Warning: Long Article.
================
The plan to mine the worldās research papers
A giant data store quietly being built in India could free vast swathes of
science for computer analysis ā but is it legal?
Carl Malamud in front of the data store of 73 million articles that he
plans to let scientists text mine. Credit: Smita Sharma for Nature (image
deleted in plain text extraction)
Carl Malamud is on a crusade to liberate information locked up behind
paywalls ā and his campaigns have scored many victories. He has spent
decades publishing copyrighted legal documents, from building codes to
court records, and then arguing that such texts represent public-domain law
that ought to be available to any citizen online. Sometimes, he has won
those arguments in court. Now, the 60-year-old American technologist is
turning his sights on a new objective: freeing paywalled scientific
literature. And he thinks he has a legal way to do it.
Over the past year, Malamud has ā without asking publishers ā teamed up
with Indian researchers to build a gigantic store of text and images
extracted from 73 million journal articles dating from 1847 up to the
present day. The cache, which is still being created, will be kept on a
576-terabyte storage facility at Jawaharlal Nehru University (JNU) in New
Delhi. āThis is not every journal article ever written, but itās a lot,ā
Malamud says. Itās comparable to the size of the core collection in the Web
of Science database, for instance. Malamud and his JNU collaborator,
bioinformatician Andrew Lynn, call their facility the JNU data depot.
No one will be allowed to read or download work from the repository,
because that would breach publishersā copyright. Instead, Malamud
envisages, researchers could crawl over its text and data with computer
software, scanning through the worldās scientific literature to pull out
insights without actually reading the text.
The unprecedented project is generating much excitement because it could,
for the first time, open up vast swathes of the paywalled literature for
easy computerized analysis. Dozens of research groups already mine papers
to build databases of genes and chemicals, map associations between
proteins and diseases, and generate useful scientific hypotheses. But
publishers control ā and often limit ā the speed and scope of such
projects, which typically confine themselves to abstracts, not full text.
Researchers in India, the United States and the United Kingdom are already
making plans to use the JNU store instead. Malamud and Lynn have held
workshops at Indian government laboratories and universities to explain the
idea. āWe bring in professors and explain what we are doing. They get all
excited and they say, āOh gosh, this is wonderfulā,ā says Malamud.
But the depotās legal status isnāt yet clear. Malamud, who contacted
several intellectual-property (IP) lawyers before starting work on the
depot, hopes to avoid a lawsuit. āOur position is that what we are doing is
perfectly legal,ā he says. For the moment, he is proceeding with caution:
the JNU data depot is air-gapped, meaning that no one can access it from
the Internet. Users have to physically visit the facility, and only
researchers who want to mine for non-commercial purposes are currently
allowed in. Malamud says his team does plan to allow remote access in the
future. āThe hope is to do this slowly and deliberately. We are not
throwing this open right away,ā he says.
The power of data mining
The JNU data store could sweep aside barriers that still deter scientists
from using software to analyse research, says Max HƤussler, a
bioinformatics researcher at the University of California, Santa Cruz
(UCSC). āText mining of academic papers is close to impossible right now,ā
he says ā even for someone like him who already has institutional access to
paywalled articles.
Since 2009, HƤussler and his colleagues have been building the online UCSC
Genome Browser, which links DNA sequences in the human genome to parts of
research papers that mention the same sequences. To do that, the
researchers have contacted more than 40 publishers to ask permission to use
software to rifle through research to find mentions of DNA. But 15
publishers have not responded or have denied permission. HƤussler is unsure
whether he can legally mine papers without permission, so he isnāt trying.
In the past, he has found his access blockedby publishers who have spotted
his software crawling over their sites. āI spend 90% of my time just
contacting publishers or writing software to download papers,ā says
HƤussler.
