NLP::Service is a RESTful web service based off Dancer to provide natural language parsing for English.





The run() function starts up the NLP::Service, and listens to requests. It currently takes no parameters. It makes sure that the NLP Engines that are being used are loaded up before the web service is ready.

It takes a hash as an argument with the following keys:


Forces the loading of all NLP models before doing anything. The value expected is anything that is not 0 or undef, to be able to do this. Example,

NLP::Service::run(force => 1);


Takes in a configuration for the internal service implementation. Currently the implementation is using Dancer, and all of these keys correspond to Dancer::Config. For more details, refer to Dancer config for the acceptable values. Example,

NLP::Service::run(config => { logger => 'console' });


The load_models() function creates all the required NLP models that are supported. This is internally called by the run() function, so the user does not explicitly need to call them. It is useful however, for explicit loading of the models, if the models need to be used in unit tests or elsewhere.

In void context it returns nothing, but in scalar context returns the number of models that were loaded, and in list context returns a hash with the keys being model names and the values being the actual references to the perl objects that represent the models. This is rarely necessary for the user to be using.

It takes a single argument which is a boolean to forcibly load the parsers or not. By default the lazy load option is assumed unless explicitly set by the user. For example, NLP::Service::load_models(1) for forced loading and NLP::Service::load_models() for lazy loading.


Multiple formats are supported in the API. Most notably they are XML, YAML and JSON. The URIs need to end with .xml, .yml and .json for XML, YAML and JSON, respectively.

GET /nlp/models.(json|xml|yml)

Returns an array of loaded models. These are the model names that will be used in the other RESTful API URI strings.

GET /nlp/languages.(json|xml|yml)

Returns an array of supported languages. Default is "en" for English.

GET /nlp/info.(json|xml|yml)

Returns a hashref of details about the NLP tool being used.

GET/POST /nlp/relations.(json|xml|yml)

The user can get a list of all the english grammatical relations supported by the NLP backend.

GET/POST /nlp/parse/$model.(json|xml|yml)

The user can make GET or POST requests to the above URI constructed by the user or their programs. The $model corresponds to one of the available models such as "en_pcfg", "en_factored", etc. The list of supported models are returned by the GET request to /nlp/models.(json|xml|yml) URI.

The return value is a Part of Speech tagged variation of the input parameter data.

The parameters needed are as follows:


One of the parameters expected is data which should contain the text that needs to be parsed and whose NLP formation of Part-of-Speech tagging needs to be returned.

GET/POST /nlp/parse.(json|xml|yml)

This performs the same function as above, but picks the default model which is en_pcfg. It expects the same parameters as above.


Copyright (C) 2011. Vikas Naresh Kumar <>

This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.

Started on 25th March 2011.