[PROPOSAL] Paper review: “A Fast Unified Model for Parsing and Sentence Understanding”

[m09]

• Title: A Fast Unified Model for Parsing and Sentence Understanding
• Author(s): Hugo and anyone interested
• Short description: This blog post will be about neural architectures that are especially well suited to handle UASTs. It will outline how hard it is for an architecture to actually be efficient to model tree structures and will show how this one provides a very interesting angle to model trees.
• Categories: Deep learning, machine learning, literature review, neural architectures, research
• Deadlines: No deadline.

Table of contents

1. Introduction to the problem of modelling trees in neural nets.
2. Prior art to the SPINN model (SPINN is the model of the paper).
3. Quick introduction to parsing natural language.
4. Transition into how this inspired SPINN.
5. Explanation of SPINN and details about its competitive points.
6. Ideas to extend it and to apply it to code.

Management

This section will be filled by @campoy.

• State: (proposed | writing | written | published)
• Scheduled:
• Link to post:

Social Media

• Wording for tweet:
• Hashtags:
• Subreddits:

[campoy]

This sounds super interesting (sorry for the delay approving)

[bzz]

🎉 @m09 please let me known if any help will be need at proof-reading the draft, etc.

Very excited for keep "paper review" post series going!

[vcoisne]

@m09 Did you get a chance to start a draft ?

[m09]

@vcoisne I did get a chance but nothing reviewable yet. I'll find time for it in the coming two weeks

[m09]

sadly reported to january since we were still swamped by format analyzer those last 2 weeks

[vcoisne]

@m09 Any updates on this draft ? Trying to plan our blog schedule for the upcoming weeks.

[vcoisne]

ping @m09

[m09]

I'll write a draft next OSD :)

[vcoisne]

@m09 Is that on your radar for next OSD on monday ?