Full-Text Indexing Workbench

http://www.simple-talk.com/sql/learn-sql-server/full-text-indexing-workbench/

/*

After you set up a full-text index on a table in a SQL Server 2005 or SQL Server 2008 database, you can perform a full-text search on the indexed columns in the table. To perform a full-text search, you can use the CONTAINS predicate or the FREETEXT predicate in your query's WHERE clause.

 

This workbench provides you with examples of how to use these predicates to perform a full-text search. (Note that SQL Server also supports two full-text functions, CONTAINSTABLE and FREETEXT table, but this workbench focuses only on the predicates.)

 

When you include the CONTAINS or FREETEXT predicate in your WHERE clause, the query engine searches the columns that are specified in the predicate arguments. These columns must be included in the full-text index that is defined on the specified table. The predicates also let you make use ofthe thesaurus that is available for any of the supported languages.

 

If you're new to full-text indexes and searches, you should first review the Simple-Talk article "Understanding Full-Text Indexing in SQL Server," published December 29, 2008. The article describes how full-text indexes are implemented in SQL Server 2005 and 2008, and provides examples of how to create those indexes.

*/

 

/*

To run the examples in this workbench, you should first set up the necessary environment to test the full-text queries. The following T-SQL statements create the StormyWeather table, populate the table, create the ftcStormyWeather full-text catalog, and then create a full text index on the table. The index is added to the ftcStormyWeather catalog.

*/

 

-- Create the StormyWeather table.

USE AdventureWorks2008 --replace with correct DB name

GO

IF OBJECT_ID (N'StormyWeather', N'U') IS NOT NULL

DROP TABLE StormyWeather

GO

CREATE TABLE StormyWeather (

  StormID INT NOT NULL IDENTITY,

  StormHead NVARCHAR(50) NOT NULL,

  StormBody NVARCHAR(MAX) NOT NULL,

  CONSTRAINT [PK_StormyWeather_StormID] PRIMARY KEY CLUSTERED (StormID ASC)

)

GO

-- Populate the StormyWeather table with data

-- that supports various full-text query types.

INSERT INTO StormyWeather (StormHead, StormBody)

VALUES('Stormy Weather Delays Travel',

  'The stormy weather made travel by motor vehicle difficult.')

INSERT INTO StormyWeather (StormHead, StormBody)

VALUES('Stormier Weather on Monday',

  'The stormier weather on Monday made vehicle travel difficult.')

INSERT INTO StormyWeather (StormHead, StormBody)

VALUES('Stormiest Weather in December',

  'December can be the stormiest month, making automobile travel difficult.')

INSERT INTO StormyWeather (StormHead, StormBody)

VALUES('Storm Grows Strong',

  'The storm is growing strong.')

INSERT INTO StormyWeather (StormHead, StormBody)

VALUES('Storms Crossing the Pacific',

  'The storms are lining up across the Pacific Ocean.')

INSERT INTO StormyWeather (StormHead, StormBody)

VALUES('Storm''s Wind Delays Travel',

  'The storm''s wind made car travel difficult on Tuesday.')

INSERT INTO StormyWeather (StormHead, StormBody)

VALUES('Storms'' Flooding Delays Travel',

  'The storms'' flooding made auto travel difficult throughout December.')

INSERT INTO StormyWeather (StormHead, StormBody)

VALUES('Children Run from Room',

  'The children often storm out of the room when upset.')

INSERT INTO StormyWeather (StormHead, StormBody)

VALUES('Boy Runs from Room',

  'The boy storms out of the room when his sister changes the channel.')

INSERT INTO StormyWeather (StormHead, StormBody)

VALUES('Girl Ran from Room',

  'The girl stormed out of the room when her brother ate the cookie.')

INSERT INTO StormyWeather (StormHead, StormBody)

VALUES('Children Running from Room',

  'The children were storming out of the room when the lights went out.')

GO

-- Create a full-text catalog and set it as the default.

CREATE FULLTEXT CATALOG ftcStormyWeather

AS DEFAULT

GO

-- Create a full-text index on the StormyWeather table.

-- Add the index to the ftcStormyWeather catalog.

