Tokenize Strings

A tokenize function is available in H2O-3, which converts strings into tokens, then stores the tokenized text into a single column, making it easier for additional processing.

Simple Tokenize Example

Below is a simple example showing strings from frames tokenized into a single column. Refer to the following demos for a more extensive demo using tokenized text in Word2Vec:

• Python: https://github.com/h2oai/h2o-3/blob/master/h2o-py/demos/word2vec_craigslistjobtitles.ipynb
• R: https://github.com/h2oai/h2o-3/blob/master/h2o-r/demos/rdemo.word2vec.craigslistjobtitles.R

library(h2o)
h2o.init()

# Create four simple, single-column R data frames by inputting values.
s1 <- as.character(as.h2o(" this is a string "))
s2 <- as.character(as.h2o("this is another string"))
s3 <- as.character(as.h2o("this is a longer string"))
s4 <- as.character(as.h2o("this is tall, this is taller"))

# Combine the datasets into a single dataset.
ds <- h2o.rbind(s1, s2, s3, s4)
ds
                            C1
1            this is a string
2       this is another string
3      this is a longer string
4 this is tall, this is taller

# Tokenize the dataset.
# Notice that tokenized sentences are separated by <NA>.
tokenized <- h2o.tokenize(ds, " ")
tokenized
      C1
1
2   this
3     is
4      a
5 string
6   <NA>

[24 rows x 1 column]

import h2o
h2o.init()

# Create four simple, single-column Python data frames by inputting values
df1 = h2o.H2OFrame.from_python({'String':[' this is a string ']})
df1 = df1.ascharacter()
df2 = h2o.H2OFrame.from_python({'String':['this is another string']})
df2 = df2.ascharacter()
df3 = h2o.H2OFrame.from_python({'String':['this is a longer string']})
df3 = df3.ascharacter()
df4 = h2o.H2OFrame.from_python({'String':['this is tall, this is taller']})
df4 = df4.ascharacter()

# Combine the datasets into a single dataset.
combined = df1.rbind([df2, df3, df4])
combined
String
----------------------------
this is a string
this is another string
this is a longer string
this is tall, this is taller

# Tokenize the dataset.
# Notice that tokenized sentences are separated by empty rows.
tokenized = combined.tokenize(" ")
tokenized.describe
C1
-------

this
is
a
string

this
is
another
string

[24 rows x 1 column]