Annodoc annotation documentation support system

This is a page of documentation created using the Annodoc system. It serves a double purpose as documentation for the Annodoc system and as a template that you can use as a starting point for creating your own documentation using annodoc.

What is Annodoc?
How does it work?
Getting started
Editing documents
Adding documents
Visualizations
Collections
Configuration
Troubleshooting

What is Annodoc?

Annodoc is a documentation support system focusing in particular on guidelines for text annotation. The system combines ease of editing using a simple plain-text like format with support for the visualization of complex, structured text annotation and close integration with version control.

A brief introduction of Annodoc [PDF] was presented at SLTC’14.

For example, consider the following fragment of documentation

Mentions of person names are annotated as PERSON

Barack Obama is the current president.
T1 PERSON 0 12 Barack Obama

(Not seeing a visualization above? See troubleshooting)

this example is generated from the following input:

Mentions of person names are annotated as [PERSON]()

~~~ ann
Barack Obama is the current president.
T1 PERSON 0 12 Barack Obama
~~~

How does it work?

The text content of Annodoc documents is in the simple Markdown format (with the option to include HTML), and the data for the visualizations is represented in any of a number of supported annotation formats such as Stanford dependency, CoNLL-X, CoNLL-U, or .ann standoff.

The final documents seen in the browser are a combination of (X)HTML and SVG for the visualizations. The primary tools for generating the final documents from the source are the Jekyll static web site generator, which converts Markdown into (X)HTML, and the brat annotation tool, which generates the SVG visualizations, with help from some Annodoc-specific extensions.

The following figure shows the primary stages of processing.

In addition to Markdown processing, Jekyll provides numerous other features such as document collection management as well as templating and simple scripting using the Liquid language, and Annodoc provides various facilities for supporting guideline development such as automatic linking of defined types.

Finally, to support the development and maintenance of complex sets of documentation, Annodoc features close integration with the Git distributed version control system through the GitHub web-based hosting service, allowing automatic conversion and publication of the most recent version of the online documentation each time the sources are updated.

For more information on these technologies, please see the following:

Jekyll: http://jekyllrb.com/
brat: http://brat.nlplab.org
Markdown: http://daringfireball.net/projects/markdown/
Kramdown (Markdown superset): http://kramdown.gettalong.org/
Liquid: http://wiki.shopify.com/Liquid
Stanford Dependency format: http://nlp.stanford.edu/software/stanford-dependencies.shtml
CoNLL-X format: http://ilk.uvt.nl/conll/#dataformat
CoNLL-U format: http://universaldependencies.github.io/docs/format.html
.ann standoff format: http://brat.nlplab.org/standoff.html
Git: http://git-scm.com/
GitHub: http://github.com

Getting started

First, you need to get a copy of the Annodoc system. If you don’t have one already, head over to the online Annodoc repository and download, clone or fork the repository. The former two options (download or clone) will give you a local copy of the system (see running locally below), while cloning the repository can allow you to work within the online version control system (see working online).

To download the system source, navigate to the Annodoc repository and click on the Download ZIP link on the right-hand toolbar. You should then unpack the downloaded .zip file in an appropriate location and move to the directory with the source. For the next steps, see running locally below.

To clone the system, you will need the Git version control system. After making sure that Git is available, navigate to the Annodoc repository and copy the “clone URL” from the right-hand toolbar. Then, run the command git clone <URL>, where <URL> is the clone URL copied previously. Finally, move to the cloned directory and see running locally below for the next steps.

To fork the system, you will need a GitHub account. Then, while logged in to GitHub, navigate to the Annodoc repository and click on the Fork button on the top right. This will give you a copy of the repository that you can work on in GitHub. You can then download or clone this copy, as above, to run the system locally, or work online (see below).

Running locally

Please note: the Annodoc system has been developed and tested in a linux environment and has so far not been specifically tested on Windows. We strongly recommend a Unix or workalike system (OSX or linux) for running the system locally.

For running Annodoc locally, you will need to have a recent version of Jekyll. In particular, version 2.0 or greater is required for collection features.

