We designed and developed a fiction book search engine that can:

• Search across over 1800 hand curated fiction novels gathered from goodreads.com

• Smartly identify author, title, content in the query

• Allow misspelling of words, phrases but still able to find the desired result

• Identify random garbage query such as “sdfnadf winsfd?d” so that the system does not return nonsense to the user

• Handle general query so that it can not only find book but also “recommend” book.

Secrets behind our search engine:

• Misspelling Correction

  • Our vocabulary in corpus is all stemmed, so is query → deal with words morphology
  • Phonetic similarity searching, e.g. “Harrrry Pttter” → “Harry Potter”
  • Jaro-Winkler and Levenshtein Edit Distance, e.g. “nehgbour” → “neighbour”

• Retrieval: Static Ranking + Dynamic query based ranking

  • Weighted BM25 with static fiction review rating score to prioritize popular novels.

• Corner Case: General query and Random query

  • Leverage the book reviews to handle general query. eg. "highly recommended fiction"
  • Identify the random query by checking if Jaro-Winkler score is less than 0.8 to any of the words in the corpus.