青空例文

#

Targeted Example Sentences from the Aozora Bunko Corpus

#

Purpose

#

This is a pretty barebones language practice service that:

• Reads in a provided vocabulary list
• Samples sentences from the Aozora Bunko Corpus containing the targeted vocabulary words
• Looks up words and kanji in the sentences for readings, definitions, etc.
• Formats the sentences and lookup information into a simple HTML email
• Sends the email to your inbox with specified frequency

Dataset

#

The dataset used for the current project was pulled from the following:

• Aozora Bunko Corpus
• Jotoba API

Usage

#

Setup is handled via a few key files:

• vocab.txt, for the vocab you want to study/get sentences for. Empty file is included in the repo as .vocab.txt, so change the name to vocab.txt on your first run.

昨日
すき焼き
を
食べました
...

• config.json, for setting email frequency and content. Empty file is included in the repo as .config.json, so change the name to config.json on your first run.

{
  "crontab": "* * * * *", // Cron expression for email frequency
  "n_vocab": 2, // number of vocab words per email
  "n_examples": 2, // number of example sentences per vocab word per email
  "min_len": 0, // shortest allowed sentence
  "max_len": 100, // longest allowed sentence
  "recipients": ["<recipient email 1>", "<recipient email 2>", ...] //
}

• .env, for your super secret email and incredibly secure password

FROM=<email login>
PASSWORD=<email password>

Notes:

• I recommend running the go binary (./aozora_reibun) as a systemd service using the provided example service file.
• It will re-query the database on each restart of the service.
• You can update the vocab and config at any time but you’ll need to restart the service to have the changes reflected.
• Email limits and security concerns will come down to and be handled by your email provider/service.