Sökta medel / Applied funding Budgetår / Budget year 2020

(1)

In19-0176:1 Mats Malm Nya vägar till det förflutna. Det litterära kulturarvet som källmaterial för humaniora och samhällsvetenskap

RIKSBANKENS JUBILEUMSFOND

Infrastruktur för forskning

Professor Mats Malm

Institutionen för litteratur, idéhistoria & religion Göteborgs universitet

Box 200

SE-405 30 Göteborg

ORCID: 0000-0003-4123-5984

Nya vägar till det förflutna. Det litterära kulturarvet som källmaterial för humaniora och samhällsvetenskap

New Paths to the Past. Literary Cultural Heritage as Source Material for the Humanities and Social Sciences

Anslagsförvaltare / Grant administrator Litteraturbanken

Startdatum / Start date: 2020- 01-01

Slutdatum / Final date: 2022- 12-31

Ämneskod /subject code 1: Litteraturstudier Ämneskod / subject code 2: Kulturstudier

Ämneskod / subject code 3: Sociologi (exklusive socialt arbete, socialpsykologi och socialantropologi)

Sökta medel / Applied funding

Budgetår / Budget year 2020 ^Lokaler

(ej del av sökt belopp)

Indirekta kostnader (ej del av sökt belopp) Indirect costs (not part of applied funding)

Lönekostnader / Salaries 0

Drift / Operating costs

3.615.000

Totalt år / Total (year) 2020 3.615.000

Personal / Staff (icke forskande personal / Non-research staff: 80%

årsarbetstid)

Disp.år / Year of PhD exam

Årsarbetstid / Annual working time

LKP / Employer's social-insurance contribution top- up

Månadslön (heltid) Monthly salary (full time)

Malm, Mats (Professor), 640510-3935 1996 10% 0% 0

Miocevic, Ljubica (Fil. dr), 771118-5004 2017 10% 0% 0

3.676.000

årsarbetstid)

3.154.600

årsarbetstid)

(2)

Total projektkostnad / Total project cost: 10.445.600

Projektledarens (Sökandens) underskrift / Project manager's (applicant's) signature

Datum / Date:

Underskrift och namnförtydligande / Signature and name:

Anslagsförvaltarens (Prefekt/motsvarande) underskrift / Grant administrator's (head of department or equivalent) signature

Datum:

Underskrift och namnförtydligande:

(3)

Budgetkommentar / Comments on the Budget

1: Digitization at Uppsala University Library (UUL) and National Archive s Unit for Digitization DIT (formerly Centre for Media Conversion):

UUL: Structuring of lists from Libris national catalogue. Control of books, creation of catalogue entries, maintenance and packing to DIT, followed by publication in Alvin. 8 600 books: 1 778 000 SEK.

Digitization of oldest and most fragile books, ca. 1 500: 2 000 000 SEK.

DIT: Digitization, handling, and transport of 7 100 books (calculated at an average of 250 pages per book and 2,87 SEK per page) and delivery of data: 5 094 250 SEK.

Total: 8 872 250 SEK.

2: Digitization of books missing or in too poor condition for UUL to handle will be ordered from other libraries: 100 000 SEK. The cost will probably be greater, but the grant administrator will cover that part.

3: Editorial work with databases for text mining:

30% of full time editorial work with database sheets adapted to the publication on-line and as downloadables, including social fees: 190 000 SEK, annually updated (2%). This is the expected minimum: if exceeded, the grant administrator will cover the rest.

4: Tools for text mining and implementation of laboratory web page: developer, 50% of full time for two years, including social fees: 390 000 SEK, annually updated.

No indirect costs are included.

5: Travel costs for participation at conferences, meetings with prospective and established partners and advisory board. 35 000 SEK per year: 105 000 SEK.

2020

Operating costs:

Digitization: 3 000 000 Travel: 35 000

= 3 035 000

Non-research staff:

Databases 30%: 190 000 Tools 50%: 390 000

= 580 000

Total = 3 615 000

(4)

2021

Operating costs:

Digitization: 3 000 000

Digitization of books elsewhere: 50 000 Travel: 35 000

= 3 085 000

Non-research staff:

Databases 30%: 193 800 Tools 50%: 397 800

= 591 600

Total = 3 676 600 2022

Operating costs:

Digitization: 2 872 000

Digitization of books elsewhere: 50 000 Travel: 35 000

= 2 957 000

Non-research staff:

Databases 30%: 197 600

= 197 600

Total = 3 154 600

(5)

Sammanfattning på svenska / Summary in Swedish

Skönlitteraturen är en unik källa till vår historia: den är inte en historisk

representation av fakta, men den gestaltar och bearbetar olika tiders samhälleliga och individuella problem och frågor. Skönlitteraturen är alltså ett centralt

källmaterial inte bara för litteratur- och språkforskare, utan också för historiker, sociologer, antropologer, idéhistoriker och en lång rad andra humanistiska och samhällsvetenskapliga forskningsinriktningar. Eller, rättare: den kan bli det, om den digitaliseras och görs tillgänglig för utforskning med avancerade verktyg.

Den internationella utvecklingen är intensiv, och det svenska materialet blir alltså utgångspunkten för innovativa internationella samarbeten som inbegriper helt nya former av teoretiskt och metodologiskt utbyte över ämnesgränserna.

Detta projekt avser att ställa ett sådant material till forskningens förfogande genom att digitalisera den svenska skönlitteratur som publicerades separat under hela 1800-talet. Materialet speglar ett sekel som kännetecknas av intensiv

kulturell och samhällelig förändring. Det utgör därmed en rik grund för analyser av utveckling och förändring på en lång rad områden.

Hela materialet görs tillgängligt för forskningen i tre portaler med olika

nyutvecklade redskap för att utforska det, och nerladdningsbart. Det blir också tillgängligt för allmänheten, skolor och bibliotek, som därmed får nya

möjligheter att närma sig och levandegöra det litterära kulturarvet.

Sammanfattning på engelska / Summary in English

Literature is a unique historical source: it treats and problematizes societal and individual issues through the stagings of the imagination. Literature – primarily prose fiction, but also drama and poetry – thus constitutes a seminal source material relevant not only to literary and linguistic researchers, but also for many other fields of the Humanities and Social Sciences. For its true potential to be realized, this material needs to be digitized and made available for exploration with advanced tools. Internationally, this potential is actively explored, which means that the Swedish material may constitute the foundation for further innovative international collaborations, including entirely new kinds of theoretical and methodological interdisciplinary development.

