Are literals with datatypes and language tags supported?

[bnouwt]

How does the KE behave when literals with datatypes and language tags are transferred. Are these correctly dealt with or are they accidently removed? Maybe make a series of unit tests to see how literals with datatypes and/or language tags are dealt with and if the behavior is consistent. For example, are literals with datatypes returned by an ANSWER KI showing up in the AskResult of an ASK KI? And how does this happen with incoming bindings?

[bnouwt]

The tests confirm that both data types ("eek"^^<http://www.w3.org/2001/XMLSchema#decimal>) and language tags ("eek"@en) are supported.👍 We did notice that because literals without datatype are automatically interpreted as being of datatype xsd:string, the xsd:string datatypes are removed from any literal that is processed by the knowledge engine.

[johankumps]

Hello Barry,

I'm running into an issue with string type literals in an ASK/ANSWER interaction. My input binding set looks like this:

{
	"value":     "\"My string value\"^^<http://www.w3.org/2001/XMLSchema#string>",
	"testValue": "\"7\"^^<http://www.w3.org/2001/XMLSchema#int>",
}

Posting this ASK KI, results in an error in the BindingValidator on the return path. The ANSWER is received and a match check is done between the incoming and outgoing bindingset:

2026-01-07 09:38:16:743 +0100 [qtp1151593579-31] ERROR InteractionProcessorImpl-KHC Managed KB - An error occurred while answering a msg: java.util.concurrent.CompletionException: java.lang.IllegalArgumentException: KB gave outgoing binding Binding [map={result="eulav gnirts yM"^^<http://www.w3.org/2001/XMLSchema#string>, input=<http://example.org/stringmanipulator/InputX>, reverseResult=<http://example.org/stringmanipulator/ReverseResultX>, anchor=<http://example.org/stringmanipulator/ReverseOperation1>, value="My string value"^^<http://www.w3.org/2001/XMLSchema#string>, testValue="7"^^<http://www.w3.org/2001/XMLSchema#int>}], but this doesn't have a matching incoming binding!

All variables from the incoming bindingset are part of the outgoing set, so that is ok but the value of the string type literals is not the same leading to the isSubBindingOf check to fail and the interaction get a failed state.

I tried to pass my string without the datatype but this fails earlier in the chain as it can't be parsed into json:

`Caused by: org.apache.jena.sparql.sse.SSE_ParseException: Encountered " "string "" at line 1, column 4.
Was expecting one of:

...

    at org.apache.jena.sparql.sse.lang.SSE_Parser.parse$(SSE_Parser.java:46)
    at org.apache.jena.sparql.sse.lang.SSE_Parser.parse(SSE_Parser.java:35)
    at org.apache.jena.sparql.sse.SSE.parse(SSE.java:500)
    at org.apache.jena.sparql.sse.SSE.parseNode(SSE.java:470)
    at org.apache.jena.sparql.sse.SSE.parseNode(SSE.java:189)
    at org.apache.jena.sparql.sse.SSE.parseNode(SSE.java:181)
    at eu.knowledge.engine.reasoner.api.Binding.put(Binding.java:50)
    at eu.knowledge.engine.smartconnector.impl.ReasonerProcessor.translateBindingSetTo(ReasonerProcessor.java:348)
    at eu.knowledge.engine.smartconnector.impl.ReasonerProcessor.executeAskInteraction(ReasonerProcessor.java:195)`

Passing non-string literals works perfectly.

Am I doing something wrong or did I trigger a bug here?

[johankumps]

I found a solution by quoting my string literal in the input binding set as follows:

{"value": ""My string value""}

Is this the way to go? In my opinion it would be better that the KE also supports string typed literals like this:

""My string value"^^http://www.w3.org/2001/XMLSchema#string"

in order to be in line with the other literal types. What do you think?

[bnouwt]

Hi @johankumps, thanks for your message!

This indeed looks like a bug. I will create a unit test and see if I can reproduce the behaviour. I also double checked whether the incoming binding set:

value: "My string value"^^<http://www.w3.org/2001/XMLSchema#string>
testValue: "7"^^<http://www.w3.org/2001/XMLSchema#int>

is part of the outgoing binding set:

value: "My string value"^^<http://www.w3.org/2001/XMLSchema#string>
testValue: "7"^^<http://www.w3.org/2001/XMLSchema#int>

and this is indeed the case.

Regards, Barry

[bnouwt]

Hi @johankumps,

I created the unit test on the 629-are-literals-with-datatypes-and-language-tags-supported-2 branch which can indeed reproduce the faulty behavior of the KE you are describing above 👍 . The problem lies in the fact described in the comment above:

We did notice that because literals without datatype are automatically interpreted as being of datatype xsd:string, the xsd:string datatypes are removed from any literal that is processed by the knowledge engine.

It looks like xsd:string datatypes are stripped from their datatype and when the BindingValidator compares the values it fails, because one has and the other hasn't got a datatype.

The KE makes use of the Apache Jena API for many of its RDF related operations, so we might be able to solve this problem there.

So, no solution yet, but at least we know its a bug and where the problem lies.

Again, thanks for reporting it!

Regards, Barry

[bnouwt]

Related to #782