Comments
The Role of Text Analytics in Customer Intelligence
You must login to participate in this chat. Please login.

nkabara - the audio is avaialble now (as of 7:23pm EST)

Blogger

Your closing statement, that there's no Easy button n text analytics, is so true!

The method you used in some of these examples, integrating text with stepwise logistic regression, is intriguing. I wonder how the results might compare with decision trees methods.

Blogger

Goutamc  Thank you for the site - will review!

Blogger

nkabra - the audio should post soon - the site has been doing so for recordings so far.

Blogger

Do we have a recording or video of the Text Analytics presentation. I missed it due to another meeting but heard it went very well....

Prospector

@pierre

please check out analytics.oskstate.edu and apply if interedted.

 

Blogger

Goutamc I may be interested in the program if possible.

Blogger

Goutamc - clensing is a big deal - there are some solutions online, like googe Refine, but it s still a manual process to place text into a tabular format.

Blogger

@Goutamc - Very true. That low hanging fruit can form at least a basic insight to what is possibly needed from the ooutside sources text.

Blogger

@BPOi0123 and @beth

Lots of editinga nd cleaning needed for tweets - some of thse can be automated. We spend about 30% of time first cleaning tweets before we can bring into text miner. tehn we often find more isues and go back and forth.

Blogger

Thanks again Goutam, great presentation! 

Data Doctor

We'll wrap up now. Listeners, thanks for tuning in. Goutam, again, thanks for the informative presentation!

Blogger

Thanks Goutam -- great examples and insight, from lots of projects over the years. 

Blogger

@Goutam: I enjoyed the talk. Thanks for the insightful use cases.

Prospector

We are at the top of the hour -- time has flown! 

Blogger

@beth

since most compamies have existing models with numeric data, the low hanging fruit is starting there and bringing in textual information and add that to teh model, IMHO.

 

Blogger

Thanks everyone for a great discussion.

Blogger

Much thanks Dr. Chakraborty for extremely informative presentation. 

Blogger

everyone

thanks for so many comments - I hope you got enough ideas from the talk. teh booh will give you more detaisla dn refernces.

 If you are interetsed in our online program, please see the web site mentioned on teh slide.. We also have a very big fulltime program and if you are interetsed in hiring studenst from my program 9as interns or full-time), do not hesitate to contact me.

Blogger

Goutam, do you think starting with existing model and supplementing that with text analytics is a good way to go (as in the couple case examples you cited)? Or does starting from scratch work better? Or is that too hard to call? 

Blogger

In the case of Chick-Fil-A, you'll have just as many people agreeing with the CEO as disagreeing, if not more. It's a basic blue state/red state split

Look at the whole #CancelColbert controversy

Beth - the idea of something someone says being held against a larger entity is problematic to begin with

@Goutam: I see! Thanks!

Prospector

GoutamC - where can companies find annotation resources? Is this a Mechanical Turk situation? Do they cull from in-house personnel?

That makes sense, Maryam, and the 'gold set' of data that's used to train a given algorithm may account for that

Thank you, Dr.Chakraborty! Very informative. Your presentatin was lucid and I was able grasp the concepts of Text analytics. I had vague idea before attending this lecture.

Prospector

@hospice

in thsio case, it was a single person that did the initial annotation and as you expect not all were good. We (me and my grdauate stduents) found several instances of wrong annotationa nd we had to go back to the company to chevk and mosify.

Yypically fora  reserach project I like to use at least 2 independent annotation with at least 85% interrate reliability

Blogger

I think the Keywords should be domain-specific, and are often determined based on the content of the dataset

Prospector

Michael i agree but i think that  cosnsistent neutral feedback should be examined maybe its better to think of neutrality on a scale

Blogger

@sethbredlove

Honestly, I use SAS almost exclusively (although my studenst sometimes uses R as well) but some of it can be done via other software as well. perhaps you should do a proof of concept project to see what's possible.

