uh, move forward in this sense as well.

462
00:51:11,220 --> 00:51:16,470
Okay, I for Enlightenment studies. Uh, this is my last, last part.

463
00:51:16,550 --> 00:51:25,470
Um, so this now, uh, for me is is a quite, uh, important, uh, slide.

464
00:51:25,620 --> 00:51:34,440
Uh, so we're trying to conceptualise, uh, between text reuse and something we call semantic similarity.

465
00:51:35,460 --> 00:51:40,080
And this has much to do with the what we are able to do with, uh.

466
00:51:41,550 --> 00:51:45,780
Embeddings. Uh, so so also if you want to say that.

467
00:51:45,810 --> 00:51:48,870
Okay, what can we do with AI and large language models.

468
00:51:49,260 --> 00:51:59,820
This has much to do with it. So if we think about parallel text uh, and uh, what is kind of overlapping or has to uh.

469
00:52:01,030 --> 00:52:06,519
What connects them in a way. So one way is to think about this kind of lexical text reuse.

470
00:52:06,520 --> 00:52:10,180
Right. That, uh, includes artefacts.

471
00:52:10,390 --> 00:52:18,360
Uh, so there's a lot of, uh, if you think about newspapers and books, uh, the, uh, advertisements for books.

472
00:52:18,370 --> 00:52:24,310
So basically, uh, imprints, uh, the one of the most repeat repeat that elements.

473
00:52:24,730 --> 00:52:28,030
There's also OCR errors. You could study those if you'd like.

474
00:52:28,540 --> 00:52:33,940
Uh, then the the reprints. Uh, that's that's an interesting question for us at least.

475
00:52:34,480 --> 00:52:38,650
Uh, you have quotations, uh, near verbatim, uh, verbatim.

476
00:52:38,650 --> 00:52:44,650
And then you have also extensive, uh, reuses and secondary quotes and that kind of things.

477
00:52:45,280 --> 00:52:53,200
But then this continuum, uh, here, uh, when you are thinking about semantic similarity, this is extremely important.

478
00:52:54,040 --> 00:52:58,540
Illusions. You can study them, but it's on. Right. Because we are not doing it at the moment.

479
00:52:59,110 --> 00:53:08,020
You can think about paraphrasing, uh, and also what is really important for us is translations full, partial, combined and so forth.

480
00:53:08,530 --> 00:53:11,350
And then also something that we call meaning matching.

481
00:53:11,620 --> 00:53:21,040
Uh, and this is, uh, really, really, uh, interesting, but at the same time maybe difficult so that,

482
00:53:21,040 --> 00:53:26,919
that when what is the meaning match and when is there like a intentionality between

483
00:53:26,920 --> 00:53:31,570
someone commenting and when are we just talking about topic or similarity?

484
00:53:32,350 --> 00:53:40,150
But this is, uh, something that it's for the US, the intellectual history and, and uh,

485
00:53:40,360 --> 00:53:48,700
humanists to sort out the methods are there, but we need to be able to define what, what we really want to do.

486
00:53:50,120 --> 00:53:57,139
So, uh. This is our sort of aim to, to make sense of,

487
00:53:57,140 --> 00:54:03,620
of what we are talking about when we are talking about tax free use and when we are talking about semantic similarity and meaning.

488
00:54:05,320 --> 00:54:10,870
So, uh, what we'd like to do. And there might be other ways also to think about it.

489
00:54:10,870 --> 00:54:16,180
So. So this kind of taxonomy of meaning or how do you want to do it?

490
00:54:16,480 --> 00:54:21,850
But the very important point is here when we are thinking about the meaning matches, uh,

491
00:54:22,450 --> 00:54:30,400
the multi uh lingual uh language models, uh, uh, transformer models, they are able to do it across languages.

492
00:54:30,910 --> 00:54:33,370
So the language barrier is not a problem anymore.

493
00:54:34,330 --> 00:54:40,120
And this is the what what probably going to change quite a quite a bit of research, uh, when we go towards the future.

