Text and Data Mining Reservation Protocol Community Group
The goal of this Group is to facilitate Text and Data Mining (TDM) Reservation Protocol in Europe and elsewhere, by specifying a simple and practical machine-readable solution, capable of expressing the reservation of TDM rights - following the rules set by the new European DSM Directive / Art.4 - and the availability of machine-readable licenses for TDM actors.
Group's public email, repo and wiki activity over time
Note: Community Groups are proposed and run by the community. Although W3C hosts these conversations, the groups do not necessarily represent the views of the W3C Membership or staff.
During its early meetings, the group worked on what TDM means in practice, the vocabulary to be used during the project, the goals and requirements for a technical solution, created several use cases and compiled a set of past and existing initiatives with a similar scope.
Then, the group agreed on three alternative technical solutions for expressing the reservation of TDM rights: one based on http headers, another based on a file hosted on the origin server, and a third based on html meta tags. These three solutions correspond to different situations and technical skills. The priority in which these techniques should be processed by TDM Agents is still to be decided.
During recent meetings, the group defined a machine readable TDM policy which details how a rightsholder can be contacted and conditions in which a TDM license can be acquired. It has been agreed by the group that TDM Policies will be defined as a profile of ODRL 2 and a detailed proposal has been written.
Issue #15, Should the vcard prefix be explicitly declared? It is closed, as it has been tested that re-defining the vcard prefix in Policy instances is not useful.
Issue #13, Which domain name for the ODRL profile used for TDM Policies? The W3C management accepts the use of the W3C domain name for identifying the properties defined in the Policy specification. We will now have to work on the details of which URL is chosen.
Issue #14, Should we merge the file on the origin server with the ODRL TDM Policy? The current comment do not allow to reach a consensus. During the dicussion it appears that the participants to the call are in favor to keep separate both structures, as it keeps a clean separation between the core deliverable of the group (related to the Article 4 of the DSM) and the optional & advanced feature (offering TDM Policies). The final choice will be done during the next meeting.
Issue #16, Property names and values of TDM-a and TDM-b The beauty contest is open, several ideas are exchanged during the call, we’ll make a choice during the next meeting.
Next steps:
The chairs will now start writing a detailed specification, using the W3C document style and tools (respect). The drafts will be discussed during the summer and we should be able to finalize it in October 2021. As soon as property names are chosen, interested parties can start prototyping, by setting TDM properties on one side, by developing test scripts on the other side. We hope being able to demo the result by October also.
In previous meetings, the group has worked on what TDM means in practice, the vocabulary to be used during the project, the goals and requirements for a technical solution, the creation of several use cases. They have also compiled a set of past and existing initiatives with a similar scope.
Then, the group discussed on how rightsholders can declare to TDM Agents the reservation of TDM Rights. Three different technical solutions were selected: one based on http headers, another based on a file hosted on the origin server, and a third based on html meta tags. These three solutions correspond to different situations and technical skills. The priority in which these techniques should be processed by TDM Agents is still to be decided.
Using one of these techniques, a rightsholder can express either that “TDM rights are not reserved”, “TDM rights are reserved” or “TDM rights are reserved but a TDM license can be acquired”.
In the latter case, it is interesting for both parties (rightsholder and TDM Actor) to create a machine readable TDM policy which details in which conditions a TDM license can be acquired.
The group is therefore now defining the model of such TDM policy.
Participants: Aziz Ndao, Tzviya, Leonard, Jean-Baptiste, Giulia, Laurent, Carlo Lavizzari, Mathilde
Agenda: discuss the draft of TDM Policy properties
The TDM Policy is currently drafted in a shared GDoc. No serialization format is chosen so far, but ODRL 2 is the best candidate so far.
The discussion focuses on some properties:
Content identifier.
STM is using DOI
We may decide that it takes the form of a URI, depending the serialization format.
But using this content id forbids using the same policy for N resources: this is a big issue.
Rightsholder name & contact info
We could add a rightsholder identifier.
Rights reservation (values are no, yes, yes + a TDM policy)
if a TDM agent has reached this policy, this is because the rights reservation value is “yes”. This value will always be “yes”, therefore its use if not obvious.
“where” property: this is initially meant to express “EU” vs “non-EU”.
