Embed Size (px)
Citation preview
Digitizando Literatura sobre Biodiversidad
(Contenido Técnico)
CONABIO, México
William Ulate, Director Técnico de BHL
17 Diciembre 2014
Digitization Workflow
Insert Smithsonian
Macaw software here
Hardware & Software
Hardware
Usando una Estación Scribe
Escaneo por Internet Archive
Northeast Regional Scanning Facility (Boston)
New Jersey Facility
Natural History Museum, London
Fedscan (Library of Congress)
Internet Archive (San Francisco)
Smithsonian Libraries
Missouri Botanical Garden (Non-Scribe operation)
Hardware & Software
Hardware
Usando una Estación Scribe
“Off-the-shelf” escaners o cámaras digitales de buena calidad
Software
Wonderfetch -> Partner Meta App
(si usan máquinas Scribe)
identifier
search_id
title
volume
creator
date
call_number
language
subject
publisher
description
page-progression
possible-copyright-
status
licenseurl
rights
duediligence
rules_scanning
rules_republishing
Hardware & Software
Hardware
Usando una Estación Scribe
“Off-the-shelf” escaners o cámaras digitales de buena calidad
Software
Wonderfetch -> Partner Meta App (when using Scribe machines)
Macaw
Software de Escaneo: Macaw
Hardware & Software
Hardware
Usando una Estación Scribe
“Off-the-shelf” escaners o cámaras digitales de buena calidad
Software
Wonderfetch -> Partner Meta App (when using Scribe machines)
Macaw
Uploading directly to Internet Archive (for example: MBG‟s Botanicus http://www.botanicus.org/)
Standards and formats to consider
The simplest way to contribute a text item to IA is currently as a single pdf file. IA creates a second pdf with a text layer, if none exist.
Items can be submitted as a stack of image files, one image per page. The files can be in JPEG2000, JPG, or TIFF format, but with strict requirements for how the files in an image stack are to be named, and the stack needs to be packed into a single .zip or .tar file before submission.
When IA (Archive.org) scans a book for a Contributing Library, they use the custom-engineered "Scribe" workstation, but for many materials, adequate images can be made with off-the-shelf scanners or good-quality digital cameras.
For best results, it is recommended to use the highest resolution your device is capable of. Most images IA processes were produced at a resolution of 300-600 ppi.
Standards and formats to consider
BHL recommends following, in part, the DLF's "Benchmark for Faithful Digital Reproductions of Monographs and Serials" (available online at http://www.diglib.org/standards/bmarkfin.htm).
Bitonal: 600 dpi, 1-bit or bitonal TIFF images
Grayscale: 300 dpi, 8-bit grayscale uncompressed TIFF, or lossless compressed image (e.g. LZW, JPEG2000 [*.jp2]).
Color: 300 dpi, 24-bit color uncompressed TIFF, or lossless compressed images (e.g. LZW, JPEG2000 [*.jp2]).
NOTE: the above specifications are the preferred ones. BHL will, however, accept lossy files. In the case of JPEG2000, files with a compression level of 85% are acceptable.
Standards and formats to consider
Currently, BHL data can be downloaded as MODS, EndNote and BibTex. See our wiki page with more information: http://biodivlib.wikispaces.com/Data+Exports#x--MODS
Title metadata as well as pagination, descriptive and page order (structural) metadata is being copied into METS files in the <biodiveristy> collection at IA.
The purpose of these METS files is to accommodate the need of our pagination data.
These METS files are pagination specific and they do not have the item/volume information included.
If bibliographic metadata for BHL content was required, it should be found in the MODS files on the Data Exports page.
Standards and formats to consider
For the future, we are looking at serving OLEF as an envelope format to share information with other BHL Nodes.
See
http://www.bhle.eu/bhl-schema/v0.3/ and
http://www.slideshare.net/HeimoRainer/bhleuropemetadataharmonisationtdwg20111018kollerwhrainer/6 )
Metadata generation and indexing strategy
Each item to be uploaded needs a unique identifier within our central repository, currently Internet Archive (archive.org) and a folder with such name is created to hold the uploaded and generated (derivative) files.
Within BHL we record metadata at 3 levels of bibliographic granularity – Title, Item & Page – as well as metadata for the Creator(s) of the title.
Metadata generation and indexing strategy
Scanned material (jp2.zip) and basic title-level metadata content (marc.xml), item-level metadata (meta.xml) and page-level metadata (scandata.xml) are uploaded to Internet Archive (IA), in the „biodiversity‟ collection.
