pipian is a tiny interface to CaboCha; a Japanese dependency structure parser. The main goal of pipian is to implement a parser for that XML output.
Usage
Installation
remotes::install_github("paithiov909/pipian")Cast dependency structure as an igraph
sentence <- "ふと振り向くと、たくさんの味方がいてたくさんの優しい人間がいることを、わざわざ自分の誕生日が来ないと気付けない自分を奮い立たせながらも、毎日こんな、湖のようななんの引っ掛かりもない、落ちつき倒し、音一つも感じさせない人間でいれる方に憧れを持てたとある25歳の眩しき朝のことでした"
df <- sentence |>
pipian::ppn_cabocha() |>
pipian::ppn_parse_xml()
head(df)
#> # A tibble: 6 × 19
#> doc_id sentence_id chunk_id token_id token chunk_link chunk_score chunk_head
#> <int> <int> <int> <int> <chr> <int> <dbl> <int>
#> 1 1 1 1 0 ふと 2 1.29 1
#> 2 1 1 2 1 振り向く…… 36 -1.17 2
#> 3 1 1 2 2 と 36 -1.17 2
#> 4 1 1 2 3 、 36 -1.17 2
#> 5 1 1 3 4 たくさん…… 4 1.93 5
#> 6 1 1 3 5 の 4 1.93 5
#> # ℹ 11 more variables: chunk_func <int>, entity <chr>, POS1 <chr>, POS2 <chr>,
#> # POS3 <chr>, POS4 <chr>, X5StageUse1 <chr>, X5StageUse2 <chr>,
#> # Original <chr>, Yomi1 <chr>, Yomi2 <chr>
g <- df |>
pipian::ppn_make_graph()
print(g)
#> IGRAPH c1835b0 DN-- 37 37 --
#> + attr: name (v/c), tokens (v/c), pos (v/c), score (e/n)
#> + edges from c1835b0 (vertex names):
#> [1] 111 ->112 112 ->1136 113 ->114 114 ->115 115 ->119 116 ->118
#> [7] 117 ->118 118 ->119 119 ->1110 1110->1136 1111->1114 1112->1113
#> [13] 1113->1114 1114->1117 1115->1116 1116->1117 1117->1131 1118->1119
#> [19] 1119->1129 1120->1121 1121->1122 1122->1123 1123->1129 1124->1126
#> [25] 1125->1126 1126->1127 1127->1128 1128->1129 1129->1131 1130->1131
#> [31] 1131->1135 1132->1133 1133->1135 1134->1135 1135->1136 1136->110
#> [37] 110 ->110
pagerank <- igraph::page_rank(g, directed = TRUE)
plot(
g,
vertex.size = pagerank$vector * 50,
vertex.color = "steelblue",
vertex.label = igraph::V(g)$tokens,
vertex.label.cex = 0.6,
vertex.label.color = "black",
vertex.label.family = "sans-serif",
edge.width = 0.4,
edge.arrow.size = 0.4,
edge.color = "gray80",
layout = igraph::layout_as_tree(g, mode = "in", flip.y = FALSE),
)