Automated SEO & Content Scripts for Marketing Success ????????

The Best Of

Jul 24, 2023, 10:53 AM

As the mod for this channel, I wanted to start off with something hopefully useful to all of you. Over the last 6mo I've been writing automation scripts for SEO, Content Marketing, PR, and Social. I've been releasing them by request only on Linkedin, but I think this is a good place to share them all at once. If you find any of them helpful, please let me know!

<https://colab.research.google.com/drive/1uDzBkt_nbhZAgxwoSZQPAv74STD9wINN#scrollTo=Havir0g_UQyT>

Instant Content Plan
<https://colab.research.google.com/drive/1uSL1mAIfa21EAHBR2LxPYJFoT6OAL75A#scrollTo=Havir0g_UQyT> ZZ

Embeddings Based Clustering and GPT3 Cluster Descriptions
<https://colab.research.google.com/drive/1lkNajbtajXQVq6I3zHS88AxtmPbGYLyG#scrollTo=o70meP0dUPC5> zz

Automatic Content Ideas Based on Intent - The Massive Power of Chained Prompts + External Data (SERPAPI) zz
<https://colab.research.google.com/drive/1OlzRj8vPYUEmiVdPKlGFdRWhy26VZD21#scrollTo=OKzhaff2vtn_>

Automatic Consumer Research and Persona Generation
<https://colab.research.google.com/drive/1nZ7-wUQ9BT9KdJqzePR4Nqt4AdZgIgMN> zz

Automatic Reddit Trend Analysis with GPT3
<https://colab.research.google.com/drive/1Zq3g5A5ZJwSTPs8gieIZ6EfkxsLTPs-0#scrollTo=Yqy2k5CrOlto> zz

Intent, Personas, and Buyer Needs by Keyword
<https://colab.research.google.com/drive/1u5wQIICk-AM975R2rG6AtALZtopkqmfy#scrollTo=A3I-rXte9q82> zz

Automatic Article Outline Generation by Analyzing The Article Text of Top Ranking Pages for a Given Keyword
<https://colab.research.google.com/drive/1Ph8KmQ4zWv5ClnO-i4L9hsQ8-9O56pev#scrollTo=xeH01QSvLl3->

Automated Subreddit and Post Title Recommendations Based on Any Article
<https://colab.research.google.com/drive/1M4SorGOFqNyiuiDmIESCC-7UOj-Qd95-#scrollTo=-UxP6j4c2de4>

Free Script - Automatic Question Expander (Free Alternative to AnswerThePublic Powered by GPT3) zz
<https://colab.research.google.com/drive/1k6AUO13urMlLIY2FeKCa0v5PnqN7Owci#scrollTo=PyQcgc3Cq52v>

Automatically Generate a Summary, Article Outline, Long-form Article, and Tweet Thread from a Youtube URL ZZ
<https://colab.research.google.com/drive/1EkJrUkf7Ofxdeg8OLpsnr877e2nenI7_#scrollTo=wJnuKnTvFVMU>

Automated Newsjacking Content Ideation
<https://colab.research.google.com/drive/1k7vOzCp0hjNjRBfIq5Y5LtuFsQqy1QVw#scrollTo=Zf6ABC0EAhhY>

Automatic Brand or Entity News Media Monitoring and Analysis by <mailto:[email protected]|[email protected]>
<https://colab.research.google.com/drive/1B8aStd1xhIfcOMfOZMYh8u7aY4NiFWQF#scrollTo=4Pvjf3SaPhX9>

Free Script - Web Crawling with Content Analysis and Anchor Text Recommendations
<https://colab.research.google.com/drive/1VLTC2DJQd2JpGIpvt9VB202F_8T02XQZ#scrollTo=fI_P5xQBvEJQ>

Automatic Content Gap Report and Analysis With Clustering and Cluster Descriptions (Public)
<https://colab.research.google.com/drive/1_KSZxDaXKz0d2U9XtGKTi9vWEglSOTcf#scrollTo=0BqpW3sPJU3t>

Automatic Schema Improvements with GPT4 (Public)
<https://colab.research.google.com/drive/1NeJbtLGCEszwAkaq9KQ_9Ai-52wmAska#scrollTo=rL-UYrcMGNY_>

Automatic Long-Form Article Generation With Semantic SEO Driven Outlines and Iteration
<https://colab.research.google.com/drive/1WXqgN7btVM_Rak05wfa1NUjP97mBeijs#scrollTo=VEYI9ytc_Jgp>

Exploring the Future of Research and Long-Form Content: Harnessing the Power of Super Large Context Window LLMs with Python and Chained Prompting!
<https://colab.research.google.com/drive/1koQgOmoo04GFkPAsWYL7l7QfECwdduRL>

Large Language Model Search Optimization (Public)
<https://colab.research.google.com/drive/1OnUves5IL9FXxVk37l6rV4stCc2HRsBf#scrollTo=dK07vK-4d4VN>