After getting a local copy of the system (see Getting started above), you can build the site from sources and serve the resulting documents by navigating to the directory with annodoc sources and using the command

jekyll serve --watch

(here, the option --watch specifies that Jekyll should watch the source documents for changes and rebuild the site when there are any.)

If the above command runs successfully, you should see something like the following:

$ jekyll serve --watch
            Source: .
       Destination: ./_site
      Generating... done.
 Auto-regeneration: enabled
    Server address: http://0.0.0.0:4000/
  Server running... press ctrl-c to stop.

Here, the value Server address (commonly http://0.0.0.0:4000/) is a URL where Jekyll is serving the site from. You should be able to see the documentation by navigating to this address using your browser. If you do not see the documentation at this address, please see troubleshooting below.

Working online

In addition to running Jekyll locally (see running locally), it is possible to use the Annodoc system entirely online, relying on the facilities of GitHub (or a workalike) for generating the site from source using Jekyll. The following assumes that the repository is stored in GitHub unless separately mentioned.

GitHub provides website hosting with Jekyll support through GitHub pages. Using GitHub pages with a GitHub repository is largely automatic: simply make sure that the source documents are held in the gh-pages branch of your repository, and the processed final documents will be automatically made available at http://username.github.io/repository, where username is your GitHub username and repository is the name of the repository.

Note: GitHub pages are publicly visible even for private GitHub repositories.

For further information setting up GitHub pages, please refer to the documentation at https://pages.github.com/.

Editing documents

You can either edit source documents directly in your favorite text editor, or via a simple online interface.

Editing locally

If you have a local copy of an Annodoc system (see Getting started above), you can edit any of the source documents directly, using any text editor.

Most typically, changes are made to one or more Markdown (.md) files. To see the effect of these changes in the final documentation, the source needs to be processed through Jekyll. (see Running locally; note that this processing is automatic with the --watch option).

If you are working with Git for version control, you should commit and push your changes to the Git repository when you are done, e.g. by running the following

git commit DOCUMENT.md
git push

(where DOCUMENT.md is the name of the changed document.)

Alternatively, you can run git commit -a to commit all changed documents.

Editing online

If you are working with a repository stored in GitHub (or similar), you can use online editing features.

Please note: the “edit page” link will only appear if the editurl site variable is set in _config.yml (see Configuration). Please make sure that your configuration is correctly set up to use these features.

When configured for online editing, the header of each Annodoc documentation page will contain an “edit page” link that leads to the GitHub online editing page, shown in the following:

This feature allows version-controlled documents to be edited directly from your browser. To get started, the only relevant parts are the large black edit area and the “Cancel” and “Commit changes” buttons at the bottom.

To give this a quick try, click on the following link: edit sandbox document. This should open a “sandbox” document in a new tab. (If this does not work, please see troubleshooting below.) After testing it out, feel free to either cancel without saving your changes, or save them into the version control system using the “Commit changes” button. You can see the resulting document here (reload to see changes, and please note it may take some time for the changes to show up.)

For experimenting with the system, we recommend using the sandbox document instead of any “real” documents.

To edit the actual documentation, first find the page you’re interested in. Then,

Click on the “edit page” link on the top
Make your changes in the GitHub editor
(Optional: add a message describing your changes in the “Commit changes” box)
Click on “Commit changes”

Finally, wait a moment for your revisions to the documentation to be compiled (normally no more than 10 seconds) and reload the documentation page to make sure they look right.

Adding documents

To add new documents into the system, simply create a new .md document and add the following front matter at the beginning of the document:

---
layout: entry
title: DOCUMENT-TITLE
---

(where DOCUMENT-TITLE is the title you wish to give to the document.)

This YAML front matter is required for Jekyll to identify the document as markdown.

If you are using the GitHub (or similar), you can create new documents simply by clicking on the + sign at the top or the repository file listing.

If you are working locally with Git for version control, you can create new documents with any text editor and then add newly created documents to the Git repository, e.g. by running the following commands:

git add DOCUMENT.md
git commit
git push

(where DOCUMENT.md is the name of the new document.)

Visualizations

As explained in the section what is annodoc?, visualizations are created using a simple syntax that identifies a section of the document as being annotation in some particular format. The system then replaces these sections with the corresponding visualizations.

