I hate to be critical about something new, but at least for the example of the Wikipedia article on DNA repair, the summarized text is sort of completely off: http://summarai.com/summarai/dna_repair.html

However the extracted tags are pretty reasonable.

It also doesn't actually do very much sentiment analysis at all unlike the title says.

It actually does, it's part of a different summarization component.

If you feed it on what it expects, it does well. Now try a NY Times opinion piece and it is always way off.

The reason is that to summarize such content you have to use words that may not be used in the text at all. Also, the summary parts may not be present in the text either.

This is good stuff.

Pardon the plug, but I have something similar, a side project of many years: http://kbsportal.com

I will do some side by side comparisons tomorrow.

Here's a weekend hacking side project, not nearly as impressive, just for the fun of it.

It summarizes a chunk of text in a single sentence.

https://github.com/mck-/oneliner

Tried using your project for the same DNA article and your server crashed.

404 Not Found nginx/1.4.6 (Ubuntu)

as you said, its a side project which would explain why the system is not ready for primetime.

I am trying to reproduce the error you reported with the article starting "Human longevity is a complex phenotype" and I don't get the runtime error. Would you please email me the text that caused the error? (email is in profile).

Try side by side comparisons with their context-control

The summary for the 10-Q is surprisingly good! SEC filings are not formatted like news articles or books which most summarize algorithms seem to be trained on, at least the ones I have looked at, and do not perform well.

Is the source code available somewhere? Or an API? I would love to use this.

Hi, one of the creators here. Yes, an API is available. One of the unique things that it does is to allow you to control the context of the summarization which we call context-controlled summarization. It's based on a few recent approaches in AI. It's currently being used as part of a larger algo system in the financial markets and legal technology space.

Great! Is there a link with information about the API? Or do I need to contact you through the website?

rpedela: you can send a mail directly to me via kasian.franks@gmail.com or info@summarai.com

What is this? apikey3_6g4agvfjdvochfub5nfpcbh...? Try http://www.aljazeera.com/news/2016/01/zika-virus-16012815173..., captures js and cdn and other junk in tags. Seems cleaning algo needs some work. Great work though.