Annotation screen goes blank after selecting a numeric text span

[david-waterworth]

I'm trying to annotate text such as TEF 4-3 Status using the NER layer. TEF is class equip, 4-3 is class id and Status is class point.

I may have messed up the file, I'm using WebAnno TSV 3.2. The issue is whenever I select tokens 4-3 and select value="id" the pane showing the document contents goes blank. There doesn't seem to be any errors in the log, and the UI is still responsive - I can export the document and it's applied the annotation but I cannot continue annotation.

My guess is it has something to do with the fact that the entity mention is text (or perhaps a numeric expression even) but it's possible I've also incorrectly generated the file (I've attached a sample).

This is an example after I reload

as soon as I click on identifier (I changed the value from id to identifier in case id was reserved)

I've attached a test file. The file loads into INCEpTION, and also https://pypi.org/project/web-anno-tsv/ so I'm assuming I've created it correctly but I don't really know.

test.tsv.zip

[reckart]

That sounds like the visualization crashes during rendering. Which browser are you using and which version of INCEpTION?

[david-waterworth]

Chrome and inception-app-standalone-0.17.2.jar

I did notice there's a newer release (inception-app-webapp-0.18.1-standalone.jar) but it won't run on my system due to a requirement for a newer java version (I'm using openjdk version "1.8.0_282") so tomorrow I'll update and see if that helps.

[reckart]

Ok. If it doesn't, please check in the Web Developer Tools of Chrome if you might see any errors in the Console view.

---

https://developers.google.com/web/tools/chrome-devtools/shortcuts

[david-waterworth]

With Java 1.11 and inception 1.8.0.282 I get the error below in the console view - it seems you're correct. I guess it still could be an issue with my file so I'll look at it a little closer

visualizer-ver-6FE6C5C69300B2B09BCF3952BBD91BFE.js:3452 Rendering terminated due to: IndexSizeError: Failed to execute 'getEndPositionOfChar' on 'SVGTextContentElement': The charnum provided (0) is greater than or equal to the maximum bound (0). Error: Failed to execute 'getEndPositionOfChar' on 'SVGTextContentElement': The charnum provided (0) is greater than or equal to the maximum bound (0). at http://localhost:8080/wicket/resource/de.tudarmstadt.ukp.clarin.webanno.brat.resource.BratVisualizerResourceReference/visualizer-ver-6FE6C5C69300B2B09BCF3952BBD91BFE.js:1356:8 at Fragment. (http://localhost:8080/wicket/resource/de.tudarmstadt.ukp.clarin.webanno.brat.resource.BratVisualizerResourceReference/visualizer-ver-6FE6C5C69300B2B09BCF3952BBD91BFE.js:1175:1) at Function.each (http://localhost:8080/wicket/resource/org.apache.wicket.resource.JQueryResourceReference/jquery/jquery-2.2.4-ver-F9EE266EF993962AD59E804AD9DEBE66.js:2:2861) at SVGTextElement. (http://localhost:8080/wicket/resource/de.tudarmstadt.ukp.clarin.webanno.brat.resource.BratVisualizerResourceReference/visualizer-ver-6FE6C5C69300B2B09BCF3952BBD91BFE.js:1174:3) at Function.each (http://localhost:8080/wicket/resource/org.apache.wicket.resource.JQueryResourceReference/jquery/jquery-2.2.4-ver-F9EE266EF993962AD59E804AD9DEBE66.js:2:2861) at a.fn.init.each (http://localhost:8080/wicket/resource/org.apache.wicket.resource.JQueryResourceReference/jquery/jquery-2.2.4-ver-F9EE266EF993962AD59E804AD9DEBE66.js:2:845) at getTextMeasurements (http://localhost:8080/wicket/resource/de.tudarmstadt.ukp.clarin.webanno.brat.resource.BratVisualizerResourceReference/visualizer-ver-6FE6C5C69300B2B09BCF3952BBD91BFE.js:1170:34) at getTextAndSpanTextMeasurements (http://localhost:8080/wicket/resource/de.tudarmstadt.ukp.clarin.webanno.brat.resource.BratVisualizerResourceReference/visualizer-ver-6FE6C5C69300B2B09BCF3952BBD91BFE.js:1198:17) at renderDataReal (http://localhost:8080/wicket/resource/de.tudarmstadt.ukp.clarin.webanno.brat.resource.BratVisualizerResourceReference/visualizer-ver-6FE6C5C69300B2B09BCF3952BBD91BFE.js:1528:13) at rerender (http://localhost:8080/wicket/resource/de.tudarmstadt.ukp.clarin.webanno.brat.resource.BratVisualizerResourceReference/visualizer-ver-6FE6C5C69300B2B09BCF3952BBD91BFE.js:3447:1)

There's also a warning appearing in my bash shell, not sure it that's related

2021-02-25 07:27:12 WARN [admin] ResourceReferenceRegistry - A ResourceReference wont be created for a resource with key [scope: de.tudarmstadt.ukp.inception.support.vue.VueBehavior; name: vue3-sfc-loader.js.map; locale: null; style: null; variation: null] because it cannot be located.

