20090921 art databanken agosti final

Literature and XML:or How to Have More Time to

Think

Donat AgostiPlazi

ArtDataBanken Stockholm, Sept 21, 2009

Who is this?What do I know about her?

Where does she live?

Who are you?What do you do?

Where are you from?

The answers are inseveral hundred million pages of printed

species descriptions in our libraries, including the descriptions redescriptions of

an estimated 1.8M species, and an estimated 50K new

(re-)descriptions annually.

Taxonomists at work ……

T. E. Lawrence: Seven Pillars of Wisdom – a triumph. 1st published for general circulation, 1935: p. 535

The traditional flux of information …

…a more or less closed, intransient system

What has this to do with XML,semantic, enhanced documents?

Access

Scanning

Pdf-conversion

(WWW)

Before antbase.org, Harvard‘s Museum of Comparative Zoology could claim to be the only location with a complete set of ant systematics publications from 1758 - present.

Through antbase.org‘s digital library, access to this body of literature is worldwide, and it is actively used (>10,000 visits in one month only).

The Biodiversity Heritage Library is currently digitizing and make accessible >100 million pages, most of them out of copyright, ie older then 1925. ........ to be finished in 2048...

Access to ant taxonomic publications through antbase.org /Smithsonian Institution, including currently the entire body of non-copyrighted publications since 1758 (>4,000 publications or 85,000 pages)

Can taxonomic work be copyrighted?

Copyright legislation is national but is based on the Berne Convention for the Protection of Literary and Artistic Works which defines a minimal standard. This international copyright standard does not require the recognition of treatments, the building stones of taxonomic publications, as works.

“work” does not mean “text”, does not mean “data”, does not mean “information”. A “work” is something more. That kind of something more has many different definitions in the various legislations, but it is always there: It may be called originality, individuality, creation, personal expression, creative shaping or anyhow else, but it is a condition for qualifying a product as a work: “Work” is an intellectual product that is in a certain sense particular, individual, original, new. (Egloff: EDIT IPR and Copyright, 2008)

Taxonomic treatments are highly structured and homogenous, part of a global >100 million page corpus growing at a rate of ca 20,000 new species descriptions per year, not counting 5 times more redescriptions. Its structure is tightly controlled by a peer review process enforcing standards, a domain specific vocabulary, not written as poem or in flowery language but scientific jargon.

Treatments do not qualify as work.

The publications including the treatments might.(Egloff: EDIT IPR and Copyright, 2008)

It is about digesting millions of pages:

>>100 M pages taxonomic literature

25M scientific publications / year25K journals

>1K with taxonomic descriptions

20K descriptions of new species / year

Is this is the access we need?!

No, we need open access to content, not the PDF per se.

It is about machines(not we) doing a great deal of the

work for us, extracting data, formulating hypothesis, ....

It is about data and information in context

„Nothing makes sense in biology except in the light of treatments“.

An example from the Neurocommons text mining pilot:

• PubMed abstracts: > 16,000,000• CNS classified abstracts: 874,727• text mining recognized: 368,688• text mining processed: 94,381

• extracted graph of 30,000+ relationships and 5,500 genes and proteins “protein-protein

interaction networks” John Wilbanks, Neurocommons

In a semantic Web environment (where machines talk to each other and do most of our work), data need to be able to talk to each other:

27,266 papers

4,563 papers41,985 papers

10,365 papers

128,437 papers

“protein-protein interaction networks” John Wilbanks, Neurocommons

Relational to Ontological Mapping

Drug

Neuron

PathologicalAgent

Receptor

Channel

inhibitsinhibits

Agent

NeuronalProperty

PathologicalChange

involvesinvolves inhibits

Compartment

has

is_located_in

is_located_in

slide courtesy of kei chung, yale

It will open up scientific literature for data mining

“protein-protein interaction networks” John Wilbanks, Neurocommons

TREATMENT Cremastogaster mimosae  Likely Diagnostically Related to: Cremastogaster tricolor  Likely Diagnostically Related to: Cremastogaster tricolor  Likely Diagnostically Related to: Cremastogaster amabilis  Likely Diagnostically Related to: Cremastogaster tricolor  Likely Diagnostically Related to: Cremastogaster amabilis  Associated with: Acacia sienocarpa Living in: Mombasa Lviing in: Tanga

It is more: it is about access to the original or

source data

The semantically enhanced treatments, extracted, stored on Plazi.org, and served in a human readable form, are linked to the underlying data: Fisher & Smith, 2008, PLoS ONE.