The EU Directive already sets rules about TDM policies in the EU for research purpose, therefore no TDM policy defined by a rightsholder is applicable to “research” in the “EU”.
Participants are wondering if going to this level of details in the policy is useful.
Both “where” and “when” are at the same conceptual level: either both are present or both absent from the TDM policy model.
Participants: Brendan, Giulia, Laurent, Claudio, Jean-Baptiste, Steve
Status:
In previous meetings, the group has worked on what TDM means in practice, the vocabulary to be used during the project, the goals and requirements for a technical solution, the creation of several use cases. They have also compiled a set of past and existing initiatives with a similar scope.
Then, the group discussed on how rightsholders can declare to TDM Agents the reservation of TDM Rights. Three different technical solutions were selected: one based on http headers, another based on a file hosted on the origin server, and a third based on html meta tags. These three solutions correspond to different situations and technical skills. The priority in which these techniques should be processed by TDM Agents is still to be decided.
Using one of these techniques, a rightsholder can express either that “TDM rights are not reserved”, “TDM rights are reserved” or “TDM rights are reserved but a TDM license can be acquired”.
In the latter case, it is interesting for both parties (rightsholder and TDM Actor) to create a machine readable TDM policy which details in which conditions a TDM license can be acquired”.
Defining what such TDM policy will be is the new task of the group.
Agenda: Presentation of RightsML as an ODRL use case; TDM policy properties.
Brendan (IPTC) presents RightsML.
RightsML is a “profile” of ODRL. “Policies” are related to time, geography and media distribution channels. RightsML 2.0 is based on ODRL 2.2 (the latest version).
ODRL is expressed in RDF and can be serialized in RDF/XML, JSON-LD or Turtle. To understand ODRL, we should start with the ODRL 2 Core Model. The “ODRL Profile Best Practices” gives hints on how we can build a TDM profile.
People can express “permissions” and define “duties” like “payment” or “attribution”. Constraints can be combined. “tickets” and “offer” are used when no “assignee” is known in advance. Possible “actions” are “derive”, “transform”, “translate”, which fit with TDM usage. Possible “contraints” are “payment”, “geospatial …” …
We could create a profile of ODRL, with a very limited set of possible actions and constraints.
It would be versioned, so that new features can be added over time and we could follow the evolution of the ODRL spec.
Q. Who is part of the ODRL group today? It is a Community Group. Renato Iannella (Monegraph) is still the guide. A core half a dozen people works of it.
Q. Is RightsML successful? RightsML has some level of uptake now. e.g. Reuters is using it. It took time… Most newspapers don’t parse RIghtsML and rely on the old fashion textual expression of rights.
Q. Are there ready to integrate software able to process ODRL? Some proofs of concept. IPTC has developed one in Python. Reuters did their own implementation.
Q. Were there some collaboration with Google? IPTC didn’t talk about RightsML with Google.
Giulia presents TDM license properties
The document was drafted from information received from participants, especially Steve from Elsevier.
Types of assignees can be commercial, non-commercial organizations, information scientist.
Elsevier is using the term “authorized users”, i.e. users authorized to apply TDM on content in a given company.
“use case” is about the kind of usage – create a knowledge graph, extracting trends … but to describe it is a machine readable langage can be difficult, because concepts are nuanced.
“data types” is about XML, PDF ….
“License data” copyrighted material, e.g. XML content, “derived data” is e.g. an index built upon XML content.
Agenda: discuss the “file at the origin” and “html meta” proposals. Discuss if the three current proposals can be seen a complementary.
Laurent, regarding the issue #10 open by Claudio: splitting html pages into identified sections is technically doable (via id attributes); associating different sections with different TDM rules is technically doable also (via XML idref attributes). But such a solution would not be html-friendly and would be complex to handle for both publishers and TDM Agents. We may think about it for a version 2 of the specification, but introducing this now would be really dangerous imho.
Claudio: I understand and agree with the complexity for a version 1.
Giulia : should http header properties be registered by a standards body?
Brendan: if we choose to name these properties with a “X-” as a prefix, we don’t need to have any registration. If not we should have our properties registered by IANA to make the solution more “standard”.
Laurent: regarding the “file” proposal, we have to choose between regexps and something simplistic like the robots solution.
