NLP Class

What's the NLP class?

NLP inherits from TextChunk.

NLP is a class designed to give access to relevant data at the different levels (document, sentence, subsentence) in an intuitive way. It allows you to perform quick data exploration, manipulation and analysis. It's also used to perform requests and can save as well as load result as JSON objects.

When a response from the API is received it's stored in a hierarchy of classes: NLP (all data) => Document => Sentence => Subsentence => Token

At each level direct access it's possible to access inferior levels i.e. nlp.sentences gives access to a list of all the Sentence in the current data, while nlp.documents[0].sentences only gives the Sentence of the first Document.

NLP is iterable and will yield Document instances.

Attributes & Properties

Name	Type	Description
documents	list of Document instances	List of all the Document instances
sentences	list of Sentence instances	Direct access to all of the Sentences instances.
subsentences	list of Subsentence instances	Direct access to all of the Subsentence instances.
tokens	list of Token instances	Direct access to all Tokens in the subsentence
fields	list of string	List of all common properties accessible at all levels (token, span, cluster, pos etc.)
client	instance of Client	Client used for performing request to Lettria's API
Common properties	depends on property	Properties allowing access to specific data (pos, token etc.)

NLP methods

Data analysis

Below is an overview list of the methods that can be used to manage data with the API.

Method	Description
add_documents()	Submits a batch of documents to API
add_document()	Submits document to API
save_data()	Saves data from json file
load_results()	Loads data from json file
reset_data()	Erase data and reinitialise object
add_client()	Adds new client / api_key
to_annotation_format()	Exports the data to the input format of our annotation platform.

add_documents()

add_documents(documents, batch_size, skip_document = False, id=None, verbose=True)

This method allows you to batch the documents in order to accelerate the calls, to be used in priority elsewhere.

Parameters:

Name	Type	Description	Optional
document	list of string	Data to be sent to the API	False
batch_size	int	Number of documents to be sent at the same time to the API. Reduce if documents are too big.	32
skip_document	bool	Whether to skip the document if there is a problem during processing	True
document_ids	list of str	List of Id to identify the documents, by default an incrementing integer is assigned.	True
verbose	bool	Whether to print additional statements about document processing.True

add_document()

add_document(document, skip_document = False, id=None, verbose=True)

This method allows you to batch the documents in order to accelerate the calls, to be used in priority elsewhere.

Parameters:

Name	Type	Description	Optional
document	string or list of string	Data to be sent to the API	False
skip_document	bool	Whether to skip the document if there is a problem during processing	True
id	str	Id to identify the document, by default an incrementing integer is assigned.	True
verbose	bool	Whether to print additional statements about document processing.True

save_results()

save_results(file = '')

Writes current results to a JSON file. If no file is specified the default path is results_X.json with X being next 'free' integer.

Parameters:

Name	Type	Description	Optional
file	string	Path of file to write in.	True

load_results()

load_results(path = 'results_0', reset = False)

Loads results from a JSON file.

Parameters:

Name	Type	Description	Optional
path	string	Path of file to load.	True
reset	bool	Whether to erase current data.	True

reset_data()

reset_data()

Erase all data inside NLP and reinitialise documents ids.

Parameters:

Name	Type	Description	Optional
file	string	Path of file to load.	True
reset	bool	Whether to erase current data.	True

add_client()

add_client(client = None, api_key = None)

Replaces current client with provided one, or creates a new client using the provided api_key.

Parameters:

Name	Type	Description	Optional
client	instance of Client class	Client instance to replace the current one.	True
api_key	string	Key to use for the new client.	True

to_annotation_format

to_annotation_format(output_file, attribute=None, filter_list = [], verbose=True):

Writes data to a file in the annotation format for lettria's platform.

Parameters:

Name	Type	Description	Optional
output_file	string	Name of the file to write to.	False
attribute	string	Attribute to be used to preselect tokens to annotate. Defaults to None.	True
filter_list	list	Filter used to compare to the chosen attribute. Defaults to [].	True
verbose	bool	Turns on/off verbosity. Defaults to True.	True