This project will provide researchers access to a large set of materials, by digitizing the Swedish literary works that were published separately (i.e., not as part of journals and newspapers) during the entire 19th century. This is a century of intense cultural and societal change, and the digitized corpus constitutes a remarkable foundation for the analysis of change and development in a wide number of areas.

The whole material will be made available to research on three web sites

providing different novel tools for exploration, and in downloadable form. It will also be available for the public, for schools and libraries, making it possible for all to approach literary cultural heritage in new ways.

(6)

Projektbeskrivning / Project description

Aim

Internationally, the development of resources, tools and approaches for text mining has been rapid. However, the use of digitized texts in Swedish Humanities research has so far been mostly limited to linguistic interests, focusing on how language is constructed rather than exploring its vast contents.

Within Swedish Social Sciences, tools for analyzing content have been more systematically explored, but mainly applied onto modern, digitally born, material.

For exploring the Swedish literary heritage and re-reading history in innovative ways, there is one major impediment: no systematic digitization of books on any scale has yet been performed in Sweden.

Swedish literary heritage 1800–1900 holds large promises for use within a wide range of disciplines, not only within the Humanities but also the Social Sciences.

It reflects a century of intense cultural and societal change, and constitutes a remarkable foundation for analyzing change and development in a number of areas. It also promises to be the venue of intense and sophisticated

interdisciplinary collaboration and development of methods, technologies and research paradigms, not least within the area of combining quantitative and qualitative methods. These statements are corroborated by the development that has characterized the field internationally. Sweden risks lagging behind this international development. However, with a focused effort of this kind, Sweden has the opportunity to take an important part in the international development and make a considerable contribution.

The aim of this project is to make a significant part of the Swedish past accessible to new kinds of research. Within the project, the bulk of Swedish literary works published separately 1800–1900 will be made available both to researchers and to the general public, together with tools for exploring the material. The intention is to pave the way for reading Swedish history anew and make Swedish research a vital part of the intense international development within Computational

Humanities and Social Sciences. The potential for methodological development and exchange over the borders of disciplines as well as nations is extraordinary.

The Swedish Literature Bank (LB) is a well-established portal for Swedish literary heritage. It supplies researchers, libraries and the community with, at present, close to 4 000 books for free use and a number of presentations and introductions. It also supplies teachers on all levels with material, introductions and exercises for their students. Literary texts provide an extremely valuable source for analysis of the past, enabling not only literary and linguistic analysis but also investigations of ideology, society, notions of identity, nationality, gender and minorities, to mention only some aspects. They also provide the perfect source for the processes of cultural heritage: how the past has been used and exploited for different needs at different times.

Uppsala University Library’s (UUL) Alvin (alvin-portal.org) is a well-established portal for cultural heritage material, which presents material in a way more

(7)

directed toward book history than LB. As UUL is the source of material for this project, all books will be made available in Alvin as well as in LB.

The materials produced within this project will profit already established or planned collaborations with a number of research environments, to a great extent through the partnership between LB and the Centre for Digital Humanities at the University of Gothenburg (CDHUGOT; http://cdh.hum.gu.se). These

collaborations include the Swedish Language Bank which provides LB:s infrastructure (http://spraakbanken.gu.se), SWE-CLARIN (http://sweclarin.se), Humlab in Umeå http://(umu.se/humlab), the networks for Digital Humanities in Uppsala and in Stockholm, and the Institute for Analytical Sociology, Linköping University (https://liu.se/organisation/liu/iei/ias), which currently runs the

Swedish Research Council project “Mining for Meaning”, aimed at tracing discourses on migration and integration in modern Sweden. Through the 19th century literary heritage, it will be possible to broaden the scope of “Mining for Meaning” onto later material and establish significant interaction as regards research questions and methods over the borders of Humanities and Social Sciences.

Another domestic collaboration of significance is the Royal Library, which has digitized a great amount of 19th century newspapers, a corpus that is ideal to place in parallel with the literary works. The planned laboratory of the Royal Library is expected to be a significant partner in the development of tools and approaches for future research.

The material will not be limited to collaborating projects, but freely available for download. All materials are downloadable from UUL:s Alvin, and LB is

developing a web page where anyone can choose among all non-restricted texts according to a number of parameters such as author, genre, time period.

Apart from providing research with entirely new venues, the project will also improve community’s access to the common past. One of the great challenges in the digital age is the imbalance between knowledge, opinions and reliable

sources. Not only will all books be possible to find and read; they will also be suggested when searches are made with common search engines such as Google, providing community with new paths to the past.

Background

While the concept of cultural heritage originally concerned spiritual/intellectual culture, it has achieved a very physical sense in the intense development of recent decades, dominated by archaeology and museology. In order to clarify, a division is often made between tangible and intangible cultural heritage, intangible being that which is incorporated not in objects but in human beings (Fairchild Ruggles and Silverman 2009). Literary heritage has the elusive character of not being possible to restrict to either the one or the other: it is tangibly transmitted through physical books, but it is intangibly actualized in humans and in collective

awareness. Thus, literature may in many respects be more adequately understood in terms of cultural memory, providing a key to the understanding of cultures past: “Intertextuality demonstrates the process by which a culture, where ‘culture’

is a book culture, continually rewrites and retranscribes itself, constantly

redefining itself through its signs. Every concrete text, as a sketched-out memory

(8)

space, connotes the macrospace of memory that either represents a culture or appears as that culture.” (Lachmann 2010, 301)  

In this constant renegotiation of culture, different texts have different status: it has been described as the “cultural working memory” which is canon, constantly actualized and processed, and the “cultural reference memory” which is the archive, passively stored and unknown (Aleida Assmann 2010). As Assmann proposes, cultural memory receives its dynamics and energy from the interaction between these two spheres: this in itself is a central object of study. Digital materials and methods provide the ideal tool for finding, mapping and analyzing these interactions between active canon and passive archive, which affect

practically all cultural aspects of society. While canon can be studied from that perspective, it is also a seminal question of power and suppression – of notions, ethnicities, class and gender. While digital materials and methods enable studying these formations, text mining projects are also known to risk repeating such structures. One vital part of avoiding such risks is to digitize all relevant material, in this case all books from the chosen period.

