Name: How text mining can help you understand your customer needs
Uploaded: 2023-06-23T11:07:30.375Z
Duration: 327.96

MIMS Text Mining V4.mp4

Data mining MIMS.vtt

English

{{00:00:00}}
Hi, I'm Piers Allington and we're
gonna talk today mostly about how text

{{00:00:04}}
analytics provides insight to companies.

{{00:00:25}}
Text mining gets a lot of
different definitions to it.

{{00:00:30}}
Text mining, text analytics,
sentiment analysis, emotion analysis,

{{00:00:35}}
natural language processing.

{{00:00:36}}
All of these things are different
labels, really, put on the simple

{{00:00:40}}
expedient of trying to turn unstructured
customer feedback, the words that

{{00:00:46}}
people use when they express themselves
in surveys or reviews or interviews.

{{00:00:50}}
Trying to turn that unstructured
verbatim content into meaningful,

{{00:00:57}}
categorised information that the
company can use as structured data.

{{00:01:05}}
Text mining works in a
variety of different ways.

{{00:01:07}}
Each company that does text
analysis does it in a different way.

{{00:01:13}}
Artificial intelligence engines have
become very much more to the fore

{{00:01:18}}
over the last five years or so, and
you're using the innate capabilities

{{00:01:23}}
of artificial intelligence, AI, in
that respect to try to make sense of

{{00:01:29}}
the variety of ways people have of
saying the same thing and categorising

{{00:01:34}}
that and making it meaningful.

{{00:01:36}}
So if you imagine any sense of
customer service or customer support,

{{00:01:42}}
care, aftercare, care management,
customer management, all those

{{00:01:47}}
things basically mean the same thing.

{{00:01:49}}
So, any text analysis engine is
trying to categorise and make sense

{{00:01:54}}
of these things, however the customer
talks about it and there is no end

{{00:01:59}}
to how customers use different words.

{{00:02:02}}
It's the power of language really,
the breadth and spread of how people

{{00:02:06}}
explain things is really what the text
analysis engine is trying to, to narrow

{{00:02:12}}
down and make sense of into meaningful
clumps, buckets, categories of data.

{{00:02:22}}
In terms of the software that's
used and how much the machine is

{{00:02:26}}
left to get on with itself, the
big mistake people make is assuming

{{00:02:31}}
that the machine has got it sussed.

{{00:02:33}}
And particularly with text analytics,
you need to have human intervention

{{00:02:38}}
and human monitoring to have a
continuous improvement process

{{00:02:42}}
because of the complexity of language.

{{00:02:45}}
And very often in language there's,
there's what's called missing data.

{{00:02:50}}
Not that the customer has missed words
out, but there's an assumption in how

{{00:02:56}}
people speak and how people write things
that you know what that means as a

{{00:03:00}}
human, but the machine will never know.

{{00:03:03}}
So if we take a simple example,
imagine you're in a pizza restaurant

{{00:03:06}}
and one of the waiters says, "Oh, the
window seat needs more doughballs."

{{00:03:12}}
Clearly, as a human, we all know
that it's the people sitting at the

{{00:03:15}}
window seat who want more doughballs.

{{00:03:18}}
The seat itself hasn't got an appetite
for carbohydrates all of a sudden.

{{00:03:23}}
So that sort of missing data
is what humans use all the

{{00:03:26}}
time to make speech efficient.

{{00:03:29}}
Machines don't have that, so you need
to continuously train the machine.

{{00:03:34}}
All AI tools work because
they've been trained by humans.

{{00:03:37}}
Text analytics has no end game.

{{00:03:40}}
It's always open-ended.

{{00:03:42}}
There is no definitive this means this,
certainly from a machine's point of view,

{{00:03:47}}
so you can't leave it to the machine.

{{00:03:53}}
The pitfalls of text
analytics are twofold, really.

{{00:03:58}}
First of all, you've gotta categorise
what it is people are talking about.

{{00:04:03}}
Add the sentiment score to
understand how positively or

{{00:04:05}}
negatively people are talking.

{{00:04:07}}
So that's the machine
processing piece of it.

{{00:04:10}}
Accuracy is twofold.

{{00:04:12}}
One is, am I actually extracting
all the things about this subject?

{{00:04:18}}
So this, first of all,
it's the machine effort.

{{00:04:21}}
The second big part is how do you
then present that to people so

{{00:04:25}}
that they can use the information
to turn it into insight.

{{00:04:29}}
Merely doing the text analytics, it's
just creating another pot of data.

{{00:04:32}}
The trick is to turn that into
insight that the company can use.

{{00:04:36}}
Using the insight genuinely to drive
the business forward is really critical.

