Word vectors data class: wordvec and embed. — as

PsychWordVec uses two types of word vectors data: wordvec (data.table, with two variables word and vec) and embed (matrix, with dimensions as columns and words as row names). Note that matrix operation makes embed much faster than wordvec. Users are suggested to reshape data to embed before using the other functions.

Usage

as_embed(x, normalize = FALSE)

as_wordvec(x, normalize = FALSE)

# S3 method for embed
[(x, i, j)

pattern(pattern)

Arguments

x: Object to be reshaped. See examples.
normalize: Normalize all word vectors to unit length? Defaults to FALSE. See normalize.
i, j: Row (i) and column (j) filter to be used in embed[i, j].
pattern: Regular expression to be used in embed[pattern("...")].

Value

A wordvec (data.table) or embed (matrix).

Functions

as_embed(): From wordvec (data.table) to embed (matrix).
as_wordvec(): From embed (matrix) to wordvec (data.table).

Download

Download pre-trained word vectors data (.RData): https://psychbruce.github.io/WordVector_RData.pdf

Examples

dt = head(demodata, 10)
str(dt)
#> Class "wordvec" [10 × 2] (inherits: "data.table")
#> $ word : "in" "for" "that" "is" "on" ...
#> $ vec : list of 10
#> - attr(*, "dims") = 300
#> - attr(*, "normalized") = FALSE

embed = as_embed(dt, normalize=TRUE)
embed
#> # embed (matrix): [10 × 300] (normalized)
#>             dim1 ...     dim300
#>  1: in    0.0530 ... <300 dims>
#>  2: for  -0.0085 ... <300 dims>
#>  3: that -0.0124 ... <300 dims>
#>  4: is    0.0037 ... <300 dims>
#>  5: on    0.0167 ... <300 dims>
#>  6: with -0.0160 ... <300 dims>
#>  7: said -0.0045 ... <300 dims>
#>  8: was   0.0126 ... <300 dims>
#>  9: the   0.0747 ... <300 dims>
#> 10: at   -0.0354 ... <300 dims>
str(embed)
#> Class "embed" [10 × 300] (inherits: "matrix")
#> - rownames(*) : "in" "for" "that" "is" "on" ...
#> - colnames(*) : "dim1" "dim2" "dim3" "dim4" "dim5" ...
#> - attr(*, "dims") = 300
#> - attr(*, "normalized") = TRUE

wordvec = as_wordvec(embed, normalize=TRUE)
wordvec
#> # wordvec (data.table): [10 × 2] (normalized)
#>     word                      vec
#>  1:   in [ 0.0530, ...<300 dims>]
#>  2:  for [-0.0085, ...<300 dims>]
#>  3: that [-0.0124, ...<300 dims>]
#>  4:   is [ 0.0037, ...<300 dims>]
#>  5:   on [ 0.0167, ...<300 dims>]
#>  6: with [-0.0160, ...<300 dims>]
#>  7: said [-0.0045, ...<300 dims>]
#>  8:  was [ 0.0126, ...<300 dims>]
#>  9:  the [ 0.0747, ...<300 dims>]
#> 10:   at [-0.0354, ...<300 dims>]
str(wordvec)
#> Class "wordvec" [10 × 2] (inherits: "data.table")
#> $ word : "in" "for" "that" "is" "on" ...
#> $ vec : list of 10
#> - attr(*, "dims") = 300
#> - attr(*, "normalized") = TRUE