A foundational assumption of the project is that quantitative methods derived from mathematics can be fruitfully combined with qualitative methods as

traditionally developed within the Humanities and the Social Sciences. This is to be regarded as one of the main challenges to future research, and will involve substantial interddisciplinary collaboration and exchange of perspectives and approaches. This process will to a considerable extent be eclectic, trying out options and solutions, but it shall constantly be paired with diligent consideration of problems such as superficial objectivity in data results and visualizations of them, obscurity what different algorithms actually effect, and institutional and theoretical implications of coupling traditional approaches with new ones in the production of knowledge (Rieder and Röhle 2012). It is widely acknowledged that research through digital materials and methods has often tended to become too focused on method, losing sight of context and theoretical framework. The ambition of the project is to not only explore possibilities, but also to map and evaluate risks and drawbacks for a theoretical understanding.

State of the art

Internationally, the amount of digitized books has reached vast proportions. At the forefront has been literary works, but the great importance of both literary works and other books as source material for a number of disciplines has been widely recognized. At the same time, the large amounts of material have made it clear that thorough mark-up, proof-reading and intense curation of texts can only be exceptions. Instead, focus has in recent years been directed toward developing technologies for unsupervised analysis of large, “raw” corpora. By now, a

number of tools for text mining and culture analytics have been tested and evaluated in different circumstances world-wide. Concerning literary texts, the perspectives have to a high degree been derived from literary studies or history of ideas, but also from history and other disciplines.

Without a doubt the best known scholar in this field is Franco Moretti, who made the expression “distant reading” known to the world. While the concept has been used in a great many contexts to designate practically any kind of text mining, it is important to note his definition: ”Distant reading: where distance […] is a

(9)

condition of knowledge: it allows you to focus on units that are much smaller or much larger than the text: devices, themes, tropes – or genres and systems”

(Moretti 2013, 48). As Moretti points out, literary studies can leave the extremely narrow canon which is surveyable by traditional reading, and bring what has been forgotten or diminished into account, thereby offering entirely new

understandings of literary cultural heritage and thus society. In accordance with this, Moretti’s focus has to a very great extent been quantitative, only rarely touching upon qualitative aspects of the books studied. This is also the nucleus of Matthew Jockers’ (2013; also Jockers and Mimno 2013) notion of

“macroanalysis”. The quantitative approach to literary and cultural history has great promise, not least since it provides magnificent opportunities to trace chronological development in very large corpora. Yet, in challenging the restricted image of the past offered by canon, further development is needed.

While the potential of methods for content analysis, applied to a variety of materials and from the viewpoint of a variety of disciplines, is substantial, two main problems with this kind of approach have been observed. First,

representativity is a crux: although the studies are based on large materials, the composition of the materials ordinarily depends on what happens to be at hand – and although the digitized material in the large languages is by now substantial, it has as a rule been selected on the basis of ideas of interest, aesthetic value and usefulness, often resulting in the furthering of already canonized male

authorships (Nowviskie 2012; Bode 2017; Bergenmar and Leppänen 2018). As the results of quantitative analysis are numbers and data that must be reliably related to other numbers and data, quantitative analysis implies high demands on representativity.

While the problem of representativity and selection can be handled in several ways, the second problem is that purely quantitative analysis not only makes itself dependent on representativity, but it also limits the possibilities of

interpretation considerably, and restricts the merging of traditional qualitative and new quantitative methods. In order to really understand the results of a

quantitative analysis, the researcher often needs to be able to go to the specific sources, control, examine, interpret individual works, passages and contexts. The blending of quantitative and qualitative approaches is the way that traditional Humanities and Social Sciences research of the qualitative kind can be furthered by digitization.

Thus, if quantitative methods can be combined with qualitative methods, the potentials of digitized sources will grow exponentially. So, for example, mapping of themes, “topic modeling”, in literary texts can be developed in ways that allow the researcher to make use of her/his domain expertise for studying large corpora on a number of different levels, ranging from the macro to the micro level via several middle ranges (Blei 2012; Tangherlini and Leonard 2013; Boyd-Graber, Hu and Mimno 2017), that is, making it possible to use powerful tools and large materials as a way to prepare, conduct and fulfil sophisticated qualitative

analysis.

While topic modeling (including different ways to refine and direct analysis such as Tangherlini and Leonard’s “sub-corpus topic modeling” (2013)) provides opportunities to follow the discussion of single topics and themes in vast materials, sentiment analysis makes it possible to add analysis of attitudes

(10)

formulated in regard to those topics, thereby mapping also how topics and themes have been appreciated – again, paving the way for closer study and analysis (Kiritchenko 2016). Yet one seminal way of tracing relations between texts and tracing ideas on another level is sequence alignment, using methods from genome research to map correspondences on the level of words and phrases rather than co-occurence (Olsen et al. 2011; Leonard and Malm 2014). A further approach of fundamental importance for understanding historical conceptions of society, life, culture as well as many other topics, is semantic change, using word embeddings (Zhang et al. 2016). The methods for shifting focus and levels on large materials are under intense development – to a great extent within the Social Sciences, often concerning web communities and other digital media. The potential for methodological development, particularly concerning quantitative and qualitative approaches, is considerable. An initiated discussion of the relationship between reading on different scales and combining quantitative and qualitative analysis for both humanistic and sociological research is to be found in English and Underwood (2016).

So, while quantitative models are often of great use, the problem of such approaches being too far distant from the texts themselves can be solved. The addition of qualitative aspects makes representativity less problematic, but on the other hand, so far practically no language area has provided its literature in full.

The only country that has succeeded in this so far is Norway, where every book printed in Norway or in Norwegian until 2000 is available (for Norwegian IP addresses) at http://bokhylla.no. If Sweden is able to provide an equally

representative material of the literary works 1800–1900, this would be, within its limitations, a larger material than Norwegian 19th century literature and would lay the groundwork for entirely new modes of comparative analysis.

