Natural language understanding (almost) from scratch

Adrian Colyer, The Morning Paper, Jul 05, 2016
Commentary by Stephen Downes

This article gives you a sense of where one branch of artificial intelligence is working today. Researchers are testing their neural network algorithms against four standard tasks in natural language analysis:

  • Part-of-speech tagging (POS), labels each word with a tag that indicates its syntactic role in a sentence
  • Chunking labels phrases or segments within a sentence with tags that indicate their syntactic role
  • Named-entity Recognition (NER) labels recognised entities within the sentence. For example, as a person, location, date, time, company
  • Semantic-role labeling (SRL) “gives a semantic role to a syntactic constituent of a sentence.” More.

The challenge here is o succeed at these tasks "without needing task-specific representations or engineering." It is often the case that you can tweak the result with a hint here or there. Anyhow, this paper describes attempts to design neural networks to attempt these tasks, and how we can score the results to see how well the network is doing.

Views: 0 today, 253 total (since January 1, 2017).[Direct Link]
Creative Commons License. gRSShopper

Copyright 2015 Stephen Downes ~ Contact: stephen@downes.ca
This page generated by gRSShopper.
Last Updated: Oct 23, 2017 12:00 a.m.