{{00:04:41}}
So the, the hurdles to using text
analytics effectively are two steps,

{{00:04:47}}
and they require two distinct mindsets
to really get the information out

{{00:04:51}}
into the business to use it properly.

{{00:04:53}}
And because of the complexity,
it can sometimes rather fall into

{{00:04:57}}
its own trap of just becoming
too complicated to use easily.

{{00:05:03}}
So it's all about the short term
response where people have given you,

{{00:05:08}}
they've taken the effort to give you
the feedback that's important to them.

{{00:05:12}}
All too often that's ignored.

{{00:05:14}}
If, if you think about your own
surveys that you filled in, when does

{{00:05:17}}
anyone ever get back to you and say,
oh, I'm sorry we dropped the ball.

{{00:05:20}}
Here's what we're gonna do about it.

Hi, I'm Piers Allington and we're
gonna talk today mostly about how text

Text mining gets a lot of
different definitions to it.

Text mining, text analytics,
sentiment analysis, emotion analysis,

All of these things are different
labels, really, put on the simple

expedient of trying to turn unstructured
customer feedback, the words that

people use when they express themselves
in surveys or reviews or interviews.

Trying to turn that unstructured
verbatim content into meaningful,

categorised information that the
company can use as structured data.

Text mining works in a
variety of different ways.

Each company that does text
analysis does it in a different way.

Artificial intelligence engines have
become very much more to the fore

over the last five years or so, and
you're using the innate capabilities

of artificial intelligence, AI, in
that respect to try to make sense of

the variety of ways people have of
saying the same thing and categorising

So if you imagine any sense of
customer service or customer support,

care, aftercare, care management,
customer management, all those

So, any text analysis engine is
trying to categorise and make sense

of these things, however the customer
talks about it and there is no end

It's the power of language really,
the breadth and spread of how people

explain things is really what the text
analysis engine is trying to, to narrow

down and make sense of into meaningful
clumps, buckets, categories of data.

In terms of the software that's
used and how much the machine is

left to get on with itself, the
big mistake people make is assuming

And particularly with text analytics,
you need to have human intervention

and human monitoring to have a
continuous improvement process

And very often in language there's,
there's what's called missing data.

Not that the customer has missed words
out, but there's an assumption in how

people speak and how people write things
that you know what that means as a

So if we take a simple example,
imagine you're in a pizza restaurant

and one of the waiters says, "Oh, the
window seat needs more doughballs."

Clearly, as a human, we all know
that it's the people sitting at the

The seat itself hasn't got an appetite
for carbohydrates all of a sudden.

So that sort of missing data
is what humans use all the

Machines don't have that, so you need
to continuously train the machine.

All AI tools work because
they've been trained by humans.

There is no definitive this means this,
certainly from a machine's point of view,

The pitfalls of text
analytics are twofold, really.

First of all, you've gotta categorise
what it is people are talking about.

Add the sentiment score to
understand how positively or

So that's the machine
processing piece of it.

One is, am I actually extracting
all the things about this subject?

So this, first of all,
it's the machine effort.

The second big part is how do you
then present that to people so

that they can use the information
to turn it into insight.

Merely doing the text analytics, it's
just creating another pot of data.

The trick is to turn that into
insight that the company can use.

Using the insight genuinely to drive
the business forward is really critical.

So the, the hurdles to using text
analytics effectively are two steps,

and they require two distinct mindsets
to really get the information out

And because of the complexity,
it can sometimes rather fall into

its own trap of just becoming
too complicated to use easily.

So it's all about the short term
response where people have given you,

they've taken the effort to give you
the feedback that's important to them.

If, if you think about your own
surveys that you filled in, when does

anyone ever get back to you and say,
oh, I'm sorry we dropped the ball.

Make it Make Sense

Explainer

Text mining has many different alias' - sentiment analysis, text analysis, natural language processing. But whatever you call it, they all share the same purpose - turning unstructured customer feedback into "meaningful, categorised information". But what does this actually mean and how do you do it?
Piers Alington, data-pioneer and consultant at Alington Consulting, joins the Streamly Studio to help make sense of text mining.

Data

Insights

Data visualization

Data Transformation

Digital Data

Data & Analytics

Insights Strategy

medium_piers.jpg

small_piers.jpg

thumbnail_piers.jpg

xsmall_piers.jpg

piers.jpg

Piers Alington

Founder and Consultant

The breadth and spread of language is neither straight-forward nor easy to group together. Alington discusses how a text mining machine can interpret and categorise the language people choose to use every day. 


Understanding the power of language

Gleaning data from text mining is just the first step of the process, next is knowing how to present that data as insight for the company. Alington explains the 2-step effort that's needed for championing text mining in the workplace. 
