Presentation from ALA Midwinter 2014 on Elsevier's new Text and Data Mining Policy


Published on

My presentation at ALA Midwinter 2014 announcing Elsevier's new Text and Data Mining Policy

Published in: Technology, Education
1 Comment
  • Great presentation, Chris
    Are you sure you want to  Yes  No
    Your message goes here
No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide
  • The requirement for licensing TDM output as CC-BY-NC only really applies to TDM Output insofar as it includes copyrightable material derived from the Elsevier content. For instance, we don’t require that any papers written about the TDM findings are published under CC-BY-NC.
  • Elsevier is one of 16 publishers signing the declaration. Note that the Elsevier policy is _not_ limited only to the EU.
  • TDM is a very large domain that encompasses many concepts. What follows are three examples of text mining-based operations on scholarly literature.
  • Content tagging is not new – we (and librarians) have done that for years – manually. Thesauri are used to classify and index publications. Text mining technologies, however, allow us to that in automated ways, covering vast corpuses very quickly and increasingly accurately.
  • Text mining is no longer purely the subject of research itself, nor just an experimental tool used by tech-savvy researchers. It is a core component of an increasing number of research projects and for many of those projects, text mining is critical for their success. An example of a project that uses text- and data mining extensively is the Blue Brain project, one of the flagship research projects in the European union and recipient of a 500 million euro (>650M US dollars) grant.
  • Here is a typical text mining workflow, courtesy of NaCTeM at University of Manchester:A corpus of source documents is collected and preparedA part-of-speech tagger runs over the documents, breaking down the sentences by grammatical syntaxTerms are normalized: plural vs. singular, alternative spellings, etc.Acronyms are identifiedA value is calculated that expresses the accuracy of a specific term’s automatic recognitionExtracted terms are consolidated into a databaseThe source documents are annotated with the extracted, normalized terms, and the terms are analyzed
  • Elsevier has supported requests from researchers for text mining access on an ad-hoc basis for several years. Demand has gradually increased as computing power increased and more tools became available. In 2012, we made investments to improve our technological support for these needs, and in 2013, ran a formal pilot to better understand researcher needs and help us design an appropriate technical and legal policy to meet those needs.
  • One of the key learnings from the pilot was that academic needs for TDM typically fall into one of both of two categories.
  • As solution for TDM is designed to address several of the key challenges researchers currently have: specifically, to make it easier to gain permission to mine and to access content in a simple and effective way suitable for text mining workflows.
  • Librarians also face challenges in supporting TDM needs of their researchers. Our new policy is designed to allow librarians to fulfill those needs quickly and simply with no additional cost.
  • Our experiences have led us to two conclusions: we need to a. support basic access for do-it-yourself text mining, and b. there is an opportunity for making text mining easier
  • Probably good to mention that you’ll talk about Prospect later.
  • Probably good to mention that you’ll talk about Prospect later.
  • Probably good to mention that you’ll talk about Prospect later.
  • Probably good to mention that you’ll talk about Prospect later.
  • We also recognize that researchers typically want to mine content across multiple publishers, so we are leading participants in a new industry initiative designed to make that easier.
  • Presentation from ALA Midwinter 2014 on Elsevier's new Text and Data Mining Policy

