As I have already blogged I have been asked by Ben Hawes at the UK Intellectual Property Office to respond to the Hargreaves report on "textmining". I shall be getting help from my OKF colleagues. The issues are, in my mind, simple:
- Legitimate human readers of the literature ("subscribers") have a right to extract factual information from the literature and have exercised this for 200 years.
- We can now do this with machines, often better than with humans. It's vastly faster and cheaper. It increases the value of the literature
- The publishers forbid us to do this and put in place legal and technical obstacles on top of normal copyright
- We are now demanding the removal of these obstacles.
This is not a negotiation, it's a statement of our absolute right. As a corollary it is an integral part of what we pay for human access so there is no reason to make any charge for this.
In essence we shall report to Hargreaves:
- Our position and the justification for it
- Whether the publishers have agreed that these are our rights. I have made them simple because they are simple. Publishers wish to appear "helpful"; this is their chance to show that they are working with us as they continually claim
We shall contact 9 publishers tomorrow through known contacts; this represents our best approach to non-repudiation. Since the publishers have no-formal mechanism for readers to make formal enquiries (in itself Institutionalised unhelpfulness to readers) it will be done through this blog and email. There is enough evidence to show that all publishers will be aware of this request and if they wish to be helpful they can.
Background and clarification
Human abstracters have for centuries abstracted from and commented on the scholarly literature and made the results public without requiring permission from publishers and/or authors. Indeed science is based on being able to do this. In our present request I shall confine the request to "facts" in scientific papers, but the permission I am asserting extends to abstracting papers and to commenting and other activities practised in the paper era. None of this requires new permissions; it is explicitly and implicitly part of current practice. If I read a paper I can write an abstract; I can also critique parts (e.g. reproduce paragraphs and comment in detail on what the authors said. Refusal to allow this is a direct attack on the integrity of science.
I do not have to be the owner of a scientific article to do this. If I borrow a journal from the public library I can sit at home and write abstracts on every paper. I would strongly urge anyone interested in abstraction, commentary, parody, etc. to make representation to Hargreaves – I personally only have time to address the extraction of scientific facts and indexing the literature.
To illustrate "facts" here is the Wikipedia article on aspirin http://en.wikipedia.org/wiki/Aspirin. The article is essentially a collection of facts (as it should be – WP is very strong on removing opinions). The fact that someone reported them in natural language does not stop them being facts. Here are some examples:
- [Aspirin] was first isolated by Arthur Eichengrün, a chemist with the German company Bayer.
- It has also been established that low doses of aspirin may be given immediately after a heart attack to reduce the risk of another heart attack or of the death of cardiac tissue.
- (the factual synthesis of aspirin from salicylic acid and acetic anhydride)
- Physical data Density 1.40 g/cm³ Melt. point 135 °C (275 °F) Boiling point 140 °C (284 °F) (decomposes) Solubility in water 3 mg/mL (20 °C)
All of these are FACTs. As is:
Synthesis of (+)-η5-Cyclopentadienyl(η4-(3aS,4S,7aS)-methyl 2,2-dimethyl-3a,7a-dihydrobenzo[d][1,3]dioxole-3a-carboxylate)cobalt(I) 18
Diene 16 (301 mg, 1.43 mmol, 1 equiv) was dissolved in dry, degassed toluene (10 mL) in a side-armed Schlenk that had been purged and refilled with argon three times. The resulting solution was added via cannula to η5-cyclopentadienylbis(ethylene)cobalt 17 (258 mg, 1.43 mmol, 1 equiv) and the mixture was stirred for 30 min at room temperature until the evolution of ethylene had ceased. The solvent was removed in vacuo and the solid residue was redissolved in a minimal amount of hexane and left to crystallise at -28°C for 48 h. Complex 18 was isolated as red-orange crystals (124 mg, 26%); m. pt. 118-120°C; [α]D +42 (c = 1, CH2Cl2); 1H-NMR (300 MHz, C6D6, Additional file 2) δ 5.44 (1H, d, J = 5.0 Hz, O-CH-), 5.09-5.04 (2H, m, -CH=CH-CH=CH-), 4.41 (5H, s, Cp-H), 3.53 (3H, s, O-CH3), 3.14 (1H, dd, J = 5.5, 1.0 Hz,=CH-C-COOCH3), 2.82 (1H, td, J = 5.0, 2.0 Hz, -O-CH-CH=), 1.42 (3H, s, C-CH3), 1.28 (3H, s, C-CH3) ppm; 13C-NMR (75 MHz, CDCl3, Additional file 2) δ 174.9, 113.9, 82.1, 80.8, 79.8, 74.6, 51.6, 48.9, 48.6, 27.1, 25.6 ppm; νmax (film) 2986, 2937, 1732, 1436, 1370, 1307, 1259, 1229, 1206, 1166, 1109, 1064, 1009, 888, 821, 762 cm-1; HRMS (+ve ESI-TOF) m/z calcd for (C15H20CoO4+H)+, 335.0688, found 335.0694. Found: C, 57.58; H, 5.76. C15H20CoO4 requires C, 57.49; H, 5.73%).