CREATE FULLTEXT INDEX ON StormyWeather(StormHead, StormBody)

KEY INDEX PK_StormyWeather_StormID

ON ftcStormyWeather

GO

 

/*

Use the CONTAINS predicate to search the columns included in the full-text index. The CONTAINS arguments must be enclosed in parentheses. Multiple columns must be separated by a comma and enclosed in parentheses. The search condition (the term or terms) must be enclosed in single quotes.

*/

 

-- Search a single column for a single term.

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE CONTAINS(StormHead, 'storm')

 

/*

The statement returns the following results:

4  Storm Grows Strong            The storm is growing strong.

6  Storm's Wind Delays Travel    The storm's wind made car travel difficult on Tuesday.

*/

 

-- Search multiple columns for a single term.

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE CONTAINS((StormHead, StormBody), 'storm')

 

/*

The statement returns the following results:

4  Storm Grows Strong            The storm is growing strong.

6  Storm's Wind Delays Travel    The storm's wind made car travel difficult on Tuesday.

8  Children Run from Room        The children often storm out of the room when upset.

 

The next statement returns the same results.

*/

 

-- Use an asterisk (*) wildcard instead of column names

-- to search all full-text columns.

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE CONTAINS(*, 'storm')

 

-- When searching multiple terms, use a comparative

-- operator, such as OR or AND, to separate the terms.

-- Individual terms should be enclosed in double quotes.

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE CONTAINS(StormHead, '"storm" OR "storms" OR "stormy" OR "stormier" OR "stormiest"')

 

/*

The statement returns the following results:

1  Stormy Weather Delays Travel   The stormy weather made travel by motor vehicle difficult.

2  Stormier Weather on Monday     The stormier weather on Monday made vehicle travel difficult.

3  Stormiest Weather in December  December can be the stormiest month, making automobile travel difficult.

4  Storm Grows Strong             The storm is growing strong.

5  Storms Crossing the Pacific    The storms are lining up across the Pacific Ocean.

6  Storm's Wind Delays Travel     The storm's wind made car travel difficult on Tuesday.

7  Storms' Flooding Delays Travel The storms' flooding made auto travel difficult throughout December.

*/

 

/*

Rather than specify multiple terms, you can use a 'prefix term' if the terms begin with the same characters.

To use a prefix term, specify the beginning characters, then add an asterisk (*) wildcard to the end of the term. Enclose the prefix term in double quotes.

The following statement returns the same results as the previous one.

*/

 

-- Search for all terms that begin with 'storm'

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE CONTAINS(StormHead, '"storm*"')

 

/*

Not all related terms can be effectively consolidated into a prefix term. For example, 'run' and 'ran' would require "r*", which would match all words beginning with 'r'. In these cases, you can specify each inflection of the word, such as '"run" OR "ran" OR "runs"', or you can use FORMSOF in your CONTAINS predicate.

*/

 

-- Specify each inflection.

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE CONTAINS(StormHead, '"run" OR "runs" OR "ran" OR "running"')

 

/*

The statement returns the following results:

8  Children Run from Room     The children often storm out of the room when upset.

9  Boy Runs from Room         The boy storms out of the room when his sister changes the channel.

10 Girl Ran from Room         The girl stormed out of the room when her brother ate the cookie.

11 Children Running from Room The children were storming out of the room when the lights went out.

 

The following statement returns the same results as the previous one.

*/

 

-- Use the FORMSOF and INFLECTIONAL keywords, along with the root word.

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE CONTAINS(StormHead, 'FORMSOF(INFLECTIONAL, run)')

 

/*

As you can see in the previous examples, the CONTAINS predicate returns an exact match, unless you specify a prefix term or FORMSOF. However, these methods will not work for different words with similar meanings, such as 'car' and 'automobile'. In these cases, you can use a thesaurus to match these types of terms.

 

For example, you can update the tsenu.xml thesaurus file (for LCID 1033) by adding the following elements:

 

    <thesaurus xmlns="x-schema:tsSchema.xml">

      <diacritics_sensitive>0</diacritics_sensitive>

        <expansion>

            <sub>car</sub>

            <sub>auto</sub>

            <sub>automobile</sub>

            <sub>vehicle</sub>

            <sub>motor vehicle</sub>

        </expansion>

    </thesaurus>

   

When you add these elements to your thesaurus file, the full-text search engine can then treat these terms the same. For example, if you search on 'car', 'auto' and 'automobile' will also be included in your search.