So far, studies of large collections of literary works have been mainly linguistic, literary or directed toward history of ideas. However, there is apparent need for structured collaborations between a variety of disciplines. As Benedict Anderson clarified when proposing the concept of “Imagined Communities”, the 19th century novel heavily contributed to conveying nationalist ideals by creating memories and imaginations about common history and identity (Anderson 1983).

Literary heritage is a rich – albeit complex – source for ideological analysis of the past, and the same applies to investigations of notions of identity, society, nationality, gender, minorities etc. If the vast body of literary heritage can be explored for such issues, it will become a prolific source material for a wide range of approaches of historical studies, not least history of mentalities, history of economics, history of ideas, sociology, ethnology, linguistics, comparative literature, and thus prepare the ground for new interactions between the

Humanities and the Social Sciences. The problem has been how to find, search and efficiently use the vast material of prose fiction. This has particular bearing on cultural heritage and cultural memory research, which has developed

considerably the past decades (e.g. Nünning/Nünning/Neumann 2010, Erll 2011, Cameron/Kenderdine 2007, Aleida Assmann 2010 and Lachmann 2010; also, in wider approaches, McCarty 2010, Bartscherer/Coover 2010). Cultural memory, as Jan Assmann puts it (1995, 132), is reflexive in three ways: it is “practice- reflexive”, i.e. focusing on common practice through proverbs, rituals etc, it is

“self-reflexive”, i.e. expands on itself, and it is “reflexive of its own image”, i.e. it

(11)

reflects the group’s self-images and social systems. This opens up for profoundly new approaches to literary heritage as a source for a wide range of disciplines – but a connection to digital resources has so far not been developed.

Significance: The project connects to a wide range of approaches within Digital Humanities and Cultural Heritage Studies, including Cultural Memory Studies, and aims at providing the foundation for systematically exploring literary heritage as a source for the Humanities and Social Sciences. The material represents not only a society and culture in dramatic development, but also a literary field and book market marked by extraordinary change. By establishing the proper source material, the project provides means to address the above mentioned issues of representativity, chronological development, the relation between quantitative and qualitative methods, and the need for opening up literary heritage to a wider range of Humanities and Social Science disciplines:

1. Creating a representative corpus of a nation’s literary works from a whole century offers a unique material to work with. A full corpus and the

representativity it entails (possible because of the limited volume of books in a small country) offers new opportunities for quantitative and qualitative

approaches, as well as comparative work with corresponding corpora such as the Norwegian. Tools for utilizing the resource will be offered, but the material will also be available as the base for using other tools and developing entirely new tools.

2. A full corpus offers new ways of studying culture and society through the lense of literature, not only as transmitted by canon, but also as formulated by the mainstream, forgotten and even suppressed authors of the period. Thus, the material opens up for a wide range of disciplines within the Humanities and Social Sciences.

3. The material offers new ways of combining quantitative and qualitative methods – a process that is already well underway, but which has significant and important development ahead, especially directed toward historical materials.

This methodological development will be very much focussed on exploring historical change.

4. Cultural memory studies and Cultural heritage studies have so far not been thoroughly connected to Digital Humanities or text mining. As a consequence of this project, we believe that new and productive ways of exploring the potential of the digitized materials will come to light.

While Sweden has been lagging behind as regards digitization of literary heritage, the proposed solution offers a concentrated effort which has the potential to place Sweden at the forefront of international research.

Method

Digitization and distribution

Through a new line of production, books will be listed and procured by UUL.

The most fragile objects will be digitized in-house, but most will be forwarded to the National Archives’ Unit for Digitization DIT (formerly Centre for Media Conversion) for digitization (transport by DIT). Lists are made through the national library catalogue Libris and UUL’s catalogues, controlled against

(12)

physical collections in UUL, with minor addition through a number of other sources. Books are taken from UUL’s collections (with some addition from other libraries) and data will be captured through scanning and photography, combined with optical character recognition of all images. Images, data and metadata are forwarded both to UUL and to LB, for publication and storage at both facilities.

Specifications

• Data will be generated as

TIFF files, 400 ppi, Adobe RGB 24 bit, compression LZW, named after Libris id.

for LB and after Alvin id for UUL, running from 1 upwards. Single pages followed by covers, scale/colour patch and index sign.

PDF files, named in the same manner. Regular PDF, Image over text, compression 94% for LB, PDF/A-1 for UUL.

ALTO files, Version 2.0, Text coordinates; words, Character formatting; full, Measurement unit; pixel.

ABBYY XML, Character attributes; true, Extended attributes; true, Character formatting; true, Coordinates with respect to original image; true

UUL: Scanned images will be stored as individual TIFF files in the highest

resolution judged reasonable for the page sizes in question, typically either 400 or 800 PPI. In addition, a PDF type A file will be generated for each volume,

containing a lower-resolution version of the pictures, as well as any OCR text generated during digitization. Metadata will be stored in records in Alvin.

LB will store TIFF files 300 PPI and PDF files with embedded OCR, as well as produce XML and plain text files for downloading.

• Data documentation will be given as metadata posts.

Alvin provides a structured metadata format for the digitized assets. The metadata creation is fully documented by the Alvin repository functions.

LB: Metadata will follow OLAC but with additional METASHARE fields. Text files will have metadata in the header.

Author/editor

Birthyear of author/editor Deathyear of author/editor Gender of author/editor Title

Year of publishing Place of publishing Edition (first or later) Language (Swedish) Type face

Genre

Provenience (library holding the original digitized and classification [signum]) Libris id of original

Libris authority id Alvin id of original

Libris id of digitized version

Number of images in digitization of original Image number for book’s page 1

(13)

Digitized by

Method/equipment for digitizing

Specification of image (resolution, colour, compression) Method for OCR

Specification of text files

Proofread or not (standard: not) PID Handle

Source: alvin.org and litteraturbanken.se

Licence (PD for image files, CC-BY for text files with metadata)

• Metadata in Alvin will follow a documented, Alvin-specific schema based on MODS in combination with EAD.

Metadata in LB: OLAC with additional METASHARE fields.

Terminologies and ontologies used in Alvin for classification and subject headings comprise

lcsh (Library of Congress Subject Index Headings) LoB (Language of Bindings)

nad (Nationell arkivdatabas) sao (Svenska ämnesord)

TGMII (Thesaurus för grafiskt material II) DDC (Dewey Decimal Classification) kssb (SAB klassifikation)

iconclass

And also, controlled vocabularies on materials, techniques, languages, countries objects and documents types.

