Text Pre-Processing: Smart Way

Overview

In this post, we will be discussing some quick and short tricks which are very effective and powerful and can save you a lot of time struggling in-text cleansing process.

Problems in text pre-processing

Making sense out of textual data is a tricky job and dirty text makes it a lot more troublesome. All of us have encountered, text related problem statements and are aware of the fact that cleaning text data is much more difficult than numerical data( that’s what I believe ).

When I say cleaning text data what is exactly am I talking about:
(1) Spelling errors
(2) Text Contractions
(3) Concatenated words

I will not be talking about the importance of Tokenization, Stemming, Lemmatization, and other pre-processing techniques.

Let’s knock them out one by one.

Spelling errors

Most of the data collected has this kind of human inserted spelling errors. Though we have some advanced NLP modeling techniques I prefer correcting these terms so that they make much more sense.

Text Contractions

As a human, we have amazing abilities to interpret complex words and sentences but machines are dumb, we need to feed them in a spoon. Text contractions are one such issue. I have seen its effectiveness in sentiment understanding problems where negation words are not neglected in pre-processing.

Concatenated words

Trends are represented by hashtags and hashtags decide the trends. It’s very important in problems like tweet classification, polarity analysis, and trend analysis. Segregating the words brings out a lot more insightful information about the topic.

Traditional Approaches

I have seen people using exhaustive list for contractions. Something like this

contractions = { 
"ain't": "am not / are not / is not / has not / have not",
"aren't": "are not / am not",
"can't": "cannot",
"can't've": "cannot have",
"'cause": "because",
"could've": "could have",
"couldn't": "could not",
"couldn't've": "could not have",
"didn't": "did not",
"doesn't": "does not",
"don't": "do not",
"hadn't": "had not",
"hadn't've": "had not have",
"hasn't": "has not",
"haven't": "have not",
"he'd": "he had / he would",
"he'd've": "he would have",
"he'll": "he shall / he will",
"he'll've": "he shall have / he will have",
"he's": "he has / he is",
"how'd": "how did",
"how'd'y": "how do you",
"how'll": "how will",
"how's": "how has / how is / how does",
"I'd": "I had / I would",
"I'd've": "I would have",
"I'll": "I shall / I will",
"I'll've": "I shall have / I will have",
"I'm": "I am",
"I've": "I have",
"isn't": "is not",
"it'd": "it had / it would",
"it'd've": "it would have",
"it'll": "it shall / it will",
"it'll've": "it shall have / it will have",
"it's": "it has / it is",
"let's": "let us",
"ma'am": "madam",
"mayn't": "may not",
"might've": "might have",
"mightn't": "might not",
"mightn't've": "might not have",
"must've": "must have",
"mustn't": "must not",
"mustn't've": "must not have",
"needn't": "need not",
"needn't've": "need not have",
"o'clock": "of the clock",
"oughtn't": "ought not",
"oughtn't've": "ought not have",
"shan't": "shall not",
"sha'n't": "shall not",
"shan't've": "shall not have",
"she'd": "she had / she would",
"she'd've": "she would have",
"she'll": "she shall / she will",
"she'll've": "she shall have / she will have",
"she's": "she has / she is",
"should've": "should have",
"shouldn't": "should not",
"shouldn't've": "should not have",
"so've": "so have",
"so's": "so as / so is",
"that'd": "that would / that had",
"that'd've": "that would have",
"that's": "that has / that is",
"there'd": "there had / there would",
"there'd've": "there would have",
"there's": "there has / there is",
"they'd": "they had / they would",
"they'd've": "they would have",
"they'll": "they shall / they will",
"they'll've": "they shall have / they will have",
"they're": "they are",
"they've": "they have",
"to've": "to have",
"wasn't": "was not",
"we'd": "we had / we would",
"we'd've": "we would have",
"we'll": "we will",
"we'll've": "we will have",
"we're": "we are",
"we've": "we have",
"weren't": "were not",
"what'll": "what shall / what will",
"what'll've": "what shall have / what will have",
"what're": "what are",
"what's": "what has / what is",
"what've": "what have",
"when's": "when has / when is",
"when've": "when have",
"where'd": "where did",
"where's": "where has / where is",
"where've": "where have",
"who'll": "who shall / who will",
"who'll've": "who shall have / who will have",
"who's": "who has / who is",
"who've": "who have",
"why's": "why has / why is",
"why've": "why have",
"will've": "will have",
"won't": "will not",
"won't've": "will not have",
"would've": "would have",
"wouldn't": "would not",
"wouldn't've": "would not have",
"y'all": "you all",
"y'all'd": "you all would",
"y'all'd've": "you all would have",
"y'all're": "you all are",
"y'all've": "you all have",
"you'd": "you had / you would",
"you'd've": "you would have",
"you'll": "you shall / you will",
"you'll've": "you shall have / you will have",
"you're": "you are",
"you've": "you have"
}

It doesn't look pretty.

For contractions also people prepare similar exhaustive lists and spelling errors we usually ignore.

A smart way to do this

I found some useful python libraries which can do this hard word in one line.

We can use Textblob to retrain your own spell-check.

Done!

What’s inside the Compound word algorithm?

A good first approximation is to assume all words are independently distributed. Then you only need to know the relative frequency of all words. It is reasonable to assume that they follow Zipf’s law, that is the word with rank n in the list of words has probability roughly 1/(n log N) where N is the number of words in the dictionary.

Once you have fixed the model, you can use dynamic programming to infer the position of the spaces. The most likely sentence is the one that maximizes the product of the probability of each individual word, and it’s easy to compute it with dynamic programming. Instead of directly using the probability we use a cost defined as the logarithm of the inverse of the probability to avoid overflows.