Jean-Baptiste : the solution should be as simple as possible. Even JS compatible regexps are so wide that they can be badly implemented. Implementing the robots.txt solution (using * and $ only) is more robust.
Rama and Claudio agree.
Laurent: ok I’ll change the “file” proposal accordingly. We’ll have to study the use of * and $ in details.
Laurent : note that in the “file” proposal, if a file does not match any rule, it is driven by default by the exception allowed by the Article 4.
Giulia, Jean-Baptiste, Rama : please add an example with 2 files treated specifically in a directory, an example with a $, an example of the difference or processing between matching files and directories.
Giulia : in robots meta (which specifies no index, no follow), does the rule apply only the html content or also to the resources referenced in the html document?
Brendan, Laurent: not sure there is a definite robots specification for that. We’ll have to check.
Laurent: warning, this is Google specific information, aimed at avoiding ambiguities. Still not sure that the robots spec gives a definitive answer.
Laurent: Therefore, do you think that the 3 solutions proposed so far are complementary and all 3 can be implemented by every TDM Agent in the world, or should we already make choices?
Giulia: the two first solutions are generic (any media): pros and cons are tied to preferences of technical providers. The third one is vertical (html only), therefore a bit different, its interest is that a publisher can embed the information without any implication of a technical provider.
Laurent: as we have a vertical solution now (for html), other vertical solutions may be proposed. A warning: each time we add an alternative solution, we force EVERY TDM Agent to implement it.
Participants: all agree that the 3 current solutions are useful and complementary, and that they are so simple that every TDM Agent can implement the 3 solutions. We’ll need an order of priority in case of conflict.
Giulia : should we also express the rightsholder name? the date of the declaration?
Claudio : if I get a “don’t mine” instruction, I’d like to know who said that and why. Therefore such metadata would be useful.
Laurent: we have a charter to provide instructions to machines for TDM purpose; semantic information is out of scope here. If publishers want to provider metadata about their content, there are standard solutions for that (json-ld + schema.org on html content, IPTC for images, XML for PDF and media content).
Next steps:
a/ We’ll see if other proposals come to the table.
b/ We will now start discussing the licence format: what publishers need to declare and which existing initiatives can be reused? could we define standard licenses ready for use? The experience of the IPTC with RightsML (based on ODRL2), EPC / Copyright Hub (also using ODRL2), STM Association, Crossref TDM service, ARDITO will be really useful for that.
Leonard: many people with a CMS don’t control their own servers. Sharepoint and other CMS are not so easy to control … Our solution should be usable for anybody putting content on the web. Therefore http headers should not be the only solution. But we must be careful and also avoid too many solutions.
Jean-Baptiste: The proposed solution would be the easiest for my company. e-distributors / aggregators control their servers. I agree that a different solution could be added for specific use cases.
Claudio: if people cannot control their servers, they should go to a proper e-distributor, who will e.g. develop a CMS plug-in.
Claudio : if you want to defend your content, the burden is on you.
Steve: what is the legal minimum of functionalities required for our solution?
Laurent: a simple flag expressing an opt-out from the exception, as defined by the Article 4. Everything else is “cherry on the cake”.
Leonard: in the proposed solution, the license is referenced via a URL. We’ll have to be careful about potential same origin issues (CORS issues). And decide if URLs can be relative or must be absolute.
Leonard: the proposal assumes that the license will be either a human readable or machine readable license. It there a use case where both will be provided by a rightsholder?
Laurent: offering both human and machine readable licenses implies a risk of discrepancy between both expressions.
Giulia: I think the need is exclusive and human readable is useful for less tech advanced rightsholders, a sort of fallback.
Claudio: I imagine the reverse; machine readable should be mandatory, human readable is a plus.
Jean-Baptiste: many French publishers will decide on a human readable license.
Leonard: As a TDM Actor, if I mis-interpret a human readable license, I am legally liable; therefore we should impose machine readable licenses.
Steve: we could define a limited set of templated licenses to ease the use of machine-readable licenses.
Leonard: maybe we miss a value TDM-a = “0.5”. Something between “TDM rights are reserved” and “”TDM rights are not reserved”.
Laurent: isn’t it what TDM-a = “2” is for? a notion of “TDM rights may be reserved or not, depending the license you acquire”?
