towards an infrastructure for mining scientific publications
TRANSCRIPT
1/38
From Open Access Metadata to Open Access Content:Towards an Infrastructure for Mining Scientific Publications
Petr KnothCORE (Connecting REpositories) project
Knowledge Media instituteThe Open University
@petrknoth, #diggicore
2/38
What is Open Access exactly?By “open access” to [peer-reviewed research literature], we mean its free availability on the public internet, permitting any users to read, download, copy, distribute, print, search, or link to the full texts of these articles, crawl them for indexing, pass them as data to software, or use them for any other lawful purpose, without financial, legal, or technical barriers other than those inseparable from gaining access to the internet itself.
[BOAI, 2002]
3/38
How to achieve OA?Two routes:• Self-archiving: Institional/Open Repositories• Open Access Journals
4/38
Why from OA metadata to OA content? • Despite large amount of OA content already available online
(Laakso & Bjork, 2012), OA content is not necessarily easily discoverable (Morrisson, 2012; Konkiel, 2012).
• Often available, but difficult to find …• Inhibiting the OA impact – accessibility, discoverability, reuse …
• Discoverability of OA content on the Web can be dramatically increased by adopting two simple principles!
5/38
Outline
1. Goals of repositories 2. The bleak truth about availability of OA metadata vs content3. Content referencing practises in repositories4. Two principles to increase visibility of OA content5. How to data mine OA aggregated data and why
6/38
Outline
1. Goals of repositories (repositories as large metadata silos)2. The bleak truth about availability of OA metadata vs content3. Content referencing practises in repositories4. Two principles to increase visibility of OA content5. How to data mine OA aggregated data and why
7/38
The primary purpose of repositories • Institutional repositories (IRs) serve a number of purposes; such
collecting and curating digital outputs, providing statistics, research excellence, etc.
• The primary goal of repositories is to open and disseminate research outputs to a worldwide audience (Crow, 2002) – SPARC’s position paper on the case for institutional repositories.
8/38
SPARC’s position paper on IRs“For the repository to provide access to the broader research community, users outside the university must be able to find and retrieve information from the repository. Therefore, institutional repository systems must be able to support interoperability in order to provide access via multiple search engines and other discovery tools. An institution does not necessarily need to implement searching and indexing functionality to satisfy this demand: it could simply maintain and expose metadata, allowing other services to harvest and search the content. This simplicity lowers the barrier to repository operation for many institutions, as it only requires a file system to hold the content and the ability to create and share metadata with external systems.”
9/38
COAR: About harvesting and aggregations …
“Each individual repository is of limited value for research: the real power of Open Access lies in the possibility of connecting and tying together repositories, which is why we need interoperability. In order to create a seamless layer of content through connected repositories from around the world, Open Access relies on interoperability, the ability for systems to communicate with each other and pass information back and forth in a usable format. Interoperability allows us to exploit today's computational power so that we can aggregate, data mine, create new tools and services, and generate new knowledge from repository content.’’
[COAR manifesto]
10/38
We need OA to content (not just metadata)
• Repositories (even the most prominent) often seen by aggregation systems as large metadata.
• OA to metadata is not disruptive. Little difference to the traditional publishing model.
11/38
Outline
1. Goals of repositories (repositories as large metadata silos)2. The bleak truth about availability of OA metadata vs content3. Content referencing practises in repositories4. Two principles to increase visibility of OA content5. How to data mine OA aggregated data and why
12/38
Study
• 83 repositories (mainly EPrints with pdf research outputs)• 1,461,016 metadata records• Ratio of metadata to content• Data acquired from CORE (Knoth & Zdrahal, 2012)
13/38
“[The institutional repository] is like a roach motel. Data goes in, but it doesn’t come out.” (Salo, 2008)
14/38
Why is this a problem?
• Lower accessibility of papers (we have them, but cannot find them)
• Text-mining• Cannot monitor growth• Loosing a strong argument for the adoption of OA!
15/38
Outline
1. Goals of repositories (repositories as large metadata silos)2. The bleak truth about availability of OA metadata vs content3. Content referencing practises in repositories4. Two principles to increase visibility of OA content5. How to data mine OA aggregated data and why
16/38
OAI-PMH and content referencing
• OAI-PMH supports representing metadata in multiple formats, but at a minimum repositories must be able to return records with metadata expressed in the Dublin Core format (OAI-PMH v2.0, 2008)
• If repositories want to satisfy the SPARC guidelines (Crow, 2002), they must provide a link to the content as part of the exposed metadata.
17/38
OAI-PMH and content referencing
The Open Research Online repository (Eprints) links directly to the resource from metadata.Cranfield repository (DSpace) identifies the resource by providing a link to a page from which the resource (if available) can be accessed.
18/38
OAI-PMH and content referencing
The OAI-PMH specification states on this topic that: “The nature of a resource identifier is outside the scope of the OAI-PMH. To facilitate access to the resource associated with harvested metadata, repositories should use an element in metadata records to establish a linkage between the record (and the identifier of its item) and the identifier (URL, URN, DOI, etc.) of the associated resource. The mandatory Dublin Core format provides the identifier element that should be used for this purpose.”
19/38
OAI-PMH and content referencing
• What is an identifier of the associated resource?Is a splash page an identifier? According to OAI-PMH examples it is:<dc:identifier>http://arXiv.org/abs/cs/0112017</dc:identifier>
• The standard is pretty weak on this aspect …
20/38
Outline
1. Goals of repositories (repositories as large metadata silos)2. The bleak truth about availability of OA metadata vs content3. Content referencing practises in repositories4. Two principles to increase visibility of OA content5. How to data mine OA aggregated data and why
21/38
The principles of the principles
• Pragmatic rather than exciting.• Generating maximum benefit for a minimum investment. • Deliberately use current standards to minimise adoption time. • Respecting differences across systems and backwards
compatibility. • Emphasizes the need for easy to use compliance mechanisms to
assist repository managers in ensuring systems interoperability.