Chris Hartgerink, a statistician who works part-time at Berlinās QUEST
Center for Transforming Biomedical Research, says he now restricts himself
to text-mining work from open-access publishers only, because āthe hassles
of dealing with these closed publishers are too muchā. A few years ago,
when Hartgerink was pursuing his PhD in the Netherlands, three publishers
blocked his access to their journals after he tried to download articles in
bulk for mining.
Some countries have changed their laws to affirm that researchers on
non-commercial projects donāt need a copyright-holderās permission to mine
whatever they can legally access. The United Kingdom passed such a law in
2014, and the European Union voted through a similar provisionthis year.
That doesnāt help academics in poor nations who donāt have legal access to
papers. And even in the United Kingdom, publishers can legally place
āreasonableā restrictions on the process, such as channelling scientists
through publisher-specific interfaces and limiting the speed of electronic
searching or bulk downloading to protect servers from overload. Such limits
are a big problem, says John McNaught, deputy director of the National
Centre for Text Mining at the University of Manchester, UK. āA limit of,
say, one article every five seconds, which sounds fast for a human, is
painfully slow for a machine. It would take a year to download around six
million articles, and five years to download all published articles
concerning just biomedicine,ā he says.
Wealthy pharmaceutical firms often pay extra to negotiate special
text-mining access because their work has a commercial purpose, says
McNaught. In some cases, publishers allow these firms to download papers in
bulk, thus avoiding rate limits, according to a researcher at a
pharmaceutical firm who did not want to be identified because they were not
authorized to talk to the media. University academics, however, frequently
restrict themselves to mining article abstracts from databases such as
PubMed. That provides some information, but full texts are much more
useful. In 2018, a team led by computational biologist SĆøren Brunak at the
Technical University of Denmark in Lyngby showed that full-text searches
throw up many more geneādisease links than do searches of abstracts (D.
Westergaard et al. PLoS Comput. Biol. 14, e1005962; 2018).
Carl Malamud and Andrew Lynn oversee the project at Jawaharlal Nehru
University in New Delhi to extract text and images from 73 million research
papers.Credit: Smita Sharma for Nature
Scientists must also overcome technical barriers when mining articles. It
is hard to extract text from the various layouts that publishers use ā
something that the JNU team is struggling with right now. Tools to convert
PDFs to plain text donāt always distinguish clearly between paragraphs,
footnotes and images, for instance. Once the JNU team has done it, however,
others will be saved the effort. The team is close to completing the first
round of extraction from the corpus of 73 million papers, Malamud says ā
although they will need to check for errors, so he expects the database
wonāt be ready until the end of the year.
A world of possibilities
Early enthusiasts are already gearing up to use the JNU depot. One is
Gitanjali Yadav, a computational biologist at Delhiās National Institute of
Plant Genome Research (NIPGR) and a lecturer at the University of
Cambridge, UK. In 2006, Yadav led an effort at NIPGR to build a database of
chemicals secreted by plants. Called EssOilDB, this database is today
scoured by groups from drug developers to perfumeries looking for leads.
Yadav thinks that āCarlās compendiumā, as she calls it, could give her
database a leg-up.
To make EssOilDB, Yadavās team had to trawl PubMed and Google Scholar for
relevant papers, extract data from full texts where they could, and
manually visit libraries to copy out tables from rare journals for the
rest. The depot could fast-forward this work, says Yadav, whose team is
currently writing the queries they will use to extract the data.
Srinivasan Ramachandran, a bioinformatics researcher at Delhiās Institute
of Genomics and Integrative Biology, is also excited by Malamudās plan. His
team runs a database of genes linked to type 2 diabetes; theyāve been
crawling PubMed abstracts to find papers. Now he hopes the depot could
widen his mining net.
And at the Massachusetts Institute of Technology (MIT) in Cambridge, a team
called the Knowledge Futures Group says it wants to mine the depot to map
how academic publishing has evolved over time. The group hopes to forecast
emerging areas of research and identify alternatives to conventional
metrics for measuring research impact, says team member James Weis, a
doctoral student at MIT Media Lab.
