Joyful Art

This past week I’ve been “burning the midnight oil”.

Every night I’m up late experimenting on an intricate convolutional neural network that may eventually find its way onto this blog, but working to 3 AM every morning isn’t conducive to writing my usual long form, so today’s post will follow in the footsteps of last week’s Sentimental Art.

My friend says I should be more “visible” on my blog, to which my reply was:

“My Gavatar is my photo, how much more visible can I be?”

Their response involves something to do with “familiarity” and “trust”… I dunno. 😛

But hey, they may be right and my readers seem to like when I release art, so… please enjoy:

 

 

 

 

I hope you enjoy my art.

If you appreciate my work and would like to say thanks, please support me over on Patreon for as little as $1 a month, thank you!

But, if all you can do is like, share, comment and subscribe… well, that’s cool too! 😉

Much Love,

~Joy

 

Sentimental Art

You’ve probably noticed I flout the traditional wisdom of having a set schedule and regularly posing on certain days though I try to post at least post once a week (usually) depending on my schedule.

Occasionally I just take a week off.

That doesn’t mean that I don’t appreciate my readers though!

Without you I’m just building one hell of an online portfolio and resume! 😉 😛

I’m always looking for new and interesting ideas to talk with you about.

The thing is, I create a LOT of content that “falls on my cutting room floor”.

Sometimes its art sometimes is code… sometimes it’s an entire post!

I cut because I run tests and decide to take the project in a different direction… or I come up with a better solution… or I talk myself out of posting about a stupid subject… sometimes I start over and draw again and again until I get it right.

But whatever the reason, there are times I still like what I did even if I couldn’t use it for it’s original purpose.

That’s what this post is about.

Here’s a few works in various stages of completion that were either cut, completely unused or otherwise I just feel like presenting now.

Let’s start with the computer screen from Testing Sentiment Polarity, this actually got used but here it is up close.

Up close you might recognize it as the merged bot polarity graph from that post. 🙂

Next is a set of test shots I created when trying several camera angles for that same post.

Notice they say “Loading…” because I actually had the computer modeled before I had the graph ready. 😛

Also, if you are in front of a computer count the number of rows of keys on your keyboard…

Now count the rows on the keyboard I modeled… 😳 😛

Guess who was rushing to model a keyboard during Xavier’s nap and totally forgot a row of keys… 😆

You might also notice I hadn’t modeled the mouse yet either. If you look closely there are 3 other things missing from these images that are in the final feature image but I’ll leave them for you to find. 😉

Up next I have an alternative version of the featured image for the Introduction to Sentiment Polarity. This almost became the style for the screens on the other Sentiment Polarity posts.

And finally… I added a little echo…  er… I mean Nuclear Neon. 😛

 

I hope you enjoy these images.

If you appreciate my work and would like to say thanks, support me over on Patreon for as little as $1 a month.

But, if all you can do is like, share, comment and subscribe… well, that’s cool too! 😉

Happy Friday!

Much Love,

~Joy

 

Writer Bot: Sentiment Polarity Testing

Today we’re going look at testing our prototype Sentiment Polarity bot.

You can find the other posts in this series here:

• Writer Bot: Sentiment Polarity
• Writer Bot: Sentiment Polarity Training

And as always you can find more content to enjoy on my Topics & Posts page.

Writer Bot: Sentiment Polarity Training

Last week we talked about Adding Sentiment Polarity to our Writer Bot.

Today we’re going to look at how to train the bot (actually 3 bots) and the code that does it.

As I mentioned last week, I wanted us to use the Brown Corpus for this but currently this is just one of my many weird hobbies… like talking to trolls and contemplating what it would be like if we put nukes in space… and I do have a Patreon if you’d like to help change that. 😉

Anyway, since the Brown Corpus is over 1 million words… and my time is limited, I opted to use a publicly available data set that I found over on Kaggle (not a sponsor, though obviously they should be! 😛 ) and you can get it here:

Dataset On Kaggle: https://www.kaggle.com/marklvl/sentiment-labelled-sentences-data-set

I will also make it available in the GitHub repo for this project which I will link to at the end of this post.

So… because we’re not making our own training data set I’m obligated to follow any stipulations the authors have placed on their work…

Their exact instructions for using their data set are:

Please cite the paper if you want to use it 🙂

So here goes: Our training data set we will be using today was created for the paper:

‘From Group to Individual Labels using Deep Features’, Kotzias et. al,. KDD 2015

On behalf of all my readers, I extend our sincere gratitude to the authors for their generosity in allowing us to use their work.

That should do it now let’s get started!

Writer Bot: Sentiment Polarity

It’s time we take the next step forward with Writer Bot!

You’ve seen the current best public results of Writer Bot in A Halloween Tale and I’ve been slowly building additional features and blogging about them (See: Writer Bot Posts).

Today we’re going to start work on Sentiment Polarity which is part of the larger concept of Sentiment Analysis.

In a nutshell, this bot can/will evaluate if a sentence is more “positive” or more “negative”. Which is a little different than ‘Emotional Analysis’ and we’ll get there!

But you are getting ahead of the class again! 😛

 

Why/How is this useful in the real world?

Well, some heads of state and politicians use similar tools to evaluate “the way the wind is blowing” on the trending twitter issues of the day to “flip-flop” er… I mean… “take the peoples position”. 😛

Some dictators use sentiment analysis to crack down on ideas they oppose and keep them from spreading online! 😦

Lots of companies use Sentiment Polarity to determine if you like (or hate) their products and some systems will even help them determine if you like them only a little 😐, A lot 🙂, ♥Love♥ them 😀  or h8te them with a passion 👿!

This can be applied to your brand as a whole by (for example) randomly surveying the last 10,000 mentions for your name on some social media platform or you can even apply it to any specific reference by a specific person of a specific item, topic or service that you offer.

Now here’s where it’s going to get interesting if you like reading… or writing. 😛

Some authors use sentiment analysis (including sentiment polarity) on their own work to act as a sort of “digital critic” or “advisor” and or “editor”.

This enables an author to evaluate what they wrote using metrics to ensure that the sentence, paragraph or chapter serves it’s exact intended purpose to the story and has the proper “tone” for the subject matter.

For example, who ever heard of an upbeat sugar-sweet super positive horror novel? I mean… if you can make that work by all means please run with it!

But… if you think about it, it’s certainly an unusual concept.

This is exactly how Writer Bot will benefit from sentiment polarity.

Sentiment Polarity provides Writer Bot the ability to evaluate and compare two or more sentences to determine which is the most/least positive or negative and if we’re evaluating more than two sentences we can even sort them from most negative to most positive.

Used properly this would enable Writer Bot to make sure that the polarity of it’s generated text meets our needs and if not, it can reject it or potentially even suggest alternatives or simply make self directed corrections in fully autonomous mode.

For now though, we will just focus on classifying the polarity of sentences we give it.