494
00:54:41,730 --> 00:54:48,480
Here's an example of what is the lexical overlap. So quote there might be this kind of OCR noise as you see here.

495
00:54:49,170 --> 00:54:52,730
And and here is an example of a semantic similarity.

496
00:54:52,740 --> 00:55:00,690
So we have a small interface where you can type in uh about women humans rights and equal to men.

497
00:55:00,990 --> 00:55:10,200
And then you can hit uh, the based on what it classifies as, uh, something that is closest to that, that that's an example of, of.

498
00:55:11,160 --> 00:55:22,110
How how how that that works. Now, maybe the most interesting thing about from this where we are now, uh, concerns this translation mining.

499
00:55:22,800 --> 00:55:33,450
So we created embeddings. Uh, so we turned the 200,000 books from echo into this, uh, embeddings, uh, that are.

500
00:55:34,290 --> 00:55:37,530
So every single book is chunked.

501
00:55:37,650 --> 00:55:41,610
Uh, so, so that they have, uh, have their own own, um.

502
00:55:43,360 --> 00:55:50,350
Uh, vector. And, and then we turned to 300,000 books that we get from garlic into similar ones.

503
00:55:50,980 --> 00:55:55,090
And then they are mapped. Uh, and, and there are different kinds of filtering steps.

504
00:55:55,420 --> 00:56:01,660
Maybe I don't talk about that, uh, now, but, uh, what we get is this type of, uh,

505
00:56:02,590 --> 00:56:09,250
as an output, uh, where you have the English book here, the beginning and the end.

506
00:56:10,030 --> 00:56:13,570
And here's the French book, uh, beginning and the end.

507
00:56:13,990 --> 00:56:18,280
Each single one book compared to each other. And then different kinds of filtering.

508
00:56:18,280 --> 00:56:26,979
What is relevant, what is not. Uh, and you can start there are these diagonals that we are looking for then, uh, to recognise that.

509
00:56:26,980 --> 00:56:30,040
What is the translation? Here's a very interesting case.

510
00:56:30,600 --> 00:56:34,390
We have a book that has a British introduction. No translation.

511
00:56:34,900 --> 00:56:38,260
Then you have a part that is translated this here in red.

512
00:56:38,590 --> 00:56:42,370
Then you have some original text the British, uh, have added here text.

513
00:56:42,830 --> 00:56:48,729
Then you have another translation part, and then you have another, uh, from a third, another book.

514
00:56:48,730 --> 00:56:51,430
There's more translation here at the end.

515
00:56:52,270 --> 00:57:00,280
Uh, so you can do I mean, all we have done this to every single one book that, that there are in these collections.

516
00:57:00,820 --> 00:57:03,040
And if you think about it, I mean that that.

517
00:57:04,940 --> 00:57:13,100
Cases that that you could start developing from there, uh, different kinds of intentionality in translation and also thinking about that.

518
00:57:13,130 --> 00:57:20,930
Uh, we are not interested only in the full translation, but, uh, different kinds of ways of, of of, of doing the translation.

519
00:57:21,350 --> 00:57:23,000
That's, that's quite interesting.

520
00:57:23,540 --> 00:57:32,440
And of course, I mean, look here, uh, you the because there are these green that go to lower our partner, uh, Philip Ginter from Turku.

521
00:57:32,750 --> 00:57:37,040
He is very angry because he wanted all of the nicely on green here on the top.

522
00:57:37,370 --> 00:57:40,160
Uh, so, so that that we could say that it works perfectly.

523
00:57:40,580 --> 00:57:47,360
Uh, but, I mean, so, so that these different kinds of evaluation methods are just as important as doing the,

524
00:57:47,840 --> 00:57:52,190
uh, development of the methods that that take us towards this direction.

525
00:57:52,850 --> 00:57:56,030
And then also, of course, there's a lot of manual annotations.

526
00:57:56,240 --> 00:58:00,469
Uh, just check that, that what is actually the translation.