Automatic Long-Form Article Generation With Semantic SEO Driven Outlines and Iteration - Now With Images!
<https://colab.research.google.com/drive/1ukB7uIHoIgY4kS_1FFi3Em730Gg32TnH#scrollTo=VEYI9ytc_Jgp>

Automatic Deep TikTok Insights with GPT and Whisper
<https://colab.research.google.com/drive/14tL3prBC70kKPl0T8raJJpjc3Cyhff_e#scrollTo=B9OxBhDpn1e->

Automatic Video Translation with LipSync
<https://colab.research.google.com/drive/1WSof1NGSQvfv8-meMjGTZaj9TkmxZ-K7#scrollTo=jtde28qwpDd6>```

Forwarded thread from another channel:

Boris Kuslitskiy

Jul 24, 2023, 10:54 AM

I guess I follow you on LinkedIn? I recognize that last one as something on my weekend to do list.

Kyle Faber

Jul 24, 2023, 10:54 AM

Super kind of you to do @kristin - thank you!

Kristin Tynski

Jul 24, 2023, 10:55 AM

Happy to help! Also, if any of you have ideas for new automations, please share them! Im looking for my next project.

Mark Barrus

Jul 24, 2023, 3:19 PM

Thank you @kristin I'll put these to good use

Eliza Morrison

Jul 25, 2023, 7:38 AM

thank you!!

Eliza Morrison

Jul 25, 2023, 7:39 AM

how does a non-technical person use this?

Eliza Morrison

Jul 25, 2023, 7:39 AM

i mean, i thought i was technical ????

Kristin Tynski

Jul 25, 2023, 9:45 AM

Most of them you just need your api keys, and then run each cell in succession!

Valentin Pletzer

Jul 27, 2023, 6:19 AM

Great collection. I wonder: Have you compared the differences in embedding results in any way? (openAI vs all-MiniLM-L6-v2) What do you personally prefer?

Kristin Tynski

Jul 27, 2023, 10:00 AM

This came out last year when the prices for embeddings were much higher that made me think (and then later after some experimenting on my own coming to a similar conclusion) that GPT's embeddings weren't much better than what you can get from open source transformer embeddings models on Huggingface.

Today OpenAI embeddings are WAY cheaper (like an order of magnitude), but I still don't really see the advantage. Their embeddings models are all old and GPT-3 based.

If they had a GPT-4 level embeddings offering, I'd certainly try it though!

I think for now using free models for embeddings is the right way to go.

Kristin Tynski

Jul 27, 2023, 10:02 AM

all-MiniLM-L6-v2 has been quite good for most of my use cases so far, but there are more specialized embeddings models that probably work better for things beyond what I have done, mainly sentence clustering.

Valentin Pletzer

Jul 27, 2023, 10:16 AM

Thanks. Sounds like I have to try all-MiniLM-L6-v2. So far I only used the openAI embeddings for sentences (mostly title + description) which works quite good for my use cases (news clustering and visiualizing)

Kristin Tynski

Jul 27, 2023, 10:20 AM

Yeah, if you find a better embedding model, let me know!

Valentin Pletzer

Jul 27, 2023, 10:27 AM

btw I saw your example using k-means which doesn’t work for my usecases because I don’t know the number of clusters beforehand. After some experimenting I landed on dbscan which works great

Kristin Tynski

Jul 27, 2023, 10:27 AM

One of my other scripts uses the elbow method for estimating clusters, it works ok, not great

Kristin Tynski

Jul 27, 2023, 10:28 AM

havent tried dbscan though

Kristin Tynski

Jul 27, 2023, 10:28 AM

Curious about your use case if you are willing to share

Valentin Pletzer

Jul 27, 2023, 10:33 AM

sure. I have two use-cases. One is to cluster evergreen-content and find opportunities by comparing our content with other sources. The other one is to visualize and cluster news. In the end a very similar approach to both.

Kristin Tynski

Jul 27, 2023, 10:35 AM

Really cool! What are you using for the visualization? Tsne? PCA? Umap? something else? Are you labeling your clusters with GPT?

Valentin Pletzer

Jul 27, 2023, 10:46 AM

So cool that someone is geeking out about the same things. ????

I’m using the Google umap js lib and the labels are entities. Every text has entities extracted via „classic“ nlp. I use dbscan on the 2-dimensional data for clustering and the centroid entity is shown in the screenshot

Valentin Pletzer

Jul 27, 2023, 10:46 AM

It’s not perfect but works reasonably well

Kristin Tynski

Jul 27, 2023, 10:48 AM

yeah, it looks like it is working great. I've used GPT to do the cluster labeling and found it can work really well for use cases where your clusters are maybe more ambiguous. You can ask it to look at everything with X cluster label and determine what it is "generally about." So even if some of your clusters have unrelated items, it can still give a great label for the entire cluster.

Valentin Pletzer

Jul 27, 2023, 10:58 AM

Cool