Search algorithm

Our proprietary search algorithm is built on top of the leading search platform, Elasticsearch. We score each document based on various factors. Our users have tools to adjust ranking manually as well.

Ranking

For example, following factors affect document ranking:

  • Self-learning algorithm: Giving higher score to documents with better click-through rates
  • Publishing time: Ranking newer news items higher in search results
  • Hidden keywords: Possibility to define artifical keywords that match to a document
  • Phrase match: Giving higher score to exact phrase matches. Returning weaker results with lower priority

Features

  • Typo-tolerance: Handling misspelled keywords with fuzzy matching. E.g. “contat information” returns results with the word “contact” properly spelled
  • Bigram matching and tokenization: For example, the URL /contact-information/ matches keywords “contact-information”, “contact information” and “contactinformation”
  • Stemming and plurals: When a language-specific stemming is enabled, the keyword “phones” matches exactly to “phone”
  • Synonyms: Possibility to define document-specific or global synonyms. E.g. “coat” matching the word “jacket”