app.py

from deepmultilingualpunctuation import PunctuationModel
import gradio as gr
import re
import metrics
        
# https://stackoverflow.com/questions/22800401/how-to-capitalize-the-first-letter-of-every-sentence
def cap(match):
    return(match.group().capitalize())

def remove_filler_words(transcript):
    
    # preserve line brakes
    transcript_hash = " # ".join(transcript.strip().splitlines())
    print('transcript_hash')
    print(transcript_hash)
    # preprocess the text by removing filler words
    # Define a list of filler words to remove
    filler_words = ["um", "uh", "hmm", "ha", "er", "ah", "yeah"]
    words = transcript_hash.split()
    clean_words = [word for word in words if word.lower() not in filler_words]
    input_text_clean = ' '.join(clean_words)
    # restore the line brakes
    input_text= input_text_clean.replace(' # ','\n')
    return input_text
    # Define a regular expression pattern that matches any filler word surrounded by whitespace or punctuation
    #pattern = r"(?<=\s|\b)(" + "|".join(fillers) + r")(?=\s|\b)"
    # Use re.sub to replace the filler words with empty strings
    #clean_input_text = re.sub(pattern, "", input_text)    

def predict(brakes, transcript):

    input_text = remove_filler_words(transcript)
    # Do the punctuation restauration
    model = PunctuationModel()
    output_text = model.restore_punctuation(input_text)

    # if any of the line brake methods are implemented,
    # return the text as a single line
    pcnt_file_cr = output_text

    if 'textlines' in brakes:

        # preserve line brakes
        srt_file_hash = '# '.join(input_text.strip().splitlines())
        #srt_file_sub=re.sub('\s*\n\s*','# ',srt_file_strip)
        srt_file_array=srt_file_hash.split()
        pcnt_file_array=output_text.split()

        print('pcnt_file_array')
        print(pcnt_file_array)
        print('srt_file_array')
        print(srt_file_array)
        # goal: restore the break points i.e. the same number of lines as the srt file
        # this is necessary, because each line in the srt file corresponds to a frame from the video
        if len(srt_file_array)!=len(pcnt_file_array):
            return "AssertError: The length of the transcript and the punctuated file should be the same: ",len(srt_file_array),len(pcnt_file_array)
        
        pcnt_file_array_hash = []
        for idx, item in enumerate(srt_file_array):
            if item.endswith('#'):
                pcnt_file_array_hash.append(pcnt_file_array[idx]+'#')
            else:
                pcnt_file_array_hash.append(pcnt_file_array[idx])
    
        # assemble the array back to a string
        pcnt_file_cr=' '.join(pcnt_file_array_hash).replace('#','\n')

    elif 'sentences' in brakes:
        split_text = output_text.split('. ')
        pcnt_file_cr = '.\n'.join(split_text)
    
    regex1 = r"\bi\b"
    regex2 = r"(?<=[.?!;])\s*\w"
    regex3 = r"^\w"
    pcnt_file_cr_cap = re.sub(regex3, lambda x: x.group().upper(), re.sub(regex2, lambda x: x.group().upper(), re.sub(regex1, "I", pcnt_file_cr)))

    n_tokens= metrics.num_tokens(pcnt_file_cr_cap)
    n_sents = metrics.num_sentences(pcnt_file_cr_cap)
    n_words = metrics.num_words(pcnt_file_cr_cap)
    n_chars = metrics.num_chars(pcnt_file_cr_cap)

    return pcnt_file_cr_cap, n_words, n_sents, n_chars, n_tokens
 
if __name__ == "__main__":

    metrics.load_nltk()

    title = "Deep Punkt App"
    description = """
<b>Description</b>: <br>
Model restores punctuation and case i.e. of the following punctuations -- [! ? . , - : ; ' ] and also the upper-casing of words. <br>
"""
    examples = [['sentences', "my name is clara i live in berkeley california"]]

    interface = gr.Interface(fn = predict,
                         inputs = [gr.Radio(["no brakes","sentences", "textlines"], value="no brakes", label="preserve line brakes"),
                                   "text"],
                         outputs=[gr.Textbox(label="Punctuated Transcript"), 
                                    gr.Number(label="Number of Words"),
                                    gr.Number(label="Number of Sentences"),
                                    gr.Number(label="Number of Characters"),
                                    gr.Number(label="Number of Tokens")],
                         title = title,
                         description = description, 
                         examples=examples, 
                         allow_flagging="never").queue(concurrency_count=2)

    interface.launch()