• Data publication: Alvin enables easy publishing of all digitized material when deemed ready for publication. Alvin can be considered a domain-specific repository for the Archives Libraries and Museums sector, maintained in cooperation between several Swedish university libraries.

LB: Data will be published at litteraturbanken.se for direct reading and for searching and mining with a number of tools directed at a wide range of

Humanities and Social Science researchers as well as the public. Simultaneously, the data is established as a corpus for equally free use but with linguistic tools at spraakbanken.gu.se/korp. Data will also be offered for download as xml files, plain text files and regular pdf files with embedded ocr.

The FAIR principles are endorsed.

• Data handling: The main bulk of data will be delivered from DIT to UUL (its originating institution) and LB simultaneously. Fragile works will be produced within UUL. Additional works will be digitized by other libraries, which store and publish the material themselves.

Digital material in Alvin is published in a shared digital repository which also serves as the primary storage of image files. If necessary image files may be easily shared.

Digital material in LB is published in a digital repository connected to the Swedish Language Bank, which is also the primary storage. XML and plain text files will be offered for easy downloading.

(14)

Persistent identifiers: UUL: All records published in Alvin receive a unique, global identifier in the form of an URN:NBN.

LB: Handle via the Swedish Language Bank. Machine Readable.

Reuse limitations: none. All digitized items in UUL/Alvin are published under a Public Domain Mark (No known copyright). Metadata is free for all types of use and re-use. For Alvin, the public domain mark is the most common (standard in this corpus) and all entries with a digitized content have a license marking.

LB: XML and plain text files with included metadata will be distributed under CC-BY, free for use and reuse. Image files Public Domain. Machine readable.

Tools

The material will be put to the disposal of text mining researchers in three ways:

A) through tools developed in LB, B) through linguistic tools supplied by the Swedish Language Bank, and C) through downloading for use with any preferred tools.

A) LB is developing a lab page where anyone will be able to use LB:s material with a (growing) number of tools. Those under construction are:

• Several basic tools for visualizing how and where different words and names occur in the material.

• Sequence alignment. Aligning strings of characters makes it possible to map

‘reuse’ of texts, for example tracing Bible quotations in the whole of the corpus.

This is useful for instance in order to trace debates on the meaning and

significance of individual Bible passages – or any kind of quotations. Another approach is to trace the literary influence of important translations such as C.A Hagberg’s Shakespeare. A third is to see to what extent an author such as August Strindberg, who always rejected the thought of having been influenced by

anyone, actually used previous literary works. A fourth is to simply run all works against one another, or selections, in order to trace intertextuality in Swedish literature. This method is bringing about a new view on intertextuality not only as something elaborate and elegant, but also as a technique of market literature, adding information to the social processes of literature (Leonard and Malm 2014;

cf. Lachmann 2010). A fifth is to turn the process around and instead of tracing similarities in different texts, trace differences in the same work as it appeared in different editions. It is well known that many works were changed, sometimes in significant respects, in later editions, and this will be a way to quickly map and find such works and thus map new aspects of literature’s aesthetic, social and political conditions. This tool is built in collaboration with the Swedish Language Bank.

• Topic modeling. While Sequence alignment traces correspondences on character and word level (allowing the user to define how strict the

correspondence to ask for), Topic modeling enables the mapping of how topics, themes, discussions spread over large materials by measuring co-occurence of words on a less literal level. The method has proven very useful for many causes, but can also be developed in several ways.

• Sub-corpus topic modeling makes it possible to curate topics from a chosen work or set of works, and then project these topics onto a larger material as a kind of trawl. The possibilities in this method are described in Leonard and

(15)

Tangherlini (2013). This tool is being developed with Leonard (Yale), Tangherlini (UCLA) and Chalmers.

• Seeded topic modeling. This technique lets the researcher define the topics instead of seeing what the computer proposes. A new topic is created with the help of a number of keywords, which makes it possible to narrow searches down significantly. This tool has been developed not least by Måns Magnusson, Aalto University, with whom we will collaborate on this. It is so new that the

experiences and testing of it are just beginning. The testing of the tool (not part of the application proper) will thus be of considerable interest to other research groups. This also pertains to the Sub-corpus topic modeling.

• Vector space models. This tool will, primarily, enable entirely new ways of tracing semantic change in large materials.

B) LB constitutes one of the linguistic corpora of the Swedish Language Bank, and can there be used with a large set of tools for computational linguistics. The tools can be seen at https://spraakbanken.gu.se/korp, where ‘Litteraturbanken’ is one of the choices in the top menu.

C) All data produced within the project, as well as much more, will be offered in suitable formats with significant metadata.

Project plan

Part 1: Cataloguing and digitization

All separate publications classified as literary (“skönlitteratur”, including novels, short stories, drama, poetry, and other kinds of prose such as literary calendars, sagas and essays) in Swedish that are not translations 1800–1900 in the national library catalogue Libris amount to ca. 10 600 books. Libris does not record all relevant publications, but can be efficiently completed by UUL’s own physical collections, which add 568 works to those registered in Libris, making the total amount ca. 11 200 books. Thus, the lacunae in Libris as well as other catalogues will be substantially filled out. The corpus is limited to separately published books: literature published in newspapers and journals is a future priority.

Of these 11 200 books, ca. 600 are available in digitized form, most of them at LB. LB has previously commissioned but not yet published all literary prose works 1880–1900 from university libraries, in total ca. 2 000 books, which leaves ca. 8 600 books to be digitized in order to cover the literary heritage of the 19th century.

Books from 1800–1829 as well as additional fragile items will be digitized at UUL (ca. 1 500 books), other works will be sent to DIT for digitization (ca. 7 100 books). The volume is calculated at an average of 250 pages per book, amounting to ca. 375 000 pages and 1 775 000 pages respectively, in total 2 150 000 pages.

In the process, optical character recognition is performed. Results are forwarded both to UUL and to LB, for publication and storage at both facilities.

All books will be available through the national catalogue Libris, where new entries are made for the digitized versions, as well as for previously missing physical books.