Semantic, enhanced treatments do the job ...

... and XML is one way to go.

XML

XML stands for EXtensible Markup Language

XML is a markup language much like HTML

XML was designed to carry data, not to display data

XML tags are not predefined. You must define your own tags

(schema)

XML is designed to be self-descriptive

XML is a W3C Recommendation

XML

Being open and non-proprietary XML is an optimal archival

format for the treatment/publication

Being a stable and rich data format, XML can be repurposed for

a variety of purposes

XML

XML application design is an art in itself .... and thus can not be

explained in 15 minutes

Plenty of resources to dive into XML on Web, eg

http://www.w3schools.com/, etc.

This means to develop a schema that models the logic content (e.g TaxonX), insert those tags that define what a word means, so a computer can understand as well. To assure, that everybody talks about the same species, the name can be linked to a reference name server

Azteca instabilisTaxonx-schema

Would then read like External schema

<tax:name><tax:xmldata> Normalization of data <dc:Genus>Azteca</dc:Genus> <dc:Species>instabilis</dc:Species> </tax:xmldata>

Azteca instabilis </tax:name>

This can also be applied to entire sections of text, such as the treatment of a species and its parts.

<tax:treatment> <tax:nomenclature> <tax:name> <tax:xid source="HNS" identifier="193329"/> <tax:xmldata> <dc:Genus>Mystrium</dc:Genus> <dc:Species>leonie</dc:Species> </tax:xmldata> Mystrium leonie </tax:name> <tax:status>n. sp.</tax:status> Fig 1 D - F </tax:nomenclature> <tax:div type="description"> <tax:p>HOLOTYPE WORKER: TL 3.95, HL 1.02, HW 0.95, CI 93, SL 1.30, SI 137, PW 0.73, ML 0.38. Mandible outer margin strongly curving to a sharp apical tooth, the apex parallel to the anterior clypeal margin. (Holotype with material in mandibles, so mandibles and anterior clypeus $ described below from paratypes.) Median clypeus....</treatment>

global unique identifiers (e.g. LSID) link up data

LSID for scientific publicationsLSID for treatmentsLSID for names (Zoobank/ HNS..)LSID for specimensLSID for DNA sequences / characters (ontologies)LSID for repositoriesGPS fixes for locations

Azteca instabilis

Would then read like

<tax:name><tax:xid source=“LSID" identifier=“urn:lsid:biosci.ohio-state.edu.osuc_concetps:13452"/> Link to external database <tax:xmldata> Normalization of data <dc:Genus>Azteca</dc:Genus> <dc:Species>instabilis</dc:Species> </tax:xmldata>

Azteca instabilis </tax:name>

We need XML-schemas, tools to convert and expose

semantically enhanced documents.

Plazi workflow: overviewPlazi deliverables

TaxonX XML schema

GoldenGate

Dspace application

Exist application

SRS

Exchange protocols (SPM, TAPIR, REST)

- Get LSID from Hymenoptera Name Server for names; ZooBank?-Add new names

- Get bibliographic Metadata from HNS (MODS)

- Get bibliographic Guids from bioguid (or EDIT?)

- Get geographic long/lat from geonames.org

Plazi workflow: GoldenGate mark up as an example

-Get Guids for - CBOL- NCBI- specimen- images- .....

Plazi Search and Retrieval Server: Access to data

TAPIR, SPM

You

You

You

human

machine

Materials examined from literature in GBIF

Plazi workflow: content

11,000 descriptions online500 publications4,500 publications

Handle, SPM and Tapir servicesFeeds into HNS and Zoobank (soon)Is harvested by GBIF, EOLSupport from GBIF, EOL, US-NSF, DFG

Does the retro mark-up process scale up to the millions of pages needed to be processed?

Only partially: Mark up takes about 5min/page: For 100 M pages = 700 man years (but it is only a first tool...)

Does the mark-up process scale up to the millions of page needed to be processed?

Only partially: Mark up takes about 5min/page: For 100 M pages = 700 man years (but it is only a first tool...); wizards can reduce the time by several factors

But: How much does it cost to digitize specimens, and what is its quality?

The cost of converting legacy publications can be avoided by

producing marked-up publications up-front

NLM/TaxonX schema allows publishers to maintain richly encoded articles whose data can be distributed and presented in multiple formats for a

variety of uses.

