Mark Zöpfgen: Software-Supported Bibliographic Recording and Linked Data

1

Lider Roadmapping Workshop

Deutsche Nationalbibliothek –Software-supported Bibliographic Recording and Linked Data

Mark Zöpfgen

Leipzig, 02.09.2014

Overview

- DNB – German National Library

- Activities in Content Extraction and Semantic Web

- MACS

- PETRUS

- Open Linked Data

- Motivation/Challenges

Lider-Roadmapping-Workshop | Leipzig | 02.09.20142

DNB – the German National Library

- central archival library and national bibliographic center for the Federal Republic of

Germany

- collect, permanently archive, comprehensively document and record

bibliographically all German and German-language publications since 1913, foreign

publications about Germany, translations of German works (German National

Bibliography)

- produces (in collaboration with other institutions the Integrated Authority File

(GND, “Gemeinsame Normdatei”)

- makes them available to the public

- develops and maintains bibliographic rules and standards for Germany

- plays a significant role in the development of international library standards.

Inventory: ~ 27,8 bibliographical units; ~ 719000 online - publications (mainly pdf

and epub)



Location in Leipzig

Location in Frankfurt

Activities in the Context of Content extractionand Semantic Web:

Contentus

CrissCross

Culturegraph

Linked Data Service

MACS

Unidissen

VIAF

PETRUS

…

For more information see:

http://www.dnb.de/DE/Wir/Projekte/projekte_node.html


MACS – Multilingual Access to Subjects I/IV

- Creation of a multilingual retrieval-vocabulary for research in bibliograpic databases.

- Links between Subject Headings of LCSH (Library of Congress Subject Headings),

RAMEAU (Répertoire d'autorité-matière encyclopédique et alphabétique unifié)

and GND (Gemeinsame Normdatei)

- In cooperation with SNB (Swiss National Library)

- Currently ~ 63000 Links wich have been imported to the GND-records

Use CasesMake data of DNB internationally available (search via LCSH/RAMEAU-subject headings)

Search in the Library of Congress /Bibliothèque de France with GND-subject headings

Possibility to overtake subject headings from bibliographical records (e.g. in case of

translations)

- Link: http://www.dnb.de/DE/Wir/Kooperation/MACS/macs.html



MACS – Multilingual Access to Subjects II/IV

– Maintenance: The links are created/updated using the LMI (Link Management

Interface). The LMI provides a web-interface, data is stored in a central

database.

– Data Export / Import: The links are exported via OAI-Interface. The import to

the CBS (Central Bibliographic Database) is currently done by script (manually

initiated)

– Planned:

Integration in the search-portal of TEL (The European Library)

Provision via linked data service (actually not integrated)

Regular update between LMI and CBS


MACS – Multilingual Access to Subjects III/IV

9

MACS – Multilingual Access to Subjects IV/IV


Petrus – Software-supported BibliographicalRecording I/V

Why software supported?

Growing number of online publications (see graphic below).

The German National Library is looking to reduce its traditional indexing operations

in areas which are no longer feasible due to the continually growing number of

publications, or are no longer necessary because of technological developments.


13525 17651

29823

112766

0

20000

40000

60000

80000

100000

120000

2007 2008 2009 2010


Petrus – Software-supported BibliographicalRecording II/V

ClassificationBased upon the DNB-”Sachgruppen” ~ first two layers of the DDC

Statistical procedure, training corpus ~ 300.000 objects with known classes (full

text and tables of content). The objects are limited to 40.000 characters.

After stemming, the data model is generated. As classifier, SVM (scalable vector

machine) is used. After the creation of the model, a 3-fold validation is executed, in

order to verify the quality.

The model can be transferred to an “endpoint”, which is a stand-alone application.

The endpoint communicates via web service-interface.

In use since January 2012; currently ~ 400 objects/day


Petrus – Software-supported BibliographicalRecording III/V

Keywording

Linguistic text analysis: language recognition, identification of sentences, words,

phrases etc.

Term matching with a dictionary which is based on the integrated authority file

(72000 subject headings), Disambiguation

Term ranking (dependant on position and frequency)

The keywording process can eventually be transferred to an “endpoint” (according

to the classification modell)

~ 80 objects/day


(1) List of publications to beprocessed

(2) Metadata to be imported out ofthe biblographic database

(3) (Full-text) objects to be importedout of the repository

(4) Transfer via a webserviceinterface

(5) Return of results(6) Storage of the results in the

bibliographic data base

Petrus – Software-supported BibliographicalRecording IV/V

Petrus – Software-supported BibliographicalRecording V/V


Appearance in the biblio-graphic record

Return of the classification software


Open Linked Data I/III

- DNB provides high quality, mainly intellectually created data.

- Authority file (GND) and National Bibliography are available in rdf-format

- Data is published under the Creative Commons Zero-License

- Currently, the data can be accessed via the Portal (for single records) or

downloaded

http://datendienst.dnb.de/cgi-bin/mabit.pl?userID=opendata&pass=opendata&cmd=login

- Target groups are research facilities and non-profit organisations as well as

commercial service suppliers (e.g. search engines, knowledge management

systems)

– Bibliographic data is highly reliable , but has a poor formal quality

(free-text fields) - High efforts for conversion

– The data was converted using Metafacture, which had been developed by

culturegraph.org. (www.culturegraph.org)


Open Linked Data II/III


Open Linked Data III/III

Bibliographic record of „Winnetou“ leads to Karl May - detail

… leads to place of birth

Coordinates

Motivation of the DNB

- Motivate external parties to work with rdf-data, e.g. linking it with other

ontologies.

- Improve search: Access by themes, browsing, unveiling relations between cultural

entities.

Technical Challenges

- Improve the accessibilty (e.g. by services - MACS)

- Search: Integrate Portal (knowledge representation, user interaction) with search

engine and linked data

–Lider-Roadmapping-Workshop | Leipzig | 02.09.201418

Questions?


Mark ZöpfgenDeutsche NationalbibliothekInformationstechnikAdickesallee 1D-60322 Frankfurt am MainTelefon: +49-69-1525-1705mailto: [email protected]://www.d-nb.de

Mark Zöpfgen: Software-Supported Bibliographic Recording and Linked Data

Data & Analytics

workshop leipzig

macs multilingual access

subjects ivivliderroadmapping

snb swiss national library

germanlanguage publications

national bibliographic

semantic web macs

data export import