Blogger

Goutam, during the lecture you mentioned that cleaning up the tweets in the fast-food example was quite an effort. Do you by chance know what percentage of project time went into that effort? And, would that be a good rule of thumb to keep in mind if you're going to be working with tweets for text/sentiment analytics?

Blogger

@goutam - can you describe further what you did with the keywords?  I get it impacts sentiment - but how did it effect remediation?

 

Prospector

statician absolutely we have cerain expectations of a brand Mc Donalds vs Ritz Carlton

Blogger

GoutamC - do the keywords have to change with each project or model?

Maryam - a neutral sentiment is definitely not optimal, from a marketing perspective. But an algorithm may flag tweets/posts as neutral only because they don't contain distinctly positive or negative terms

@tonymjr -- thanks for digging out the earnings. I'm not sure I like the lesson here for CEOs. They seem to have carte blanche on what they can say without ramification. 

Blogger

@tony

soem of teh NDA prohibits me from saying exactly what was used in text analysis - but we found certain keywords spoken by customers were very good for incerasing the lift of the predictive model.

Blogger

@Maryam, so if the experience is linked to the brand then they are really just one in the same? 

Data Doctor

Great talk, really interesting and great links! Thanks!

Data Doctor

BPOI2013 - having humans read everything is prohibitively expensive. These algorithms can make the workload more manageable

Yes, GREAT talk. The case studies are very helpful in understanding the topic

Prospector

Great talk, thanks professor Chakraborty!

Prospector

@Goutam: To build a model you need the data to be annotated by domain experts. How do you evaluate the accuracy of the data annotation task? How many annotators per data instance?

Prospector

statitician yes the experieince is part of the brand and eventually if the expereinces are nuetral overall the customer simply not buy from that brand because there is nothing unique or special about them. Another brand can lure them away with a better expereince.

Blogger

Thank you, Goutam.  Your talk was very informative.

 

Prospector

great lecture - I like the way Goutam tied these concepts to real-world case studies

@ Goutam,  do you have any software recommendations for a small or midsized company that are just starting text analysis?  Or do you recommend outsourcing?

Data Doctor

Re collections. what was the remediation that provided the lift?  How was it determined through the text analysis?

 

Prospector

data doctor its interesting they are doing a complete change of direction in their new strategy this year to try and address this issue.

Blogger

@Maryam, but is the experience linked to the brand? will the customer make a distinction? 

Data Doctor

I agree. We need TIME. 

Becuase we have "tons of data" doesn't mean it's easy

Prospector

"I lunch at Chick-Fil-A every Tuesday and Thursday" would be flagged as neutral

@Michael, if they do not feel neutral about the brand then what does it indicate? 

Data Doctor

How do your tools compensate for typos, mixed languages (Spanglish) and text shorthand? A lot of data might not be caught, do you think there is probably better value in having analysts read all posts?

Prospector

maybe not about the brand but about the experience they had

Blogger

@ Maryam, I think it does show it hurt profits, because they are still having a problem two years later. After the start of the scandle, the good comments went away and never really came back. The damage seems to be permanent. 

Data Doctor

That's true, Maryam, but the sentiment being categorized as 'neutral' doesn't mean the poster feels neutral about the brand

I can really appreciate Goutam's slide #27.  Time to do this right is of the essence.

Prospector

Michael if a customer is neutral he can easily defect to a competitor for price location, etc. A loyal customer will not be driven by these factors solely they will patronize a brand they evangelize.

Blogger

I like that Michael states that text analysis is never easy.   Many execs do not understand analytics and don't support projects.   This varies greatly over indusitries. 

Data Doctor

Tonymjr Press being good is all relative - I need to review that article, but... it may have been a sales growth, but may not have been growth for the long term.

Blogger

I think neutral means that the algorithm can't tell whether it's positive or negative

@Hospice, I would think that neutral is customers no longer having a favorite inclination for your company, but thats just my take on it. 

Data Doctor

@Maryam: Does neutral mean I don't care about the company anymore?

Prospector

How do you mean, Maryam?