df = data.frame(token=LETTERS, D1=1:26/10000, D2=26:1/10000)
as_embed(df)
#> # embed (matrix): [26 × 2] (NOT normalized)
#>          dim1 ...     dim2
#>  1: A  0.0001 ... <2 dims>
#>  2: B  0.0002 ... <2 dims>
#>  3: C  0.0003 ... <2 dims>
#>  4: D  0.0004 ... <2 dims>
#>  5: E  0.0005 ... <2 dims>
#>  6: F  0.0006 ... <2 dims>
#>  7: G  0.0007 ... <2 dims>
#>  8: H  0.0008 ... <2 dims>
#>  9: I  0.0009 ... <2 dims>
#> 10: J  0.0010 ... <2 dims>
#> 11: K  0.0011 ... <2 dims>
#> 12: L  0.0012 ... <2 dims>
#> 13: M  0.0013 ... <2 dims>
#> 14: N  0.0014 ... <2 dims>
#> 15: O  0.0015 ... <2 dims>
#> 16: P  0.0016 ... <2 dims>
#> 17: Q  0.0017 ... <2 dims>
#> 18: R  0.0018 ... <2 dims>
#> 19: S  0.0019 ... <2 dims>
#> 20: T  0.0020 ... <2 dims>
#> 21: U  0.0021 ... <2 dims>
#> 22: V  0.0022 ... <2 dims>
#> 23: W  0.0023 ... <2 dims>
#> 24: X  0.0024 ... <2 dims>
#> 25: Y  0.0025 ... <2 dims>
#> 26: Z  0.0026 ... <2 dims>
as_wordvec(df)
#> # wordvec (data.table): [26 × 2] (NOT normalized)
#>     word                    vec
#>  1:    A [ 0.0001, ...<2 dims>]
#>  2:    B [ 0.0002, ...<2 dims>]
#>  3:    C [ 0.0003, ...<2 dims>]
#>  4:    D [ 0.0004, ...<2 dims>]
#>  5:    E [ 0.0005, ...<2 dims>]
#>  6:    F [ 0.0006, ...<2 dims>]
#>  7:    G [ 0.0007, ...<2 dims>]
#>  8:    H [ 0.0008, ...<2 dims>]
#>  9:    I [ 0.0009, ...<2 dims>]
#> 10:    J [ 0.0010, ...<2 dims>]
#> 11:    K [ 0.0011, ...<2 dims>]
#> 12:    L [ 0.0012, ...<2 dims>]
#> 13:    M [ 0.0013, ...<2 dims>]
#> 14:    N [ 0.0014, ...<2 dims>]
#> 15:    O [ 0.0015, ...<2 dims>]
#> 16:    P [ 0.0016, ...<2 dims>]
#> 17:    Q [ 0.0017, ...<2 dims>]
#> 18:    R [ 0.0018, ...<2 dims>]
#> 19:    S [ 0.0019, ...<2 dims>]
#> 20:    T [ 0.0020, ...<2 dims>]
#> 21:    U [ 0.0021, ...<2 dims>]
#> 22:    V [ 0.0022, ...<2 dims>]
#> 23:    W [ 0.0023, ...<2 dims>]
#> 24:    X [ 0.0024, ...<2 dims>]
#> 25:    Y [ 0.0025, ...<2 dims>]
#> 26:    Z [ 0.0026, ...<2 dims>]

dd = rbind(dt[1:5], dt[1:5])
dd  # duplicate words
#> # wordvec (data.table): [10 × 2] (NOT normalized)
#>     word                      vec
#>  1:   in [ 0.0703, ...<300 dims>]
#>  2:  for [-0.0118, ...<300 dims>]
#>  3: that [-0.0157, ...<300 dims>]
#>  4:   is [ 0.0070, ...<300 dims>]
#>  5:   on [ 0.0267, ...<300 dims>]
#>  6:   in [ 0.0703, ...<300 dims>]
#>  7:  for [-0.0118, ...<300 dims>]
#>  8: that [-0.0157, ...<300 dims>]
#>  9:   is [ 0.0070, ...<300 dims>]
#> 10:   on [ 0.0267, ...<300 dims>]
#> ! 5 duplicate words: use `unique()` to delete duplicates
unique(dd)
#> # wordvec (data.table): [5 × 2] (NOT normalized)
#>    word                      vec
#> 1:   in [ 0.0703, ...<300 dims>]
#> 2:  for [-0.0118, ...<300 dims>]
#> 3: that [-0.0157, ...<300 dims>]
#> 4:   is [ 0.0070, ...<300 dims>]
#> 5:   on [ 0.0267, ...<300 dims>]