A career unlocking copyright
Malamud only recently had the idea of extending his activism to academic
publishing. The founder of a non-profit corporation called Public Resource,
based in Sebastopol, California, Malamud has focused on buying up
government-owned legal works and publishing them. These include, for
instance, the state of Georgiaās annotated legal code, European toy-safety
standards and more than 19,000 Indian standards for everything from
buildings and pesticides to surgical equipment.
Because these documents are often a source of revenue for government
agencies, some of them have sued Malamud, who has argued back that
documents which have the force of the law cannot be locked behind
copyright. In the Georgia case, a US appeals court cleared him of
infringement charges in 2018, but the state appealed, and the case is with
the US Supreme Court. Meanwhile, a German court ruled in 2017 that the
publication of toy standards by Public Resource, including a standard on
baby dummies (pacifiers), was illegal.
But Malamud has enjoyed victories, too. In 2013, he filed a lawsuit in a US
federal court asking the Internal Revenue Service (IRS) to publish the
forms it collected from tax-exempt non-profit organizations ā data that
could help to hold these organizations to account. Here, the court ruled in
Malamudās favour, prompting the IRS to release the financial information of
thousands of non-profit organizations in a machine-readable format.
In early 2017, aided by the Arcadia Fund, a London-based charity that
promotes open access, Malamud turned his attention to research articles.
Under US law, works by US federal government employees cannot be
copyrighted, and Public Resource says it has found hundreds of thousands of
academic articles that are US government works and seem to defy this rule.
Malamud has called for such articles to be freed from copyright assertions,
but itās not clear whether that would hold up in court. He has posted his
preliminary results online, but has put further campaigning on hold,
because the project prompted him to take on a wider mission: democratizing
access to all scientific literature.
Opportunity in India
A trigger for this mission came from a landmark Delhi High Court judgment
in 2016. The case revolved around Rameshwari Photocopy Services, a shop on
the campus of the University of Delhi. For years, the business had been
preparing course packs for students by photocopying pages from expensive
textbooks. With prices ranging between 500 and 19,000 rupees (US$7ā277),
these textbooks were out of reach for many students.
Rameshwari Photocopy Services in New Delhi was taken to court for copying
parts of textbooks, and won.Credit: Sajjad Hussain/AFP/Getty
In 2012, Oxford University Press, Cambridge University Press and Taylor and
Francis filed a lawsuit against the university, demanding that it buy a
license to reproduce a portion of each text. But the Delhi High Court
dismissed the suit. In its judgment, the court cited section 52 of Indiaās
1957 Copyright Act, which allows the reproduction of copyrighted works for
education. Another provision in the same section allows reproduction for
research purposes.
Malamud has a long association with India: he first travelled there as a
tourist in the 1980s, and he wrote one of his first books, on database
design, on a houseboat in Srinagar. And around the same time that he heard
about the Rameshwari judgment, he had come into possession (he wonāt say
how) of eight hard drives containing millions of journal articles from
Sci-Hub, the pirate website that distributes paywalled papers for anyone to
read. Sci-Hub itself has lost two lawsuits against publishers in US courts
over its copyright infringements, but despite those judgments, some of its
domains are still working today.
Malamud began to wonder whether he could legally use the Sci-Hub drives to
benefit Indian students. In a 2018 book about his work called Code Swaraj,
co-authored with Indian tech entrepreneur Sam Pitroda, Malamud writes that
he imagined showing up on Indian campuses in the equivalent of an American
taco truck, ready to serve the articles up to those who wanted them.
Ultimately, he zeroed in on the idea of the JNU text-mining depot instead.
(Malamud has also helped to set up another mining facility with 250
terabytes of data at the Indian Institute of Technology Delhi, which isnāt
in use yet.) But he is cagey about where the depotās articles come from.
Asked directly whether some of the text-mining depotās articles come from
Sci-Hub, he said he wouldnāt comment, and named only sources that provide
free-to-download versions of papers (such as PubMed Central and the
āUnpaywallā tool). But he does say that he does not have contracts with
publishers to access the journals in the depot.