Huffington Post

In the latest sign that all press is good press, Leon Stafford of the Atlanta Journal-Constitution reported that Chick-fil-A's sales soared 12 percent, to $4.6 billion, in 2012.

http://www.huffingtonpost.com/2013/01/31/chick-fil-a-sales-2012_n_2590612.html

Prospector

No easy button in text analytics!  AGREED!!!

Prospector

neutral can be the death of loyalty

Blogger

What does that show us? What is a neutral message? "I am having lunch at Chick-Fil-A?" or "I'd like to hold my next PFLAG meeting at the local Chick-Fil-A?"

as we're all fast food customers it's intriquing what analytyics are showing about us and in our social comments about companies.

Data Doctor

@bulk: What insight do we get from "neutral" sentiment?

Prospector

I have to think that an event like this is a great case study for this type of analytics. really the perfect storm. 

Data Doctor

Chic-fil-A is still having problems and those comments were made in 2012.  It shows how social media can keep scandel and boycotts alive, rather than be quickly forgotten.  Once it's on the net, it is forever. 

Data Doctor

Henson & muppets? I'm missing the connection... although I didn't follow all of the conversation that followed from the CEO's statement.

Blogger

data doctor i think they can provide meaningful data if the company is willing to listen and change

Blogger

I will keep it to myself as long as nobody web scrapes the posts on this blog. ;)

Prospector

the fast food case, is an example of being prepared. This analysis could not have been done 6 months  after the event

Prospector

@sascertificate, I wont tell if you wont....

Data Doctor

Yeah... naming the company as the center of the concept map is probably not a good idea. ;)

Prospector

I'm not so sure the fast food industry will see customers providing a lot of unstructured data voluntarily that might be valid or useable comments and it would be good to hear more about those possibilities and how the industry will deal with that.

Data Doctor

He must be talking about Chic fil Lay. Big media mess. 

Data Doctor

yes and they are using it Chick fila is changing its entire business model based on its research

Blogger

this chicken restaurant seems to have become synonymous with that controversy, hasn't it? I'd say the sentiment persists.

the name is in the title of the article lol

Prospector

bulk - it depends on how much data the company has on that particular debtor

But in all honesty I can see the fast food industry really bennifiting by applying analytics, given the nature of the business. 

Data Doctor

Quick google "second largest fast food chicken restaurant" LoL..... just kidding... 

Data Doctor

The fast food industry seems to be a bit more difficult to implement the data without causing some privacy issues or public relations issues among customers, but it will be interesting to see how the data might well predict some outcomes.

Data Doctor

Automatic convertion of speech into text is not always accurate. Is the consertion done by human agents? 

Prospector

the collections case is very exciting. 

Prospector

I was always wondered if they really recorded every call.  Only because if you ask an agent to bring up that past recording, they always seem to avoid it. 

Data Doctor

In your collection case, is the remediation in scripts used or tone of the collection agent?  

Prospector

data dacotor ia agree sentiment is always a key variable with text

Blogger

@Michael, are there varying ways to calculate that metric that you saw in Orlando? 

Data Doctor

Predicting financial industry delinquent customer behavior through agent-customer phone interaction would seem to lead to some probable scoring errors from inexperienced or not well trained agents.

Data Doctor

It  also can show how the agent at the call agents response affects customer behavior. 

Data Doctor

Collecting debts from delinquent customers -- Hmmm!

Prospector

"Likelihood to pay" is a very critical metric - I attended a session on this back at Analytics 2013 in Orlando

lyndon I have seem off the shelf software that works well its becoming a growing and more accurate area especialy for call centers

Blogger

@Lyndon, that is something I was wondering as well, that is not something I am familar with in industry. 

Data Doctor

Value of  case 2 model: By understanding why customers call it makes easier to predict (and plan to reduce) call volumes in a contact centre. 

Prospector

seth exactly before that person tells their 10 friends and posts it on their facebook page and the expereince becomes its own ad

Blogger

How was call center data converted from voice to text? Voice recognition/parsing developed in-house or off the shelf software? How accurate?