Part 2: Insertion into databases and publication

(16)

The books will be part of Uppsala University’s portal Alvin and LB databases.

As regards Alvin, publication is included in the price. As regards LB, editors will create database sheets with metadata adapted to the publication on-line and as downloadable files. The expected minimum time for this is 10 min. per book. As some books will be more complicated, 0.9 years’ editorial work is anticipated.

Part 3: Tools

LB will present a laboratory page where users can define different selections or use the whole material for different kinds of analysis. One seminal function for exploring patterns in vast materials is to trace the connections between different works – both by mapping verbatim correspondences for exposing influences, citations and plagiarisms (sequence alignment), and by tracing discussions and wider influences on diverse topics (topic modeling). The laboratory page will be developed in collaboration mainly with CDHUGOT, Chalmers, Yale University and University of California Los Angeles. The plan is also to do this in

conjunction with the Swedish Royal Library’s projected laboratory.

As a corpus within the Swedish Language Bank

(https://spraakbanken.gu.se/korp), LB is already possible to use with a wide range of linguistic tools, which will be expanded with time. This is included in the Swedish Language Bank’s ordinary development.

(See Method)

Part 4: Collaborations and research initiatives

The material provides great opportunities for cutting-edge research and cross- disciplinary development of methods and approaches with national and

international research groups.

Apart from overarching project management, Mats Malm will work with collaborations and research initiatives on different venues, not restricted to the ones here mentioned.

Ljubica Miočević will collaborate with Malm in this and be working with detailed coordination of the production line.

The CDHUGOT, recently appointed a Knowledge Center within SWE-CLARIN with a particular focus on historical text collections, will be a vital liaison with existing and emerging research initiatives.

The digitization and delivery of data will be made in three years, 2020–2022, and all material will have been published for open use by mid 2023. The tools and laboratory web page will be finished, in a state prepared for addition of modules, by the end of 2021.

Risk: 250 pages per average book is an estimation that might prove false: in such case, the administrator will make sure to produce the rest. Regarding digitization and publication through LB and Alvin, the problems that might arise are not considered to be able to create worse effects than delays. In case the publication of the full corpus is delayed, the plan will still be fulfilled. Regarding the

development of tools and cooperations, which are not funded by this application, it is projected that some venues may have to be reconsidered while other

(17)

possibilities arise. There is no anticipation that the result should be less than described in the plan.

As the material never involves depictions less than 120 years old, no ethical issues need to be handled. The material includes a number of books that are not out of copyright and thus cannot yet be published: all such conditions will be carefully monitored. Books that are not yet out of copyright will be stored until the date it is legal to publish them.

The project will profit from the expertise of an advisory board, with which questions concerning infrastructural practicalities as well as research

development and collaborations will be handled. The advisory board consists of Marianne Gullberg, Professor of Psycholinguistics and director of Lund

University Humanities Lab.

Peter Hedström, Professor of Analytical sociology and head of Institute for Analytical Sociology, Linköping University.

Lorna Hughes, Professor in Digital Humanities, University of Glasgow.

Peter Leonard, Director of the Yale University Digital Humanities Laboratory.

Relation to existing infrastructures

Nationally, all digitized works flow into two existing infrastructures, LB and Alvin. They also flow into the national library catalogue Libris as new entries (and thus become readily available not least for the customers of public libraries).

LB is developed within the Swedish Language Bank and constitutes one of its corpuses for computational linguistics. As the Swedish Language Bank is the centre of SWE-CLARIN, initiated to make linguistic resources and tools

available to greater ranges of researchers, LB resources become available there, as well. Currently, preparations are being made for connecting LB to K-samsök (http://www.ksamsok.se). Through CDHUGOT, with a coordinating function in SWE-CLARIN, connections are built with infrastructures such as the Royal Library’s projected laboratory.

Internationally, works will flow through K-samsök to Europeana. A selection of works flows into COST action European Literary Text Collection

(https://www.cost.eu/actions/CA16204), while other material will be delivered to the Nordic Digital Humanities Laboratory (under construction).

Research’s need for infrastructure

The need for this kind of resources and tools is significant, to say the least. It affords entirely new approaches to history and cultural heritage, paving the way for new understandings and questioning of old conceptions. It opens up literary cultural heritage as a research source for practically all Humanities and Social Science disciplines, and thus provides the ground for entirely new kinds of cross- disciplinary exchange and development of methods and perspectives.

This development is well on its way internationally, and Swedish researchers have the opportunity to take an important position in it. Many technological advances have already been made, and make it all the more efficient to explore digitized books. Internationally, there are still very few examples of a so full and

(18)

representative corpus as the one planned here, which means that it can provide results that are of considerable interest outside of Sweden. While the material is domestic, the field is exceptionally international and cooperative.

The need for strengthening Swedish infrastructure for the Humanities has become more and more obvious in recent years. This has been reflected in a number of seminars and calls. In a Report on Infrastructure for the Humanities, a group led by Patrik Svensson and including the applicant has pin-pointed the needs for strategic work in order to promote the Humanities in Sweden. One infrastructural resource that will be seminal for a great number of approaches not only within the Humanities, but also the Social Sciences, is the books that carry the nation’s collective memory. The books of primary importance are the literary works, which constitute the source to individual and social notions, ideals, dreams, wishes, criticisms and problematizations – always, in different ways and to differing extent, pertaining to the society of their time (cf. the above argument on cultural memory). The literary works published separately in Swedish are the first priority and the one concerned in this application. Thereafter, the works of other periods, literary works published in newspapers and journals, literary works translated from other languages into Swedish, and literary works translated from Swedish into other languages will be prioritized. For a number of approaches, all of these materials can be very efficiently studied in parallel with newspapers.

Relation to international research

Developing resources and tools of the kind this application concerns is a fervent field over the world, and the international cooperation is intense. Although there are many texts available for data mining in the large languages, one fundamental problem is that they are primarily constituted of mainstream and canonized texts:

that is, large-scale analysis will cover the canonized but not the entire literary system. The comparatively small amount of Swedish literary texts makes it possible to digitize all texts, and thus to perform much more representative

analyses that will be of considerable general impact: Swedish is an ideal test-case and the results of this kind of work would be highly generalizable, of

considerable interest to parallel research in other language areas.

