[English] Multiple sentences grouped into one when a newline character is used

[dsenkyr]

How to reproduce the behaviour

# Sentences (13 in total) separated by '\n' / source: https://data.mendeley.com/datasets/7zbk8zsd8y/1 | g24-unibath.txt
text = "As a depositor, I want to deposit and maintain datasets through a simple web interface, so that I don\'t need to install and learn new software to deposit.\nAs a depositor, I want to have a user interface that is familiar to me, so that I feel like all the University systems are joined up.\nAs a depositor, I want to deposit and maintain datasets through Pure, so that I have a single onestop shop for managing my research outputs.\nAs a depositor, I want to deposit and maintain datasets through Virtual Research Environments and other workflow tools, so that I can continue to work with tools with which I\'m familiar.\nAs a depositor, I want to deposit the files that I have, so that I don\'t have to spend a lot of time finding the right version and converting to the right format.\nAs a depositor, I want to place data under an embargo, so that my right of first use is protected, and I can fulfil my confidentiality responsibilities.\nAs a depositor, I want to apply licenses to datasets, so that my IP rights are protected appropriately.\nAs a depositor, I want to allow my collaborators privileged access to datasets, so that we continue to have a productive relationship.\nAs a depositor, I want to deposit arbitrarily large files, so that I am not limited in what files I can and cannot deposit.\nAs a depositor, I want to link datasets to publications in Opus, so that both my data and publications are more easily discovered.\nAs a depositor, I want to mint DOIs for my data, so that it can be discovered and cited more easily and citations can be tracked so that I can receive credit.\nAs a depositor, I want to have metadata automatically filled from other University systems and remembered from previous deposits, so that I don\'t have to waste time reentering the same information.\nAs a depositor, I want to link to data stored in external repositories, so that I can store my data in an appropriate repository but still register it with the University and I don\'t have to deposit my data in multiple places."
doc = nlp(text)

index = 1
for sentence in doc.sents:
    print(f"=========================================== {index}")
    print(sentence)

    index += 1

Current output

As you can see, multiple sentences are grouped together in the second recognized sentence by spaCy.

=========================================== 1
As a depositor, I want to deposit and maintain datasets through a simple web interface, so that I don't need to install and learn new software to deposit.
=========================================== 2

As a depositor, I want to have a user interface that is familiar to me, so that I feel like all the University systems are joined up.
As a depositor, I want to deposit and maintain datasets through Pure, so that I have a single onestop shop for managing my research outputs.
As a depositor, I want to deposit and maintain datasets through Virtual Research Environments and other workflow tools, so that I can continue to work with tools with which I'm familiar.
As a depositor, I want to deposit the files that I have, so that I don't have to spend a lot of time finding the right version and converting to the right format.
As a depositor, I want to place data under an embargo, so that my right of first use is protected, and I can fulfil my confidentiality responsibilities.
As a depositor, I want to apply licenses to datasets, so that my IP rights are protected appropriately.
As a depositor, I want to allow my collaborators privileged access to datasets, so that we continue to have a productive relationship.
As a depositor, I want to deposit arbitrarily large files, so that I am not limited in what files I can and cannot deposit.
As a depositor, I want to link datasets to publications in Opus, so that both my data and publications are more easily discovered.
As a depositor, I want to mint DOIs for my data, so that it can be discovered and cited more easily and citations can be tracked so that I can receive credit.   
As a depositor, I want to have metadata automatically filled from other University systems and remembered from previous deposits, so that I don't have to waste time reentering the same information.
=========================================== 3


=========================================== 4
As a depositor, I want to link to data stored in external repositories, so that I can store my data in an appropriate repository but still register it with the University and I don't have to deposit my data in multiple places.

Expected output

When I use a space instead of a newline character, the output is as expected.

=========================================== 1
As a depositor, I want to deposit and maintain datasets through a simple web interface, so that I don't need to install and learn new software to deposit.       
=========================================== 2
As a depositor, I want to have a user interface that is familiar to me, so that I feel like all the University systems are joined up.
=========================================== 3
As a depositor, I want to deposit and maintain datasets through Pure, so that I have a single onestop shop for managing my research outputs.
=========================================== 4
As a depositor, I want to deposit and maintain datasets through Virtual Research Environments and other workflow tools, so that I can continue to work with tools with which I'm familiar.
=========================================== 5
As a depositor, I want to deposit the files that I have, so that I don't have to spend a lot of time finding the right version and converting to the right format.
=========================================== 6
As a depositor, I want to place data under an embargo, so that my right of first use is protected, and I can fulfil my confidentiality responsibilities.
=========================================== 7
As a depositor, I want to apply licenses to datasets, so that my IP rights are protected appropriately.
=========================================== 8
As a depositor, I want to allow my collaborators privileged access to datasets, so that we continue to have a productive relationship.
=========================================== 9
As a depositor, I want to deposit arbitrarily large files, so that I am not limited in what files I can and cannot deposit.
=========================================== 10
As a depositor, I want to link datasets to publications in Opus, so that both my data and publications are more easily discovered.
=========================================== 11
As a depositor, I want to mint DOIs for my data, so that it can be discovered and cited more easily and citations can be tracked so that I can receive credit.   
=========================================== 12
As a depositor, I want to have metadata automatically filled from other University systems and remembered from previous deposits, so that I don't have to waste time reentering the same information.
=========================================== 13
As a depositor, I want to link to data stored in external repositories, so that I can store my data in an appropriate repository but still register it with the University and I don't have to deposit my data in multiple places.

Your Environment

• spaCy version: 3.4.1
• Platform: Windows-10-10.0.19044-SP0
• Python version: 3.10.6
• Pipelines: en_core_web_trf (3.4.0)

[polm]

Sorry this isn't working well for you. Our pretrained components use data that doesn't contain newlines, so they don't always handle them well. We use whitespace augmentation when training to help with this, but for sentence boundaries in particular, the models have probably learned that only a period followed by a space terminates a sentence.

In this case I would preprocess your text to convert newlines to spaces, or, if your needs are more complicated, train a sentence recognizer.

[dsenkyr]

No worries, and thank you for your answer. Preprocessing is a fine solution for my use case.