    1. 1. Facilitating Text and Data Mining ALA Midwinter 26 Jan 2014 Chris Shillum VP, Platform and Content Elsevier A&G Research Markets
    2. 2. Outline • • • • • Our new policy What is Text and Data Mining (TDM)? Timeline of TDM at Elsevier Practicalities Outlook 2
    3. 3. Elsevier’s New Text and Data Mining Policy Researchers at academic institutions can text mine subscribed content on ScienceDirect for noncommercial purposes via the ScienceDirect APIs Access is granted to faculty, researchers, staff and students at the subscribing institution Text mining output can be shared publically under these conditions 1. May contain "snippets" of up to 200 characters of the original text 2. Should be licensed as CC-BY-NC 3. Should include DOI link to original content Corporate and other subscribers • Your Elsevier Account Manager will be happy to discuss options to meet your needs Open access content • Text and Data mining permission are determined by the author's choice of user license. This information is detailed in the individual articles 3
    4. 4. Policy Aligned with the Recent STM Declaration on TDM 4
    5. 5. Text and Data Mining – What is it? Data mining: “the computational process of discovering patterns in large data sets involving methods at the intersection of artificial intelligence, machine learning, statistics, and database systems” Text mining: essentially a subset/“input step” to data mining: turning text into structured data for analysis, via application of natural language processing (NLP) and analytical methods. Typical text mining tasks include text categorization, text clustering, concept/entity extraction, production of granular taxonomies, sentiment analysis, document summarization, and entity relation modeling (i.e., learning relations between named entities). 5
    6. 6. TDM Examples – Entity Tagging and Linking 6
    7. 7. TDM Examples – Information Extraction 7
    8. 8. TDM Examples – Information extraction 8
    9. 9. TDM Examples – Natural Language Modelling 9 9
    10. 10. Text mining is Becoming an Essential Tool Blue Blain project • Reconstructing the brain piece by piece and building a virtual brain in a supercomputer • Received €500M (~$685m) of EU funding “The success of the Blue Brain project depends on very high volumes of standardized, high quality experimental data covering all possible levels of brain organization. Data comes both from the literature (via the project’s automatic information extraction tools) and from experimental work conducted by the project itself.” 10
    11. 11. Typical TDM Workflow Source: Sophia Ananiadou/National Center for Text Mining, University of Manchester, UK 11
    12. 12. Supporting TDM at Elsevier 2006 … • Began to support ad-hoc TDM access requests from customers • Low but consistently increasing level of interest from early adopters as computing power increases and tools get better • First Content Mining policy published • New APIs and Content Syndication Service rolled out to 2012 provide better technical solutions for TDM content access • Pilot with ~30 academic customers to better understand needs and define future policy 2013 • New Text and Data Mining policy for academic customers announced 2014 12
    13. 13. TDM Pilot Learnings – Use Cases Most academic TM requests fall in one or both of these categories: 1. Answering a specific research question 2. Building a new data resource for the community • How long does it take for concepts in STM literature to reach general media? • What is the relationship between the research and consulting commitments of economics and finance professors? • What are the characteristics of subjects in social psychology experiments? • An HIV mutation database for which mutations found in literature are mapped to the underlying database sequence • A database on growth and alimentation of fishes, and develop a fish classification to identify new species for aquaculture • A database with the electrophysiological properties of diverse neuron types 13
    14. 14. TDM Pilot Learnings – Researcher Challenges Technical • Obtaining necessary infrastructure • Having to deal with different formats from content providers • Sourcing and understanding TDM technology Functional • Fine-tuning pipeline, curating output, representing output meaningfully Logistical/Legal • Gaining access to the needed content • Gaining permission to mine the content 14
    15. 15. TDM Pilot Learnings – Library Challenges Expertise • Understanding specific TDM-based projects well enough to assess implications & offer advice to patrons Legal • Understanding and tracking what is allowed for what resources • Negotiating permissions with multiple providers • Ensuring academic freedom is protected Financial • Concerns about any additional costs • Understanding how TDM affects usage figures for the library 15
    16. 16. TDM Pilot – Conclusions 1. Elsevier can offer enhanced value to ScienceDirect customers by including basic text and data mining access rights in subscription agreements 2. Self-service access to content for TDM via our APIs meets the needs of most researchers in academia 3. There is demand for services beyond basic content access that make text mining easier 16
    17. 17. TDM Access: What do Institutions Need to do? • TDM access clause will be part of standard ScienceDirect subscription agreement for new academic customers and upon renewal • For existing agreements, an add-on contract amendment is available – just contact your Elsevier Account Manager • After signing institutional agreement/amendment, access to our API key registration page for your researchers will be enabled for your institution’s IP address range 17
    18. 18. TDM Access: What do Researchers Need to do? 1. Register at 18
    19. 19. TDM Access: What do Researchers Need to do? 2. Accept a simple click-through agreement 19
    20. 20. TDM Access: What do Researchers Need to do? 3. Obtain an Elsevier API Key 20
    21. 21. TDM Access: What do Researchers Need to do? 4. Use API Key to retrieve full text of journal articles and book chapters via the Elsevier API • Elsevier XML and plain-text formats supported • We are looking into supporting specialized textmining friendly XML formats 5. Process retrieved full-text through text mining tools/workflow of choice Documentation available on developer portal 21
    22. 22. Why an API? Can’t we just crawl • We are providing separate channels for machine-tomachine and human access to content • This helps us to maintain site performance, availability and reliability for everyone • Other major web information sources have similar policies • Wikipedia: • Twitter: • PubMed Central: • People who text-mine prefer APIs!! 22
    23. 23. Accessing an HTML page 23
    24. 24. Accessing an XML document 24
    25. 25. Supporting Cross-publisher TDM Prospect is a new service from CrossRef providing two components to address the issue of text and data mining scholarly literature across multiple publishers: • The “Prospect Common API” (PCAPI) can be used to access the full text of content identified by CrossRef DOIs across publisher sites and regardless of their business model. • The “Prospect License Registry” (PLR) can (optionally) be used by researchers and publishers as an efficient mechanism to provide “click-through” agreement of proprietary TDM licenses. • Both components are free to use by researchers and the public 25
    26. 26. Prospect Workflow 26
    27. 27. Elsevier and Prospect Elsevier is the first publisher to fully integrate with the Prospect beta system: • Researchers may read and agree to the Elsevier TDM clickthrough agreement via the Prospect License Registry • Researchers may use a Prospect token to access Elsevier content through the Prospect Common API rather than using the Elsevier-specific API Key and Elsevier API • Content is available in the same formats as the Elsevier API 27
    28. 28. Outlook: Beyond Basic Access Text Mining as a Service • Pilot with NaCTeM to integrate their tools with Elsevier content • Hosted in the cloud • Avoids the need for researchers to build and maintain TDM infrastructure • Ability to define and execute TDM workflows in a graphical environment 28
    29. 29. Takeaway Points • Researchers at academic institutions can now textmine subscribed Elsevier content for noncommercial purposes at no additional cost • Contact your Elsevier Account Manager if you are interested • Elsevier is collaborating with customers and industry partners to make text mining easier • More info at: 29
    30. 30. Thankyou! @cshillum This work is licensed under a Creative Commons Attribution 4.0 International License. 30