This section provides examples of various ways to create visualizations.

Visualization basics

There are a number of ways to add visualizations into a document. The most straightforward one is to wrap annotations in a format FORMAT with the lines ~~~ FORMAT (before) and ~~~ (after). For example,

~~~ ann
Barack Obama is the current president.
T1 PERSON 0 12 Barack Obama
~~~

gives the following visualization:

Barack Obama is the current president.
T1 PERSON 0 12 Barack Obama

Above, the FORMAT value ann specifies that the source content is in the .ann standoff format. Currently, the following formats are supported:

.ann standoff format: http://brat.nlplab.org/standoff.html
Stanford Dependency format: http://nlp.stanford.edu/software/stanford-dependencies.shtml
CoNLL-X format: http://ilk.uvt.nl/conll/#dataformat
CoNLL-U format: http://universaldependencies.github.io/docs/format.html

Some more simple examples follow.

A single tree in the Stanford Dependency format can be embedded using the following syntax:

~~~ sdparse
Dogs run
nsubj(run, Dogs)
~~~

which results in this embedded visualization:

Dogs run
nsubj(run, Dogs)

Parses represented in the CoNLL-X format can be embedded as illustrated by the following example:

~~~ conllx
1    Dogs   dog    _    NNS    _    2    nsubj
2    run    run    _    VBP    _    0    ROOT
~~~

gives

1    Dogs   dog    _    NNS    _    2    nsubj
2    run    run    _    VBP    _    0    ROOT

Finally, parses in the CoNLL-U format can be embedded as shown below:

~~~ conllu
1    They    they    PRON    PRN    Case=Nom|Num=Plur            2    nsubj    _    _
2    buy     buy     VERB    VBP    Num=Plur|Per=3|Tense=Pres    0    root     _    _
3    books   book    NOUN    NNS    Num=Plur                     2    dobj     _    _
~~~

resulting in

1    They    they    PRON    PRN    Case=Nom|Num=Plur            2    nsubj    _    _
2    buy     buy     VERB    VBP    Num=Plur|Per=3|Tense=Pres    0    root     _    _
3    books   book    NOUN    NNS    Num=Plur                     2    dobj     _    _

The remainder of this section provides further details for each supported format and additional information on options for embedding visualizations.

.ann standoff format

This section provides information on the .ann standoff (Ann) format. For full documentation, please refer to http://brat.nlplab.org/standoff.html.

Ann format: annotation primitives

The following example illustrates some of the basic Ann format annotation primitives that are supported in visualizations. The annotation

~~~ ann
Sony formed a joint venture with Ericsson, a company based in Sweden.
T1 Organization 0 4	Sony
T2 MERGE-ORG 14 27	joint venture
T3 Organization 33 41	Ericsson
E1 MERGE-ORG:T2 Org1:T1 Org2:T3
T4 Country 62 68	Sweden
R1 Origin Arg1:T3 Arg2:T4
A1 Name T4
~~~

is visualized as follows:

Sony formed a joint venture with Ericsson, a company based in Sweden.
T1 Organization 0 4	Sony
T2 MERGE-ORG 14 27	joint venture
T3 Organization 33 41	Ericsson
E1 MERGE-ORG:T2 Org1:T1 Org2:T3
T4 Country 62 68	Sweden
R1 Origin Arg1:T3 Arg2:T4
A1 Name T4

This example involves text-bound annotations, an attribute annotation, an event annotation, and a relation annotation. All annotations occupy a single line of the input and begin with a unique identifier. The remaining structure varies depending on the annotation primitive.

Ann format: text-bound annotations

Text-bound annotations identify a span of text using (start, end) offsets and assign it a type (note that the marked text is repeated for reference):

T1 Organization 0 4	Sony
T2 MERGE-ORG 14 27	joint venture
T3 Organization 33 41	Ericsson
T4 Country 75 81	Sweden

Text-bound annotations can be used, for example, to mark mentions of specific named entities in text (Sony, Ericsson and Sweden above) or the “triggers” of event annotations (joint venture).