[reckart]

It could also be a bug in the rendering code - we had a similar issue on Firefox recently: https://github.com/inception-project/inception/issues/1849

[reckart]

The warning in the bash log you can ignore. That happens when Chrome is trying to load the source map file which we do not have for the vue3-sfc-loader.

[david-waterworth]

What I've found works is to remove the whitespace from my input file i.e.

Text=EXF-LD 1_1

1-1 0-3 EXF 1-2 3-4 - 1-3 4-6 LD 1-4 6-7 1-5 7-8 1 1-6 8-9 1-7 9-10 1 _

Becomes (removed token 1-4)

Text=EXF-LD 1_1

1-1 0-3 EXF 1-2 3-4 - 1-3 4-6 LD 1-4 7-8 1 1-5 8-9 1-6 9-10 1 _

I also read that certain chars need to be escaped (https://webanno.github.io/webanno/releases/3.4.5/docs/user-guide.html#_reserved_characters) but when I attempted this (i.e. token 1-5) the file won't load. It's not clear to me if I'm supposed to escape both the text (EXF-LD 11 -> EXF-LD 1_1) and/or the token (1-5 -> _) and does that mean I have to update the offsets. But it actually seems to be working better with no escaping, just removal of whitespace tokens.

[reckart]

INCEpTION (and WebAnno for that matter) do not support tokens that start or end (or consist exclusively) of whitespace.

[david-waterworth]

OK That's probably my issue then. Still not 100% sure how to properly escape the text but what I've done (not escape anything) does seem to be working.

[reckart]

Can you make an example where you thing something might need to be escaped?

[david-waterworth]

I've attached an example, token 1-5 is an underscore. It's in the list of reserved characters \,[,],|,_,->,;,\t,\n,* if I'm reading correctly.

I have many examples which contain _ or | (and some which are \

Text=EXF-LD 1_1

1-1 0-3 EXF 1-2 3-4 - 1-3 4-6 LD 1-4 7-8 1 1-5 8-9 1-6 9-10 1 _

test.tsv.zip

[reckart]

I have dropped the text "_ * \n \t [ ]" (mind that \n is a linebreak and \t is a tab) into the Tsv3XSerializer and this is what it produces:

#Text=_ * 
#Text= \t [ ]
1-1 0-1 _   
1-2 2-3 *   
1-3 8-9 [   
1-4 10-11   ]   

So the tab is escaped in the #Text line. The line break is realized as an actual line break starting a new #Text line. In the token lines, the characters are not escaped. It looks like in the text, only tab, line feed, form feed, carriage return, backspace and backslash are escaped. However, in the annotation columns, the special characters mentioned in the documentation are escaped.

[reckart]

Hm, yes, the escaping is only in the #Text lines - but in the text column (i.e. third column), there is no escaping at all:

        // Write unit text
        aOut.print(doc.getJCas().getDocumentText().substring(aUnit.getBegin(), aUnit.getEnd()));
        aOut.printf(FIELD_SEPARATOR);

I guess this is something to be changed in a future iteration of the format and to be properly documented.

[david-waterworth]

Thanks. I don't think [ and ] are reserved characters. By my reading of the documentation I linked to above, (comma) is and since they're displaying the reserved characters in a comma-separated list the actually reserved comma is enclosed to [ ] to try and remove confusion (but I didn't originally notice so it didn't work).

So I think the list of reserved characters is supposed to be

\,|_;* \t and \n (along with the -> character?)

But as we've seen Tsv3XSerializer (is this what you use to parse the files) doesn't escape most of these characters.

[reckart]

The Tsv3XSerializer is part of the code that does the serialization - as is the function below from the Java class Escaping which handles the escaping of the feature column values. Here you can see exactly what is escaped in which way.

    public static String escapeValue(String aValue)
    {
        return StringUtils.replaceEach(aValue,
                new String[] { "\\", "[", "]", "|", "_", "->", ";", "\t", "\n", "*" },
                new String[] { "\\\\", "\\[", "\\]", "\\|", "\\_", "\\->", "\\;", "\\t", "\\n",
                        "\\*" });
    }

[reckart]

Assuming the that issue was resolved as there was no more feedback and thus closing the issue.