[FREE] Vector Spinner: free, powerful data-driven spinner

One more spinner for you guys.

Vector Spinner was my experiment in creating a spinner which doesn't use a thesaurus - instead, it looks for synonyms for both entire phrases and individual words using an approach inspired by automatic translation, based around NLP analysis of gigabytes of training data (English only, unfortunately). Easy-to-use, customizable, handles a variety of topics, and most of the results are readable (though errors might stack).

Features

• completely data-driven: trained on a database of actual English-language reviews and articles on a variety of topics
• tries to find replacements for both entire phrases and individual words (you can customize the balance between the two, or only pick one option)
• customizable in terms of uniqueness v. similarity to original, and depth of search
• should achieve between 70 and 90% uniqueness with standard settings (with very high 'phrases' setting and low 'quality', it sometimes approaches 100%)
• includes an open-class 'keywords' parameter, which can further guide results towards certain topics
• easy to add potential updates: if you think vocabulary for your content niche is missing, let me know and I might throw in more data to try and balance it out
• ridiculously easy to use

Things to bear in mind

• it's a bit slow: it may take up to a minute to spin stuff, at least initially (see the notes on performance)
• the spintax is very randomized, so it will always generate some gibberish or redundancy - make sure you play around with the settings/keywords!
• the examples generates are only a random subset and ranked by uniqueness, I recommend generating examples from the spintax yourself
• best substitutions are added first, so setting 'phrase' and 'word' spin settings lower will also generally improve coherence at the cost of uniqueness
• 'quality' setting corresponds to similarity to original words, but not necessarily coherence when spinning individual words - setting 'phrase' spin very high and 'word' spin very low may produce more varied content that's still readable

I had bigger plans for it originally, but ran into some difficulties on the way and won't have as much time to work on it from now on, so... enjoy, IT'S FREE!

Thanks to all the people who sent me feedback during the beta test. Let me know if you still encounter any bugs or problems, especially if you use the API.

Available as web app:

• http://jespersen.pythonanywhere.com/vector

Also available as an API:

• http://jespersen.pythonanywhere.com/vector_api_doc

The website doesn't collect any data, display ads, or log information.

VirusTotal check just in case (0/62):

• https://www.virustotal.com/pl/url/936b4107cc0c16f593716ec3c780c35f89af0e7981370b2f9c5a45f01dcd1004/analysis/1426192128/
• https://www.virustotal.com/pl/url/291a5e3d6fbb77fcb811d0f31d35b47f5a3733c3b34d1e78e04c4553fcf96478/analysis/1426192165/

I might still update some things, so go ahead and share any suggestions, complaints, etc.

It's a pity this didn't succeed as a project. It seems to output quite different phrases than an ordinary thesaurus spinner, thus increasing uniqueness, but the quality is pretty bad. I tried increasing quality and it did came up with some interesting spins however.

Thanks for the share, this might come in handy!

Thanks for sharing!

It's a pity this didn't succeed as a project. It seems to output quite different phrases than an ordinary thesaurus spinner, thus increasing uniqueness, but the quality is pretty bad. I tried increasing quality and it did came up with some interesting spins however.

Click to expand...

Yeah, you definitely need to play around with the settings (I may have made the defaults too random). Try the default first (4/8 for everything), then something like: quality at 6/8, 'spin phrases' at 8/8, 'spin words' at 1/8 + relevant keywords for the same content... compare the two and see if it makes any difference. The results also depend on the actual content. It works best for technical-style articles, worse for casual language.