If you rename the folder e.g. when you want to rename the corpus you can't open the documents after you renamed the folder.

The error message indicates that the SDocument node in the SaltProject file is not updated

  <nodes xsi:type="sCorpusStructure:SCorpus">
      <labels xsi:type="saltCore:SFeature" namespace="salt" name="SNAME" value="ACED0005740006636F72707573" valueString="corpus"/>
      <labels xsi:type="saltCore:SElementId" namespace="graph" name="id" value="ACED000574000C73616C743A2F636F72707573" valueString="salt:/corpus"/>
    </nodes>
    <nodes xsi:type="sCorpusStructure:SDocument">
      <labels xsi:type="saltCore:SFeature" namespace="salt" name="SNAME" value="ACED000574000F636F727075735F646F63756D656E74" valueString="corpus_document"/>
      <labels xsi:type="saltCore:SElementId" namespace="graph" name="id" value="ACED000574001C73616C743A2F636F727075732F636F727075735F646F63756D656E74" valueString="salt:/corpus/corpus_document"/>
      <labels xsi:type="saltCore:SFeature" namespace="salt" name="SDOCUMENT_GRAPH_LOCATION" value="ACED000574004A66696C653A2F686F6D652F74686F6D61732F61746F6D69632D776F726B73706163652F53656365646765546573742F636F727075732F636F727075735F646F63756D656E742E73616C74" valueString="file:/home/thomas/atomic-workspace/SecedgeTest/corpus/corpus_document.salt"/>
    </nodes>

Atomic Version is v0.1.9

Reproduce: Start from command line with -consoleLog arg, create referent, drag to Referents view.

When the Annotation Perspective is open, and one of either Level or Sentence View are closed, click actions trigger above exception...

I tried to open a *.salt file (describing an SDocumentGraph) that I had generated with SaltNPepper. I pointed the Pepper Module Import/SaltXMLImporter to the directory that contained the *.salt file. This is what happened:

Exception in thread "PepperModuleController[SaltXMLImporter]" de.hu_berlin.german.korpling.saltnpepper.salt.saltCommon.exceptions.SaltResourceNotFoundException: Cannot load Object, because the file '/home/arne/saltProject.salt' does not exist.
    at de.hu_berlin.german.korpling.saltnpepper.salt.impl.SaltFactoryImpl.load(SaltFactoryImpl.java:148)
    at de.hu_berlin.german.korpling.saltnpepper.salt.impl.SaltFactoryImpl.loadSCorpusGraph(SaltFactoryImpl.java:273)
    at de.hu_berlin.german.korpling.saltnpepper.salt.impl.SaltFactoryImpl.loadSCorpusGraph(SaltFactoryImpl.java:257)
    at de.hu_berlin.german.korpling.saltnpepper.salt.saltCommon.sCorpusStructure.impl.SCorpusGraphImpl.load(SCorpusGraphImpl.java:542)
    at de.hu_berlin.german.korpling.saltnpepper.pepperModules.saltXML.SaltXMLImporter.importCorpusStructure(SaltXMLImporter.java:66)
    at de.hu_berlin.german.korpling.saltnpepper.pepper.pepperFW.impl.PepperModuleControllerImpl.realImportCorpusStructure(PepperModuleControllerImpl.java:572)
    at de.hu_berlin.german.korpling.saltnpepper.pepper.pepperFW.impl.PepperModuleControllerImpl.run(PepperModuleControllerImpl.java:427)
    at java.lang.Thread.run(Thread.java:744)

Kind regards,
Arne

Refactor parser for annotation command in console class so that the user can define a namespace just like in the graphical editor

When user double-clicks on saltProject.salt, the AtomicSaltEditor is being opened on the file.

Disallocate editor for this filename & handle in AtomicSaltEditor.java.

1. Besteht die Möglichkeit, die Relationen zu benennen?
2. Besteht die Möglichkeit, Annotationsebenen und ein Tagsets zu definieren, sodass man nicht wiederholt „xy:abc“ beim Benennen der Knoten eingeben muss.
3. Ideal, aber nicht notwendig, wäre noch eine Art Aufräumfunktion, die bei sehr komplexen Konstituentenstrukturen die annotierten Strukturen platzsparend anordnet.

As reported by @Annotation-123. Details to follow.

