Swinburne University of Technology

Aland Astudillo

### How to use GROBID to extract text from PDF

<figure>
<img
src="https://cdn-images-1.medium.com/max/1024/1*uyD7hkSBZGCOn4wqu2TgMw.png" />
<figcaption>Created using DALL.E on 13 Feb 2024</figcaption>
</figure>

### Author

* Aland Astudillo (ORCID: [0009-0008-8672-3168](https://orcid.org/0009-0008-8672-3168))

GROBID is a powerful and useful tool based on machine learning that can
extract text information from PDF files and other files to a structured
format.

One of the key challenges in knowledge mining from academic articles is
reading the content of PDF files. The ability to efficiently and
accurately read the content of PDF documents using automated tools and
transform this data into a structured model provides many opportunities
for the integration of PDF collections to expert systems, and
AI-assisted data pipelines such as chatbots.

In this article, we will discuss GROBID capabilities, explain how to
install it, and explore how to use it in a Python script or Jupyter
Notebook.

### **What is GROBID?**

GROBID stands for GeneRation of Bibliographic Data and is a machine
learning library for extracting, parsing, and re-structuring raw
documents such as PDF into structured XML/TEI encoded documents. It has
a particular focus on technical and scientific publications. TEI stands
for Text Encoding Initiative (<https://tei-c.org/Guidelines/>). It
defines and documents a markup language for representing the structural,
renditional, and conceptual features of texts.

### How it works?

GROBID is a machine-learning library that extracts information from PDF
files. The goal of GROBID is to help the text mining process,
information extraction, and semantic analysis of scientific publications
by transforming them into machine-friendly, structured, representations
with predictable structure.

Despite the large amount of PDF files and their extensive use among all
fields, it is still a big challenge to process them because they are
encoded in a vast variety of different publisher formats, often
incomplete or inconsistent. In short, PDF files are difficult to exploit
and use.

In a nutshell, GROBID uses a cascade of "sequence labelling models" to
parse a document. Each model from the collection of models is in charge
of identifying a small amount of labels to classify areas and
information in the document. The final models produce 55 labels where
each piece of identified information from the document will be
structured. More information about how this library has been developed
and the machine learning models are in this link:
<https://grobid.readthedocs.io/en/latest/Principles/>

### How to install GROBID

Here we explain the installation of the GROBID tool as a service in a
Linux machine (Ubuntu for example). The same steps apply to a Windows
system using the Windows Subsystem for Linux tool (WSL version 2).

For Linux, in the terminal with super user (sudo) rights create a folder
and once inside, proceed to download the last version of GROBID using
the wget tool. The current version of GROBID is v 0.8.0. After that,
unzip the file:

    > wget https://github.com/kermitt2/grobid/archive/0.8.0.zip
    > unzip 0.8.0.zip

For Windows, you can use Docker containers to install and run the GROBID
service locally. More documentation is available at
<https://grobid.readthedocs.io/en/latest/Run-Grobid/>

Note: At the time of writing this article, GROBID requires JAVA(JDK) and
it works fine with JDK version 1.11 up to version JDK 1.17. Other recent
JDK versions should work correctly. To check your JAVA installation, in
the terminal use the following command:

    > java -version

If you do not have any JAVA installed, proceed to install JAVA using the
following line in the terminal.

    > sudo apt install default-jdk

It is a good practice to check Java after installation to ensure the
right version of Java is installed and ready to operate on your system.

### **How to run GROBID**

Once you downloaded and unzipped the "*grobid"* file, proceed to the
folder and run the grewed file. In the terminal ensure to use "./"
before the command "*grewed".*

    > cd grobid-0.8.0

    > ./grewed run

The first time that you run the service, it will download and install
all the required dependencies and then it will run the service. It will
be completed by printing a message "**:grobid-service:run"** as
presented in the following screenshot.

<figure>
<img
src="https://cdn-images-1.medium.com/max/920/1*gEsGJ87IUzEovXwcFRDCzg.png" />
</figure>

### **How to use GROBID**

After running the GROBID service, you can connect to the created server
in the following direction using your browser:

<http://localhost:8070/>

This follows the default configuration, but you can change the port in
the config file. The image below shows how it looks the first time that
you access the web layout.