JP2.zip: The compressed JP2 images (Compression Quality 15) that IA will use for delivering pages to the Read Online feature following a very specific naming convention for the filenames: Master images files named with local library identifier + 4-digit sequence number (with no gaps).
MARC.xml: The MARC record for the title from the library catalog in MARCXML format
Title, *Abbreviation, *Creator, Description, Publisher, Start Date Published, End Date Published, Local Library Identifier, *OCLC Number, *ISSN, *ISBN, *Call Number, *Subject, *Language, Date Created, Date Last Modified, *Foreign Keys
Metadata generation and indexing strategy
META.xml: The item level information (even redundant with the title-level information) including the title, author, publisher, copyright information, digitizing sponsor, date published, type of item, and who originally uploaded it. IA may also update this XML file with information as it processes the pages of the item.
Barcode, Sequence, Local Library Identifier, +Start Volume, End Volume, +Start Date, End Date, *Language, Scanning Institution, *Scanning Contributor, *Scanning Sponsor, Date Created, Date Last Modified
SCANDATA.xml: An XML file (scandata.xml) recording information about each page image (handSide, cropBox, original width & height, etc. )
FileName, Sequence, *Page number, *Page Type, Year, Volume, IssuePrefix, Issue, Date Created, Date Last Modified
Metadata generation and indexing strategy
CREATOR: A “Creator” is defined as a person or company responsible for the creation of the Title.
Name, *Role, Date of Birth, Date of Death, Biography
A detailed description of the contents of each one of these files and the whole process of Uploading content to IA is available at: http://biodivlib.wikispaces.com/Upload
Metadata generation and indexing strategy
Internet Archive runs the OCR process and generates “derivative files” that include:
The resulting files of the OCR process with ABBYY FineReader (djvu, djvu.txt, djvu.xml, abby.gz)
A 100x152 pixel GIF with a looping, animated thumbnail of the first 20 pages of a book.
The presentation version on BHL in PDF format.
The MARC record in binary and XML formats.
And others ( for a more detailed description you can see http://biodivlib.wikispaces.com/Download+All+File+Types+and+Descriptions )
Metadata generation and indexing strategy
The metadata from new items included in the BHL collection is included in the database and indexed to be used in searches through the Portal and API services.
Periodically, the OCR pages are ran through taxonomic names services to mine for new taxa names like TaxonFinder (ubio.org) or GNRDS (Global Names resolution tools and services: resolver.globalnames.org) soon.
Taxa names are added to the database and written back into Internet Archive (names.xml)
Online Platform
Capture System
Scribe machines
Macaw
Publication
BHL Portal
BookViewer
PDF Generator
Online Platform
Publication
BHL API (biodivlib.wikispaces.com/Developer+Tools+and+API)
The BHL Application Programming Interface (API) is a set of REST-like web services that can be invoked via HTTP queries (GET/POST requests) or SOAP.
Responses can be received in one of three formats: JSON, XML, or XML wrapped in a SOAP envelope.
We are currently developing a new API v3, closer to a RESTful design than previous versions, using resource-centric URLs (where possible) and GET/PUT/POST/DELETE verbs.
Online Platform
Publication
Data Exports (biodivlib.wikispaces.com/Data+Exports)
Online Platform
Management
BHL Admin Dashboard
Admin Functions (Alert Message, Image Server, Collections, Institutions, Languages, Page Types, PDF Requests, Segment Types)
Library Functions (Titles/Items/Segments /Pagination/Authors)
Science Functions (Names (Taxa) on a Page)
Library Statistics (Titles/Items/Pages/Names/Segments/Items with Segments, Names, Pages with Names)
Growth Statistics (Titles/Items/Pages/Names/Segments new this Month/Year)
Online Platform
Management
BHL Admin Dashboard
PDF Generation Statistics (Generated: 174,162)
Internet Archive Harvesting Statistics (Complete: 119,125 items)
BioStor Harvest Statistics (Published: 11,126 as of Aug. 29, 2013)
DOI Assignment Statistics (DOI Approved: 57,338 as of Aug 29, 2013)
Web Traffic Statistics (API v2, OpenURL)
Reports (Item Pagination, Title Import History, Character Encoding Problems, DOIs by Institution, Monographic Contributions, Items by Contributor)
Deduplication
• We try to avoid duplication where possible
• Tools
• Serials = Scanlist
• Monographs = Monographic deduper
• Check the BHL before you send for scanning
• We do our best but duplication happens
• Post-digitization, we merge titles as necessary
Online Platform
Management
Monographic Deduping Tool
The MBLWHOI Library has been working on a tool that assists with de-duplicating the monographs that BHL members are sending to IA for scanning.
