Apache Tika : Settting classpath for opennlp models on tika-server

I can't seem to set the classpath for the tika-server so that the opennlp models are detected correctly.

I've followed the instructions here:

https://wiki.apache.org/tika/TikaAndNER

(substituting app for -server, seen as that looked like it contained everything required)

I have created the following folder structure

tika
`-- tika-ner-resources
    `-- org
        `-- apache
           `-- tika
              `-- parser
                  `-- ner
                    `-- opennlp
                       |-- ner-location.bin
                       |-- ner-organization.bin
                       `-- ner-person.bin

Running:

java -classpath tika/tika-ner-resources -jar tika-server-1.18.jar --config /etc/tika-config.xml -enableUnsecureFeatures -h 0.0.0.0

and issuing {{ curl -v -XPUT --data-binary @test.pdf http://localhost:9998/tika --header "Accept: text/plain" --header "Content-Type: application/pdf"}}

results in

INFO going to load, instantiate and bind the instance of org.apache.tika.parser.ner.opennlp.OpenNLPNERecogniser WARN Couldn't find model from org/apache/tika/parser/ner/opennlp/ner-location.bin using class loader INFO LOCATION NER : Available for service ? false WARN Couldn't find model from org/apache/tika/parser/ner/opennlp/ner-organization.bin using class loader INFO ORGANIZATION NER : Available for service ? false WARN Couldn't find model from org/apache/tika/parser/ner/opennlp/ner-date.bin using class loader INFO DATE NER : Available for service ? false WARN Couldn't find model from org/apache/tika/parser/ner/opennlp/ner-money.bin using class loader INFO MONEY NER : Available for service ? false WARN Couldn't find model from org/apache/tika/parser/ner/opennlp/ner-person.bin using class loader INFO PERSON NER : Available for service ? false WARN Couldn't find model from org/apache/tika/parser/ner/opennlp/ner-percentage.bin using class loader INFO PERCENT NER : Available for service ? false WARN Couldn't find model from org/apache/tika/parser/ner/opennlp/ner-time.bin using class loader INFO TIME NER : Available for service ? false INFO org.apache.tika.parser.ner.opennlp.OpenNLPNERecogniser is available ? false INFO going to load, instantiate and bind the instance of org.apache.tika.parser.ner.regex.RegexNERecogniser INFO org.apache.tika.parser.ner.regex.RegexNERecogniser is available ? false INFO Number of NERecognisers in chain 0

The only thing that seems to work is re-packing the jar by adding the contents of the tika/tika-ner-resources directory (i.e. org/blah/blah/*.bin). The curl command then executes without any issues. I've also tried almost every combination of setting the classpath.

Does anyone have any ideas ?

Solution

For anyone else having issues, the following command worked for me by removing the -jar and manually specifying the TikaServerCli class

java -classpath tika/tika-ner-resources/:tika-server-1.18.jar  
org.apache.tika.server.TikaServerCli --config /etc/tika-config.xml -enableUnsecureFeatures -h 0.0.0.0