<figure>
<img
src="https://cdn-images-1.medium.com/max/1003/1*N5URo-rzuOb_1Cu15yEKEA.png" />
</figure>

The TEI menu will allow you to choose among the different services to
call and to upload a file with the button "Select file". The following
functionalities are available as a service:

- Header extraction and parsing
- References extraction and parsing
- Citation contexts recognition and resolution
- Full-text extraction and structuring
- PDF coordinates
- Parsing of references
- Parsing of names
- Parsing of affiliation and address
- Parsing of dates
- Consolidation and resolution of the extracted bibliographical
  references
- Extraction and parsing of patent and non-patent references in patent
  publications
- Extraction of funders and funding information

Once you upload a PDF document and choose the required service, you can
press the Submit button

<figure>
<img
src="https://cdn-images-1.medium.com/max/995/1*m3zDNnvFesFQuQ7Bv0U3yA.png" />
</figure>

After a few seconds, the web will show you the XML structure of the
output depending on what service you have selected. You can download the
resulting XML structure in an XML file. The image below shows the result
of processing the full-text document for a given paper. For example, in
the image below we used the article *Pan et al., 2024 Unifying Large
Language Models and Knowledge Graphs: A Roadmap. arXiv:2306.08302*
downloaded from arXiv (<https://arxiv.org/abs/2306.08302>)

<figure>
<img
src="https://cdn-images-1.medium.com/max/993/1*xTfpktk_4nH1dOPqoETvkg.png" />
</figure>

### **How to use GROBID Python Client**

The GROBID Python client is a handy tool that can process efficiently
and concurrently a set of PDF files in a given folder using GROBID as a
service without requiring the web option. It includes a command line for
processing PDFs and writing the results in a given folder. It allows us
to use this tool inside Python scripts. To use this client, It assumes
that you are running GROBID service, as we saw in the previous steps.

The repository of this client is here
<https://github.com/kermitt2/grobid_client_python/blob/master/grobid_client/grobid_client.py>

To install the GROBID Python client to use in your scripts or Jupyter
Notebook, in your Python environment, you can use the pip manager in a
given environment:

    !pip install grobid_client_python

If you want to use this client in a script or a Jupyter Notebook, you
must import the GROBID client class. Remember that it assumes that the
GROBID service is already running.

    from grobid_client.grobid_client import GrobidClient

    client = GrobidClient(config_path="./config.json")

Once the client connects to the service it will show you the following
message.

    "GROBID server is up and running"

### **How to analyse a bunch of PDF documents**

After checking the connection, you can call the process method to
analyse a bunch of documents in a folder. The process method has 3
required parameters and other optional parameters. In the following
example we are using the service "**processFulltextDocument**" for
processing all the pdf files in the folder "**resources/test_pdf**" and
get the xml files output on the folder "**resources/test_out/**".
Additionally, we set the consolidate citation option and the TEI
coordinates to "true".

    client.process("processFulltextDocument", 
         "./resources/test_pdf", 
     output="./resources/test_out/", 
     consolidate_citations=True, 
     tei_coordinates=True, 
     force=True)

In the defined folder, the output will be the XML files for each PDF
file with the re-structured information.

### **How to analyse a single PDF file**

For analysing one PDF file at a time, you can use the method
**process_pdf**, with similar parameters as the **process** method:

    service_name = "processFulltextDocument"

    pdf_file = "./pdf_examples/input/0046d83a-edd6-4631-b57c-755cdcce8b7f.pdf"

    rsp = client.process_pdf(service_name, pdf_file, 
                             generateIDs=True, 
                             consolidate_header=True, 
                             consolidate_citations=True, 
                             include_raw_citations=True, 
                             include_raw_affiliations=True, 
                             tei_coordinates=True, 
                             segment_sentences=True)

The following text snippet shows the top section of the "rsp" variable
that holds the results from the PDF processing.