In order to promote Scandinavian and international collaboration, the

organization Digital Humanities in the Nordic Countries was started in 2015. The initiative proved to fill an urgent need and was a great success: the organization’s fourth annual conference was held in Copenhagen 6–8 March 2019 with well over 200 participants. DHN is an associated organization to the European Association for Digital Humanities and Alliance of Digital Humanities

Organizations, and thus connects to a wide range of international researchers and projects.

During a period at Oslo University as Professor II for strengthening competence in Digital Humanities, the applicant established a close collaboration with the National Library’s digital department.

Among the number of different contacts with international research, the most important ones are listed in the following section.

(19)

International research cooperation

The Norwegian National Library has digitized all books printed in Norwegian or in Norway until 2000, amounting to over 500 000 books. While Sweden is very far from that kind of resource, with a representative material of the literary works of the 19th century, comparative studies of an entirely new kind will be enabled.

Preparatory experiments have been made.

We cooperate with the Danish project “Measuring Modernity: Literary and Social Change in Scandinavia (1870–1900)”. As can be seen, the corpus concerned in this application fits perfectly, and the Scandinavian comparative perspective is very promising as the Norwegian material is already at hand.

Further Scandinavian collaborations include the Aarhus–Gothenburg–Oslo–

Helsinki network Nordic Digital Humanities Laboratory (NDHL), which involves a number of American Digital Humanities experts in developing text and tool resources. The ambition is to establish a common portal for digitized texts and tools for processing them. LB takes a special position here, in offering a great number of controlled, proof-read texts which can be used as gold standard in developing tools.

LB and CDHUGOT particularly closely collaborate with Yale Digital Humanities Lab (http://dhlab.yale.edu) and the University of California Los Angeles

(scandinavian.ucla.edu) in developing tools for text mining.

In the UGOT Centre for Critical Heritage Studies

(https://criticalheritagestudies.gu.se), the applicant and the leader of CDHUGOT Cecilia Lindhé closely collaborate with the UCL Centre for Digital Humanities Centre of University College, London (https://www.ucl.ac.uk/digital-humanities/) in devising new approaches to digital cultural heritage.

In the European research collaboration and repository European Literary Text Collection, LB is one of the providers of hundreds of novels for new kinds of comparative analysis of literary genre and style in a chronological perspective.

With the Centrum für reflektierte Textanalyse (CRETA) at Stuttgart University, we have commenced planning for studies of the imprint of Swedish authors in the German-speaking areas. We also hope to broaden the scope into studying Danish and Norwegian literature’s impact abroad through the above mentioned

collaborations.

Participants’ relevant competencies and merits

Mats Malm is a professor of Comparative Literature at the University of

Gothenburg, and the head of the Swedish Literature Bank. He founded and was the first head of the Centre for Digital Humanities at the University of

Gothenburg and Digital Humanities in the Nordic Countries.

Ljubica Miočević is a Ph.D. in Comparative Literature, specialized in 19th century Swedish literature, and editor at LB. She has also studied engineering physics at the Royal Institute of Technology in Stockholm.

Current employments

Mats Malm is a professor at the Dept. of Literature, History of Ideas and Religion at the University of Gothenburg.

(20)

Ljubica Miočević, Ph.D., is an editor at LB (80%) and at the Swedish Society for Belles-Lettres (20%).

Reception, maintenance and development in long term

LB: Data are stored and made available through http://litteraturbanken.se. Long- term storage is guaranteed through the Swedish Language Bank, University of Gothenburg. Data are stored redundantly on Swedish Language Bank servers and regular long-term backups are made and stored offline in a separate fireproof cell.

Backup is made every night and saved for at least three months. In addition, regular long-term backups are made and stored offline in a separate fireproof cell.

Recovery can be made by authorized systems administrators.

UUL: Data in Alvin are safely stored according to the Uppsala university principles for data security and is backed up and mirrored in a separate storage unit. See http://regler.uu.se/Detaljsida/?contentId=41950 (in Swedish only).

The data will be accessible through persistent identifiers and according to FAIR principles.

OCR will be of comparatively poor quality for the first decades of the material.

Still, it will be useful, and the images will be re-run when OCR technology has improved enough to provide better machine readability.

Plan for making infrastructure available to researchers and community

The material will be available through three already existing infrastructures: LB (litteraturbanken.se), Alvin (alvin-portal.org/) and, through LB, the Swedish Language Bank’s KORP (spraakbanken.gu.se/korp/). While Alvin focusses on book history and artefacts as such, LB also presents facsimiles of the books but focusses on OCR:ed texts for text mining of different kinds, and the Swedish Language Bank provides linguistic tools. Besides the possibility of simply finding and reading the texts, LB will offer a number of tools developed in collaboration with The Swedish Language Bank, Chalmers University of Technology, Yale University and University of California, Los Angeles. In addition, the texts will be made possible to download in different formats for anyone wishing to use other tools. Texts will also be distributed to international platforms for Scandinavian and European comparative research. The expected effect is that not only researchers, but also students at University and other levels will have substantial options for new kinds of studies.

All works will be catalogued in the national library database Libris, which makes them available to readers and libraries also in this way. LB is devised to invite the community by not only offering books, but also introductions and orientations of different kinds, for the users to find their ways to different materials. The

relevance of this project is not restricted to research: it also opens up cultural heritage to the public. All works will be available for anyone to read on the web site, and they will be easily searchable for specific, private research mapping, for instance, the fictional or real events of a particular place or town, person, family, name, etc. In this way, the material offers new approaches to the cultural heritage and society, adding to – for instance – the rapidly growing field of genealogical research. Through other funding, the Swedish Literature Bank will curate

(21)

selections of the material as epub files for the benefit of public libraries.

Financing from grant administrator

The grant will be administered by the Swedish Literature Bank, who finances the costs for

• Project management, coordination and collaborations at a minimum of 10% of full time for Malm and Miočevic.

• Indirect costs for editorial work in LB and development through CDHUGOT.

• Maintenance of the infrastructure, not included in budget.

• Additionally, the digitization and handling of 2 000 more books, part of the project but not included in this application.

(22)

Referenser / References

Anderson, Benedict (1983), Imagined Communities. Reflections on the Origin and Spread of Nationalism, London / New York.