Blogger

Would like to participate, but my company has tje audio blocked  :(

Content blocked by your organization
Reason:This Websense category is filtered: Internet Radio and TV.

Prospector

@aru, that is a great point. 

Data Doctor
I guess what we are looking for constructive neg comments. Things we can actually work on.
Data Doctor

 i think the key is in the timing changing the experience as soon as possible

Blogger

companies also reinforce being negative. if you complain, you may get something free. if you say something nice, you proabably get a sales pitch LOL

Prospector

the statistics are in line with human behavior a negative expereince will generate more negative comments positive experiences are expected

Blogger

@Goutham: How are the features of the model defined? 

Prospector
Also, people tend to be much negative on the net.
Data Doctor

Who me, complain? NEVER. ;-)

Blogger

i agree I don't think we can be anoymous anymore opt in is even becoming a grey area

Blogger

If the classification is done by company personal, the classification might be biaised.

Prospector

Scoring customer comments into negative and positive attributes seems easy enough but how to correlate when there's  more negative over positive comments in certain circumstances that might not predict sentiments of all customers accurately.

Data Doctor

this sounds like the JULIA system that Huffington Post put into place to filter site comments before they get to human moderators

Michael I don't believe God I could ever truly be anonymous always be pieced back together.
Data Doctor

It would seem the financial and oil and gas industry might be the industries that can immediately use unstructured data most readily as they see a huge growth in incoming data to utilize

Data Doctor

Michael, I don't generally think there's some insidious plot to preach anonymized data but not really do it completely, but I think there's real concern among lots of people that feel if somebody wants to work hard enough at it, a person can be linked to his or her anonymized data.

Blogger

@Michael: I don't believe in data 'anonymization', especially when the data is shared with a third party

Prospector

maybe its hard to find in one person because of the way analysts are educated--there is a divide between those that are structured gurus and those that specialize in unstructured

Blogger

@Michael, if we are talking about anonymizing data then I think it needs to start at the source, I don't think we can count on any other process doing it. 

Data Doctor

The iceberg analogy seems rather apt and the personality disorder requirement of analysts seems pretty on target as well!

Data Doctor

Multiple Personality Desorder  --- It means having many skills, right?

Prospector

@Beth, Yeah it is..... OMG! 

Data Doctor

modeling, business domain, hadoop - sounds like a couple of weeks of training

And that's a disturbing picture on slide 10!

Blogger

I think every analysis in the "big data" world is broken down to 90/10

Prospector

Beth, HH, do you think these entities can be trusted to 'anonymize' data and look at it in aggregate, or is there another tier that has personal profiles on individuals?

@Beth, I think the metaphor is a great fit, just not with this beautiful picture. LoL

Data Doctor

I love the iceberg metaphor in this context!~

Blogger

I really would not call the bottom of that ice berg ugly, its sort of majestic actually..... ;-)

Data Doctor

hospice sometimes beavior is the opposite of what consumers tell us they will do

Blogger

I agree with Beth - as long as I don't have anything to hide, per se

But, let me add, I think we're only at the beginning of how much companies will come to know about us -- and the future can seem kind of scarily Big Brotherish.

Blogger

but people don't like "negative" information being used

 

Prospector

@Hospice -- I'm not happy about it, but I'm not incensed about it either. 

Blogger

Customer Intelligence provides a detailed understanding of the experience customers have in interacting with a company, and allows predictions to be made regarding reasons behind customer behaviors.

So consumer behavior is not always the important insight?

Prospector

@hospice_houngbo - I think it depends on the company. People like recommendations from Amazon and Netflix

 

Prospector

@Hospice, I think as long as its transparent what is there to be unhappy about? If its not transparent then that is a different story. 

Data Doctor

Stakeholders can also be members of the public in case of govt orgs... Passengers in case of public transport sys...

Blogger

@ipolepalli - if you are at work; the audio may be blocked. if you have data on your phone you can go to site and listen

 

Prospector

I meant,  as customers, how many of you are "happy" about the way some companies gather data on you? 

Prospector

As customers, how many of you are about the way some companies gather data on you? 