Is it legal?
Malamud says that where he got the articles from shouldnāt matter anyway.
The data mining, he says, is non-consumptive: a technical term meaning that
researchers donāt read or display large portions of the works they are
analysing. āYou cannot punch in a DOI [article identifier] and pull out the
article,ā he says. Malamud argues that it is legally permissible to do such
mining on copyrighted content in countries such as the United States. In
2015, for instance, a US court cleared Google Books of copyright
infringement charges after it did something similar to the JNU depot:
scanning thousands of copyrighted books without buying the rights to do so,
and displaying snippets from these books as part of its search service, but
not allowing them to be downloaded or read in their entirety by a human.
The Google Books case was a test of non-consumptive data mining, says
Joseph Gratz, an IP lawyer at the law firm Durie Tangri in San Francisco,
California, who represented Google in the case and has previously
represented Public Resource. Even though Google was displaying snippets,
the court ruled that the text was too limited to amount to infringement.
Google was scanning authorized copies of books (from libraries in many
cases), even though it did not ask permission. Copyright holders might
argue that if Sci-Hub or other unauthorized sources supplied the JNU depot,
the situation would be different from the Google Books case, Gratz says.
But a case involving unauthorized sources has never been argued in American
courts, making it hard to predict the outcome. āThere are good reasons why
the source shouldnāt matter, but there may be arguments that it should,ā
says Gratz.
The question of the facilityās legality in the United States might not even
be relevant, because international researchers would be getting results
from a depot that sits in India, even if they are accessing it remotely. So
Indian law is likely to apply to the question of whether it is legal to
create the corpus, says Michael W. Carroll, a professor at the American
Universityās Washington College of Law in Washington DC.
Here, Indiaās copyright laws might help Malamud ā another reason why the
facility is in New Delhi. The research exemption in section 52 means that
the JNU data depotās actions would be considered fair under Indian law,
argues Arul George Scaria, an assistant professor at Delhiās National Law
University. Not everyone agrees with this interpretation, however. Section
52 allows researchers to photocopy a journal article for personal use, but
doesnāt necessarily allow the blanket reproduction of journals as the JNU
depot has done, says T. Prashant Reddy, a legal researcher at the Vidhi
Centre for Legal Policy in New Delhi. That entire articles arenāt shared
with users does help, but the mass reproduction of text used to create the
database puts the facility in āa legal grey zoneā, Reddy says.
Risky business
When Nature contacted 15 publishers about the JNU data depot, the six who
responded said that this was the first time they had heard of the project,
and that they couldnāt comment on its legality without further information.
But all six ā Elsevier, BMJ, the American Chemical Society, Springer
Nature, the American Association for the Advancement of Sciences and the US
National Academy of Sciences ā stated that researchers looking to mine
their papers needed their authorization. (Springer Nature publishes this
journal; Natureās news team is editorially independent of its publisher.)
Malamud acknowledges that there is some risk in what he is doing. But he
argues that it is āmorally crucialā to do it, especially in India. Indian
universities and government labs spend heavily on journal subscriptions, he
says, and still donāt have all the publications they need. Data released by
Sci-Hub indicate that Indians are among the worldās biggest users of their
website, suggesting that university licences donāt go far enough. Although
open-access movements in Europe and the United States are valuable, India
needs to lead the way in liberating access to scientific knowledge, Malamud
says. āI donāt think we can wait for Europe and the United States to solve
that problem because the need is so pressing here.ā
Nature 571, 316-318 (2019) doi: 10.1038/d41586-019-02142-1
Updates & Corrections
Correction 19 July 2019: An earlier version of this feature used the term
āfair useā inappropriately ā the term isnāt relevant under Indian law.
========
Dr P Vyasamoorthy
040-27846631 / 09490804278 / +918179386472
--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