In addition to simple (start, end) spans, discontinuous span annotations are also supported using the syntax START END[;START END[...]]. For example,

~~~ ann
Barack and Michelle Obama
T1 PERSON 0 6;20 25 Barack Obama
T2 PERSON 11 19 Michelle
~~~

gives

Barack and Michelle Obama
T1 PERSON 0 6;20 25 Barack Obama
T2 PERSON 11 19 Michelle

Ann format: attribute annotations

Attribute annotation associate either a binary flag (e.g. IsName) or a key-value pair (e.g. Tense, Past) with another annotation, such as an entity mention (textbound) or an event annotation.

A1 IsName T4
A2 Tense E1 Past

Ann format: relation annotations

Relation annotations relate two other annotations and have a type (e.g. Origin).

R1 Origin Arg1:T3 Arg2:T4

Ann format: event annotations

Event annotations associate any number of annotations in specific roles (e.g. Theme, Cause), identifying the event with a type (e.g. MERGE-ORG) and a textbound annotation expressing it in text.

E1 MERGE-ORG:T2 Org1:T1 Org2:T3

Ann format: other annotation primitives

The .ann standoff format additionally supports equivalence relations, normalization annotations and comment (or note) annotations. These annotation primitives are presently not supported in Annodoc.

Stanford Dependency format

This section provides details on the Stanford Dependency (SD) format.

SD format basics

Basic SD format visualizations consist of a single line of text followed by any number of typed dependencies between words using the simple format TYPE(FROM, TO). For example, the input

~~~ sdparse
The quick brown fox jumped
det(fox-4, The-1)
amod(fox-4, quick-2)
amod(fox-4, brown-3)
nsubj(jumped-5, fox-4)
~~~

gives

The quick brown fox jumped
det(fox-4, The-1)
amod(fox-4, quick-2)
amod(fox-4, brown-3)
nsubj(jumped-5, fox-4)

SD format: ambiguous tokens

If your example has several instances of the same token, you can use their position to refer to the exact token. In the following example can-5 refers to the fifth token of the sentence, can.

~~~ sdparse
I can can the can .
nsubj(can-3, I)
aux(can-3, can-2)
det(can-5,the)
dobj(can-3,can-5)
punct(can-3,.)
~~~

will result in this visualization

I can can the can .
nsubj(can-3, I)
aux(can-3, can-2)
det(can-5,the)
dobj(can-3,can-5)
punct(can-3,.)

SD format: part-of-speech tags

Part-of-speech (POS) tags are optional and use the format text/POS.

~~~ sdparse
POS/NNP tags/NNS can/MD be/VB attached/VBN to/TO tokens/NNS ./.
nn(tags, POS)
nsubjpass(attached, tags)
aux(attached, can)
auxpass(attached, be)
prep(attached, to)
pobj(to, tokens)
~~~

POS/NNP tags/NNS can/MD be/VB attached/VBN to/TO tokens/NNS ./.
nn(tags, POS)
nsubjpass(attached, tags)
aux(attached, can)
auxpass(attached, be)
prep(attached, to)
pobj(to, tokens)

If the source text contains any literal slashes (/), these can be escaped using backslash.

 ~~~ sdparse
 \\/\\ escapes/VBZ :/: \\o\//\\o\/
 nsubj(escapes, \)
 ~~~

\\/\\ escapes/VBZ :/: \\o\//\\o\/
nsubj(escapes, \)

Multiple lines of text

The literal sequence \n in the SD input text is interpreted as a newline. (This sequence should be separated by space from the rest of the input.)

~~~ sdparse
One line \n and another.
~~~

gives:

One line \n and another.

CoNLL-X format

This section provides information on the visualization of the CoNLL-X format.

CoNLL-X format basics

The CoNLL-X format is a format for representing dependency parses, developed as a variant of previous related formats for the 2006 CoNLL-X shared task on multi-lingual dependency parsing.

Each line in the CoNLL-X format specifies information relating to one token (or word). For example, the following defines two words

~~~ conllx
1    Dogs   dog    _    NNS    _    2    nsubj    _    _
2    run    run    _    VBP    _    0    ROOT     _    _
~~~

visualized as

1    Dogs   dog    _    NNS    _    2    nsubj    _    _
2    run    run    _    VBP    _    0    ROOT     _    _

The definitions of the various space-separated fields are provided in the following.