Prospector

I have tried three different browsers and also installed latest flash player. it did not work

 

Prospector

Deciding when the predictive models may be improved with the use of adding unstructure text data may be a real challenge to get it right without lots of planning time I would think

Data Doctor

hospice exactly and with the nature of viral this explosion can easily take on alife of its own on the web

Blogger

Just thinking of how this all converges with enterprise managment solutions like onbase which are taking companies to a paperless office so all their data is digital. 

Data Doctor

@kq4ym - I think that depends on the question you're trying to answer, right? Ask the question, then determine what data could help you answer it. Think in that context rather than internal/external, structure/unstructured.

Blogger

This is very true and I've seen this in my life time.  At first, businesses did not go out and obtained external data.  Marketing decisions were often made on hunches, rather than information. 

Data Doctor

@Maryam: That's right! The type of unstructured data has changed as well, with more input from customers 

Prospector

@raficus -- the audio player should appear in the window about Goutam's picture. If you don't see it, perhaps try another browser.

Blogger

I wonder what the differences may be in determining which to use: external, internal data, and how to choose among the structure and unstructured for the best in reliable predictability vs. cost of  programming time and effort 

Data Doctor

Audio link, please...

 

Prospector

@ Hospice - I would agree with that statement.  We lived in a much simpler world back then. There wasn't social media and ect only just a few year ago. 

Data Doctor

I think there's a whole ecosystem of consultants and service providers to handle that challenge, kq4ym

@Seth, the only problem I see is that my wife might threaten to kill me if I try to take on another educational goal while still working on my phd. LoL

Data Doctor

Hospice I think the refernece is toward traditional analytics however today unstructured has expanded due to social media. We have always had unstructured data but it is expanding...

Blogger

I am sorry I do not see it

 

Prospector

I do not see any audio link....

 

Prospector

Yes, I also love to have some type of credit or certification on my resume as well. 

Data Doctor

We  grew up with structured data?

Prospector

hKp://watson.okstate.edu/

marke/nganaly/cs/ doe snot work

 

Prospector

@lpolepalli, you should see an audio player in the window above Goutam's picture

Blogger

Beth, that is a good endorsment in my book! 

Data Doctor

With the meteoric rise in customer text data availabe, I would think keeping current with the ability to store, maintain and manipulate this data may be a challenge for smaller companies without the experience needed to use the predictive analytics

Data Doctor

Is it started? I cannot see any audio button

 

Prospector

Hello everyone.  Seth here. 

Data Doctor

I've met a lot of OSU "analytics" students over the years, and all have positive things to say about the program (and Professor Chakraborty, as well)

Blogger

Yeah, 12 credit hours could make appealing and it could be a nice talking point on the old resume. 

 

Data Doctor

@Beth: Is the book free?

Prospector

The Oklahome State online graduate certificate in marketing analytics for either professional with a technical or even a non-technical background looks like a pretty cool idea with only a 12 hour minimum hour class load required.

 

Data Doctor

Anybody have a copy of his book?

Blogger

Customer intelligence (CI) is the process of gathering and analyzing information regarding customers; their details and their activities, in order to build deeper and more effective customer relationships and improve strategic decision making.  (Wiki)

Prospector

Hey everybody - welcome to the class

Very nice and detailled slides!

Prospector

Hi everybody. As a reminder, the audio player will appear in the window above at start time. If it does not play automatically, please click the left arrow. If streaming stops, you might have to re-click -- or try another browser.

Blogger

It's nice to find a goup that gets excited about analytics :-)

Prospector

I can't wait, really looking forward to it! 

Data Doctor

Hi everybody. We'll be starting today's lecture at 2 p.m. ET. I'm looking forward to the lecture, too!

Blogger

Looking forward to hearing your take on text analytics for marketing!

Blogger

I am looking forward to this chat, and hearing how text analytics has advanced. 

Blogger


INFORMATION RESOURCES
ANALYTICS IN ACTION
CARTERTOONS
VIEW ALL +
QUICK POLL
VIEW ALL +