22/38
Principle 1 – Content referencing
Open repositories should always establish a link from the metadata record to the item the metadata record describes using a dereferencable identifier pointing to the version held in the repository. The dereferencable identifier should be provided in the appropriate metadata element in the used metadata format (i.e. dc:identifier in the case of Dublin Core).
23/38
Implications: Principle 1 – Content referencing
• Repositories can use different standards to deliver metadata over OAI-PMH (DC, METS, MPEG-21 DIDL)
• Identifier must resolve (be actionable) to the object it identifies• In the case of DC, if more identifiers are present, use the first
identifier as the identifier of the object• Should resolve to the version of the object in the local repository• Similarity with RIOXX identifier field
• The principle is easily applicable in the OA domain: each item can be freely resolved
24/38
Open access statistics and principle 1
• Only dereferencable items are OA
• Increases stats acuracy• Avoids anecdotal
situations (e.g. 23,380 Dark Items)
25/38
Principle 2 – Content accessibility to machines
Open repositories must provide universal access to machines with the same level of access as humans have. It is the role of open repositories to allow machines harvest the entire content of the repository in a reasonable time to enable harvesting systems to acquire and maintain up-to-date information about the repository content.
26/38
Example from Arxiv.org
• Googlebot: unrestricted• Yahoo/MSN: can
reharvest in 6 months• Researchers: access
denied
27/38
Implications: Principle 2 – Content accessibility to machines
• Accessibility of repository content by machines• Enabling reuse through new services, such as those relying on
text-mining
• Open Repositories should not discriminate, except for abusive behavior
• Presumption of innocence
28/38
Validation tools• Key to adoption – Repository managers should not be left alone• Repository Analytics
29/38
Outline
1. Goals of repositories (repositories as large metadata silos)2. The bleak truth about availability of OA metadata vs content3. Content referencing practises in repositories4. Two principles to increase visibility of OA content5. How to data mine OA aggregated data and why
30/38
CORE API
Enables external systems to interact with OA data (JSON or XML)• Search, download metadata and cotent• Content recommendation• Citation references• Statistics• …
31/38
Data dumps• About 11.5 million records• Over 1 million full-texts• Cleaned and enriched with additional information• Distributed as two large zip files: metadata + full-texts
32/38
Examples of usage
• Author disambiguation• Mining URLs from papers to detect trends• Tagging of chemical compounds for image retrieval• Citation analysis• Content recommendation• Detecting collaboration patterns of scientific communities• Monitoring of OA growth
• Any form of text or data mining …• API useful for services and data dumps for offline experiments
33/38
Why to use it?• It is only OA, thus you can legally mine it …• You can redistribute it: essential for reproducible research• Very large and growing• Kept up-to-date• Ability to rerun experiments with new data• All research content will soon be OA (UK HEFCE policy)• Status of a UK national aggregator• 0.5 million monthly visits, but only 150k six months ago
34/38
Why to use it?
• Open infrastructure for open science• Not owned or managed by a for profit company => Ability to run
your own services = new opportunities and no give away of your research to commercial companies
35/38
Conclusions• Visibility of OA content can be significantly improved by adoption
two principles:1) Dereferencable identifiers - Open Repositories provide open
access to content and not just metadata2) Machine access – Open Repositories should provide free access
to content (for anybody and mainly researchers)• Compliance validation tools are needed to support repositories
• Researchers who want to mine content or build services that can rely on aggregators to acquire datasets
• Researchers can deploy their solutions, not just rely on commercial providers.
36/38
Thank you!
Open access needs open repositories and open science
37/38
References 1/2
[BOAI, 2002] Budapest Open Access Initiative. (2002) http://www.opensocietyfoundations.org/openaccess/boai-10-recommendations [Crow, 2002] Crow, R. (2002). The case for institutional repositories: a SPARC position paper. ARL Bimonthly Report 223.[Knoth & Zdrahal, 2012] Knoth, P. and Zdrahal, Z. (2012) CORE: Three Access Levels to Underpin Open Access, D-Lib Magazine, 18, 11/12, Corporation for National Research Initiatives, http://dx.doi.org/10.1045/november2012-knoth [Konkiel, 2012] Konkiel, S. (2012) Are Institutional Repositories Doing Their Job? https://blogs.libraries.iub.edu/scholcomm/2012/09/11/are-institutional-repositories-doing-their-job/ [Laakso & Bjork, 2012] Laakso, M., & Björk, B. C. (2012). Anatomy of open access publishing: a study of longitudinal development and internal structure. BMC Medicine,
10(1), 124.
38/38
References 2/2
[Morrison, 2012] Morrison, Louise (2012) 5 reasons why I can’t find Open Access publications. http://mmitscotland.wordpress.com/2012/08/06/5-reasons-why-i-cant-find-open-access-publications-2/ [OAI-PMH v2.0, 2008] The Open Archives Initiative Protocol for Metadata Harvesting Version 2.0 (OAI-PMH), Impementation Guidelines (2008). http://www.openarchives.org/OAI/openarchivesprotocol.html [ResourceSync draft, 2013] ResourceSync protocol draft. 2013 http://www.niso.org/workrooms/resourcesync/ [Salo, 2008] Salo, D. (2008). Innkeeper at the roach motel. Library Trends, 57(2), 98-123.[Van de Sompel et al, 2004] Van de Sompel, H., Nelson, M. L., Lagoze, C., & Warner, S. (2004). Resource harvesting within the OAI-PMH framework. D-lib magazine, 10(12), 1082-9873.