Skip to main content

Challenges in Machine Generation of Analytic Products from Multi-Source Data Proceedings of a Workshop (2017) / Chapter Skim
Currently Skimming:

4 Session 3: Machine Learning from Natural Languages
Pages 14-19

The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.


From page 14...
... . Commonly deployed machine learning methods include sequence labeling, neural nets, and distant learning, while features could be linguistic or based on similarity, popularity, gazeteers, ontologies, or verb triggers.
Read the entire page →
From page 15...
... Cross-lingual sentiment models are used to leverage parallel and comparable corpora in low resource languages -- multiple transfer model scenarios are created, target languages are engaged, and transfer models are created. These models, then, can be used to create sentiment analysis for languages with no labeled data.
Read the entire page →
From page 16...
... and the powerful hardware that are available. Radev remarked that many of the most fundamental natural language processing tasks are classification problems -- for example, optical character recognition, spelling correction, part-of-speech tagging, word disambiguation, parsing, named entity segmentation and classification, sentiment analysis, and machine translation.
Read the entire page →
From page 17...
... Radev noted that manually built ontologies of world knowledge are gradually fading, and the future is in automatically learned resources. Tom Dietterich, Oregon State University, commented that automatically structuring the architecture is important since deep networks require writing a program and that it is important to identify useful models and reusable program components for this new style of differentiable programming.
Read the entire page →
From page 18...
... Across industry, these machine learning methods may be used for call center analytics, advertisement placement, customer service analytics, and problematic dialogue identification. Stent added that there is a large gap between what researchers do and what industry does in the context of building conversational agents, which will hopefully begin to be addressed by the technology capabilities discussed in this workshop.
Read the entire page →
From page 19...
... Further innovation could come from providing the community with shared portals for evaluation and crowdsourcing of data. Stent hopes that work on world knowledge induction and situational awareness will also be sponsored in the United States, as such work is already under way abroad.
Read the entire page →

Key Terms

  • machine learning
  • language processing
  • neural networks
  • deep learning
  • learning methods

  • sentiment analysis
  • low resource
  • resource languages
  • world knowledge
  • machine translation

  • neural network
  • recurrent neural
  • learning techniques
  • processing community
  • fake news

  • percent accuracy
  • supervised learning
  • reinforcement learning
  • distant learning
  • streaming data

  • radev explained
  • conversational agents
  • methods include
  • dragomir radev
  • mckeown noted

  • stent hopes
  • radev described
  • advanced analytics
  • speech recognition
  • mckeown explained


    • This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
    More information on Chapter Skim is available.