get_context.py

def get_context(text, list_of_tokens, context_span=20):
    import re
    context = []
    for token in list_of_tokens:
        all_occurences_indices = [m.start() for m in re.finditer(token, text)]
        for index in all_occurences_indices:
            left_index = max(index - context_span, 0)
            right_index = min(index + context_span, len(text))
            substring = text[left_index: right_index].strip()
            
            if left_index == 0:
                substring = ' '.join(substring.split(' ')[:-1])
            elif right_index == len(text):
                substring = ' '.join(substring.split(' ')[1:])
            else:
                substring = ' '.join(substring.split(' ')[1:-1])
            context.append(substring)
    return context


# Test
text = """Twitter said Tuesday that Trump's tweets about mail-in votin did not violate the company's rules because they don't explicitely discourage people from voting. But, the company said, the label offers context surrounding Trump's claims"""
get_context(text, ['Trump'], context_span=30)

# [ "Twitter said Tuesday that Trump's tweets about mail-in", "offers context surrounding Trump's claims" ]