问题描述
我有一个函数,它接受两个数组包含两个文本的标记/单词,并给出余弦相似性值,显示两个文本之间的关系。
函数接受数组$ tokensA(0 => house,1 => bike,2 => man)和数组$ tokensB(0 => bike,1 => house,2 => car)作为浮点值。
通常可以索引具有任意维数的数据, MySQL 空间能力仅限于几何$ c $如果您的向量 2 $> c $ c> -dimensional 和,您可以对其进行规范化,然后执行以下操作:
- 符合你的差异的角度数量的两倍
- 找到与每个扇区中心具有给定余弦差的向量的 MBR li>
- 查找 MBR 中的所有向量
- 对精确的差异进行精细过滤。
然而,在这种情况下,最好只是预先计算值的角度,并用平滑的 B-Tree 索引。
I have a function which takes two arrays containing the tokens/words of two texts and gives out the cosine similarity value which shows the relationship between both texts.
The function takes an array $tokensA (0=>house, 1=>bike, 2=>man) and an array $tokensB (0=>bike, 1=>house, 2=>car) and calculates the similarity which is given back as a floating point value.
function cosineSimilarity($tokensA, $tokensB) { $a = $b = $c = 0; $uniqueTokensA = $uniqueTokensB = array(); $uniqueMergedTokens = array_unique(array_merge($tokensA, $tokensB)); foreach ($tokensA as $token) $uniqueTokensA[$token] = 0; foreach ($tokensB as $token) $uniqueTokensB[$token] = 0; foreach ($uniqueMergedTokens as $token) { $x = isset($uniqueTokensA[$token]) ? 1 : 0; $y = isset($uniqueTokensB[$token]) ? 1 : 0; $a += $x * $y; $b += $x; $c += $y; } return $b * $c != 0 ? $a / sqrt($b * $c) : 0; }
If I want to compare 75 texts with each other, I need to make 5,625 single comparisons to have all texts compared with each other.
Is it possible to use MySQL's spatial columns to reduce the number of comparisons?
I don't want to talk about my function or about ways to compare texts. Just about reducing the number of comparisons.
MySQL's spatial columns
- You create spatial columns with: CREATE TABLE abc (clmnName TYPE)
- possible types are listed here
- here is how I select the data later [e.g. MultiPointFromText() or AsText()]
- You insert values like this: INSERT INTO clmnName VALUES (GeomFromText('POINT(1 1)'))
But how do you use this for my problem?
PS: I'm looking for ways to reduce the number of comparisons with algorithms in this question. Vinko Vrsalovic told me that I should open another question for the spatial features.
While R-Trees in general can index data with arbitrary number of dimensions, MySQL spatial abilities are only limited to Geometry types (2 dimensions).
If your vectors are 2-dimensional and you can normalize them, then do the following:
- Split the circle into twice the number of angles which fit your differences
- Find the MBR of vectors with given cosine difference from the center of each sector
- Find all vectors within the MBR
- Do the fine filtering for exact difference.
In this case, however, it will be better just to precaculate the angle of the value and index it with a plain B-Tree index.
这篇关于加速文本比较(特征向量)与空间MySQL功能的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持!