[Journey] 1 million UVs/month in 12 months using AI generated content. Let's do it!

[prvncqr]

So use bs4 and paraphrase every paragraph separately. What is newspaper advantage over bs4?

Because newspaper3k is what I use and it returns everything with no paragraphs iirc. And when I paraphrase I lose the paragraphs too

We can easily split paragraphs by using some simple methods. Think for some time, you will get a definite solution. I have used newspaper3k to create paraphrased articles with proper segments without any issues.

 

[13385]

Fantastic journey!

Do you have anything in place to stop brandnames from getting into your content. Urls are easy to block, but brandnames and branded websites can be tricky I found.

 

[spectrejoe]

Another question, not sure if you check but isn't keyword density for your base keyword like really high if you add FAQ based on people also ask? (since every single question will always repeat at least a few words). And how are you getting the content to answer those questions? Are you simply getting the snippets from them & paraphrasing?

 

[rampzeus]

You're right, newspaper3k does have the paragraphs, they're just lost when I run my code through NLTK to separate the text into sentences so I can paraphrase, that's when the paragraphs are lost.

Think I may have figured out how to fix this now that u both called it out lol

Grab article with newspaper3k, split article on the '/n/n' to get the paragraphs. Then loop each paragraph to the paraphraser.

Newspaper3k has not been updated for a while so i am currently checking out alternatives like news-please.

Beter to have no images then same repeating one every 5-10 articles. It's hard to scrape good non copyright images in certain niches.

 

[Sartre]

Beter to have no images then same repeating one every 5-10 articles.

Use text on images to make them unique. Or something like that. Symbols maybe too? You can manipulate a lot of things.

 

[Niffo]

Use text on images to make them unique. Or something like that. Symbols maybe too? You can manipulate a lot of things.

Nah, just to some basic image manipulation. (Turn, crop, mirror, etc.) And try to filter out images with watermarks and logos.

The last one is the hard part. I can filter like 80-90% by filtering images containing text. If anyone has a better approach I would love to know.

 

[spectrejoe]

My issue with image manipulation is that Nider is with python 3.6 which is annoying as hell.

I can write text in the image with Pillow but it's a mess to make it fit perfectly within every single image since all images have different sizes... Plus nider makes it a lot easier & better visually than Pillow

 

[desperado596]

We can easily split paragraphs by using some simple methods. Think for some time, you will get a definite solution. I have used newspaper3k to create paraphrased articles with proper segments without any issues.

Grab article with newspaper3k, split article on the '/n/n' to get the paragraphs. Then loop each paragraph to the paraphraser.

Newspaper3k has not been updated for a while so i am currently checking out alternatives like news-please.

Beter to have no images then same repeating one every 5-10 articles. It's hard to scrape good non copyright images in certain niches.

You are scraping paragraphs but how are you combine it ? one paragraphs of each site ? is it gonna be readeble how are you going to make it difference to other paragraphs. ?

 

[Archemike]

looks like quillbot just takes args via http

https://quillbot.com/api/paraphrase...y=false&fthresh=-1&inputLang=en&quoteIndex=-1
It's also not blocked with login. can't you just rotate ip per set ammount of characters and use infinitely inside a script with curl? where was the rate limiting?

 

[Sartre]

My issue with image manipulation is that Nider is with python 3.6 which is annoying as hell.

I can write text in the image with Pillow but it's a mess to make it fit perfectly within every single image since all images have different sizes... Plus nider makes it a lot easier & better visually than Pillow

You might have problems with installing Nider/Pillow on Python 3.7+:

Linux:

• sudo apt-get install python-dev libjpeg-dev libfreetype6-dev zlib1g-dev
• do the above before installing pillow. if you already installed then uninstall and pip install --no-cache-dir -r requirements.txt

Windows:

• python -m pip download nider==0.5.0
• python -m pip install --no-deps -c constraints.txt nider==0.5.0
• python -m pip install colorthief
• python -m pip install -r requirements.txt
• Problems? pip uninstall pillow, do these steps again.

constraints.txt file contains 1 line:
pillow>=9.0.0

 

[hwp994108018]

this I don't understand. You could easily train a model which predicts with 99% if a site is an autoblog/spam site.

When the page base is hundreds of billions, 1% of pages mean that a lot of "non-spam" content is missed.
In addition, it seems that this has something to do with stock prices, and the prosperity of the network (even if false) will benefit stock prices

 

[desperado596]

You might have problems with installing Nider/Pillow on Python 3.7+:

Linux:

• sudo apt-get install python-dev libjpeg-dev libfreetype6-dev zlib1g-dev
• do the above before installing pillow. if you already installed then uninstall and pip install --no-cache-dir -r requirements.txt

Windows:

• python -m pip download nider==0.5.0
• python -m pip install --no-deps -c constraints.txt nider==0.5.0
• python -m pip install colorthief
• python -m pip install -r requirements.txt
• Problems? pip uninstall pillow, do these steps again.

constraints.txt file contains 1 line:
pillow>=9.0.0

You are scraping paragraphs but how are you combine it ? one paragraphs of each site ? is it gonna be readeble how are you going to make it difference to other paragraphs. ?

Could you help with that ?