CoNLL-X format field definitions

The CoNLL-X format defines 10 fields separated by space (strictly TAB characters in the original definition):

ID: Token counter, starting at 1 for each new sentence.
FORM: Word form or punctuation symbol.
LEMMA: Lemma or stem of word form.
CPOSTAG: Coarse-grained part-of-speech tag.
POSTAG: Fine-grained part-of-speech tag.
FEATS: Unordered set of syntactic and/or morphological features.
HEAD: Head of the current token.
DEPREL: Dependency relation to the HEAD.
PHEAD: Projective head of current token.
PDEPREL: Dependency relation to the PHEAD.

The current implementation of the visualization only uses the ID, FORM, CPOSTAG, HEAD and DEPREL attributes.

CoNLL-U format

This section provides information on the visualization of the CoNLL-U format.

CoNLL-U format basics

The CoNLL-U format is a format for representing dependency parses, developed in the Universal Dependencies project as a revision of the CoNLL-X format.

Most lines in CoNLL-U format specify the annotation of a single word or token. The format uses blank lines to separate sentences (as in CoNLL-X), and allows also comment lines beginning with the hash character (#). The following example illustrates these:

~~~ conllu
# this is one sentence
1    Dogs   dog    NOUN    NNS    _    2    nsubj    _    _
2    run    run    VERB    VBP    _    0    ROOT     _    _

# this is a second sentence
1    Cats   cat    NOUN    NNS    _    2    nsubj    _    _
2    sleep  sleep  VERB    VBP    _    0    ROOT     _    _
~~~

# this is one sentence
1    Dogs   dog    NOUN    NNS    _    2    nsubj    _    _
2    run    run    VERB    VBP    _    0    ROOT     _    _

# this is a second sentence
1    Cats   cat    NOUN    NNS    _    2    nsubj    _    _
2    sleep  sleep  VERB    VBP    _    0    ROOT     _    _

CoNLL-U format field definitions

The CoNLL-U format defines 10 fields separated by space (strictly TAB characters in the original definition):

ID: Word index, integer starting at 1 for each new sentence; may be a range for tokens with multiple words.
FORM: Word form or punctuation symbol.
LEMMA: Lemma or stem of word form.
CPOSTAG: Google universal part-of-speech tag from the universal POS tag set.
POSTAG: Language-specific part-of-speech tag; underscore if not available.
FEATS: List of morphological features from the universal feature inventory or from a defined language-specific extension; underscore if not available.
HEAD: Head of the current token, which is either a value of ID or zero (0).
DEPREL: Universal Stanford dependency relation to the HEAD (root iff HEAD = 0) or a defined language-specific subtype of one.
DEPS: List of secondary dependencies (head-deprel pairs).
MISC: Any other annotation.

For the complete documentation on the CoNLL-U format, please see http://universaldependencies.github.io/docs/format.html.

Alternative visualization syntaxes

As an alternative to the ~~~ syntax, you can use the equivalent HTML tag form, where the format specification appears as the value of the class attribute. For example,

<div class="sdparse">
Dogs run
nsubj(run, Dogs)
</div>

gives

Dogs run nsubj(run, Dogs)

This form is more flexible in allowing e.g. additional attributes to control aspects of the visualization. For example,

<div class="sdparse" id="simple-example-parse" tabs="yes">
Dogs run
nsubj(run, Dogs)
</div>

gives

Dogs run nsubj(run, Dogs)

(note the edit tab on the top right of the visualization.)

The Kramdown variant of Markdown provides an additional syntax for specifying attributes called Attribute List Definitions (ALDs). Using ALDs, the above example can be alternatively written as

~~~
Dogs run
nsubj(run, Dogs)
~~~
{:.sdparse}

giving

Dogs run
nsubj(run, Dogs)

ALDs can be used to specify arbitrary attributes. For example, the id and tabs attributes can be specified as in the following:

~~~
Dogs run
nsubj(run, Dogs)
~~~
{:#simple-example-parse-2 .sdparse tabs="yes"}

which gives

Dogs run
nsubj(run, Dogs)

For full details on this feature, see http://kramdown.gettalong.org/syntax.html#attribute-list-definitions.