Everything in this paragraph is factual – what was done, what was observed, what was measured. And our software can extract 95% of the meaning from this in a few seconds, whereas many final year undergraduates might struggle.
Factual information is frequently contained in graphs, tables, images, speech and video. Therefore "text-mining" is a subset of information-mining and I shall use that term. Indeed our software can understand simple human spoken discourse about chemical reactions and extract the facts.
Alicia Wise from Elsevier wants to know what I want to do with the content. There is no reason why I should have to justify what I do to Elsevier, but here it is:
I want to extract as many facts as I can from the scientific literature and publish them (as CC0) for me and others to do science with, to build new scientific tools and improve the quality of science.
It is my right. There is absolutely no reason why anyone should need to involve the publisher in information-mining. I have legally mined 200,000 scholarly documents without requiring help or permission from the publisher. I strongly urge anyone thinking of information-mining to explore what, if anything they need the publisher for. Scientists should not have to ask permission not should they have to "use the publisher API" and they should never have to pay.
Legitimate publisher concerns about information-mining
There is only one valid reason for liaising with the publisher – the possibility of server overload. This is a negligible problem if done responsibly – for example if one allows a short pause between each download request (I use 1 second, but I'm willing to be informed of best practice.
publisher concerns about information-mining
I suspect the following concerns:
- Peter Murray-Rust will steal and publish "our" content. I find this deeply offensive. I have been confronted by publisher robots which, in essence, announce: "You are illegally downloading content; we have cut off the journal supply to the whole University; you will have to justify to a senior member of the university why we should reconnect you". [I did nothing illegal or anticontractual – these robots act at the slightest trigger].There is no debate. PM-R is guilty until proved innocent. It is demeaning to be confronted by colleagues who accuse you of having their It is publisher HADOPI except it is "1 strike and you are out". It is publisher institutionalism at its worst. An unbelievable arrogance that the scientific world is out to defraud publishers. And to say "oh, it's not you but it's your graduate students" is worse. If there is anything that underpins science it is the need for ethical behaviour and almost all scientists are highly ethical. If unethical behaviour is detected they are severely reprimanded by the community and it may be the end of their career. To accuse scientists of being thieves (even if you accept that sharing copyrighted papers is theft) is inexcusable
- Peter Murray-Rust and his robots will find errors in our papers. I hope that no-one is afraid of this. It is the purpose of science to find errors and our robots are better than humans at finding many types of error. The publishers' refusal to allow us to validate the literature is damaging science, not enhancing it.
Peter Murray-Rust will create disruptive technology that will seriously disturb our cosy monopolies. I think this is the real crux. Elsevier forbad me to data-mine chemistry because it threatens their cosy data monopolies. Here's what they said in 2010-12
[An Elsevier staff member] have not been able to get a clear story on the Tetrahedron supplementary content mining. There are two opposing views on this (roughly: ScienceDirect is fine with it, and Reaxys is not), and it is not clear what the resolution is. I am very sorry about this, and will keep trying to get a coherent response out of the multifarious monster that is a big company. As soon as I know, you'll know.
Interpretation. Elsevier run a chemical database where they abstract information from the literature (Reaxys) which probably has a revenue stream of several hundred million USD [ACS do the same ("Chemical Abstracts", CAS) and estimates are in the range 200-500 million USD]. So to preserve their monopoly they prevent me mining information. Is it a real threat? One chemist against Elsevier? Yes. Because I have many people who think the same way.
Other walled gardens include bibliography and citations. It's possible to extract both of these robotically and we have the technology to do this. But Scopus and World of Science will be disrupted by this.
What Elsevier and other publishers should do, and what we should do.
The Elsevier contract states (http://lists.okfn.org/pipermail/open-science/2011-April/000724.html )
The CDL/ Elsevier contract includes [@ "Schedule 1.2(a)
General Terms and Conditions "RESTRICTIONS ON USAGE OF THE LICENSED
PRODUCTS/ INTELLECTUAL PROPERTY RIGHTS" GTC1]
"Subscriber shall not use spider or web-crawling or other software programs,
routines, robots or other mechanized devices to continuously and automatically
search and index any content accessed online under this Agreement. "
It forbids me to do anything. The answer is simple:
SCRAP THE ANTI-MINING CLAUSE
[Aside: How my University and any other University could meekly sign this without a titanic public fight is beyond me.]
If Elsevier don't scrap it I'll urge Universities to take them to court. It's against natural justice. We've paid enough for the subscription – we should be allowed our natural rights to do whatever we want.
Then the robots:
TURN THE ROBOTS OFF
The robots have no benefit to the subscriber and are deeply insulting.
I am prepared to agree that we should be considerate in our crawling. I have been very considerate so far, verbally agreed it with at least two publishers. It's insulting to suggest that Universities are incapable of writing robots.
That's it. We demand our rights and will – in one area – agree to abide by a common technical protocol for information mining. There is no other legitimate reason for denying us.
And this applies to ALL toll-access publishers.