cometaNLP

A NLP Comments Extraction and Text Analysis tool for Italian, Dutch, and English social media comments.

Table of contents

• Installation
• TextAnalyzer
• Demo TextAnalyzer
• DataWrapper
• Demo DataWrapper
• Static methods
• Authors
• Q&A

Installation

Windows

py -m pip install cometaNLP

Unix\MacOS

python3 -m pip install cometaNLP

.ipynb

!pip install cometaNLP

Supported languages

TextAnalyzer

TextAnalyzer has one main behavior:

    """A class that operates on the text level to extract data and metadata from a comment. It applies a set of tranfromartion to it based on
    boolean arguments and returns a dictionary contatining the relevant information. If visualization is set to true, it 
    
    It applies a set of transformation on the string level.
    The user can choose which transformations
    to apply and which to not. If visualize is set to True, it also shows a simple visualization of the pos-tags.
    
    Args:
        self: reference to the current instance of the class
        input_string: Any string
        remove_punctuation (bool): Optionally choose whether to keep (False) or remove (True) punctuation from the input text
        remove_emojis (bool): Optionally choose whether to keep (False) or remove (True) emojis from the input text
        remove_stopwords (bool): Optionally choose whether to keep (False) or remove (True) stopwords from the input text
        TTR (bool): Optionally choose whether to calculate (True) or not (False) the input text TTR
        CFR (bool): Optionally choose whether to calculate (True) or not (False) the input text CFR
        lemmatize (bool): Optionally choose whether to lemmatize (True) or not (False) the input text
        pos (bool): Optionally choose whether to pos-tag (True) or not (False) the input text
        visualize (bool): If set to true, visualizes the pos-tags using nltk color schemes.
    Returns:
        dictionary: a dictionary containing the relevant data and metadata for the input text
    """

Demo TextAnalyzer

from cometaNLP import TextAnalyzer

analysis = TextAnalyzer(language='english')

analysis('My name is Leonardo @Leonardo #name', visualize = False)

#Output

{'n_hashtags': 1,
 'n_urls': 0,
 'n_user_tags': 1,
 'n_emojis': 0,
 'clean_comment': 'My name is Leonardo',
 'TTR': 1.0,
 'CFR': 2,
 'pos': ['NOUN', 'NOUN'],
 'tokens': ['name', 'Leonardo']}

DataWrapper

(1) data_wrapper

    """A class method that applies a series of tranformations to csv and tsv files
    and returns a dictionary
    
    data_wrapper() reads .csv and .tsv file, applies the functions and methods within the
    the `TextAnalyzer` module to the column contaning the comments/text in the dataframe
    in meaningful order, and finally converts it into a dictionary by index according to the following format:
    {index -> {column -> value}}. The dictionary contains relevant information
    for each comment in the dataset.
    To make the function as comprehensive as possible, the user is asked to enter
    whether the csv/tsv file has a header or not and the index of the column
    on which they wish to apply the transformations. The column should be the one containing the comments\text
    data that the user whishes to analyze.
    
    Args:
        self: reference to the current instance of the class
        file_type (str): A string (csv/tsv)
        file_path (str): A string containing a file path to a csv/tsv file
        data_frame (bool): If set to true, the function returns additionally pandas DataFrame object
                           rather than a dictionary
    
    Returns:
        output (dict): A nested dictionary {index -> {column -> value}} containing
                       relevant data and metadata for each comment in the input dataframe
    """

(2) data_wrapper_summary

    """A class method that returns the relevant data comparison based on grouping 
    comparison (e.g., X v. Y) rather than for each comment individually.
    
    data_wrapper_summary() is built upon the data_wrapper method. If visualize is set to True, it also shows a simple visualization of all the
    summarized data. It compares average number of emojis, hashtags,
    urls, user tags, length, type-token ratio, content-function ratio for
    two classes of comments.
    Args:
        self: reference to the current instance of the class
        file_type (str): A string (csv/tsv)
        file_path (str): A string containing a file path to a csv/tsv file
    
    Returns:
        tuple: a tuple of values.
    """

Demo DataWrapper

from cometaNLP import TextAnalyzer

datawrap = DataWrapper(language='Italian')

datawrap.data_wrapper(file_path="C:/Users/.../file.tsv", file_type='tsv')