527
00:58:00,470 --> 00:58:07,460
So the between commentary on Bible obviously isn't a translation most of the times.

528
00:58:07,970 --> 00:58:12,050
Uh, but uh, just as something on the same topic.

529
00:58:12,440 --> 00:58:16,850
So, so these kind of things you need to sort out, it's a probabilistic approach.

530
00:58:17,180 --> 00:58:25,220
So we're not claiming that is 100%. But here if you see a clear semantic and intellectual similarity check, it's a translation.

531
00:58:26,220 --> 00:58:32,910
And then for us, kind of what is important is also developing a kind of taxonomy of translations.

532
00:58:33,510 --> 00:58:40,590
So, so, uh, going a little towards the Linnaean direction that we are not saying that these are set in stone,

533
00:58:40,950 --> 00:58:47,400
but we want to kind of organise knowledge and think about the world in a way that we we are doing different and we have the

534
00:58:48,180 --> 00:58:56,489
we are brave enough to to say that there are some kind of taxonomies that organise the knowledge in this humanities world.

535
00:58:56,490 --> 00:59:02,550
And even when, uh, this class and that class are individual and they can be studied,

536
00:59:03,210 --> 00:59:07,590
you know, as objects of their own, but this may be some kind of similarity between objects.

537
00:59:08,070 --> 00:59:11,640
So, so that's also something, something that we've been developing.

538
00:59:13,080 --> 00:59:16,090
And then we get to the meaning of meaning. So, Rousseau.

539
00:59:17,270 --> 00:59:21,860
Man is born free and everywhere. He's in chains as it's translated into English.

540
00:59:22,400 --> 00:59:29,390
Uh, so what is interesting is the reinterpret interpretation used in misuse in different contexts.

541
00:59:30,310 --> 00:59:33,540
This is on the plus side. We did this together with Ana.

542
00:59:34,300 --> 00:59:38,410
Uh, so, like, really that the interest in.

543
00:59:39,280 --> 00:59:43,959
I don't want to say that everything is reuse or the meaning is reuse or.

544
00:59:43,960 --> 00:59:47,140
But but oftentimes the meaning comes from that.

545
00:59:47,800 --> 00:59:54,370
Uh, and so first it's uh, taken as a moderately in criticism or criticism towards Mandeville.

546
00:59:54,370 --> 01:00:02,730
So there's vices uh, commercial society will lead us French Revolution suggestion to take arms against monarchy.

547
01:00:02,740 --> 01:00:06,160
Uh, then it's a claim against slavery. Some people said that.

548
01:00:06,160 --> 01:00:11,230
Well, clearly it's, uh, uh, you can use it in the pro-slavery way as well.

549
01:00:11,350 --> 01:00:18,400
Uh, and then, of course, uh, him being criticised, that principle doesn't apply to women.

550
01:00:19,150 --> 01:00:24,550
But so what we can do algorithmically. So we start separating engagement.

551
01:00:24,610 --> 01:00:29,230
Now, now here the the meaning, uh, comes through the engagement.

552
01:00:30,280 --> 01:00:40,659
So we see that that okay, here is not, uh, it's not a text you use, but, um, similar meaning.

553
01:00:40,660 --> 01:00:50,410
And there's an engagement to Rousseau and there's then a context where it's understood and those contexts are shifting.

554
01:00:51,130 --> 01:00:55,450
So a little bit about the question where we started that what is intellectual history?

555
01:00:56,320 --> 01:01:00,610
But why aren't we. In intellectual history.

556
01:01:00,620 --> 01:01:04,730
Interested in theory anymore? I mean, we really should be.

557
01:01:05,360 --> 01:01:13,970
And there's such good opportunities here to start thinking about this also for the people who don't want to be computational.

558
01:01:14,810 --> 01:01:20,840
And that that I mean, it's it's a it's a silly thing that if we think that everybody needs to be computational.

