“An early diagnostics
might save your Business”
What is sentimental
Analysis?
Sentiment analysis refers
to the use of natural language processing, text analysis, to
systematically identify, extract, quantify, and study effective states and
subjective information. Sentiment analysis is widely applied on reviews, survey
responses and social media for applications that range
from marketing to customer service.[Wikipedia]
Sentimental Analysis is not
something new, but, since the rise of the social media, Brands are referring to
this as the need of time. They require to understand wider view of the
customers’ feelings, opinion and attitude about the Brand.
Almost everywhere customer have
access to the social media forums, Facebook, Twitter etc. where Brands publish
about their products and services and Customers’ express their opinions,
feedback, expectations and emotions which describe their behaviors and attitude
towards the Brand.
What do the Organizations need to do?
Organizations are not limited to providing goods,
it can be a FMCGs, Education, Services, Politics and Government etc.
Organizations need to listen to the customers’
feedback and mine their sentiments. Organizations need the ability to translate these
insights collected from social data because it has shown that the social
sentiments of the customers are associated with the customer retention,
reputation of the Organizations, stock and many more critical business aspects. Below
are the key areas of focus for any business
- Identify
Priorities – Which sentiment they need to focus first?
Organizations need to
address the Negative Sentiments at
the earliest, so as they can manage the customer’s issue before it impacts
others or customer leaves. We know that, the cost of getting a new customer is
4 time more than retaining the existing customer.
“Make
sure you’re responding to any unhappy customers as quickly as possible.”
- Identify
Critical Factors / Perspective that impact the Business most.
In a Telecom Sector, if customers
are talking about a Network Coverage or
Low Service in any particular area
of large customer base with higher revenue, it might be the issues that needs
to be fixed at the earliest to avoid any drops in revenue.
- Focus on Organizations Image
Organizations spending
on corporate image is always at the higher side, they need to know if the
customers are really associated with the image – If the Organizations is spending
saying they are “Reliable” if the customer is not believing that the Organizations products or services are Reliable? than its time to address these sentiments affecting the Organization's image.
“Use sentiment reporting is to see
the response to certain campaigns, launches, or events”
- Competitors
Position
Organizations must pay
attention where do the competitor stand and how they can assure that they are
offering and attending their customer either competitively or better.
“Use
sentiment analysis to measure and report on how your competitors are talked
about on social media”
Challenges of Sentimental
Analysis Product
Customer feedback on the social
media is complex in nature, it gets more complex when people use the language
in arbitrary ways and mixing the different language, dialects and spelling the
words differently. Addition of emotion-icons ( J
L ) also changes the
meanings. It’s really a big deal to teach the machines to interpret this kind
of data.
Accuracy of the Sentiments is a key factor, it
requires both the human intelligence as well as adaptive approach to learn and improve
with time to assure quality of accuracy. Apart from the complexities discussed,
reviewing the results and adapting to different methodologies to assure the
precision is as good as possible is big ask.
Sentimental Analysis also require
the algorithm to cover the full sentence for evaluation, many algorithms are
looking for buzz words or certain sentences to evaluate, full coverage /
traversing of the data is a key factor for preciseness.
What Services we can Offer ?
We have used the NLP
(Natural Language Processing) algorithms which address all the above business
needs and complexities of Sentimental Analysis. But as we know, human are very intuitive in
communication, and develop certain new term, words, expression which a human
can understand easily but machines do need to be trained. We have developed an
adaptive learning mechanism which helps the algorithm to learn with time
adopting to such changes.
Current Product is trained on
Urdu and English used for the Telecom domain data of all Pakistan Telecom
Operators and precision is >80% which will increase gradually (2-4 months)
to reach 90% with adaptive approach.
Some of the major milestones of
our product are as below
- 1Data collection is automated from all major Social Media Channels (Facebook, Twitter, eBay, Amazon etc.)
- 2Product can read data from any website, journal etc.
- Product is trained on local dialects of Urdu, written in English and can detect misspells and local jargons
- We ensure that the coverage is 100%, even if we cannot build the complete context with all the data, algorithm will send notification to adaptive learning or human interaction if not solved by self learning.
Sample results based on actual data for Telecom domain in Pakistan
