Ask HN: What cool NLP project are you working on?

28 points by tangled_zans ↗ HN
Hey guys, I'm quite keen to get more into NLP, so would anyone want to share what they've been hacking to give me some inspiration?

12 comments

[ 4.8 ms ] story [ 45.4 ms ] thread
A natural language home automation interface to control all the electronics in my home.
Ooo, nice one :D What does it control so far?
A web app that helps English language learners with reading and pronunciation:

https://lexical.io

It tries to find relevant multilingual information from Wikipedia and Wiktionary when you click on words in a submitted text.

Eventually I want to expand it to other target languages beyond English.

I worked on a project (customer recommendation stuff, not that interesting in and of itself) that used Stanford's CoreNLP.

Due to the project not being written in Java, we obviously needed a way to communicate with CoreNLP without the cost of calling java for every job.

The solution in the end was to fork an existing (Java) project that wraps CoreNLP and exposes it over HTTP.

The fork adds:

- support for JSON (the original was XML only)

- packages it for Debian

- defaults to "today" for relative dates

- adds a utility class 'RegexNERValidator' to allow testing/quoting of mapping files for CoreNLP's RegexNER, to allow checking a file can be used by RegexNER before the main CoreNLP process is restarted.

The result is at https://github.com/Koalephant/StanfordCoreNLPHTTPServer

Please note: I do not usually work in Java, so I'm well aware there are likely better ways to achieve some/many/all things this project does. If you feel inclined to improve it, send a PR (preferably with some indication of why its an improvement, if its not an obvious bug/feature improvement).

Wow, thanks for sharing! I did look at CoreNLP before and the fact that it's in Java has put me off, so this will be useful :)
You're welcome, I'd love to hear what you end up building.
At Vesper.ai we worked on an application that read email and recommended which messages users should followup on.
Bias detector in python for NYT articles using Vader sentiment analysis and the textacy library. Rates articles positive, negative, and neutral sentiments and how intense each one is. Although the majority of the heavy lifting is covered in the library, it's open source on github and reading the code (the actual main operations are less than 500 lines) is teaching me some cool NLP techniques, especially when it comes to rule-based analysis.