Amazon Comprehend is an NLP service that identifies the language of the text; extracts key phrases, places, people, brands, or events; understands how positive or negative the text is; and automatically organizes a collection of text files by topic.

Would you recommend this product?
11 Reviews4.1/5
I spent a few minutes trying to figure out how this might be useful from a marketing perspective, and here's what I came up with: I recently came across a copy-writing 'hack' of sorts, which involves collecting a large number of negative reviews from a similar product/service to which you are attempting to write sales copy for and using a simple word-density analyser to determine which are the main pain-points that the feedback revolves around. That way, in theory, you can ensure that you're speaking to your audiences actual pains, as opposed to your assumptions of what they are. This could be used for the same purpose, only on steroids 💪🏼
Upvote (29)Share
@thecoppinger Exactly the perspective I was looking at this with. Now, I'll wait for someone to list an iteration of this with their own tool that does that specifically...
Here's a screenshot of what Amazon Comprehend looks like/does by the way: Can't wait to experiment with this on - could provide enormous benefit and insights for our users
Other providers like and are great at this too. See how Recast automatically find entities like Amazon and even identify "north" and "south": This is what we use for, working well I can tell you!
If you'd like to play with this technology locally, Stanford CoreNLP is a great open-source framework (GPL license) for NLP processing:
@imkmf Let's see a performance comparison.
@imkmf @brennanerbz @martinhn I've played around with a few different packages (CoreNLP + some derivatives, Indicoio, etc.) and was very pleased with the keyword and sentiment features of Comprehend. The sentiment feature especially seems far more accurate.
@imkmf @geteighty @jrwallenberg what’s remarkable is that not much has changed since the 1990s when the Stanford parser was born. We’re still only doing basic entity and sentiment analysis. Where’s the meaning?
@imkmf @geteighty @brennanerbz I can't speak for how sophisticated it was back then, but I find this stuff super helpful. I can barely write Python, but I'm able to automatically parse through thousands of customer reviews and extract what they're talking about and how they feel about it with a decent level of accuracy - using completely free resources!
This has the potential to make NLP somewhat more mainstream among tech companies. Curious to see what the adoption will look like.
Isn’t there a way to help teach it to recognize new entities? Without that, moving into any sort of vertical, like legal, would be impossible. SpaCy for Python produces about the same entities as in the example, and has a simple training model.
@martinhn I was successful in using Spacy with machine learning in the area of Legal Docs.