Decipher Text Insights and related Business Use Cases

The objective is to provide crisp information about possibilities in Text analytics and how it can be leveraged in business scenarios

“I got to work on this 500 page of tender, prepare a report in a day. How am I going to do it as still the last one is not sorted out; its submission date has already reached …”

Presales Team member (Legal/Business)

“we are getting hundreds of reviews for our apps hosted in app store… it’s too much to do with part time resource to understand what’s happening there ….”

App Owner

“Can someone give me insights on posts we are getting on twitter, please make sure I get them on a daily basis ….”

Head, Digital Channels

“Looks like we are missing some vital information in most of the warranty claims, I am getting a feeling that pre-defined data points are not giving enough information to get to the bottom of the reasons for so many frequent claims ….”

Warranty Claims Manager

“Looks like this IT request is for password reset and similar to what David marked lastly for another one but am not sure ….”

IT Service Desk Executive

All the above information from different channels holds qualitative information about existing customers as well prospects and the problems they are facing and hence an opportunity for improvement. The problems are pointing to not so easily available information hidden deep inside text. In some cases, problem gets augmented due to the size of text to be processed whereas for some its standardization of analysis which varies based on human skillset as well experience. Solutions to most of these problem statements are available or can be developed with implementation of Artificial intelligence (AI), powered by Machine Learning (ML) or (and) Deep Learning (DL).

Solutions for above listed examples comes under Text analytics domain and can be clubbed under one of the three major categories, also will try to brief at higher level what different solutions are available. Detailed solution approach we plan to cover for each category in subsequent posts

Classification:

It’s all about labeling a series of words or sentences to a more concise & meaningful one. This has lots of applications in different industries starting from labeling a service desk request to categorize the indirect parts consumed in manufacturing plants for costs optimization.

This is indeed the simplest type of text analytics from algorithmic perspective. It has become more mature over a period with the advent of machine learning algorithms used in tandem with NLP engines

Sentiment Analysis:

As mentioned in previously listed examples, Teams or Owner responsible for Apps as well social media channels are struggling to find out how is the customer experience post launch of new products or apps with the addition of new features; whether it’s getting good or bad or just ok. The information flow is almost 24 X 7 and may also go multilingual. Hence there is a greater need to have a system which can provide real time sentiment analysis of feedback received from customers by business from multiple social media platforms and possibly accommodate multiple languages.

All AI platforms provides out of box sentiment analysis as well there are many open source solutions available to get this done without any binding to any platform. All these are quite matured & efficient and powered by models developed using ML.

Summarization or Text Insights:

Online Feedbacks / Blogs / Posts

What these social media platforms as well similar apps do with data captured by user in form of tweets or posts or online feedback? What’s special in information keyed in by users can help business do better? Answer lies in serving customers by offering personalized service to each customer or prospect by developing a better understanding about them with the help of their reviews, feedback, views etc.

Legal Contract Documents Analysis

There is another area where quite a lot of information in structured manner flows out between business partners, although the volume of information generated is nowhere near to end users but from business impact perspective, it is quite a significant one. Here we are talking about Legal Business contract documents exchange between business partners which have quite significant value in commercial perspective. Considering the real-world timeline expectations and overall resource availability in general, it could be quite stressful for team to process them considering any lapse can result in significant loss to business.

Warranty Claims Reviews

Similarly, in case of warranty claims, a significant information gets lost between transformation of raw data to structured one by Service Agents or other manual means. Delving more into informal text provided by end user can provide more insights i.e. if there is any pattern or trend coming up, and can help in making more informed decisions

Text analytics with help of Deep Learning & NLP algorithms are handy in this scenario to provide summary or insights. Using them one can bring out insights of varied degree from the content (text).