Text Analysis : a brief introduction
You are communicating all the time!
Be it with your family, neighbours, manager, customer, client, pets, alien, spirits, self, or with anyone in your parallel universe.
These communications can happen either in the form of a ‘speech’, ‘text’ or ‘gestures’. And every single one of them carries an insight, be it relevant or not.. Right!?
Now, all those human conversations that you can avail in the form of text including those which can convert from speech-to-text, can be processed by a machine to extract relevant insights.
Wait, what about Gestures? — No!
Pet’s Bow-wow? — No No!!
So, what is Text analysis?
Text analysis is Natural Language processing (NLP) that processes unstructured text(s) to automatically extract and classify relevant information out of them using different techniques such as Topic extraction, Sentiment analysis, Aspect classification, Named entity extraction, and more. These texts could have been generated in different sources such as survey responses, emails, product reviews, tweets, support tickets, etc.
A business may use this to extract specific information from a text such as company information, classify competitors, keywords, or names. They may want to categorise their texts with tags and also classify it with a sentiment (positive, neutral or negative).
Do ‘I’ need to perform text analysis?
Irrespective of your conscious need, you ALWAYS perform text analysis.
When your sibling accidentally hits you and says, “Don’t tell mom about this!”
Your brain automatically processes sentence and extracts different factors out of it, such as:
Sentiment: Negative (Positive for me cuz I can use this to threaten my sibling :P)
Named entity: Mom
Topic Category: A usual Sibling thing!
See!! You do perform text analysis, don’t you?
Okay, then what about text analysis at work?
Many times, you have lesser texts and hence may opt to process them manually. However, incase of numerous unstructured texts, let the ‘Machine’ do it for you!
It has tons of feedback flowing in every minute from multiple sources, all of them carrying some or the other insight which are useful for the Facebook’s stakeholders (esp. the Management team) w.r.t. different aspects. Let’s say UI/UX, feature requests, bugs, sales, content, etc.
All these texts are unstructured and ‘supposedly’ human-written (unless a cat or a robo did the typing).
So, processing such a mammoth amount of data by a team of humans will be very time consuming, costly and inconsistent (as interpretations may vary due to multiple-humans processing the data). This could take Forever..
That’s why, in such a scenario, “MACHINE’’ becomes your best friend!
Talking about Machine learning in simple words w.r.t. text analysis, you can train a machine to perform such text analysis for you.
It’s the same as you would teach a human, using a set of data to let it learn ‘How’ to analyse a piece of text. Then use this ‘model’ (machine’s brain that just learnt ‘how’ to perform the text analysis) to run other texts through it and generate the desired structured output within a few seconds unlike manual processing. Also, the machine gets more wiser when you feed it with more data for training, giving you a higher accuracy rate overtime.
“Who said ‘Practice just makes a Man perfect!’ Even machines get better..”
Well, now that you have a brief idea about text analysis, learn more about its different techniques and methods of text analysis in our next blog!
Written by: Aishwarya Prasad