Problems when deploying web service (NLprot)

[martavillegas]

We have deployed a WS as wrapper for a command line tool (NLProt) that reads PDF and returns a JSON (we understood that XMI was recommended but not compulsory).

you can run it:
curl -F cas=@input-test.pdf inb.bsc.es/service/openminted/process

input-test.pdf

When running the app in test.openminted we get the following message:

Failed
Unable to locate named workflow

[gkirtzou]

@martavillegas
the error you have mentioned is a know issue in the test.openminted.eu for the software that is registered as application (and is reported to the responsible tech team in the redmine, see issue 840)

Independent of that, we need first to insure that the software you provide is OMTD compliant to the specifications and guidelines for web services. The OMTD specifications and guidelines for a web services require that both the input and output are corpora of UIMA CAS XMI files. Furthermore, in order to deserialize correct the output of the web service it requires a typesystem, which is available in the maven central repository.

Given the metadata you have provided us, in order for your software to run within OMTD platform, you would need to make it OMTD compliant. In other words,

1. you need to generate UIMA CAS XMI files as output, not JSON
2. you need to define what typesystem the XMI CAS file use so as to be deserialized correctly and
3. make that typesystem available via the maven central repository.

For more details see here:

1. Guidelines : https://guidelines.openminted.eu/guidelines_for_providers_of_sw_resources/sharing-web-services.html
2. Specifications : https://openminted.github.io/releases/processing-web-services/1.0.0/specification

[martavillegas]

Thanks! just two questions:

• we understood that apps differ from components in that they did not to consume/return XMI. Is this right? (that's why we changed to app)

• can we consume PDF files directly?

[gkirtzou]

@martavillegas To answer your first question what is the difference between a component and an application and what an application or a component can consume/produce, the OMTD interoperability specification say that :

• components that support (i.e. produce & consume) data in the XML Metadata Interchange (XMI) format; in the case of components that support other formats, appropriate readers and writers for converting XMI messages from and to the components’ format must also be provided
• applications that support as input file formats that are used for publications (e.g. PDF, PubMed XML, plain text etc.) or XMI.

In general, both components and applications can consume/produce whatever they want. For the components it is advisable to offer other components converters (readers/writers from/to XMI) for interoperability reasons. ie to allow a TDM expert to combine components from different providers/architectures.

Note that an application can be create by two different ways:

1. Via the Workflow editor, where a TDM expert combines interoperable components or
2. By registering a single component as an application.

In this second case, where a single component is registered as application, the way the OMTD workflow service handles them makes the assumption that the application consumes XMI, and thus always executes before the component-application the following two components:

1. the omtd importer so as to fetch the data from the OMTD store to OMTD workflow executor
2. and the PDF reader component to transform PDF files to XMI.

As for the second question, since you have decided to offer your software as a web service, you need to make your component compliant not to the generic specifications, but to the web service ones. Which means that you need to:

1. ensure that it uses UIMA CAS XMI messaging system as specified at https://openminted.github.io/releases/processing-web-services/1.0.0/specification and
2. upload the typesystem that the web service deploys in Maven.

So to answer your second question, you cannot read PDF files directly if you are offering your software as web service. If you want to read PDF files directly, you could register your software as component of docker image and create an application using the workflow editor so that you can execute your software. The instructions how you can build a workflow can be found here:
https://openminted.github.io/releases/workflow-editor/1.0.0/workflow

[martavillegas]

Hi Katerina, we are back....

Our NLProt service reads and delivers MXI with DKPro Core type system. in this case do we need to uoload the typesystem (we understand it is already uploaded).

Thanks!

[gkirtzou]

@martavillegas

Our NLProt service reads and delivers MXI with DKPro Core type system. in this case do we need to uoload the typesystem (we understand it is already uploaded).

That's great! In order to be compliant you need to update the metadata of the application/component by adding the typesystem details, as the guidelines mention:

Please, note that in the metadata record, you must include
(1) the URL where the web service can be executed in the element distributionLocation, e.g.
<distributionLocation>http://nactem.ac.uk/api/openminted/chebi</distributionLocation>
(2) the typesystem details in the the module componentDependencies/typesystem and more
specifically in the elements resourceName and resourceIdentifier (Maven coordinates), e.g.
<resourceName>Chebi Curation Type System</resourceName>
and
<resourceIdentifier resourceIdentifierSchemeName="maven">mvn:uk.ac.nactem.uima:ChebiCurationTypeSystem:0.1</resourceIdentifier>

Then you could try again to call you web service and see whether it works as expected.

[martavillegas]

Sorry Katerina, but in case we use default type system, namely:

xmlns:type="http:///de/tudarmstadt/ukp/dkpro/core/api/metadata/type.ecore"
xmlns:type2="http:///de/tudarmstadt/ukp/dkpro/core/api/segmentation/type.ecore"

do we need to upload it in Maven?

[reckart]

The DKPro Core type system is split over various "API" artifacts on Maven: http://search.maven.org/#search%7Cga%7C1%7Cde.tudarmstadt.ukp.dkpro.core.api

[galanisd]

@reckart just said the one thing that I wanted to say

The DKPro Core type system is split over various "API" artifacts on Maven: http://search.maven.org/#search%7Cga%7C1%7Cde.tudarmstadt.ukp.dkpro.core.api

The typesystem that you use is already uploaded to maven central. So you have just specify in the metadata its coordinates. E.g. like

a. chebi web service

https://services.openminted.eu/landingPage/application/11dc2e5e-1cd8-40bf-a46f-df17664b3c77
->
mvn:uk.ac.nactem.uima:ChebiCurationTypeSystem:0.1

https://mvnrepository.com/artifact/uk.ac.nactem.uima/ChebiCurationTypeSystem/0.1

b. the OGER web service
https://services.openminted.eu/landingPage/application/71345d18-297f-4ac5-b4de-38ef3cacbe75
->
mvn:de.tudarmstadt.ukp.dkpro.core:de.tudarmstadt.ukp.dkpro.core.api.ner-asl:1.9.1

https://mvnrepository.com/artifact/de.tudarmstadt.ukp.dkpro.core/de.tudarmstadt.ukp.dkpro.core.api.ner-asl/1.9.1

[martavillegas]

Thanks guys with the types!!

I updated the metadata with types (two coordinated from maven) but now when I try to run the application I get the following error message:

System error executing service (Server responded: #: 2 schema violations found)

[gkirtzou]

@martavillegas Would be possible to send the xml file of the metadata with us, to check them?

[martavillegas]

Sure!
2c448b7c-6fe2-451f-b9d3-1426330251ef (2).zip

[galanisd]

We have checked the workflow engine (Galaxy); it seems that the Galaxy wrapper that is required has been generated as it should. The wrapper contains the command that will be executed from the generic web service executor that we have designed. The command seems correct and the executor was designed to support more that one typesystems. I remember also that I have tested it in such a case and it worked.

Also we checked Galaxy (with @gkirtzou) and it seems that nothing has been executed the last few days. So I assume (not sure 100%) that the error in the Registry side, in particular in the point where the workflow execution is initiated. I think that we should ask @antleb.

Please could you test the same app also in services.openminted.eu? test.openminted usually is not as stable as services.

Dimitris

[martavillegas]

I registered the tool in services. When I run it I get:

There was a problem running the application. Try again in a while. (Unable to locate named workflow)