The application is ready for use and it‟s entirely web-based, requiring no client or user configuration.
The monographic deduper acts as a master database that contains records for all of the monographs that any BHL partner institution has scanned.
Online Platform
Management
Monographic Deduping Tool
In addition, there is a process also in place that allows for material ingested from the Internet Archive, but not contributed by a BHLpartner institution, to be added to the deduper database.
Ultimately, the Monographic deduper database should be seen as living record of accountability that communicates to staff collaborating in the BHL network, a partner‟s promise to digitize a particular monographic title.
Online Platform
Management
Serials Bid List
It is a catalogue that allows users to browse and search Serials titles held by BHL member institutions using advanced filtering.
Technical Group at MBG
Mike Lichtenberg Developer
Trish Rose-Sandler Data Analyst
William Ulate Technical Director
Technical Support
MBG IT Division
Manage servers, systems and telecommunications.
Installs software needed
And others:
MBL
Smithsonian
Internet Archive
BHL-Australia
BHL-Europe
Technical Advisory Group
Fire
wa
ll
Images (JP2) PDF Coordinate-based OCR XML metadata
BHL Architecture: Window Seat Ed.
BHL DB
Internet Archive
Storage
Logic
APIs UI Data
Exports
Access
Data Transform Utilities
Geocoding
Name Finding
Projects
Global Names
Art of Life
Purposeful Gaming
Digging into Data
Scientific Name Extraction
TaxonFinder algorithm in production since 2008
More than 100 million candidate name strings
More than 1.5 million unique, verified names
Available through UI, APIs, Data Exports & Internet Archive
New collaboration with Global Names project
Improved algorithm, better precision & recall
More data with TaxonFinder and Neti Neti!
http://gnrd.globalnames.org/
Taxon Names
BEFORE
Name Instances 101,591,803 101,288,804
Unique Names 7,498,554 7,464,924
Verified Names 1,905,507 1,902,803
EOL Names 63,130,350 62,963,582
EOL Pages 13,579,868 13,532,684
AFTER
Name Instances 151,222,182 150,066,425
Unique Names 29,246,382 29,091,767
Verified Names 10,153,165 10,109,540
EOL Names 87,791,695 87,135,089
EOL Pages 15,466,713 15,342,867
Article-level metadata
Chapter-level metadata
Treatment-level metadata
Part-level metadata
Articles in the BHL UI
See also:
Related Titles
Digitization workflow
1. Titles vs. Items vs. Segments
2. Metadata we need: • MARC for book and journal titles • Volume information • Page data
BHL Term Titles Items Segments
Library Term Book or Journal
Titles
Volume, Piece Articles, Book
chapters,
Meaning Conceptual unit Object Section of
consecutive pages
Art of Life
Art of Life
Art of Life
Art of Life
Art of Life
Art of Life
Macaw
https://github.com/cajunjoel/macaw-book-metadata-tool
Reviewing Metadata
Reviewing Metadata
Manually built:
1,714 sets
89,457 images
Purposeful Gaming
*E.xvi�c�piteI von c. cXx.WptdvonfnrWmn bu�fbe;bcn.5 am cix bIa � S &3rn~ 41X a�m cv(f b1air�'o�et ert oiensr �; �', :�hlrfc�c wa ff�4am.diug bist a 6aiw~s ff oJrJtwt nof bL4ecImt& blfafra mem b t wag `wr 4 cn wiu 4 e8t5m.ed bvUratflb ck wuo, ma144'*4I bttE5rmbebt =rt3'kn am4ra tif vrmr Waff C * t6rmnli an `tn�ciblatGteaM w ?ffoaifrn w4wmeu nu weib e , wpiteI voE5teiri ct c ober gtUcr cit cm` 91 cLi biar J ' >bSciatl�Oiff ;Bruet wacfttc n qmcx b1a bl: bt5c lttmtt bb9 lkr w.llr#e iti ncn xoa ff cu :r trtuft *e t � B Rn "� trv W1Rt' ?Cm c blas waIwutr Ober �ci ti 1V Ces ' wt gbtiemwwajfu tpctt, afferain 9 c: b�titbfof �r f eran m rs bra wlg auig4;f aer�m *mc vrt blatcabtfm wfru an'deg~m rt blas Iaum bwWt� run f ncmai b14ianf tJobrrfan ebrut4net vnber Brwt Ober awawi*m.crriii btafwfm uww c on$ 'it ttu wttkc 5,10 $ m~C fca trc* cx u W�e�&mcyfbq4 Mabtt mmw rc a iiu bc Jcn ncI.end.*, blat s. a\ u:�rprd3 rw4ftf wm c ii,+ ttCC tn wa frr9fr orfab fcfbt enb c optiti bt -r9 ceDa ttDcn i34M sn Sem i
OCR Improvements
Gaming
Transcription
OCR Improvements
Transcription
Purposeful Gaming
Looking at…
Crowdsourcing Markup & Annotation
Purposeful Gaming DIGITALKOOT
Joint project run by the National Library of Finland and Microtask to index the library's enormous archives so that they are searchable on the Internet for easier access to the Finnish cultural heritage.
