NLP Hacks for Writers

While putting together NLP in Action we found some handy NLP tools to make writing a little easier, more fun. In addition to employing a document summarizer based on Dirichlet Allocation to help us compose summary and introductory sentences, grammar and spell-checkers, code linters, etc, there are easy-to-use online implementations of sophisticated NLP tools that all writers can use. Here are 4 that we went back to time and again.

Thomas Mikolov’s Word2Vec

• Word2Vec Javascript implementation for 10k common words
• 7 Word2Vec models including GoogleNewsNegative300
• Gensim’s demo of the GoogleNews Word2Vec model

I loved using Word2Vec to find synonyms for technical and business terms, rather than using a thesaurus. It was great to see what machines think of words rather than going to dictionary.com where you only see what human linguists think. The mix of synonyms was different from what you’d get in a conventional thesaurus. And there were a lot more modern technical terms than what you’d find in WordNet or similar human-curated word similarity models. Plus we could do math on words to compose “synonyms” from combinations of other words, something no conventional thesaurus (or even Princeton’s WordNet can do. Check out our book, just released today for more on the magic of word2vec.

Princeton’s WordNet

WordNet is great for validating more conventional English words to make sure their formal definition similarity matches their co-use statistics in Google News articles.

Google Trends

And Google Trends was great for calibrating my impression of global trends for various NLP approaches, like the resurgence of deep learning and AI perhaps being driven by data-based statistical approaches becoming more viable at the expense of grammar-based approaches.

Google’s Ngram Viewer

In addition to being a massive source of 5-grams for unsupervised training of word embeddings and multi-word expressions or named-entity recognition, the N-gram viewer is also great for just figuring out the most common way to say something. I used it to learn about the phrase “beneath the words” (analogous to “between the lines”) and discovered that it’s actually a pretty popular new way to talk about the subtlties of word meaning.

NLP Tools for Coders

These tools are great for authors that also know a little bit of python.