    ('./pdf_examples/input/0046d83a-edd6-4631-b57c-755cdcce8b7f.pdf',
     200,
     '<?xml version="1.0" encoding="UTF-8"?> 
    <TEI xml:space="preserve" xmlns="http://www.tei-c.org/ns/1.0"  
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"  
    xsi:schemaLocation="http://www.tei-c.org/ns/1.0 https://raw.githubusercontent.com/kermitt2/grobid/master/grobid-home/schemas/xsd/Grobid.xsd"  
    xmlns:xlink="http://www.w3.org/1999/xlink"> 
      <teiHeader xml:lang="en"> 
        <fileDesc> 
          <titleStmt> 
            <title level="a" type="main" xml:id="_nrS3kCw">Multi-contact functional electrical stimulation for hand opening: electrophysiologically driven identification of the optimal stimulation site</title> 
            <funder ref="#_ZTFTHCu"> 
              <orgName type="full">German Federal Ministry for Education and Research [BMBF</orgName> 
            </funder> 
            <funder ref="#_xTkqs2w"> 
              <orgName type="full">German Research Council</orgName> 
            </funder> 
            <funder ref="#_dV4wQd4 #_dVJJnCd"> 
              <orgName type="full">Federal Ministry for Education and Research</orgName> 
            </funder> 
          </titleStmt> 
          <publicationStmt> 
            <publisher>Springer Science and Business Media LLC</publisher> 
            <availability status="unknown">
       <p>Copyright Springer Science and Business Media LLC</p> 
            </availability> 
            <date type="published" when="2016-03-08">2016-03-08</date> 
          </publicationStmt> 
          <sourceDesc> 
            <biblStruct> 
              <analytic> 
                <author> 
                  <persName coords="1,56.69,234.15,85.88,9.71"><forename type="first">Cristiano</forename>
                  <surname>De Marchis</surname></persName> 
                  <email>cristiano.demarchis@uniroma3.it</email> 
                  <affiliation key="aff0"> 
                    <note type="raw_affiliation"><label>1</label> Division of Functional and Restorative Neurosurgery, Department of Neurosurgery, Eberhard Karls University, Otfried-Mueller-Str.45, 72076 Tübingen, Germany</note> 
                    <orgName type="department" key="dep1">Division of Functional and Restorative Neurosurgery</orgName> 
                    <orgName type="department" key="dep2">Department of Neurosurgery</orgName> 
                    <orgName type="institution">Eberhard Karls University</orgName> 
                    <address> 
                      <addrLine>Otfried-Mueller-Str.45</addrLine> 
                      <postCode>72076</postCode> 
                      <settlement>Tübingen</settlement> 
                      <country key="DE">Germany</country> 
                    </address> 
                  </affiliation> 
                </author> 
    …

(Note: The result was truncated)

As you can see, the information is organised hierarchically, using
standardised tags \<tag\> to identify each type of information of the
publication. For instance, the tag *\<titleStmt\>* contains sub-tags
holding the title of the article.

    <title level="a" type="main" xml:id="_nrS3kCw">
    Multi-contact functional electrical stimulation for hand opening: electrophysiologically driven identification of the optimal stimulation site
    </title> 

Other tags contain information about authors, affiliations, and so on.
For example, the identification of the organisation is in the tag
*\<orgName\>*

    <orgName type="full">
    German Federal Ministry for Education and Research [BMBF
    </orgName> 

Information about the authors is inside the tags *\<author\>* containing
details of each author, complete name (\<forename), last name
(*\<surname\>*), email (*\<email\>*), affiliations (*\<affiliation
key="aff0\"\>*), and all other information like the address of the
institution to which the author is affiliated (*\<address\>*), and all
the information contained in the article about the author.

    <author> 
      <persName coords="1,56.69,234.15,85.88,9.71">
        <forename type="first">Cristiano</forename>
        <surname>De Marchis</surname>
      </persName> 
      <email>cristiano.demarchis@uniroma3.it</email> 
      <affiliation key="aff0"> 
      <note type="raw_affiliation"><label>1</label> Division of Functional and Restorative Neurosurgery, Department of Neurosurgery, Eberhard Karls University, Otfried-Mueller-Str.45, 72076 Tübingen, Germany</note> 
        <orgName type="department" key="dep1">Division of Functional and Restorative Neurosurgery</orgName> 
        <orgName type="department" key="dep2">Department of Neurosurgery</orgName> 
        <orgName type="institution">Eberhard Karls University</orgName> 
        <address> 
          <addrLine>Otfried-Mueller-Str.45</addrLine> 
          <postCode>72076</postCode> 
          <settlement>Tübingen</settlement> 
          <country key="DE">Germany</country> 
        </address> 
      </affiliation> 
    </author>

In the same way, we can retrieve information on the citations of the
article in the tag for references. For example, an article in the list
of references has the following structure

