4 ms·

Amazon Comprehend – Natural Language Processing

31 points by jeffbarr 9y ago

Xeoncross 9y ago
I was hoping that AWS Comprehend would have actual text comprehension. This seems to be just entity extraction and noun phrases. Basic Python Spacy.io or NLTK will get you this far. That said, I'm still glad they built it. Every time a new company offers a library or API like this I picture a new startup having an extra tool to enter a market.
- igravious 9y ago
  True true. That said, this takes a lot of the coding out of your hands. Presumably Microsoft and Google and IBM have similar cloud offerings. It's really hard to evaluate the quality of these offerings. I'd love if a website existed that had a cross platform analysis of these natural language processing site – basically compare and contrast these cloud offerings against rolling your own. It would have to use a few data-sets to get an all round picture.
  - dragonwriter 9y ago
    > Presumably Microsoft and Google and IBM have similar cloud offerings. Google certainly does: https://cloud.google.com/natural-language/ https://cloud.google.com/natural-language/
    - igravious 9y ago
      Thank you. Any others that you know of?
  - hallman76 9y ago
    > Presumably Microsoft and Google and IBM have similar cloud offerings. Yes, IBM has a similar product [1] along with a suite of APIs all under the "cognitive" umbrella[2]. Same [3] for Microsoft [4]. [1] https://www.ibm.com/watson/services/natural-language-classifier/ https://www.ibm.com/watson/services/natural-language-classif... [2] https://www.ibm.com/watson/developercloud/doc/index.html https://www.ibm.com/watson/developercloud/doc/index.html [3] https://azure.microsoft.com/en-us/services/cognitive-services/directory/lang/ https://azure.microsoft.com/en-us/services/cognitive-service... [4] https://azure.microsoft.com/en-us/services/cognitive-services/ https://azure.microsoft.com/en-us/services/cognitive-service...
    - igravious 9y ago
      Awesome. Any others that you know of?
- timr 9y ago
  "I was hoping for skynet. This is just a killer robot. You can put a machine gun on a lawnmower and get this far." (I'm kidding, obviously. This is not new tech, but having a cloud API for it does make life easier for a lot of tasks.)
  - killjoywashere 9y ago
    I suspect the folks tasked with building skynet would actually love the funding it requires to get a machine gun and a lawnmower in the same fiscal year.
  - Xeoncross 9y ago
    Not quite, I mean it's literally just three lines of code to get entities: > doc = nlp(u'Apple is looking at buying U.K. startup for $1 billion') > for ent in doc.ents: > print ent.text https://spacy.io/usage/spacy-101 https://spacy.io/usage/spacy-101 They even have an awesome visualizer you can run locally: https://demos.explosion.ai/displacy/ https://demos.explosion.ai/displacy/
    - timr 9y ago
      Yes, lots of open-source tools exist, like NLTK. Now, put it in production at non-trivial scale.

31 points by jeffbarr 9y ago

I was hoping that AWS Comprehend would have actual text comprehension. This seems to be just entity extraction and noun phrases. Basic Python Spacy.io or NLTK will get you this far. That said, I'm still glad they built it. Every time a new company offers a library or API like this I picture a new startup having an extra tool to enter a market.

igravious 9y ago
True true. That said, this takes a lot of the coding out of your hands. Presumably Microsoft and Google and IBM have similar cloud offerings. It's really hard to evaluate the quality of these offerings. I'd love if a website existed that had a cross platform analysis of these natural language processing site – basically compare and contrast these cloud offerings against rolling your own. It would have to use a few data-sets to get an all round picture.
- dragonwriter 9y ago
  > Presumably Microsoft and Google and IBM have similar cloud offerings. Google certainly does: https://cloud.google.com/natural-language/ https://cloud.google.com/natural-language/
  - igravious 9y ago
    Thank you. Any others that you know of?
- hallman76 9y ago
  > Presumably Microsoft and Google and IBM have similar cloud offerings. Yes, IBM has a similar product [1] along with a suite of APIs all under the "cognitive" umbrella[2]. Same [3] for Microsoft [4]. [1] https://www.ibm.com/watson/services/natural-language-classifier/ https://www.ibm.com/watson/services/natural-language-classif... [2] https://www.ibm.com/watson/developercloud/doc/index.html https://www.ibm.com/watson/developercloud/doc/index.html [3] https://azure.microsoft.com/en-us/services/cognitive-services/directory/lang/ https://azure.microsoft.com/en-us/services/cognitive-service... [4] https://azure.microsoft.com/en-us/services/cognitive-services/ https://azure.microsoft.com/en-us/services/cognitive-service...
  - igravious 9y ago
    Awesome. Any others that you know of?
timr 9y ago
"I was hoping for skynet. This is just a killer robot. You can put a machine gun on a lawnmower and get this far." (I'm kidding, obviously. This is not new tech, but having a cloud API for it does make life easier for a lot of tasks.)
- killjoywashere 9y ago
  I suspect the folks tasked with building skynet would actually love the funding it requires to get a machine gun and a lawnmower in the same fiscal year.
- Xeoncross 9y ago
  Not quite, I mean it's literally just three lines of code to get entities: > doc = nlp(u'Apple is looking at buying U.K. startup for $1 billion') > for ent in doc.ents: > print ent.text https://spacy.io/usage/spacy-101 https://spacy.io/usage/spacy-101 They even have an awesome visualizer you can run locally: https://demos.explosion.ai/displacy/ https://demos.explosion.ai/displacy/
  - timr 9y ago
    Yes, lots of open-source tools exist, like NLTK. Now, put it in production at non-trivial scale.