Steve: outside of the EU, I want to reserve TDM rights for commercial companies. If TDM Actors pay a fee they get specific TDM rights. How will people be able to express such a thing?
Giulia: when we study the license format, we’ll have to find a format which allows conditional information (if the use is commercial; at this location; before this date).
Conclusion
For the next call, the co-chairs will prepare a “Proposal based on a file hosted on the origin website” and if possible a “Proposal based on meta properties in html documents”.
We will use the same anti-bikeshedding TDM-a and TDM-b properties.
The Charter is approved unanimously by the participants. Community members who didn’t attend the meeting are kindly invited to send comments, if any, before the next meeting.
Review of the goals and requirements
During the discussion, it is made clear that the requirements are related to the technical specification, not the implementations of the specification; said differently, rightsholders will have a flexible use of the different features of the specification. For instance they will be able to express a licence only if available, and choose between expressing their licences via human readable information **or** machine-readable information.
To make it clearer, the Core requirements will be introduced via the phrase “The technical specification shall”.
Note: It is nevertheless important to note that TDM Actors will have to implement every feature in their TDM Agents, therefore we’ll have to be very careful not to multiply options.
Because they cannot be verified against a technical solution, what was called “High level requirements” is now called “Goals”.
Because “Primary” and “Secondary” Requirements was not clear, these are renamed “Core Requirement” (what the technical solution must fulfill) and “Additional Requirements” (which will complement Core Requirements in specific cases). The second slot is currently empty, the Group may decide if items should be added here during our work.
The Goals and the Requirements are approved unanimously by the participants. Community members who didn’t attend the meeting are kindly invited to send comments, if any, before the next meeting.
Review of the Vocabulary
One of the comments made on the Goals was that “publishers” was used but not defined. A simple definition was therefore added to the vocabulary. This definition would be more accurate for distributors of digital publications on the Web. Calling them “publishers” is non perfect but still usable in our context, where we refer to publishing as the act of making content available to a public.
The Vocabulary is approved unanimously by the participants.
Workplan
How do we plan our work, avoid multiple threads of discussion and foster participation?
The way to go, agreed during the call, is to proceed in sequence, focussing on one solution at a time, from the simplest to more elaborated solutions. Next steps will be:
Study a solution based on http headers;
Study a solution based on a central file hosted on a Web server hosting Web resources;
Study a solution based on the embedding of metadata in Web resources.
As inspiration can come from past and existing initiatives (https://github.com/w3c/tdm-reservation-protocol/blob/main/docs/initiatives.md), we will start by studying the robots.txt initiative and the co-chairs, with the help of the participants, will try to schedule presentations of other initiatives by their experts . For instance, RightsML (IPTC) and Assert your rights (Copyright Hub) could be the subject of such presentations.
How will participants be able propose technical solutions for discussion within the group?
Different solutions are discussed: Google Docs, Github Wiki and raw Github documents in markdown. The issue with Google Docs is that they are separate from the Github framework in which we are working. The issue with Github Wiki pages is that history of changes are difficult to see and Issues or Pull Requests cannot be synchronized with changes.
decision of the frequency and preferred date-time for the next calls.
Many thanks to the participants who have posted on the mailing list a short introduction about themselves and their interest for the project. All participants are invited to do so.
The following decisions were taken:
the group will have a call every two weeks, on Tuesdays, at 11 am EST = 17:00 CET/CEST (= 4 pm UTC in Winter, 3 pm UTC in Summer).
we’ll discuss asap how we deal with the winter/summer time change.
we’ll use the same Zoom details we used for the first call. This information must be requested to the co-chairs.
the CG home page (https://www.w3.org/community/tdmrep/) is a great starting point to find information: posts (like this one), Github link, Mailing List.
the use of IRC and scribes is not decided: this can be complex for newcomers, therefore we’ll talk about this later only.
participants are invited to comment on the current open issues 1 to 5.
participants are invited to read and comment as issues (*) the documents provided on Github as source for discussion.
the co-chairs will prepare for the next meeting a proposal of evolution of the vocabulary and requirements, using as a source the current open issues 1 to 5.
the co-chairs will add a page to the documentation, describing the robots.txt technology for non technicians.
Therefore the next call is planned on Tuesday 09 Marchat 11 am EST = 16 pm UTC = 17:00 CET.
PS: TDM actors, please join the discussion; the door is more than open.