    <biblStruct coords="8,319.75,636.26,214.89,6.62;8,319.75,645.27,210.17,6.62;8,319.75,654.28,118.20,6.62" xml:id="b0">
      <analytic>
        <title level="a" type="main">The influence of functional electrical stimulation on hand motor recovery in stroke patients: a review</title>
        <author>
          <persName xmlns="http://www.tei-c.org/ns/1.0">
            <forename type="first">Fanny</forename>
            <surname>Quandt</surname>
          </persName>
        </author>
        <author>
          <persName xmlns="http://www.tei-c.org/ns/1.0">
            <forename type="first">Friedhelm</forename>
            <forename type="middle">C</forename>
            <surname>Hummel</surname>
          </persName>
        </author>
        <idno type="DOI">10.1186/2040-7378-6-9</idno>
      </analytic>
      <monogr>
        <title level="j">Experimental & Translational Stroke Medicine</title>
        <title level="j" type="abbrev">Exp & Trans Stroke Med</title>
        <idno type="ISSNe">2040-7378</idno>
        <imprint>
          <biblScope unit="volume">6</biblScope>
          <biblScope unit="issue">1</biblScope>
          <biblScope unit="page">9</biblScope>
          <date type="published" when="2014-08-21"/>
          <publisher>Springer Science and Business Media LLC</publisher>
        </imprint>
      </monogr>
    </biblStruct>

In which we can identify the title tag (*\<title level="a"
type="main"\>*), authors information (*\<author\>*), DOI (*\<idno
type="DOI"\>*), and the journal details (*\<monogr\>*) of the referenced
article.

Finally, to get each piece of information from this structure for
further analysis, the next steps involves extracting the required
information based on the tags of the XML structure, which can be done
with different tools to match these tags in the text. That said, the
text is ready for mining all the information.

### **Conclusion**

We have explored the important aspects of GROBID. With its simple
installation process user interface, and scripts, we can put GROBID to
work easier than ever. Whether you are facing mountains of PDFs or
struggling to tame unruly bibliographic data, GROBID is your trusty
companion on the path to research efficiency. Download GROBID today and
unlock the door to a world of streamlined workflows, saving time, and
deeper insights from your research endeavours.

**Materials for further reading:**

- <https://grobid.readthedocs.io/en/latest/>
- <https://github.com/kermitt2/grobid_client_python>

![](https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=67df995b16fa){width="1"
height="1"}


How to use GROBID to extract text from PDF  Author  Aland Astudillo (ORCID: 0009-0008-8672-3168)  GROBID is a powerful and useful tool based on machine learning that can extract text information from PDF files and other files to a structured format. One of the key challenges in knowledge mining from academic articles is reading the content of PDF files.

How to use GROBID

Cameron Neylon

*[![goods](http://cameronneylon.net/wp-content/uploads/2015/11/goods-300x204.jpg){.alignleft
.size-medium .wp-image-19303 decoding="async" loading="lazy" width="300"
height="204"
srcset="https://cameronneylon.net/wp-content/uploads/2015/11/goods-300x204.jpg 300w, https://cameronneylon.net/wp-content/uploads/2015/11/goods.jpg 703w"
sizes="(max-width: 300px) 100vw, 300px"}](http://cameronneylon.net/wp-content/uploads/2015/11/goods.jpg)This
is an approximate rendering of my comments as part of the closing panel
of "[The End of Scientific Journal? Transformations in
Publishing](https://royalsociety.org/events/2015/11/transformations-in-publishing/)"
held at the Royal Society, London on 27 November 2015. It should be read
as a reconstruction of what I might have said rather than an accurate
record. The day had focussed on historical accounts of "journals" as
mediators of both professional and popular research communications. A
note of the meeting will be published. Our panel was set the question of
"will the journal still exist in 2035".*

Over the course of 2015 I've greatly enjoyed being part of the series of
meetings looking at the history of research communications and
scientific journals in the past. In many cases we've discovered that our
modern concerns, today the engagement of the wider public, the challenge
of expertise, are not at all new, that many of the same issues were
discussed at length in the 17th, 18th and 19th centuries. And then there
are moments of whiplash as something incomprehensible streaks past:
[Pietro
Corsi](http://www.history.ox.ac.uk/faculty/staff/profile/corsi/index.html)
telling us that dictionaries were published as periodicals; [Aileen
Fyfe](https://risweb.st-andrews.ac.uk/portal/da/persons/aileen-fyfe%2859c3c852-a83d-497c-a5a6-2d3a0b045857%29.html)
explaining that while papers given to at Royal Society meetings were
then refereed, the authors could make only make "verbal" not
intellectual changes to the text in response; [Jon
Topham](http://www.leeds.ac.uk/arts/profile/20042/464/jonathan_topham)
telling us that chemistry and physics were characterised under
literature in the journals of the early 19th century.