NLM/Taxonx XML Document

Print

NLM/Taxonx XML Document

PDF

Print

NLM/Taxonx XML Document

HTML

PDF

Print

NLM/Taxonx XML Document

HTML

SPM /RDF

PDF

Print

SPM /RDF

SPM /RDF

NLM/Taxonx XML Document

HTML

SPM /RDF

PDF

Print

Database

HTML /Species Page

HTML

SPM /RDF

SPM /RDF

HTML /Species Page

HTML /Species PageEg. EOL, scrathpads

NLM/Taxonx XML Document

HTML

SPM /RDF

PDF

Print

Database

HTML /Species Page

LSID resolver

HTML

SPM /RDF

SPM /RDF

HTML /Species Page

HTML /Species PageEg. EOL, scrathpads

NLM/Taxonx XML Document

HTML

SPM /RDF

PDF

Print

Database

HTML /Species Page

Google

Dataminig, ...

LSID resolver

HTML

SPM /RDF

SPM /RDF

HTML /Species Page

HTML /Species PageEg. EOL, scrathpads

Semi-automatically generated semantic, enhanced

e-publications are the only way to describe the missing 10 M species, and

to deal with an increasing flood of data.

ms submission(„Taxon-x-version“)

new ms alertPosting for review

Edited ms

Revised msPublication: pdf

Publication: hard copy

Publication database(„taxon-x-version“)

ontology

bibliography

analysis & ms preparation

ZooBank / NS

Character DB

Specimen DB

Description DB

Distribution DB

Char. Matrix DB

Phyl. Tree DB

Char-state Im.

Specimen Im.

Habitat Image

Leg. Publicat.

Tax

on D

B

New Data

feedback

Accepted ms

New taxon alert

The future of publications: The publication semiautomaticall generated

Word MS

DB

Input forms

export

export

convert NLM taxpubIndesign

NLM taxpub

author

author

author

publisher

publisher

publisher

Journal authoring and production workflow

Ctd.

NLM/Taxonx XML Document

HTML

SPM /RDF

PDF

Print

Database

HTML /Species Page

Google

Dataminig, ...

LSID resolver

HTML

SPM /RDF

SPM /RDF

HTML /Species Page

HTML /Species PageEg. EOL, scrathpads

Ctd.

Word MS

DB

Input forms

export

export

convert NLM taxpubIndesign

NLM taxpub

author

author

author

publisher

publisher

publisher

Journal authoring and production workflow:

What do we miss?

available

prototypes

to be developed

Where do we stand?

2008LSIDs, external links

Where do we stand?

2008LSIDs, external links, XML

Where do we stand?

2008

Where do we stand?

2009

LSIDs, external links, external data via doi, export services

Where do we stand?

2009LSIDs, external links

Recommendations:

Individual levelAssure that all you do is open access• Understand copyright – be not afraid of copyright• Self archive (the Green Road)• Create content for the Web

Self archive (the Green Road): UNIZ as one of the global leaders in self archiving

Recommendations:

Individual levelAssure that all you do is open access• Understand copyright – be not afraid of copyright• Self archive (the Green Road)• Don‘t sign any contracts giving away rights• Talk to your scientific societies and museum to adopt a policy to

at least allow self archiving• Demonstrate the power of access through inovative research

projects and data: Research will be the only motivation to change law and build up infrastructure

OECD Declaration for Access to Research Data from Public Funding

(Spring 2007)

How to implement this?

Recommendations:

• Assure that all you do is open access• Understand, adopt and propagate an adequate copyright policy• Talk to your scientific societies and museum to adopt a policy to

at least allow self archiving• Talk to your publishers to move into XML publishing• Support the emergence of standards and transfer protocols

Recommendations: (ctd.)• Science policy has to change to build and maintain the

necessary cyberinfrastructure, similarly to the building of libraries

• Prospective publications must be structured to allow machines to read and understand them.

• Copyright must be adjusted to accomodate ingenious and best useage of our shared knowledge, such as using Creative Commons licencies or applying the principles of the Conservation Commons.

• Sharing data has to become standard practice between scientists

Recommendations: (ctd.)• Science policy has to change to build and maintain the

necessary cyberinfrastructure, similarly to the building of libraries

• Prospective publications must be structured to allow machines to read and understand them.

• Copyright must be adjusted to accomodate ingenious and best useage of our shared knowledge, such as using Creative Commons licencies or applying the principles of the Conservation Commons.

• Sharing data has to become standard practice between scientists

antbase.org: Freier Zugang als Grundlage…

http://plazi.org

Thank you very much!

Donat Agosti

agosti@plazi.org