Cf. e.g., http://stackoverflow.com/questions/5728651/eclipse-plugin-to-handle-same-file-extension-with-different-editors, http://www.eclipse.org/forums/index.php/t/106327/,

The latter terminology makes it clearer that the target sections needn't be syntactic sentences at all.

Check how Eclipse does this. When, where, and how are projects loaded, etc.

Get Atomic implemented on Mars

Add Pepper API once it makes sense in the implementation flow

In Atomic 0.1.9 werden jetzt zwar SPAN und STRUCT nach RelANNIS exportiert. Die Pointing-Relationen werden aber (von struct-Knoten zu Token oder zwischen zwei Span-Knoten) nicht übertragen.

Ich kann lediglich erkennen, dass beim PAULA-Export folgender Fehler nach wie vor besteht, der dann auch bei der Konvertierung mit SnP angegeben wird: Eine pointing-Relation von einem Strukturknoten zu einem Token wird so angegeben:

[Spitze Klammer] rel id="sPointingRel2" xlink:href="null#structure1" target="nolayer.corpus_document.tok.xml#sTok13"/>

hinter "xlink:href="" müsste der Dateinanem erscheinen, und nicht null stehen.

Pointing-Relationen zwischen Span-Annotationen -also zwischen zwei Einheiten, die aus mehr als einem Token bestehen, werden gar nicht expoertiert.

Beste Grüße

TreetaggerImporter erzeugt keine Datei, die sich im graphischen Editor öffnen lässt.

Wenn ich etwas aus Exmaralda importiere, funktioniert das bei einem Satz, aber mein Korpus (1 Text, 3300 token) lässt sich nicht öffenen. ATOMIC reagiert beim Öffnen nach 5 Minuten immer noch nicht.

When exporting a project with the SaltXMLExporter, the following folder structure is created:

target dir
|__ salt:
      |__corpus
           |__corpus_document.salt
|__saltProject.salt

This points to a URI being resolved for use as export target (e.g., "salt:/corpus/corpus_document"). This could be a bug in the SaltNPepper API.

And wrap Salt elements (´a la EMF Editor -> .diag)

Currently, for creating a new project you have to right click on the empty space of the "Navigation" panel and select the right context menu entry. On the first startup this can be difficult to find (the user basically only sees an empty space). One could imagine to

• have a menu entry under "File" (this is an Ubuntu Unity bug with the global menu, see http://stackoverflow.com/questions/19452390/eclipse-menus-dont-show-up-after-upgrading-to-ubuntu-13-10 for a solution)
• a button in the toolbar
• a link on a Eclipse Welcome Page alike start page with instructions how to start

to make the first startup easier (even if the user did not read the manual).

Since the atomic console is interactive it would be more intuitive to use the help instead of --help.

Especially for historic corpora - where punctuation isn't standardized/intrdoduced - the NLP/Java API-based sentencers don't work well.

Hence, introduce a sentencer based on regex/char lists which the corpus-informed user can instrumentalize to achieve correct sentencing.

E.g., on having processed an annotation command such as
"e n1 t1 x:y",
print feedack to command-line:
"Created dominance relation with annotation x:y from node 1 to token 2".

Include assertions before feedback.

Cc @Annotation-123

Issue: token-to-token relations are always displayed, independent of level selection. Let controller query levels to display and pick relations respectively.

Introduce bendpoints for relations.

This would need:

• a preference sheet where namespaces can be read into / manually defined
• a specific colour on the annotation (also set via prefs page)
• a view to show namespace in use + their colour

Refactor parser for annotation command in console class so that the user can define a namespace just like in the graphical editor

1. Das Exportieren nach RelANNIS funktionieren bei mir nicht. Der Exporter läuft immer weiter – ohne Ergebnis.
2. Bei Exporten von Syntax-Bäumen nach PAULA tritt bei mir folgender Fehler auf:
   Im struct-Dokument werden die untergeordneten Relationen falsch ausgegeben. Sie enthalten nach dem id-Wert und den Anführungszeichen noch weitere Anführungszeichen (1x zu viel).

  struct id="structure1">
  rel id="sDomRel1" " xlink:href="nolayer.corpus_document.tok.xml#sTok4"/>
  rel id="sDomRel2" " xlink:href="nolayer.corpus_document.tok.xml#sTok3"/>
  /struct>

3. Ich habe die überflüssigen Anführungszeichen entfernt und habe PAULA mit saltnpepper dann nach ReLANNIS konvertiert. In ANNIS wurden dann aber die Bezeichnungen der Knoten nicht angezeigt.