#Output
{0: {0: 6847,
  1: "@matteorenzi ...all'invasione di questi animali, e forse è un offesa agli animali, stranieri che arrivano in Italia e solo in Italia?...",
  2: 1,
  'n_hashtags': 0,
  'n_urls': 0,
  'n_user_tags': 1,
  'clean_comments': "all'invasione di questi animali e forse è un offesa agli animali stranieri che arrivano in Italia e solo in Italia",
  'n_emojis': 0,
  'tokenized_comments': ['all',
   'invasione',
   'di',
   'questi',
   'animali',
   'e',
   'forse',
   'è',
   'un',
   'offesa',
   'agli',
   'animali',
   'stranieri',
   'che',
   'arrivano',
   'in',
   'Italia',
   'e',
   'solo',
   'in',
   'Italia'],
  'length': 21,
  'TTR': 0.8095238095238095,
  'CFR': 0.8888888888888888,
  'stop_words_removed': ['invasione',
   'animali',
   'forse',
   'offesa',
   'animali',
   'stranieri',
   'arrivano',
   'Italia',
   'solo',
   'Italia'],
  'lemmatized_comments': ['invasione',
   'animale',
   'forse',
   'offesa',
   'animale',
   'straniero',
   'arrivare',
   'Italia',
   'solo',
   'Italia'],
  'POS_comments': ['NOUN',
   'ADJ',
   'ADJ',
   'ADJ',
   'NOUN',
   'NOUN',
   'NOUN',
   'NOUN',
   'NOUN',
   'NOUN']},
 1: {0: 2066,
  1: 'È terrorismo anche questo, per mettere in uno stato di soggezione le persone e renderle innocue, mentre qualcuno... https://t.co/GFAvh0QLe5',
  2: 0,

Static methods

The TextAnalyzer module offers a set of @staticmethods that can be used independently from its main class methods:

• load

    """A method that reads .csv and .tsv files.
    
   
    Args:
        file_type (str): The type of file containing the data (csv/tsv)
        file_path (str): The path to the file where the file is stored.
      
    Returns:
        df (object): A pandas DataFrame object
        
    
    """
  

• count_hashtags

    """A function to be run on comments. It returns the number of hashtags.
    
    Args:
        text (str): Any string
    
    Returns:
        count (int): A count of the number of hashtags contained in the string
    """
  

• count_url

    """A function to be run on comments. It returns the number of urls.
    Args:
        text (str): Any string
    Returns:
        count (int): A count of the number of urls
    """
  

• count_user_tags

    """A function to be run on comments. It returns the number of user tags.
    Args:
        text (str): Any string
    Returns:
        count (int): A count of the number of user tags
    """
  

• preprocessor

    """A function to be run on the comments through apply to clean them.
    
    The function applies a series of transformation to the comments. Hashtags,
    urls, and user tags are removed.Digits and leading/trailing spaces are also removed.
    
    Args:
        text (str): Any string
    
    Returns:
        txt (str): The input text without hashtags, urls, etc.
    
    
    """
  

• punctuation_removal

    """A function to be run on comments. It returns comments without punctuation.
    
    Args:
        text (str): Any string
    Returns:
        txt (str): The input text without punctuation
    """
  

• count_emoji

    """A function to be run on comments. It returns the number of emojis.
    
    Args:
        text (str): Any string
    Returns:
        count (int): A count of the number of emojis
    """
  

• demojizer

    """A function to be run on comments. It returns the number of urls.
    
    Args:
        text (str): Any string
    Returns:
        txt (str): The input text without emojis
    """
  

• comment_length

    """A function to be run on comments. It returns the length of the comments.
    
    Args:
        text (str): Any string
    Returns:
        count (int): The comment length
    """
  

• word_counts

    """A function to be run on comments. It returns a dictionary containing the word counts.
    
    Args:
        l (list): Any list of strings
    Returns:
        counts (dict): A word-counts dictionary
    """
  

• type_token_ratio

    """"A function to calculate type-token ratio.
    
    A function to calculate the type-token ratio on the words in a string. The type-token
    ratio is defined as the number of unique word types divided by the number
    of total words. ATTENTION: requires the COMETA.word_counts() to run.
    
    Args:
        text (str): Any string
    
    Returns:
        
        float: A float expressing the comments TTR
    """
  

• visualize_pos

    """A function used to visualize POS-tagged comments.
    visualize_pos() is a void function. If 'visualize = True',
    a displacy visualization will appear on top of the TextAnalyzer
    get_data() output.
    
    Args:
        tokens (list): Any list of strings
    
    Returns:
        NonValue-Returning
    
    """
  

Authors

• Leonardo Grotti

Q&A

Q: Why does cometa start from distribution 0.0.3?

A: The first two distributions of the package were erased due to some naming conflicts.

Q: Why has version 0.0.3a0 been yanked?

A: The version was meant to fix additional bugs but failed in doing so. As such, it was yanked.