Assmann, Aleida 2010), “Canon and Archive”, A Companion to Cultural Memory Studies, ed.

Astrid Erll and Ansgar Nünning, Berlin/New York. 97–107.

Assmann, Jan (1995), “Collective Memory and Cultural Identity” (1988) New German Critique 65.

125–133.

Bartscherer, Th. & R. Coover eds. (2010), Switching Codes. Thinking Through Digital Technology in the Humanities and the Arts, Chicago.

Blei, David M., “Topic Modeling and Digital Humanities”, Journal of Digital humanities 2:1, 2012.

Boyd-Graber, Jordan, Yuening Hu and David Mimno (2017), ”Applications of Topic Models”, Foundations and Trends in Information Retrieval 11: 2–3. 143–296.

Cameron, Fiona & Sarah Kenderdine eds. (2007), Theorizing Digital Cultural Heritage. A Critical Discourse, Cambridge, MA.

English, James F. and Ted Underwood (2016), “Shifting Scales: Between Literature and Social Science”, Modern Language Quarterly 77:3. 277–295.

Erll, Astrid (2011), Memory in Culture, transl. by Sara B. Young, Basingstoke.

Fairchild Ruggles, D. and Helaine Silverman (2009), “From Tangible to Intangible Heritage”, Intangible Heritage Embodied, Dordrecht etc. 1–14.

Jockers, Matthew (2013), Macroanalysis: Digital Methods and Literary History, Urbana.

Jockers, Matthew and David Mimno, “Significant Themes in 19th-Century Literature, Poetics, Dec 2013. 750–769.

Lachmann, Renate (2010), “Mnemonic and Intertextual Aspects of Literature”, A Companion to Cultural Memory Studies, red. Astrid Erll, Ansgar Nünning, Berlin/New York. 301–310. 

McCarty, Willard (2010), Text and Genre in Reconstruction. Effects of Digitization on Ideas, Behaviours, Products and Institutions, Cambridge.

Moretti, Franco (2013), Distant Reading, London.

Nowviskie, B. (2012). “What do girls dig?”, M. K. Gold (Ed.), Debates in the Digital Humanities, Minneapolis, MN: University of Minnesota Press.

Nünning, Vera & Nünning, Ansgar & Neumann, Birgit eds. (2010), Cultural Ways of Worldmaking.

Media and Narratives, Berlin.

Olsen, Mark, Russell Horton and Glenn Roe (2011), “Something Borrowed: Sequence Alignment and the Identification of Similar Passages in Large Text Collections”, Digital Studies / Le Champ numérique 2.1.

Rieder, Bernhard and Theo Röhle (2012), “Digital methods: Five Challenges”, Understanding Digital Humanities, ed. David M. Berry. 68–80.

Tangherlini, Timothy R. & Peter Leonard (2103), “Trawling in the Sea of the Great Unread: Sub- Corpus Topic Modeling and Humanities Research”, Poetics 41:6. 725–749.

Underwood, Ted (2013), Why Literary Periods Mattered: Historical Contrast and the Prestige of English Studies, Palo Alto.

Zhang, Y, A. Jatowt, S. S. Bhowmick, and K. Tanaka. “The past is not a foreign country: Detecting semantically similar terms across time”, IEEE Transactions on Knowledge and Data Engineering, 2016.

(23)

Projektdeltagarnas publikationer / Project participant's publications

Mats Malm

Malm, Mats 2019, ”Digitaliserat litterärt kulturarv?”, Humanists and the Digital Toolbox. In honour of Christian-Emil Smith Ore, ed. Martin Doerr et al, Oslo, 77–87.

Leonard, Peter, Mats Malm 2016, ”Marknadens intertextualitet. Kulturarv och återbruk 1840–

1900”, Spänning och nyfikenhet. Festskrift till Johan Svedjedal, ed. Gunnel Furuland, Andreas Hedberg, Jerry Määttä, Petra Söderlund and Åsa Warnqvist, Hedemora, 28–36.

Kokkinakis, Dimitrios, Mats Malm 2015, ”Detecting Reuse of Biblical Quotes in Swedish 19th Century Fiction using Sequence Alignment”, Corpus-based Research in the Humanities workshop (CRH).

Kokkinakis, Dimitrios, Ann Ighe, Mats Malm 2015, ”Gender-Based Vocation Identification in Swedish 19th Century Prose Fiction using Linguistic Patterns, NER and CRF Learning”,

Proceedings of the Fourth Workshop on Computational Linguistics for Literature (Clfl). Co-located with the NAACL/HLT. Denver, Colorado, USA.

Malm, Mats 2015, ”Digitala texter och forskningsfrågor”, Kungl. Vitterhets Historie och Antikvitets Akademien Årsbok 2015, Stockholm, 95–106.

Ljubica Miočević

Miočević, Ljubica 2018. ”’Jag vill förtälja om...’ Tal och skrift i Nicander och Norlings spökhistorier”, i Böckerna i borgen. Ett halvsekel i Roggebiblioteket, Acta Bibliothecae regiae Stockholmiensis 90, Stockholm: Kungliga biblioteket, 85–100.

Miočević, Ljubica 2017. Fantasiens morgonrodnad. En studie i Clas Livijns romaner, diss.

Stockholm, Lund: ellerströms, 461 pp.

Miočević, Ljubica 2015. ”Perspektiv på Clas Livijns handskrifter. Något om tillkomsten av Spader Dame”, Kladd, utkast, avskrift. Studier av litterära tillkomstprocesser, red. Paula Henrikson & Jon Viklund, Skrifter utgivna av Avdelningen för litteratursociologi vid Uppsala universitet, 68, 2015, 149–179. Peer reviewed.

Miočević, Ljubica 2013. ”Bilder från Boo. Karl August Nicander – En Magus i Spådomskonsten”, Samlaren. Tidskrift för litteraturvetenskaplig forskning, 2012 (133), Uppsala: Svenska

litteratursällskapet, 166–222. Peer reviewed.

Miočević, Ljubica 2013. ”’What’s difference?’ On Language and Identity in the Writings of Aleksandar Hemon”, Languages of Exile. Migration and Multilingualism in Twentieth-Century Literature, eds Axel Englund & Anders Olsson, Exile Studies, 13, Oxford: Peter Lang, 55–79.