1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
|
---
categories: ["book review"]
title: "[Book review] Relevant Search"
date: 2021-05-06T16:35:08+07:00
tags: [book, review, search, programming, algorithm]
book: "Relevant Search: With applications for Solr and Elasticsearch"
authors:
- Doug Turnbull
- John Berryman
isbn: 9781617292774
genres: [programming]
lang: en
layout: books
---
# Summary
## The search relevance problem
Given an increasingly large amount of information, it is infeasible for users
to retrieve what they needed. Relevance scoring is therefore essential for
search engines.
In general, the relevance engineers have to identify the most important
features describing the content, the user, or the search query, transfer those
features to the search engine, then measure what's relevant to the search by
crafting signals and finally balance the weights of the signals to rank the
results.
Unfortunately, it is a challenging problem. Each search application
serves a different type of content and thus has different expectation for
relevance. Consequently, there is no silver bullet to solve this problem.
Even the academic field that thoroughly study this problem, information
retrieval is not a one-size-fit all solution. Relevance is strongly tied with
the field and the application purpose.
## Tackling the problem
The book approaches the problem first by a top-down analysis of how a typical
search engine works. It then shows how a search query is processed by the
search engine. After providing basic knowledge of how search work, the authors
give some examples of relevance score tuning and show how it helps improving
the relevance of the search results. Not stopping at the technical view, the
authors also approach the problem from business view: they note that
interdiscipline collaboration is important in order to define and increase
relevance.
# Comments
## What I like
The book approaches the problem from various views: business view, algorithmic
view, and practical view (giving examples). The book accentuates the diversity
of problems and thereby encouraging readers to critically think of their own
problems. While it suggests that search results should be influenced by
sponsors, it also notes that without balance that will as well lead to failure.
## What I don't like
Its structure is somewhat unclear and flow to me. I think some chapters can be
re-ordered so it's more logical. Also, I find weighing sponsors' priorities
over customers' unethical, but that is probably just a harsh truth in this
society rather than the authors' view.
|