So if we are to answer the exam question we need to address the charge
that [Vanessa
Heggie](http://www.birmingham.ac.uk/staff/profiles/applied-health/heggie-vanessa.aspx)
gave us in the first panel discussion. What has remained the same? And
what has changed? If we are to learn from history then we need to hold
ourselves to a high standard in trying to understand what it is (not)
telling us. Prediction is always difficult, especially about the
future...but it wasn't [Niels
Bohr](http://en.wikipedia.org/wiki/Niels_Bohr "Niels Bohr"){.zem_slink
target="_blank" rel="wikipedia"} who first said that. A Dane would
likely tell us that "det er svÃ¦rt at spÃ¥, isÃ¦r om fremtiden" is a
quote from Storm Peterson or perhaps [Piet
Hein](http://en.wikipedia.org/wiki/Piet_Hein_%28scientist%29 "Piet Hein (scientist)"){.zem_slink
target="_blank" rel="wikipedia"}, [but it probably has deeper
roots](http://chaosbook.blogspot.co.uk/2010/06/lundskovdk-citater.html).
It's easy to tell ourselves compelling stories, whether they say that
"everything has changed" or that "it's always been that way", but
actually checking and understanding the history matters.

So what has stayed the same? We've heard throughout today the importance
of groups, communities. Of authors, of those who were (or were trying to
be) amongst the small group of paid professors at UK universities. Of
the distinctions between communities of amateurs and of professionals.
We've heard about language communities, of the importance of who you
know in being read at the Royal Society and of the development of
journals as a means of creating research disciplines. I think this
centrality of communities, of groups, of *clubs* is a strand that links
us to the 19th century. And I think that's true because of the nature of
knowledge itself.

Knowledge is a slippery concept, and [I've made the argument
elsewhere](http://cameronneylon.net/blog/the-limits-on-open-why-knowledge-is-not-a-public-good-and-what-to-do-about-it/),
so for now I'll just assert that it belongs in the bottom right quadrant
of [Ostrom](https://en.wikipedia.org/wiki/Elinor_Ostrom)'s
categorisation of goods. Knowledge is non-rivalrous â€" if I give it to
you I still have it â€" but also excludable â€" I can easily prevent you
from having it, by not telling you, or by locking it up behind a
paywall, or simply behind impenetrable jargon. This is interesting
because [Buchannan](https://en.wikipedia.org/wiki/James_M._Buchanan)'s
work on [the economics of
clubs](https://en.wikipedia.org/wiki/Club_good) show us that it is
precisely the goods in this quadrant which are used to sustain clubs and
make them viable.

The survival of journals, or of scholarly societies, disciplines or
communities, therefore depends on how they deploy knowledge as a *club
good.* To achieve this deployment it is necessary to make that
knowledge, the club good, less exclusive, and more public. What is nice
about this view is that it allows us, to borrow Aileen Fyfe's language,
to talk about "public-making" ([Jan
Velterop](http://theparachute.blogspot.co.uk/) has used the [old
term](https://en.wiktionary.org/wiki/publicate) "publicate" in a similar
way) as an activity which includes public engagement, translation, and
â€" to [Rebekah
Higgitt](https://www.kent.ac.uk/history/staff/profiles/higgitt.html)'s
point â€" education as well as scholarly publishing as we traditionally
understand it as overlapping subsets of this broader activity.

But what has changed? I would argue that the largest change in the 20th
century was one of scale. The massive increase in the scale and
globalisation of the research enterprise, as well as the rise of
literacy meant that traditional modes of coordination, within scholarly
societies and communities, and beyond to interested publics were
breaking down. To address this coordination problem we took knowledge as
a club good and privatised it, introducing copyright and intellectual
property as a means of engaging corporate interests to manage the
coordination problem for us. It is not an accident that the scale up,
the introduction of copyright and IP to scholarly publishing, and
scholarly publishing becoming (for the first time) profitable all
co-incide. The irony of this, is that by creating larger, and clearly
defined markets, we solved the problem of market scale that troubledÂ 
early journals that needed to find both popular and expert audiences, by
locking wider publics out.