Other features

Editing controls

Controls for visualization editing and information is accessible in elements with the attribute tabs="yes" (or any other non-empty value):

<div class="ann-annotation" tabs="yes">
Barack Obama is the current president.
T1 PERSON 0 12 Barack Obama
</div>

This gives:

Barack Obama is the current president. T1 PERSON 0 12 Barack Obama

You can click on the tab on the top right to edit the visualization, but note that the edits are not saved anywhere as there’s no server. This is mostly useful to build and debug examples.

Unicode

Everything is unicode-compliant.

~~~ sdparse
ロボットは 東大に  入れる か 。/。
nsubj(入れる, ロボットは)
nommod(入れる, 東大に)
~~~

ロボットは 東大に  入れる か 。/。
nsubj(入れる, ロボットは)
nommod(入れる, 東大に)

Collections

Jekyll collections are a recently introduced Jekyll feature that allows sets of related documents to be grouped together. Collections support various operations that specifically benefit guideline development, such as automatic listings and generation of “merged” documents.

As an illustrative example, this Annodoc repository contains two collections named entity and relation with a few example documents. Using the Liquid language, this allows e.g. a listing of documents in a specific collection to be created as follows:

{% for e in site.entity %}
* {{ e.title }}
{% endfor %}

which produces the following:

ORGANIZATION
PERSON

Alternatively, we can produce “merged” documents by importing the contents of documents in collections. The following example shows the contents of documents in the entity collection up to the first  marker (if any):

{% for e in site.entity %}
* {{ e.content | split:"<!-- details -->" | first }}
{% endfor %}

which gives

References to human groups that have been established for a specific purpose are annotated as ORGANIZATION.
References to single individual humans as well as groups of humans are annotated as PERSON.

We can similarly generate a listing of all collections, linking all documents in those collections:

{% for c in site.collections %}
* <b>Collection</b>: {{ c.label }}
  {% for d in c.docs %}
  * [{{ d.title }}]({{ d.url | remove_first:'/' }}): {{ d.shortdef }}
  {% endfor %}
{% endfor %}

which gives

Collection: entity
- ORGANIZATION: human groups established for a purpose
- PERSON: individual human or group of humans
Collection: posts
Collection: relation
- FAMILY: familial relationship between humans

To add documents to a collection, simply add them to the corresponding directory (e.g. _type/ for the collection type). To add new collections to the system, first create the directory, and then add the corresponding entry to collections in _config.yml (see Configuration below).

It is recommended to organize (non-overlapping) sets of related documents into collections. To get the most out of your Jekyll collections, please refer also to the Liquid language documentation at http://wiki.shopify.com/Liquid.

Linking to collection entries

It is often necessary to cross-reference parts of documentation from others. To make creating such links easier, Annodoc provides support for linking to collection entries.

The syntax for creating such links is simple: just wrap the name of any collection entry as in [NAME](). For example, to create a link to the documentation for the PERSON type, use

[PERSON]()

which gives the following link

PERSON

Alternatively, you can use empty HTML anchor tags (<a>, </a>), as in

<a>PERSON</a>

(These two alternative forms result in identical links.)

For cases where the same label (name) occurs in several collections, it is possible to disambiguate the target by including the collection name in the format COLLECTION/ENTRY, for example

[entity/PERSON]()

(or, equivalently, <a>entity/PERSON</a>) which produces

entity/PERSON

(Note that the text of the generated link only includes the entry label.)

Either of these forms can be used in running text, for example

[FAMILY]() relation arguments have type [PERSON]().

gives

FAMILY relation arguments have type PERSON.

Cases where linking failed are visually marked to identify the issue:

[no-such-type]()

is shown as

no-such-type

For all of the forms listed above, the text of the link matches the linked type. To get a different text, use the syntax [text](TYPE). For example,

[people](PERSON)

and

[people](entity/PERSON)

both generate the same link: people and people

This form can also be written in HTML as

<a href="PERSON">people</a>

Which again gives an identical link: people

Note that this automatic processing only applies to <a> elements that have no href, id, or name attributes or whose href attribute matches a simple pattern (e.g. lacking http:// and .html). This means that standard HTML anchors and links can be used normally in Annodoc documents without triggering this feature.

