Tokenization and normalization for high recall in all languages in Elasticsearch | Hacker News Reader