The internet and the web also changed everything, but its not the cost
of reproduction that most matters. The critical change for our purpose
here is the change in the economics of discovery. As part of our
privatisation of knowledge we parcelled it up into journals, an
industrial broadcast mechanism in which one person aims with as much
precision as possible to reach the right, expert, audience. The web
shifts the economics of discovering expertise in a way that makes it
viable to discover, not the expert who knows everything about a subject,
but the person who just happens to have *the right* piece of knowledge
to solve a specific problem.

These two trends are pulling us in opposite directions. The industrial
model means creating specialisation and labelling. The creation of
communities and niches that are, for publishers, markets that can be
addressed individually. These communities are defined by credentialling
and validation of deep expertise in a given subject. The ideas of
micro-expertise, of a person with no credentials having the key
information or insight radically undermines the traditional dynamics of
scholarly group formation. I don't think it is an accident that those
scholarly communities that [MichÃ¨le
Lamont](http://scholar.harvard.edu/lamont/home) identifies as having the
most stable self conception have a tendency to being the most
traditional in terms of their communication and public engagement.
Lamont identifies history (but not as [Berris
Charnley](http://berrischarnley.com/) reminded me the radicals from the
history of science here today!) and (North American analytical)
philosophy in this group. I might add synthetic chemistry from my own
experience as examples.

It is perhaps indicative of the degree of siloing that I'm a trained
biochemist at a history conference, telling you about economics â€" two
things I can't claim any deep expertise in â€" and last week I gave a
talk from a cultural theory perspective. I am merrily skipping across
the surface of these disciplines, dipping in a little to pull out
interesting connections *and no-one has called me on it\*.* You are
being forced, both by the format of this panel, and the information
environment we inhabit, to assess my claims not based on my PhD thesis
topic or my status or position, but on how productively my claims and
ideas are clashing with yours. We discover each other, not through the
silos of our disciplinary clubs and journals, but through the networked
affordances that connect me to you, that in this case we could trace
explicitly via Berris Charnley and [Sally
Shuttleworth](http://www.english.ox.ac.uk/about-faculty/faculty-members/victorian-period/shuttleworth-professor-sally).
That sounds to me rather more like the 19th century world we've been
hearing about today than the 20th century one that our present
disciplinary cultures evolved in.

This restructuring of the economics of discovery has profound
implications for our understanding of expertise. And it is our cultures
of expertise that form the boundaries of our groups â€" our knowledge
clubs â€" whether they be research groups, disciplines, journals,
discussion meetings or scholarly societies. The web shifts our
understanding of public-making. It shifts from the need to define and
target the expert audience through broadcast â€" a one-to-audience
interaction â€" to a many-to-many environent in which we aim to connect
with the right person to discover the right contribution. The importance
of the groups remains. The means by which they can, and should want to
communicate has changed radically.

The challenge lies, not in giving up on our ideas of expertise, but in
identifying how we can create groups that both develop shared
understanding that enables effective and efficient communication
internally but are also open to external contributions. It is not that
defining group boundaries doesn't matter, it is crucial, but that the
shape and porosity of those boundaries needs to change. Journals have
played a role throughout their history in creating groups, defining
boundaries, and validating membership. That role remains important, it
is just that the groups, and their cultures, will need to change to
compete and survive.

We started 2015 with the idea that the journal was invented in 1665.
This morning we heard from Jon Topham that the name was first used in
the early 19th century, but for something that doesn't look much like
what we would call a journal today. I believe in 20 years we will still
have things called journals, and they will be the means of mediating
communications between groups, including professional scholars and
interested publics. They'll look very different from what we have today
but their central function, of mediating and expressing identity for
groups will remain.

\* *This is not quite true. [Martin Eve](https://www.martineve.com/) has
called me on skipping too lightly across the language of a set of
theoretical frameworks from the humanities without doing sufficient work
to completely understand them. I don't think it is co-incidental that
Martin is a cultural and literary scholar who also happens to be a
technologist, computer programmer and deeply interested in policy design
and implementation, as well as the intersection of symbolic and
financial economies.*


<em>
 This is an approximate rendering of my comments as part of the closing panel of “The End of Scientific Journal? Transformations in Publishing” held at the Royal Society, London on 27 November 2015. It should be read as a reconstruction of what I might have said rather than an accurate record. The day had focussed on historical accounts of “journals” as mediators of both professional and popular research communications.
</em>

Rogue Scholar Posts

How to use GROBID

The end of the journal? What has changed, what stayed the same?