(*) Github offers “issues” but also “discussions”. We could decide to use the latter, maybe simpler for non technicians.
Dear participants in the Text and Data Mining Reservation Protocol CG (TDMrep in short), Doodle has delivered its oracle:
The kickoff of the project will happen on Tuesday 23 Feb, 16:00 UTC (local datetime via dateandtime) and will last one hour. This time is not great for Asian participants, therefore the co-chairs will contact them before the meeting to get their inputs on the topics to be discussed.
We will use Zoom for the call, please contact the co-chairs to get details.
decision of the frequency and preferred date-time for the next calls.
Please read these documents before the call. Please also drop 3 lines presenting yourself and your interest for the CG on the public-tdmrep@w3.org mailing list, so that we don’t spend too much time on the initial roundtable.
Dear participants in the Text and Data Mining Reservation Protocol CG (TDMrep in short), it has taken some time for the co-chairs before inviting you to a first call; this is simply because we were waiting a sufficient number of participants representing TDM actors. We are currently 13 participants in this CG, and we are expecting other people to join now that we have reached ACReps from organizations which have a foot in TDM.
Therefore it is now time to have a kickoff call, and Giulia and I propose to have it between Feb. 15th and 23th.
We will use Zoom for the call, please contact the co-chairs to get details (it is bad practice to let that in a blog post).
Before the call, you can read the documents which have been put in place in the project Github space. The most important is certainly the set of requirements which should be discussed and – if possible – validated during the first call. A proposed charter for the CG is also to be discussed and validated asap.
The repository also includes a proposal for a common vocabulary, a tentative to define TDM and some useful extracts of the EU DSM Directive (Articles 3 and 4). All good reads.
Giulia and I hope to meet you all at this kickoff call 🙂
In addition to their significance in the context of scientific research, text and data mining techniques (TDM) are widely used both by private and public entities to analyse large amounts of data (including copyright protected content like text, images, video etc.) in different areas of life and for various purposes, including for government services, complex business decisions and the development of new applications or technologies.
In a digital environment, TDM usage of copyright protected works can be subject to different terms and conditions, depending on the legal framework. In generic terms, an act of reproduction is required before TDM can be applied on content accessible on the Web; international laws stipulate that such act of reproduction is subject to authorization by rightsholders. So far, analyzing and processing the terms and conditions of a website, contacting rightsholders, seeking for permission and concluding licensing agreements require time and resources.
In such context, a machine-readable solution which streamlines the communication of TDM rights and licenses available for online copyrighted content is necessary to facilitate the development of TDM applications and reduce the risks of legal uncertainty for TDM actors. Such a solution, that shall rely on a consensus by rightsholders and TDM actors, will optimize the capacity of TDM actors to lawfully access and process useful content at large scale.
The EU Directive 2019/790, better known as the DSM Directive (DSM meaning Digital Single Market), introduces two exceptions or limitations to the rights of rightsholders on lawfully accessible content, for reproductions and extractions for the purposes of TDM:
In its Article 3, a mandatory exception for research organisations and cultural heritage institutions which carry out TDM for the purposes of scientific research.
In its Article 4, an exception for any organisation willing to carry out TDM for any purpose other than scientific research, including commercial purposes, which applies on the condition that the use of content for TDM has not been expressly reserved by their rights holders in an appropriate manner, such as machine-readable means.
These TDM exceptions apply to TDM usage in the European Union in relation to content from European and foreign rightsholders. Outside of the EU, where the DSM legislation does not apply, the said exception does not apply: exclusive rights of right-holders to authorize acts of reproduction are maintained. In such cases, no TDM can be performed without the explicit authorisation of these rightsholders: in these countries, the absence of a reservation of rights by rightsholders cannot be considered as an implicit authorization to reproduce copyrighted content for TDM purpose, and advocating fair use or a similar rule is legally uncertain, as these actions are judged on a case-per-case basis.
The “opt-out” mechanism introduced by the DMS Directive is therefore a real opportunity for TDM actors and publishers across countries to define a machine-readable technique able to express not only if TDM rights on specific Web content are reserved or not, but also how rightsholders can be contacted and which licenses are available, if any. This is a tremendous help for TDM actors from all countries looking for legal certainty.
w3c/tdm-reservation-protocol