 

After you update a thesaurus file, you might need to reload it. In SQL Server 2008, you can use the sp_fulltext_load_thesaurus_file system stored procedure to reload the thesaurus file after you've updated it. In SQL Server 2005, you must restart the full-text search service.

 

Now when you search on car and specify the FORMSOF and THESAURUS keywords, your search will return all rows that include any of the terms specified in the thesaurus file.

*/

 

-- Use the term as is to return an exact match.

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE CONTAINS(StormBody, 'car')

 

/*

The statement returns the following results:

6  Storm's Wind Delays Travel     The storm's wind made car travel difficult on Tuesday.

*/

 

-- Use the FORMSOF and THESAURUS keywords to use the thesaurus files.

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE CONTAINS(StormBody, 'FORMSOF(THESAURUS, car)')

 

/*

The statement returns the following results:

1  Stormy Weather Delays Travel   The stormy weather made travel by motor vehicle difficult.

2  Stormier Weather on Monday     The stormier weather on Monday made vehicle travel difficult.

3  Stormiest Weather in December  December can be the stormiest month, making automobile travel difficult.

6  Storm's Wind Delays Travel     The storm's wind made car travel difficult on Tuesday.

7  Storms' Flooding Delays Travel The storms' flooding made auto travel difficult throughout December.

*/

 

/*

You can also use the NEAR keyword between terms in your search condition to specify words or phrases that must be near to each other.

*/

 

-- Use near to return rows in which 'travel' is near forms of 'storm'.

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE CONTAINS(StormBody, '"storm*" NEAR travel')

 

/*

The statement returns the following results:

1  Stormy Weather Delays Travel   The stormy weather made travel by motor vehicle difficult.

2  Stormier Weather on Monday     The stormier weather on Monday made vehicle travel difficult.

3  Stormiest Weather in December  December can be the stormiest month, making automobile travel difficult.

6  Storm's Wind Delays Travel     The storm's wind made car travel difficult on Tuesday.

7  Storms' Flooding Delays Travel The storms' flooding made auto travel difficult throughout December.

*/

 

-- You can use NEAR to chain together multiple terms.

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE CONTAINS(StormBody, '"storm*" NEAR travel NEAR Tuesday')

 

/*

The statement returns the following results:

6  Storm's Wind Delays Travel    The storm's wind made car travel difficult on Tuesday.

*/

 

/*

The full-text queries in the examples here are relatively straightforward. However, these queries can get quite complicated. Fortunately, the CONTAINS  predicate supports a number of methods that let you simplify your queries (such as the prefix term). In addition, you can use the FREETEXT predicate to simplify your queries even more. FREETEXT treats each word in a phrase as a separate term and automatically finds different inflections for that term and applies the appropriate thesaurus files.

*/

 

-- Define each form of a term in your search condition.

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE CONTAINS(StormBody, '"weather" OR "flood" 

  OR "flooding" OR "car" OR "auto" OR "automobile" OR "vehicle" OR "motor vehicle"')

 

/*

The statement returns the following results:

1  Stormy Weather Delays Travel   The stormy weather made travel by motor vehicle difficult.

2  Stormier Weather on Monday     The stormier weather on Monday made vehicle travel difficult.

3  Stormiest Weather in December  December can be the stormiest month, making automobile travel difficult.

6  Storm's Wind Delays Travel     The storm's wind made car travel difficult on Tuesday.

7  Storms' Flooding Delays Travel The storms' flooding made auto travel difficult throughout December.

 

The following two statements return the same results as the previous one.

*/

 

-- When possible, use prefix terms, thesaurus files, or other devices

-- to simplify your full-text queries.

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE CONTAINS(StormBody, '"weather" OR "flood*" OR FORMSOF(THESAURUS, car)')

 

-- You can also use FREETEXT when applicable to search for terms.

SELECT StormID, StormHead, StormBody FROM StormyWeather

WHERE FREETEXT(StormBody, 'weather flood car')

 

/*

That's all there is to using the CONTAINS and FREETEXT predicates. Keep in mind that CONTAINS is more precise than FREETEXT. And, of course, a full-text search can be much more complex than shown in the examples here. Be sure to check out SQL Server Books Online for more details about both of these predicates.