Either to "Atomic Sentence Editor" or to "Atomic Span Editor" or to "Atomic Graph Editor"

As requested in #32 (#32 (comment)), create shortcuts to create new nodes in the Annotation Graph Editor, preferrably using "CTRL + ..." sequences.

When running Atomic on Linuxes utilizing the Unity GUI, menus do not include the keyboard shortcut text which should appear to the right of the menu items (e.g., Quit CTRL+Q).

This is due to the fact that Eclipse does not (fully) support the Unity Global Menu, where menus appear in the WM GUI rather than the application. Therefore, key bindings do work in Atomic on Unity Linux (i.e., Ubuntu), but the shortcut is not shown.

However, Unity can be forced to drop use of the Global Menu for applications. To do so, start the application with the prefix UBUNTU_MENUPROXY=0, as in UBUNTU_MENUPROXY=0 ./atomic.

You can start Atomic as above to see keyboard shortcuts in the menus.

This will be fixed by including a custom starter which will be included in Linux-targeted products, i.e., a shell script that can be used instead of the default launcher file atomic. This shell script will do nothing but call UBUNTU_MENUPROXY=0 ./atomic.

As reported by @Annotation-123. Details to follow.

Beim Einsatz der Console ist mir wiederholt ein Fehler aufgefallen: Ich wollte beispielsweise die Kante D9 mit x:y versehen und habe "a d9 x:y" eingegeben. Annotiert wurde x:y aber bei Kante D10. Der Fehler tritt nicht mehr auf, wenn man das Projekt speichert und neu öffnet. Es könnte sein, dass der Fehler dadurch bedingt wird, dass ich vorher eine oder mehrere Kanten gelöscht habe. Es scheint so zu sein, dass die Console mit einem falschen Wert oder einer falschen Zahl arbeitet, die beim Schließen bzw. erneuten Öffnen berichtigt wird. Sollte ich nochmal auf den Fehler treffen, schaue ich mir alles nochmal genauer an und leite es weiter.

This is due to the order in which model children of GraphPart are added to the list of model children. When a child is added but targets of its edges aren't yet rendered, the respective edge points into deep space, or 0x0.

Please put all files into atomic folder before zipping, otherwise unzipping can mess up the target folder.

This happens even though the VM argument -XstartOnFirstThread is passed on (cf. product configuration).

.workspace/Activator.java, line 31.

Some users may not be familiar with the concept of workspace

For larger corpora (which may take long to load in the Annotation Graph Editor (AGE)), it should be possible to navigate smaller portions of the corpus.

Navigation should be possible by

1. sentences
2. selection of text

For 1., the corpus documents must be run through sentence detection. This should be possible a) in the New Project Wizard, b) on import (extra option "document contains sentence spans"), c) when opening a corpus document already existing in the workspace for which sentences haven't been detected.
Sentence detection via Apache OpenNLP ("official" & own models), regex, and third-party sentence detectors (via extension point) should be possible.
Sentences are added to the model as ranges of indices, and as sorted lists of tokens via SProcessingAnnotations on the SDocumentGraph.
Working on these sentences are the following UI elements.

• A consecutively numbered list of sentences with checkboxes next to them. This view serves as a) access point to a sentence's ID (= consecutive number) and b) a way to select sentences to display in the AGE. The view also provides a Deselect all button.
• A group of three menu elements (a button <, a textfield displaying the sentence ID, a button >), which makes it possible to navigate through the sentences by ID (use up and down button, enter sentence ID).
• A set of key bindings (e.g., PgUp, PgDown to go to next, previous sentence).
• A set of AtomicAL commands for navigation.

For 2., a view should be provided displaying the whole corpus text, which is selectable. The selection will be displayed in the AGE.

As requested by @Annotation-123.

As of ffc4185, tokens IDs are assigned in the order in which sentences are selected for display in the Sentence View. I.e., if sentence 5 with 10 tokens is selected, those will get IDs T1-T10. If then, sentence 1 with 10 tokens selected, its tokens will be T11-T20. If both sentences are displayed simultaneously, the order will be T11, T12 .. T19, T20, T1, T2 .. T9, T10.

This is not intuitive and may lead to issues when creating annotations for the whole corpus text.

Hence, token IDs should be in natural ascending order based on their index in the text. This could be done, e.g., in the sentencing step.

Cf. http://www.eclipse.org/forums/index.php/t/64088/