559
01:01:20,840 --> 01:01:26,180
Of course not. But I mean that these, these tools and opportunities need to be.

560
01:01:27,360 --> 01:01:35,610
You know, they need to embed what we are doing. Uh, because otherwise we are just looking at Mandeville and Hume and, you know, so.

561
01:01:35,610 --> 01:01:38,910
So that we can still do it. But but but let's move forward.

562
01:01:42,290 --> 01:01:45,550
Okay, I'll skip that. All right.

563
01:01:45,560 --> 01:01:48,710
Oh, so let me just say.

564
01:01:50,390 --> 01:02:00,500
Conclusions. So also with respect that if I'm thinking that use of computation in kind of historical humanities,

565
01:02:01,400 --> 01:02:06,889
it was for a long, long time and still I mean summed up by its mixed methods enterprise.

566
01:02:06,890 --> 01:02:16,320
Right. So you use algorithms to spot patterns, and then you go to the classic close reading to make the interpretation.

567
01:02:16,500 --> 01:02:19,530
So that that's that's quite often what people have been doing.

568
01:02:20,480 --> 01:02:26,870
But I don't think that the next step forward. It doesn't have to be another algorithmic novelty.

569
01:02:27,290 --> 01:02:30,410
I think with this, this kind of stuff, we have quite a bit already.

570
01:02:31,590 --> 01:02:37,650
So the tools are here, but then the theoretical growth inside that that will take us forward.

571
01:02:38,190 --> 01:02:46,440
So instead of kind of chasing for the next quick mix, uh, genuinely scientific mindset.

572
01:02:47,520 --> 01:02:51,780
That we are really blending the domain knowledge and rigorous data analysis,

573
01:02:52,320 --> 01:02:59,250
and then this kind of large scale infrastructure gains would be much higher than than what we have here, for example.

574
01:03:00,090 --> 01:03:08,160
So then the ad hoc fixes, remember this kind of one off topic models that people wanted to do maybe still,

575
01:03:08,550 --> 01:03:17,010
but they postpone the kind of bigger ambition. And, uh, so really sustained collaboration between historians and, uh,

576
01:03:17,020 --> 01:03:21,220
data science and, and so forth that I think a lot of people are pushing towards that.

577
01:03:22,030 --> 01:03:28,210
So then the AI comes to serve us, and we are not just some kind of mascots in other people's projects.

578
01:03:30,670 --> 01:03:38,500
So also maybe this kind of computational hermeneutics that should push the machine towards the.

579
01:03:40,790 --> 01:03:43,790
Or beyond the preliminary pattern hunting.

580
01:03:43,850 --> 01:03:47,540
Uh, or maybe the modelling. Modelling history.

581
01:03:47,570 --> 01:03:52,490
What would this mean? And what do we want to do? So. So this should, uh.

582
01:03:53,640 --> 01:03:58,050
Lead us to reshape. Uh, also the interpretive act.

583
01:03:59,990 --> 01:04:06,070
And, uh, so the interpretation is no longer something when the computer is completely put aside and then,

584
01:04:06,080 --> 01:04:12,620
okay, I'll think about it and make my own biased, uh, view of of things.

585
01:04:13,610 --> 01:04:18,230
So iteratively and, uh, kind of in concert going, going forward.

586
01:04:19,370 --> 01:04:23,689
And I really don't think that this will eclipse or the use of AI that that would

587
01:04:23,690 --> 01:04:29,660
eclipse anything from take anything away from us in a in a more substantial sense.

588
01:04:30,750 --> 01:04:42,520
And it will give us, uh, we sharpen the understanding of historical change and still rely on the our interpretive judgement in a,

589
01:04:42,530 --> 01:04:49,440
in a much bigger, bigger, grander scale. Uh, as part of, uh, this kind of integrated interdisciplinarity.

590
01:04:50,720 --> 01:04:54,170
And of course, I mean, we will change, uh, as, as we move along.

591
01:04:54,170 --> 01:04:59,060
But that's how the world works. The got stop here.