Configuration

Different aspects of the way a set of documentation is presented are controlled by various configuration files and settings. The following are the most important.

_config.yml

The Jekyll top-level configuration file _config.yml controls many aspects of the conversion of the source documents into the final site as well as how Jekyll serves the site. This file also configures the collections that are defined for the documentation. Full documentation for _config.yml is available from http://jekyllrb.com/docs/configuration/.

_includes/header.html, _includes/footer.html

These HTML files are automatically attached to every page on the site. Edit the HTML content of these two files to customize the look of the documentation, add navigation, etc. These files are standard HTML.

css/main.css, css/style-vis.css

The CSS files found in the css subdirectory control the style (look) of the main documentation (main.css) and the visualizations (style-vis.css). It is not necessary to modify these in basic usage. If changes to the style of the documentation pages or the visualization are needed, please refer to e.g. http://www.w3.org/Style/CSS/Overview.en.html for more information.

lib/local/config.js

This JavaScript file holds the configuration for the brat visualization component. The configuration specifies, for example, the colors, labels, label abbreviations, and line styles to use for visualizing various categories of annotation. For documentation on the contents and syntax of this file, refer to http://brat.nlplab.org/configuration.html.

Troubleshooting

The following sections provide instructions for troubleshooting various possible issues with the system.

Troubleshooting: basic visualization

Are the annotation visualizations working? To check, see if the following visualization and image look (broadly) similar.

Visualization:

Barack Obama is the current president.
T1 PERSON 0 12 Barack Obama

Image:

If yes, the basic visualization support is OK. If no (e.g. if instead of a visualization text such as T1 PERSON 0 12 Barack Obama is shown), then try the following:

Make sure you are using a supported browser. The visualizations require SVG support. We recommend recent versions of Chrome or Safari. For IE, version 9 or newer is required.
Check that JavaScript is enabled. The system requires JavaScript to generate the visualizations and will fail without it.
Check the JavaScript console for errors. (The console is accessed in different ways depending on your browser.)
Check that coderay is not running on the server. If active, coderay may pre-empt embedded visualization processing by replacing the annotation blocks. The _config.yml option kramdown: enable_coderay: false should disable coderay.
Report the issue to the developers: email sampo.pyysalo@gmail.com

Troubleshooting: local usage

If you are having trouble running the system locally, please check the following:

Check that you have a recent version of Jekyll by running jekyll -v. Version 2.0 or greater is recommended.

Some debian-based linux distributions are currently shipping a dated version of Jekyll. If you are using such a system, you may wish to consider using gem instead of apt-get to install Jekyll (see e.g. http://jekyllrb.com/docs/installation/#install-with-rubygems).

Troubleshooting: editing online

If the “edit page” links do not appear on Annodoc documentation pages, please make sure editurl site variable is set in _config.yml (see Configuration).

If the “edit page” link appears but does not lead to a GitHub online editing page, please check the following:

GitHub online editing is working as intended. Click on any document in your GitHub repository and click on the pen icon to test this feature.
editurl is correctly set. Try comparing the generated “edit page” links with the URL used for the GitHub online editing page.
You are allowed to edit the repository (owner or collaborator) and you are logged in to GitHub.

Table of contents

What is Annodoc?

How does it work?

Getting started

Running locally

Working online

Editing documents

Editing locally

Editing online

Adding documents

Visualizations

Visualization basics

.ann standoff format

Ann format: annotation primitives

Ann format: text-bound annotations

Ann format: attribute annotations

Ann format: relation annotations

Ann format: event annotations

Ann format: other annotation primitives

Stanford Dependency format

SD format basics

SD format: ambiguous tokens

SD format: part-of-speech tags

Multiple lines of text

CoNLL-X format

CoNLL-X format basics

CoNLL-X format field definitions

CoNLL-U format

CoNLL-U format basics

CoNLL-U format field definitions

Alternative visualization syntaxes

Other features

Editing controls

Unicode

Collections

Linking to collection entries

Configuration

Troubleshooting

Troubleshooting: basic visualization

Troubleshooting: local usage

Troubleshooting: editing online