 

[Archemike]

You are scraping paragraphs but how are you combine it ? one paragraphs of each site ? is it gonna be readeble how are you going to make it difference to other paragraphs. ?

Could you help with that ?

Something like splice arrays based on keywords into your template do that the ordering stays in line with the articles theme. you get top 5 serps, paraphrase, ensure they're similar in a list of the article and then reorganize accordingly. You're probably scraping as a blob of a string. This is hard to parse and reorganize. Recategorize at top of script.

 

[hwp994108018]

no point, just put it on a server and post using WordPress REST API. easy AF.

I got two rtx 3080...

GPT is garbage. We use own stuff/other services.

100% true.

sorry I've been too busy. Real life got in the way too much (kids and stuff)


Google cannot afford the computational power to do that. I'm 100% positive of that. Google "new york seo expert" and check what's #1. Almost 3 months up. (hint: it's lorem ipsum + the phrase "New York SEO Expert" inbetween + black hat links).

Instead or proxies I use a SERP API. I don't want to advertise any company here, you can Google it, there's a dozen to choose from. Costs me $10-20 to get enough scraping done to research and construct a big website. So nice not having to play around with proxies and handle errors.

recently I found out a fantatic and super fast Python library called yake, which I use to extract keywords from paragraphs/articles/long-tails and compare them with keywords and construct articles accordingly. That's what we use now. Used machine learning before, very little difference. https://pypi.org/project/yake/

Also Levenshtein distance is always a useful algo for many things! https://pypi.org/project/python-Levenshtein/

Another site is picking up nicely!!!! So far haven't experienced any problems with Google.

View attachment 198610

congratulations!
Also, how many automated websites are you running now after starting this thread?

In general, websites enter an explosive period at the beginning of the fourth month.

If all goes well and you create more than 30 websites, I bet you can reach 1 million traffic per month in 6 months.

Also, it is a bit difficult to monetize such websites, and perhaps you should do some preparation for this in advance.

How do you get keywords, I looked at threads, you focus on PPA keywords, but it doesn't seem to specifically describe how to get keywords.

I wish you an early date towards your goal.

 

[Alexion]

Any idea how to remove extra html tags from newspaper3k with article_html?

I need <h> and <p> along with <li> etc, but don't want the divs with their class. It takes up character count and it adds up when I'm using deepL api.

Looking for a programmatically way, otherwise I can load them in google docs and it does that for me.

Any thoughts on this?

 

[spectrejoe]

Any tips for improving the text uniqueness? Other than simply paraphrasing, what else can you do to make sure it's 100% unique/doesn't trigger duplicate content issues

 

[spectrejoe]

Any idea how to remove extra html tags from newspaper3k with article_html?

I need <h> and <p> along with <li> etc, but don't want the divs with their class. It takes up character count and it adds up when I'm using deepL api.

Looking for a programmatically way, otherwise I can load them in google docs and it does that for me.

Any thoughts on this?

Just iterate through these tags with beautifulsoup and save the text inside each <h> or <li> into a list, then when iterating through your text phrases if it contains the text from inside the <h> or <li> you add those tags to it

 

[Sartre]

congratulations!
Also, how many automated websites are you running now after starting this thread?

In general, websites enter an explosive period at the beginning of the fourth month.

If all goes well and you create more than 30 websites, I bet you can reach 1 million traffic per month in 6 months.

Also, it is a bit difficult to monetize such websites, and perhaps you should do some preparation for this in advance.

How do you get keywords, I looked at threads, you focus on PPA keywords, but it doesn't seem to specifically describe how to get keywords.

I wish you an early date towards your goal.

I've prepared for the monetization part in several ways. I'm sorry if I'm not revealing everything.

I get keywords by scraping PAAs. I also analyze trends (looking at global DNS queries) and search for websites that are growing rapidly and steal their keywords using an algorithm I've coded recently.

I've got over 20 websites right now. I'm 100% hitting the initial goal in a few months.

Hey guys I got an idea. Tell me a niche and if I like it I will make a website with 1000 articles just for fun for the sake of this thread and publish the url here PG-13 stuff only please

 

[spectrejoe]

I've prepared for the monetization part in several ways. I'm sorry if I'm not revealing everything.

I get keywords by scraping PAAs. I also analyze trends (looking at global DNS queries) and search for websites that are growing rapidly and steal their keywords using an algorithm I've coded recently.

I've got over 20 websites right now. I'm 100% hitting the initial goal in a few months.

Hey guys I got an idea. Tell me a niche and if I like it I will make a website with 1000 articles just for fun for the sake of this thread and publish the url here PG-13 stuff only please

Go for a competitive niche like weight loss

 

[GVald]

I've prepared for the monetization part in several ways. I'm sorry if I'm not revealing everything.

I get keywords by scraping PAAs. I also analyze trends (looking at global DNS queries) and search for websites that are growing rapidly and steal their keywords using an algorithm I've coded recently.

I've got over 20 websites right now. I'm 100% hitting the initial goal in a few months.

Hey guys I got an idea. Tell me a niche and if I like it I will make a website with 1000 articles just for fun for the sake of this thread and publish the url here PG-13 stuff only please

Good stuff man; Why not focus on just a few sites tho? That way, you're using the previously built authority for all the new articles