.
Purposeful Gaming DIGITALKOOT
Launched on Feb 8 2011, nearly 110 000 participants completed over 8 million word fixing tasks by Nov 29 2012
DigiTalkoot enabled volunteers to participate in this fixing work by playing games.
.
Purposeful gaming and BHL: engaging the public in improving and
enhancing access to digital texts
IMLS Grant Program: National Leadership Grants for Libraries
Partners:
Missouri Botanical Garden
Harvard University
Cornell University
New York Botanical Garden
P.I.: Trish Rose-Sandler, Missouri Botanical Garden
Dates: Dec 2013 – Nov. 2015
Project objectives and benefits
Test new means of crowdsourcing to support the enhancement of content in BHL
Demonstrate if digital games are an effective tool for analyzing and improving digital outputs from OCR and transcription
Benefits of gaming include:
improved access to content by providing richer and more accurate data;
an extension of limited staff resources; and
exposure of library content to communities who may not know about the collections otherwise.
OCR Improvements
German text interpreted by the OCR process as:
“unb auf ben ©elnrgen be6 fublic{)en”
AOCR Improvements
Different resulting texts from parsing the phrase: “und auf den Gebirgen des südlichen Deutschlands”
(“and on the mountains of southern Germany”)
IA OCR OCR 2 Transcriptio
n 1
Transcriptio
n 2
1 unb und und und Ok
2 den ben den den Ok
3 ©elnrgen ©ebirgen Bebirgen Gebirgen X
4 be6 des de5 des Chk
5 fublic{)en fublichen Füdlichen Südlichen X
6 £)eittfc{)(anb6 Deutfchlanbs Deutfchlands Deutschland
s X
Purposeful Gaming
iDigBio‟s aOCR Hackathon
Improve OCR parsing of labels with clear metrics (datasets, output formats, scoring algorithm)
Libraries of regular expr. to clean up each field (different error correction for latitude/longitude coordinates than personal names or herbarium catalog numbers)
Tool for classifying segments of the image before submitting to OCR
Do a first pass of OCR to clean images before sending them to a second, 'real' pass of OCR
iDigBio‟s CITScribe Hackathon
1. Interoperability betweenpublic participation tools and biodiversity data systems,
2. Transcription quality assessment/quality control (QA/QC) and the reconciliation of replicatetranscriptions,
3. Integration of optical character recognition (OCR) into thetranscription workflow
4. User engagement
NfN & iDigBio‟s CITScribe Hackathon
Jason Best‟s DarwinScore
Ben Brumfield‟s Handwriting Gibberish Detector
Dictionaries to improve crowdsourcing consensus (e.g., names of collectors, scientific names)
Word Clouds created using n-gram scoring, faceting, and Solr for indexing + Carrot2 for specimen selection (visualize and explore of the use with a word of interest from the word cloud) and a data cleaning step (highlight infrequent words by the system).
NESCent EOL-BHL Research Sprint
There is no place like home: Defining “habitat” for biodiversity science
Robert D. Stevenson
UMass Boston, Dept. of Biology, 100 Morrissey Blvd., Boston, MA 02125-3393
Carl Nordman (Natureserve) and
Evangelos Pafilis
Hellenic Centre for Marine Research, P.O. Box 2214, Heraklion, 71003, Crete, Greece
NESCent EOL-BHL Research Sprint
Assessing Risk Status of Mexican Amphibians Through Data Mining.
