[ihrishi]

I have been lurking on this forum for a long time and have been active on the two Harrier threads on and off. There is an amazing amount of information in these threads. Since I am learning to code again after years of being hands off, I thought it might be a good idea to run some analytics on the Tata Harrier thread.

So, ladies and gentlemen, what I did, is scraped all the comments (in this and the original Harrier thread) and extracted the entities mentioned in the comments. Entities usually are - person, location, organization, work of art, consumer good etc. Think 'nouns'. The first step is to see visually a summary of the entities and the frequency with which they are mentioned. I have filtered out a few obvious ones like - 'Tata', 'Harrier', 'car', 'one', 'point', 'lot', 'post', 'vehicle'

I will be looking at extracting more information from the text - including the sentiments of each of the above entities. It's slow work and I work on this over the weekends so kindly bear with me. Also let me know if you would like to see any more insights from the data! Will give it a go.

PS - Pls, do let me know if this is not relevant - will not post any more!



----------------------------------------------------------- ------------------------------------------------------- ---------------------------------------------------------

Continuing where I left off - the next step was to make more sense of the data.

So I did the following:

1. Used Google NLP API to extract the entities out of the entire lot of comments.
2. Extracted the Sentiments and their Relevance to the entire comment
3. Calculated product of Sentiment & Relevance
4. Lemmatized the entity names to reduce duplicates
5. Picked up the top 10000 positive entities and plotted them in a word cloud
6. Picked up the top 10000 positive entities and plotted them in a word cloud
7. Picked up the top 10000 negativeentities and plotted them in a word cloud

Note:

• These comments are from day 1 - hence they include BS4 feedback as well as BS6 feedback for the following threads:
• https://www.team-bhp.com/forum/offic...al-review.html (Tata Harrier : Official Review)
• https://www.team-bhp.com/forum/india...-69-lakhs.html (The BS6 Tata Harrier with 6-speed AT. EDIT: Launched @ 13.69 lakhs)
• Following words have been filtered since they did not add any value: 'Tata', 'Harrier', 'car', 'one', 'point', 'lot', 'post', 'vehicle', 'people', 'thing', 'something', 'reason'

Positive Entities


Negative Entities

[amarnani13]

Originally Posted by ihrishi

DISCLAIMER - Completely off topic - Mods - Kindly help if you feel if this is not relevant.

So, ladies and gentlemen, what I did, is scraped all the comments (in this and the original Harrier thread) and extracted the entities mentioned in the comments. Entities usually are - person, location, organization, work of art, consumer good etc. Think 'nouns'. The first step is to see visually a summary of the entities and the frequency with which they are mentioned. I have filtered out a few obvious ones like - 'Tata', 'Harrier', 'car', 'one', 'point', 'lot', 'post', 'vehicle'

I will be looking at extracting more information from the text - including the sentiments of each of the above entities. It's slow work and I work on this over the weekends so kindly bear with me. Also let me know if you would like to see any more insights from the data! Will give it a go.

[/i]

Dear ihrishi, I for one, enjoyed the analysis. I guess what is not coming up (or I failed to notice) is anything related to after sales service. You probably need to use some common word for that in your coding / library. Btw, what software are you using for the word cloud?

[nainan]

Originally Posted by ihrishi

I thought it might be a good idea to run some analytics on the Tata Harrier thread.

Extremely spot on analysis Rishi. If we combine price & pricing that would form the largest tag group here. Which in my mind is the biggest indicator of what should concern Tata the most about the Harrier. The Hexa and the Creta its biggest competitors? Which means that irrespective of segment and other considerations most people here compare/contrast Harrier with these two SUVs. Looks like Auto and AWD also have equal importance.

In my hypothesis the Harrier pricing against the Hexa & Creta and to a lesser extent the Compass is the overriding consideration most people here have period and no amount of premium tagging is diminishing that. I wish Tata listens and takes this feedback if not in the Harrier at least the Gravitas

[ihrishi]

Originally Posted by amarnani13

Dear ihrishi, I for one, enjoyed the analysis. I guess what is not coming up (or I failed to notice) is anything related to after sales service. You probably need to use some common word for that in your coding / library. Btw, what software are you using for the word cloud?

Thanks - I am using the WordCloud library in Python to render the image.
I had a look at the data and the word service appears 52 times and hence is very tiny. look under seater word in the bottom row - just below e and a.

[steadfast]

Originally Posted by ihrishi

Continuing where I left off - the next step was to make more sense of the data.

Does each color represent a specific subgroup ?

Mod Note - Please quote only the relevant bits of a post.

[ihrishi]

Originally Posted by steadfast

Does each color represent a specific subgroup ?

No - the color is from the tool's default color palette.

[steadfast]

Originally Posted by ihrishi

No - the color is from the tool's default color palette.

