Ask HN: Books about full text search?

232 points by sopromo 3 years ago | 31 comments

I would love to learn more about FTS at a very low level and I'm looking for books to read more on that topic. Any good suggestions ?

binarymax 3 years ago |

“Relevant search” by Doug Turnbull and John Berryman, published by Manning, is THE best book to get started with tuning search engines.

I’be been a search engineer for >10 years and this is always the first book I recommend.

https://www.manning.com/books/relevant-search

softwaredoug 3 years ago | |

Awe thanks Max <3

deanebarker 3 years ago | | |

Before I read your book, I thought, "I know all about search!"

After I read it...

"...I knew NOTHING about search."

No book has ever knocked me off my pedestal so brutally and so thoroughly.

aliswe 3 years ago | | |

For a moment I thought you were Doug Cutting

ssn 3 years ago |

Three reference textbooks are available openly:

* Introduction to Information Retrieval, http://informationretrieval.org/

* Information Retrieval in Practice, http://www.search-engines-book.com/

* Entity-Oriented Search, https://eos-book.org/

Modern Information Retrieval is also a classic reference. Not openly available but some contents are (were?) available online. Their site seems to be down but the Internet Archive has a copy.

Additional resources here:

* https://nlp.stanford.edu/IR-book/information-retrieval.html http://web.archive.org/web/20220708135205/http://grupoweb.up...

firebones 3 years ago | |

I am biased, but building the Intro to Information Retrieval chapters in your favorite language, bit by bit, is really good to get the feel of the tradeoffs for index capabilities.

100k 3 years ago |

At a general audience level, "Index" is on my list to read. It covers the invention of the index up to digital search engines. https://www.nytimes.com/2022/02/09/books/review-index-histor...

"Introduction to Information Retrieval" is a textbook which is available online https://nlp.stanford.edu/IR-book/ Here's a review: http://glinden.blogspot.com/2009/02/book-review-introduction...

Another textbook which IMHO is a bit lower level is "Information Retrieval: Implementing and Evaluating Search Engines". The book website is down for me right now, but you can find it on Amazon here: https://www.amazon.com/Information-Retrieval-Implementing-Ev...

Another commenter linked to "Relevant Search", which is great if you want to learn how to effectively use a search engine to improve relevance (as opposed to how to implement a search engine). It's old, but another book in that vein that was really helpful for me earlier in my career is Lucene in Action: https://www.amazon.com/Lucene-Action-Second-Covers-Apache/dp...

driscoll42 3 years ago | |

Going to second the rec on "Index", it's a very understandable, well researched book that the general audience or even a skilled practitioner would enjoy.

DamonHD 3 years ago |

Managing Gigabytes

https://books.google.co.uk/books/about/Managing_Gigabytes.ht...

Old but good!

CoolestBeans 3 years ago | |

Came here to recommend Managing Gigabytes as well. People these days are managing far more than gigabytes but the fundamental ideas remain useful.

dekhn 3 years ago | |

Check out the first review on the Amazon page. Norvig read it around the time he started at Google.

francoisprunier 3 years ago |

Not a book, but this paper from 2019 covers a lot of ground and reviews the different topics extensively: https://tonellotto.github.io/publication/fntir/fntir_main.pd...

pixelmonkey 3 years ago |

Take a look at my post “Lucene: The Good Parts”—

https://blog.parse.ly/lucene/

The book mentioned there is Lucene in Action.

And then this YouTube presentation by a Lucene/Elasticsearch committer will give you a nice overview of some related algorithms—

https://youtu.be/eQ-rXP-D80U

brudgers 3 years ago |

Not a book but Hellerstein’s CS186 from 2015 starting with Lecture 17 gave me a basic understanding (I think).

Playlist https://youtube.com/playlist?list=PLhMnuBfGeCDPtyC9kUf_hG_Qw...

Also from that lecture series, the low level is always IO. One disk read tends to dwarf n^2 in-memory algorithms.

And IO is all about tuning caches and hardware for the specific structural relationships in the data, the way in which it is accessed, and the hardware everything runs on.

Good luck.

MonkoftheFunk 3 years ago |

Hotz... Is that you... Trying to learn to improve Twitter search? ;)

fiedzia 3 years ago |

https://www.manning.com/books/relevant-search

Also "taming text"

arooaroo 3 years ago | |

Manning also have a book on Lucene, the library that powers Solr and ElasticSearch. IIRC the book covered how Lucene actually works under-the-good and would therefore act as a good reference on the subject in general.

gardenfelder 3 years ago | |

Taming Text is about building a question-answering system; it came out about the time Watson came online; it's not a plan, rather a cookbook of experiments using Apache products like Solr and OpenNLP, but is a great tutorial on how question answering works.

vdfs 3 years ago |

Lucene in Action, good introduction to Lucene, which can be helpful to learn ElasticSearch (most used FTS these days)

_tom_ 3 years ago | |

Lucene in Action covers Lucene 3.0, and is from 2010. Current version is 9.4.2. So much has changed.

tgv 3 years ago |

Check the literature of open courses on Text Retrieval. E.g. https://stanford.edu/class/cs276/

Beefin 3 years ago |

series of tutorials and comparisons that aim to teach a foundations about vector search:

https://vectorsearch.dev/

cb321 3 years ago |

It's all in the Nim programming language, but if you prefer reading code or running diffs then you might get a vague sense of (some) low level nuts & bolts from: https://github.com/c-blake/nimsearch

User23 3 years ago |

Is there some better alternative to Knuth-Morris-Pratt or Boyer-Moore? Both can easily be adapted to regular expression matching and as far as I know there’s no faster algorithm that doesn’t do preprocessing.

Beefin 3 years ago |

Stanford's NLP course:

https://www.youtube.com/playlist?list=PLoROMvodv4rOSH4v6133s...

leeseonwook 3 years ago |

123

leeseonwook 3 years ago | |

456

unixhero 3 years ago |

Just use Postgres fulltext Search, its good enough http://rachbelaid.com/postgres-full-text-search-is-good-enou...

johnthescott 3 years ago | |

for postgres, i highly recommend the rum index over the core fts. rum is written by postgrespro, who also wrote core fts and json indexing in pg.

    https://github.com/postgrespro/rum

rum handles +20mil pdf pages, interactively.

SPBS 3 years ago | | |

Pleasantly surprised that RUM is just a drop-in replacement for the built-in GIN index, you can still use Postgres' native FTS operations with it.

unixhero 3 years ago | | |

Sounds very interesting. Never heard of rum, thank you for suggesting it.