[[relevance-conclusion]]
=== Relevance Tuning Is the Last 10%
In this chapter, we looked at a how Lucene generates scores based on TF/IDF.
Understanding the score-generation process((("relevance", "controlling", "tuning relevance"))) is critical so you can
tune, modulate, attenuate, and manipulate the score for your particular
business domain.
In practice, simple combinations of queries will get you good search results.
But to get _great_ search results, you'll often have to start tinkering with
the previously mentioned tuning methods.
Often, applying a boost on a strategic field or rearranging a query to
emphasize a particular clause will be sufficient to make your results great.
Sometimes you'll need more-invasive changes. This is usually the case if your
scoring requirements diverge heavily from Lucene's word-based TF/IDF model (for example, you
want to score based on time or distance).
With that said, relevancy tuning is a rabbit hole that you can easily fall into
and never emerge. The concept of _most relevant_ is a nebulous target to hit, and
different people often have different ideas about document ranking. It is easy
to get into a cycle of constant fiddling without any apparent progress.
We encourage you to avoid this (very tempting) behavior and instead properly
instrument your search results. Monitor how often your users click the top
result, the top 10, and the first page; how often they execute a secondary query
without selecting a result first; how often they click a result and immediately
go back to the search results, and so forth.
These are all indicators of how relevant your search results are to the user.
If your query is returning highly relevant results, users will select one of
the top-five results, find what they want, and leave. Irrelevant results cause
users to click around and try new search queries.
Once you have instrumentation in place, tuning your query is simple. Make a change,
monitor its effect on your users, and repeat as necessary. The tools outlined in this
chapter are just that: tools. You have to use them appropriately to propel
your search results into the _great_ category, and the only way to do that is with
strong measurement of user behavior.
- Introduction
- 入门
- 是什么
- 安装
- API
- 文档
- 索引
- 搜索
- 聚合
- 小结
- 分布式
- 结语
- 分布式集群
- 空集群
- 集群健康
- 添加索引
- 故障转移
- 横向扩展
- 更多扩展
- 应对故障
- 数据
- 文档
- 索引
- 获取
- 存在
- 更新
- 创建
- 删除
- 版本控制
- 局部更新
- Mget
- 批量
- 结语
- 分布式增删改查
- 路由
- 分片交互
- 新建、索引和删除
- 检索
- 局部更新
- 批量请求
- 批量格式
- 搜索
- 空搜索
- 多索引和多类型
- 分页
- 查询字符串
- 映射和分析
- 数据类型差异
- 确切值对决全文
- 倒排索引
- 分析
- 映射
- 复合类型
- 结构化查询
- 请求体查询
- 结构化查询
- 查询与过滤
- 重要的查询子句
- 过滤查询
- 验证查询
- 结语
- 排序
- 排序
- 字符串排序
- 相关性
- 字段数据
- 分布式搜索
- 查询阶段
- 取回阶段
- 搜索选项
- 扫描和滚屏
- 索引管理
- 创建删除
- 设置
- 配置分析器
- 自定义分析器
- 映射
- 根对象
- 元数据中的source字段
- 元数据中的all字段
- 元数据中的ID字段
- 动态映射
- 自定义动态映射
- 默认映射
- 重建索引
- 别名
- 深入分片
- 使文本可以被搜索
- 动态索引
- 近实时搜索
- 持久化变更
- 合并段
- 结构化搜索
- 查询准确值
- 组合过滤
- 查询多个准确值
- 包含,而不是相等
- 范围
- 处理 Null 值
- 缓存
- 过滤顺序
- 全文搜索
- 匹配查询
- 多词查询
- 组合查询
- 布尔匹配
- 增加子句
- 控制分析
- 关联失效
- 多字段搜索
- 多重查询字符串
- 单一查询字符串
- 最佳字段
- 最佳字段查询调优
- 多重匹配查询
- 最多字段查询
- 跨字段对象查询
- 以字段为中心查询
- 全字段查询
- 跨字段查询
- 精确查询
- 模糊匹配
- Phrase matching
- Slop
- Multi value fields
- Scoring
- Relevance
- Performance
- Shingles
- Partial_Matching
- Postcodes
- Prefix query
- Wildcard Regexp
- Match phrase prefix
- Index time
- Ngram intro
- Search as you type
- Compound words
- Relevance
- Scoring theory
- Practical scoring
- Query time boosting
- Query scoring
- Not quite not
- Ignoring TFIDF
- Function score query
- Popularity
- Boosting filtered subsets
- Random scoring
- Decay functions
- Pluggable similarities
- Conclusion
- Language intro
- Intro
- Using
- Configuring
- Language pitfalls
- One language per doc
- One language per field
- Mixed language fields
- Conclusion
- Identifying words
- Intro
- Standard analyzer
- Standard tokenizer
- ICU plugin
- ICU tokenizer
- Tidying text
- Token normalization
- Intro
- Lowercasing
- Removing diacritics
- Unicode world
- Case folding
- Character folding
- Sorting and collations
- Stemming
- Intro
- Algorithmic stemmers
- Dictionary stemmers
- Hunspell stemmer
- Choosing a stemmer
- Controlling stemming
- Stemming in situ
- Stopwords
- Intro
- Using stopwords
- Stopwords and performance
- Divide and conquer
- Phrase queries
- Common grams
- Relevance
- Synonyms
- Intro
- Using synonyms
- Synonym formats
- Expand contract
- Analysis chain
- Multi word synonyms
- Symbol synonyms
- Fuzzy matching
- Intro
- Fuzziness
- Fuzzy query
- Fuzzy match query
- Scoring fuzziness
- Phonetic matching
- Aggregations
- overview
- circuit breaker fd settings
- filtering
- facets
- docvalues
- eager
- breadth vs depth
- Conclusion
- concepts buckets
- basic example
- add metric
- nested bucket
- extra metrics
- bucket metric list
- histogram
- date histogram
- scope
- filtering
- sorting ordering
- approx intro
- cardinality
- percentiles
- sigterms intro
- sigterms
- fielddata
- analyzed vs not
- 地理坐标点
- 地理坐标点
- 通过地理坐标点过滤
- 地理坐标盒模型过滤器
- 地理距离过滤器
- 缓存地理位置过滤器
- 减少内存占用
- 按距离排序
- Geohashe
- Geohashe
- Geohashe映射
- Geohash单元过滤器
- 地理位置聚合
- 地理位置聚合
- 按距离聚合
- Geohash单元聚合器
- 范围(边界)聚合器
- 地理形状
- 地理形状
- 映射地理形状
- 索引地理形状
- 查询地理形状
- 在查询中使用已索引的形状
- 地理形状的过滤与缓存
- 关系
- 关系
- 应用级别的Join操作
- 扁平化你的数据
- Top hits
- Concurrency
- Concurrency solutions
- 嵌套
- 嵌套对象
- 嵌套映射
- 嵌套查询
- 嵌套排序
- 嵌套集合
- Parent Child
- Parent child
- Indexing parent child
- Has child
- Has parent
- Children agg
- Grandparents
- Practical considerations
- Scaling
- Shard
- Overallocation
- Kagillion shards
- Capacity planning
- Replica shards
- Multiple indices
- Index per timeframe
- Index templates
- Retiring data
- Index per user
- Shared index
- Faking it
- One big user
- Scale is not infinite
- Cluster Admin
- Marvel
- Health
- Node stats
- Other stats
- Deployment
- hardware
- other
- config
- dont touch
- heap
- file descriptors
- conclusion
- cluster settings
- Post Deployment
- dynamic settings
- logging
- indexing perf
- rolling restart
- backup
- restore
- conclusion