If color does not represent a category then most probably size of words would define number of instances of each world. Where bigger size is directly proportional to higher number of incidences and emotional intelligence attached to that keyword.
Is this a correct assumption. Does this NLP algorithm work this way ?

[ihrishi]

Originally Posted by steadfast

If color does not represent a category then most probably size of words would define number of instances of each world. Where bigger size is directly proportional to higher number of incidences and emotional intelligence attached to that keyword.
Is this a correct assumption. Does this NLP algorithm work this way ?

That is correct - the size of the words defines the number of times it has been repeated in the comments.
From an emotional - sentiment - perspective, I have actually segregated the positive and negative ones in different images; so that has no impact on the size.

[steadfast]

Originally Posted by ihrishi

That is correct - the size of the words defines the number of times it has been repeated in the comments.
From an emotional - sentiment - perspective, I have actually segregated the positive and negative ones in different images; so that has no impact on the size.

Great job indeed 👍

Looking at two figures, it seems that the two Harrier threads have a lot of noise i.e. lot of inputs from people who either do not own a Harrier or wish for skies in a Harrier at that price or even just bash the product by talking less or acknowledging the strengths of Harrier.

The key insights shared by official review and experience by real owners do not represent the major keywords from NLP algorithm. This is bad!

Even more dangerous situation - if experts at Tata perform a similar exercise on our Team-Bhp thread, they could come up with a misleading analysis.

[ihrishi]

Originally Posted by steadfast

Great job indeed ��
The key insights shared by official review and experience by real owners do not represent the major keywords from NLP algorithm. This is bad!
Even more dangerous situation - if experts at Tata perform a similar exercise on our Team-Bhp thread, they could come up with a misleading analysis.

True - and some of the comments by the expert reviewers were really shortsighted. A simple example is the location of the USB and the Aux port. Almost everyone cribbed about the location being inaccessible and hard to find.
What nobody thought about is - how many times do you really need to plug and unplug the cable from the port? It is probably just once till the cable lasts. Rest of the time you are going to be just plugging the other end of the cables in the phone!
So while it may be a chore for someone living with the car for a day, it is perfect for an owner!!

[Gannu_1]

Originally Posted by ihrishi

PS - Pls, do let me know if this is not relevant - will not post any more!

Dude, this is incredible!

You should do this for some of the popular threads here on Team-BHP. Would be interesting.

[Aditya]

CarCloud: I did a sentiment Analysis of the Tata Harrier launch & review threads - Posts moved to a new thread.

[CEF_Beasts]

Originally Posted by ihrishi

True - and some of the comments by the expert reviewers very really shortsighted. A simple example is the location of the USB and the Aux port. Almost everyone cribbed about the location being inaccessible and hard to find.
What nobody thought about is - how many times do you really need to plug and unplug the cable from the port? It is probably just once till the cable lasts. Rest of the time you are going to be just plugging the other end of the cables in the phone!
So while it may be a chore for someone living with the car for a day, it is perfect for an owner!!

Amazing analysis!

Regarding keeping the USB slotted in the slot could lead to drainage of battery after switching off the engine, I don't know if the Harrier also keeps supplying power through the USB. I am asking because in my EcoSport i had left the USB plugged in after switching off the car and removing my phone and after 2-3 days when I would crank the car it would not due to complete draining of the battery. After these 2 cases I realised my mistake and never keep the USB slotted.

[ihrishi]

Originally Posted by Gannu_1

Dude, this is incredible!

You should do this for some of the popular threads here on Team-BHP. Would be interesting.

Actually - that is the idea behind this 'project'. Let me refine this model and pipeline and then can apply it on any thread within Team-Bhp.

Mods: Is it ok if I post subsequent improvements and seek feedback for the fine tuning of the model?

Originally Posted by CEF_Beasts

Amazing analysis!

Regarding keeping the USB slotted in the slot could lead to drainage of battery after switching off the engine, I don't know if the Harrier also keeps supplying power through the USB. I am asking because in my EcoSport i had left the USB plugged in after switching off the car and removing my phone and after 2-3 days when I would crank the car it would not due to complete draining of the battery. After these 2 cases I realised my mistake and never keep the USB slotted.

Thank you!

Regarding the USB being plugged in - I meant only the USB cable plugged in, but not really connected to anything full time. Did even that cause the battery to be drained?
Also the Harrier slot seems to switch off when the engine is off. Will try this out next time just to be sure.

[CEF_Beasts]

Originally Posted by ihrishi

Thank you!

Regarding the USB being plugged in - I meant only the USB cable plugged in, but not really connected to anything full time. Did even that cause the battery to be drained?
Also the Harrier slot seems to switch off when the engine is off. Will try this out next time just to be sure.

Yup, only keeping the USB slotted in without any device attached it drained. I have observed this in my Creta as well.