r - Using textstat_simil with a dictionary or globs in Quanteda
问题描述
I looked into the documentation, but as far as I understand, there is now way to use the textstat_simil
function with a dictionary or globs. What would be the best way of approaching something like the below?
txt <- "It is raining. It rains a lot during the rainy season"
rain_dfm <- dfm(txt)
textstat_simil(rain_dfm, "rain", method = "cosine", margin = "features")
Do I need to use tokens_replace
to change "rain*" to "rain", or is there another way to do this? In this case, stemming would do the trick, but what about cases where that is not feasible?
解决方案
这是可能的,但首先您需要使用dfm_lookup()
. (注意:还有其他方法可以做到这一点,例如标记化然后使用tokens_lookup()
, or tokens_replace()
,但我认为查找方法更直接,这也是您在问题中提出的问题。
另请注意,对于特征相似性,您必须拥有多个文档,这解释了为什么我在这里添加了两个。
txt <- c("It is raining. It rains a lot during the rainy season",
"Raining today, and it rained yesterday.",
"When it's raining it must be rainy season.")
rain_dfm <- dfm(txt)
然后使用字典将带有“rain*”的全局匹配(默认)转换为“rain”,同时保留其他功能。(在这种特殊情况下,您是正确的,dfm_wordstem()
可以完成同样的事情。)
rain_dfm <- dfm_lookup(rain_dfm,
dictionary(list(rain = "rain*")),
exclusive = FALSE,
capkeys = FALSE)
rain_dfm
## Document-feature matrix of: 3 documents, 17 features (52.9% sparse).
## 3 x 17 sparse Matrix of class "dfm"
## features
## docs it is rain . a lot during the season today , and yesterday when it's must be
## text1 2 1 3 1 1 1 1 1 1 0 0 0 0 0 0 0 0
## text2 1 0 2 1 0 0 0 0 0 1 1 1 1 0 0 0 0
## text3 1 0 2 1 0 0 0 0 1 0 0 0 0 1 1 1 1
现在,您可以计算“rain”的目标特征的余弦相似度:
textstat_simil(rain_dfm, selection = "rain", method = "cosine", margin = "features")
## rain
## it 0.9901475
## is 0.7276069
## rain 1.0000000
## . 0.9801961
## a 0.7276069
## lot 0.7276069
## during 0.7276069
## the 0.7276069
## season 0.8574929
## today 0.4850713
## , 0.4850713
## and 0.4850713
## yesterday 0.4850713
## when 0.4850713
## it's 0.4850713
## must 0.4850713
## be 0.4850713
推荐阅读
- python - 如何连接到boto3中的区域
- android - 单击通知时检查活动是否在屏幕上?(使用意图)
- android - 如何让 OnCompleteListener 在 Kotlin 中等待?
- api - Binance Websocket 深度通道对给定 ID 的更新过多
- python - 散景 categorical_bar_stacked
- php - 我该如何解决这个关于 webdriver 的错误?
- json - 将大对象序列化为 JSON 字符串
- prometheus - 普罗米修斯中的刮擦间隔和评估间隔
- javascript - 使用 jquery 返回搜索结果页面
- scala - 函数式缩放 - 如何避免可选映射的深度嵌套