Esther Quintero and Bárbara Ayala
National Commission for Knowledge and Use of Biodiversity (CONABIO)
and
Anne Thessen
Marine Biological Laboratory and Arizona State University
Planning for global change: using species interactions in conservation
Nicole F. Angeli, Emma P. Gomez, Margot A. Wood,
Applied Biodiversity Sciences Program, Texas A&M University, College Station, Texas
Tweet me @auratus_nicole
and
Javier Otegui
University of Colorado-Boulder
There is no place like home: Defining “habitat” for biodiversity science
Robert D. Stevenson
UMass Boston, Dept. of Biology, 100 Morrissey Blvd., Boston, MA 02125-3393
Carl Nordman (Natureserve)
Evangelos Pafilis
Hellenic Centre for Marine Research, P.O. Box 2214, Heraklion, 71003, Crete, Greece
http://epafilis.info/ , [email protected]
Evolution in the usage of anatomical concepts in the biodiversity literature
Todd Vision ([email protected]), Prashanti Manda ([email protected]), and Dongye Meng ([email protected])
University of North Carolina at Chapel Hill
NESCent EOL-BHL Research Sprint
Evolution in the usage of anatomical concepts in the biodiversity literature
Todd Vision ([email protected]),
Prashanti Manda ([email protected]), and
Dongye Meng
University of North Carolina at Chapel Hill
Some preliminary observations…
Our API seemed to work fine
Access via a taxon (or a group), for example:
“I want to harvest all pages with names from this taxon (Chordata) or this common name (Vertebrate)”.
Groups started getting results after 2.5 days.
The structure of BHL was explained so researchers could understand the title, item, page and part levels and define what they wanted. Ex: one group was looking for terms in the titles and the parts‟ titles.
Some others said they would Harvest the OCR from IA although they will not be able to harvest the text on a page by page granularity (only item level).
NESCent EOL-BHL Research Sprint
There is no place like home: Defining “habitat” for biodiversity science
Robert D. Stevenson
UMass Boston, Dept. of Biology, 100 Morrissey Blvd., Boston, MA 02125-3393
Carl Nordman (Natureserve) and
Evangelos Pafilis
Hellenic Centre for Marine Research, P.O. Box 2214, Heraklion, 71003, Crete, Greece
Mining Biodiversity
Mining Biodiversity
Mining Biodiversity: Enriching Biodiversity Heritage with Text Mining and Social Media
One of the international projects that won in the third round of the 2013 Digging Into Data Challenge
Promote the development of innovative computational techniques to apply into big data in the humanities and social sciences
The National Centre for Text Mining (UK)
Missouri Botanical Garden (US)
Dalhousie University's Big Data Analytics Institute (Canada)
Social Media Lab (Canada)
MiBIO: Mining Biodiversity
1. Automatic error correction of OCR text errors.
2. Crowdsource annotation of legacy texts with semantic metadata.
3. Adapt text mining techniques to extract terminology, entities and significant events automatically and to track terminology evolution over time.
4. Use Interactive visualization techniques to help users manage search results through next generation browsing capabilities, assisted by a semantic similarity network of important terms and entities.
5. Design of a social media layer, serving as an environment for diverse users to interact and collaborate on science, public education, awareness and outreach.
MiBIO: Mining Biodiversity
Crowdsource Markup
Display text Species Profile Model category
General/summary TaxonBiology
Geographic range Distribution
Habitat Habitat
Food sources and feeding behavior TrophicStrategy
Physical description (general) Description
Physical description (detailed
morphology)
DiagnosticDescription
Visit to NaCTeM, Feb. 17, 2014
NaCTeM‟s Biodiversity- relevant tools
ANNNOTATION PLATFORM
Remote Processing Workflows processed on remote machines. No attendance needed
Workflows GUI for creating single-flow and multi-branch workflows
Workflow Designer
User Interaction Annotation Editor allows for making changes while processing Annotator/Curator
Web S
erv
ice
Third-party applications
Processing Components Data (de)serialisation, search engines, NLP, NER, etc.
Developers
Workflows view
Processes View
Documents view
Workflow editor
Workflow as a Web service
Workflow as a Web service
http://argo.nactem.ac.uk/test/services/webservice/314
INPUT
OUTPUT
NAMED ENTITY RECOGNISERS AND NORMALISERS
✔
✔ ✔
✔
✔
Automatically recognised named entities
Linking to external dictionaries
Species and habitat recognition
EVENT EXTRACTORS
Events: associations between entities
SEMANTIC SEARCH
TERM EXTRACTION
Ryerson University SocialLab‟s Netlytic.org
http://miningbiodiversity.com/ http://miningbiodiversity.org/
Thank you William Ulate
BHL Technical Director
Missouri Botanical Garden
Skype: william_ulate_r
Thank you!
And thanks to Bianca Crowley for the workflow slides