*/

  • 0
    点赞
  • 0
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
1 目标检测的定义 目标检测(Object Detection)的任务是找出图像中所有感兴趣的目标(物体),确定它们的类别和位置,是计算机视觉领域的核心问题之一。由于各类物体有不同的外观、形状和姿态,加上成像时光照、遮挡等因素的干扰,目标检测一直是计算机视觉领域最具有挑战性的问题。 目标检测任务可分为两个关键的子任务,目标定位和目标分类。首先检测图像中目标的位置(目标定位),然后给出每个目标的具体类别(目标分类)。输出结果是一个边界框(称为Bounding-box,一般形式为(x1,y1,x2,y2),表示框的左上角坐标和右下角坐标),一个置信度分数(Confidence Score),表示边界框中是否包含检测对象的概率和各个类别的概率(首先得到类别概率,经过Softmax可得到类别标签)。 1.1 Two stage方法 目前主流的基于深度学习的目标检测算法主要分为两类:Two stage和One stage。Two stage方法将目标检测过程分为两个阶段。第一个阶段是 Region Proposal 生成阶段,主要用于生成潜在的目标候选框(Bounding-box proposals)。这个阶段通常使用卷积神经网络(CNN)从输入图像中提取特征,然后通过一些技巧(如选择性搜索)来生成候选框。第二个阶段是分类和位置精修阶段,将第一个阶段生成的候选框输入到另一个 CNN 中进行分类,并根据分类结果对候选框的位置进行微调。Two stage 方法的优点是准确度较高,缺点是速度相对较慢。 常见Tow stage目标检测算法有:R-CNN系列、SPPNet等。 1.2 One stage方法 One stage方法直接利用模型提取特征值,并利用这些特征值进行目标的分类和定位,不需要生成Region Proposal。这种方法的优点是速度快,因为省略了Region Proposal生成的过程。One stage方法的缺点是准确度相对较低,因为它没有对潜在的目标进行预先筛选。 常见的One stage目标检测算法有:YOLO系列、SSD系列和RetinaNet等。 2 常见名词解释 2.1 NMS(Non-Maximum Suppression) 目标检测模型一般会给出目标的多个预测边界框,对成百上千的预测边界框都进行调整肯定是不可行的,需要对这些结果先进行一个大体的挑选。NMS称为非极大值抑制,作用是从众多预测边界框中挑选出最具代表性的结果,这样可以加快算法效率,其主要流程如下: 设定一个置信度分数阈值,将置信度分数小于阈值的直接过滤掉 将剩下框的置信度分数从大到小排序,选中值最大的框 遍历其余的框,如果和当前框的重叠面积(IOU)大于设定的阈值(一般为0.7),就将框删除(超过设定阈值,认为两个框的里面的物体属于同一个类别) 从未处理的框中继续选一个置信度分数最大的,重复上述过程,直至所有框处理完毕 2.2 IoU(Intersection over Union) 定义了两个边界框的重叠度,当预测边界框和真实边界框差异很小时,或重叠度很大时,表示模型产生的预测边界框很准确。边界框A、B的IOU计算公式为: 2.3 mAP(mean Average Precision) mAP即均值平均精度,是评估目标检测模型效果的最重要指标,这个值介于0到1之间,且越大越好。mAP是AP(Average Precision)的平均值,那么首先需要了解AP的概念。想要了解AP的概念,还要首先了解目标检测中Precision和Recall的概念。 首先我们设置置信度阈值(Confidence Threshold)和IoU阈值(一般设置为0.5,也会衡量0.75以及0.9的mAP值): 当一个预测边界框被认为是True Positive(TP)时,需要同时满足下面三个条件: Confidence Score > Confidence Threshold 预测类别匹配真实值(Ground truth)的类别 预测边界框的IoU大于设定的IoU阈值 不满足条件2或条件3,则认为是False Positive(FP)。当对应同一个真值有多个预测结果时,只有最高置信度分数的预测结果被认为是True Positive,其余被认为是False Positive。 Precision和Recall的概念如下图所示: Precision表示TP与预测边界框数量的比值 Recall表示TP与真实边界框数量的比值 改变不同的置信度阈值,可以获得多组Precision和Recall,Recall放X轴,Precision放Y轴,可以画出一个Precision-Recall曲线,简称P-R
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值