dm = as_embed(dd)
dm  # duplicate words
#> # embed (matrix): [10 × 300] (NOT normalized)
#>             dim1 ...     dim300
#>  1: in    0.0703 ... <300 dims>
#>  2: for  -0.0118 ... <300 dims>
#>  3: that -0.0157 ... <300 dims>
#>  4: is    0.0070 ... <300 dims>
#>  5: on    0.0267 ... <300 dims>
#>  6: in    0.0703 ... <300 dims>
#>  7: for  -0.0118 ... <300 dims>
#>  8: that -0.0157 ... <300 dims>
#>  9: is    0.0070 ... <300 dims>
#> 10: on    0.0267 ... <300 dims>
#> ! 5 duplicate words: use `unique()` to delete duplicates
unique(dm)
#> # embed (matrix): [5 × 300] (NOT normalized)
#>            dim1 ...     dim300
#> 1: in    0.0703 ... <300 dims>
#> 2: for  -0.0118 ... <300 dims>
#> 3: that -0.0157 ... <300 dims>
#> 4: is    0.0070 ... <300 dims>
#> 5: on    0.0267 ... <300 dims>

# more examples for extracting a subset using `x[i, j]`
# (3x faster than `wordvec`)
embed = as_embed(demodata)
embed[1]
#> # embed (matrix): [1 × 300] (NOT normalized)
#>          dim1 ...     dim300
#> 1: in  0.0703 ... <300 dims>
embed[1:5]
#> # embed (matrix): [5 × 300] (NOT normalized)
#>            dim1 ...     dim300
#> 1: in    0.0703 ... <300 dims>
#> 2: for  -0.0118 ... <300 dims>
#> 3: that -0.0157 ... <300 dims>
#> 4: is    0.0070 ... <300 dims>
#> 5: on    0.0267 ... <300 dims>
embed["for"]
#> # embed (matrix): [1 × 300] (NOT normalized)
#>           dim1 ...     dim300
#> 1: for -0.0118 ... <300 dims>
embed[pattern("^for.{0,2}$")]
#> # embed (matrix): [6 × 300] (NOT normalized)
#>             dim1 ...     dim300
#> 1: for   -0.0118 ... <300 dims>
#> 2: form   0.0835 ... <300 dims>
#> 3: force  0.1641 ... <300 dims>
#> 4: forms -0.0206 ... <300 dims>
#> 5: forum -0.1309 ... <300 dims>
#> 6: forth -0.1074 ... <300 dims>
embed[cc("for, in, on, xxx")]
#> ✖ 1 word not found: "xxx"
#> # embed (matrix): [3 × 300] (NOT normalized)
#>           dim1 ...     dim300
#> 1: for -0.0118 ... <300 dims>
#> 2: in   0.0703 ... <300 dims>
#> 3: on   0.0267 ... <300 dims>
embed[cc("for, in, on, xxx"), 5:10]
#> ✖ 1 word not found: "xxx"
#> # embed (matrix): [3 × 6] (NOT normalized)
#>           dim5 ...    dim10
#> 1: for -0.0182 ... <6 dims>
#> 2: in   0.0693 ... <6 dims>
#> 3: on   0.0062 ... <6 dims>
embed[1:5, 5:10]
#> # embed (matrix): [5 × 6] (NOT normalized)
#>            dim5 ...    dim10
#> 1: in    0.0693 ... <6 dims>
#> 2: for  -0.0182 ... <6 dims>
#> 3: that -0.1104 ... <6 dims>
#> 4: is   -0.1328 ... <6 dims>
#> 5: on    0.0062 ... <6 dims>
embed[, 5:10]
#> # embed (matrix): [8000 × 6] (NOT normalized)
#>                          dim5 ...    dim10
#>    1: in               0.0693 ... <6 dims>
#>    2: for             -0.0182 ... <6 dims>
#>    3: that            -0.1104 ... <6 dims>
#>    4: is              -0.1328 ... <6 dims>
#>    5: on               0.0062 ... <6 dims>
#> -----                                     
#> 7996: salesperson      0.0172 ... <6 dims>
#> 7997: computation      0.1465 ... <6 dims>
#> 7998: psychotherapy    0.2539 ... <6 dims>
#> 7999: equations       -0.4082 ... <6 dims>
#> 8000: psychotherapist  0.0198 ... <6 dims>
embed[3, 4]
#> that [dim4] 
#>    0.050293 
embed["that", 4]
#> that [dim4] 
#>    0.050293

Word vectors data class: `wordvec` and `embed`.