Question

This is impressive!  I have never done keyword research like this.This is impressive! I have never done keyword research like this. Have you?  
https://searchengineland.com/how-to...29912

Answer 1

Using machine learning for doing a keyword research sounds like overkill. It's not that complex

Comments

@legroom = more time for something else.  

---

Yeah ya still need to sort through all of them lol. If it included number of searches, KD, and whether it was a money keyword for sorting purposes, now THAT would be helpful :)

---

If you have scaling bottle necks it's useful. Little local client just starting out it's probably burning to much profit to execute

Answer 2

I dono. Might be a great Co Worker ;-D

Answer 3

This is great but I feel that there would only be a few niches it would work for. Reason being is that a lot of top ranking sites for niches are awfully put together but still rank well, ie the web design niche ‍♂️

Answer 4

Keywords are 1998 fella you need topical clusters!  

Comments

Of course, there is a move by Google towards entities, but I keep finding "flaws" for things which I would not have expected. An interesting example I noted was, if you search for "Priti Patel husband", Alex Sawyer was returned (as at a few days ago; this morning in London I get different results) with the image carousel returned high up in the searches, and the first Alex Sawyer who came up was an English actor (not her husband, who is a marketing consultant as well :) ), alongside other politicians who ARE more relevant to the search. To me it would make sense that major news sites would the most regularly crawled and updated, as they need to be live for breaking news and I would have imagined this sort of search to yield more precise results, especially given how much she's been in the news lately, but it wasn't. This makes me have doubts in the idea of totally dropping at least occasional "isolated" keyword/ keyphrase research. In quite a few cases, I've seen things that make me feel that direct keyword/ keyphrase research isn't totally out yet. I prefer to tread cautiously in each case, and would rather use more than one method and review as much data as possible just to be sure, for at least some cases. Just, your comment made me think this.  

Answer 5

What would be the point? If you're not selling the same products you're just spinning CPU cycles.  

Comments

@puiia I suppose. If you have a large team of content writers you need to fill the pool of editorial assignments. If you don't have the production capability then finding all these keywords is overkill.  

Answer 6

That's pretty cool :) Also I would definitely recommend checking out Word2Vec / Glove pre-trained word embeddings for more advanced keyword clustering models. They require no labelling / training data as the word embeddings for each word contain the semantic context in a vectorized format. Unsupervised clustering methods include: - Hierarchical clustering - Kmeans / Db Scan

Comments

@lorenzetti is that an app ?  

---

@coucal I'm making a few free tools to do this. Feel free to subscribe here: sempioneer. com

---

When I clicked sign up

---

@lorenzetti cheers will check it out

---

@coucal thanks for the warning I'll be sure to check this out!  

---

@lorenzetti I was on IPhone btw