[問題] text2vec 怎麼取 context vectors

看板R_Language作者augustana (微小的希望)時間8年前 (2017/07/26 18:09)推噓0(0推 0噓 0→)

留言0則, 0人參與討論串1/1

[問題類型]: 程式諮詢(我想用R 做某件事情，但是我不知道要怎麼用R 寫出來) [軟體熟悉度]: 入門(寫過其他程式，只是對語法不熟悉) [問題敘述]: 我想利用text2vec將文字向量化不過我沒辦法抓到他的word vector跟context vector 感覺是因為這兩個向量是在glove這個environment中private的區塊所以怎麼寫都抓不出來當我輸入glove$的時候, R自己跑出來的選單都只有public的東西 environment如下, 其中w_i, w_j就是我想要抓的向量 <GloVe> Inherits from: <word_embedding_model> Public: clone: function (deep = FALSE) dump_every_n: 0 dump_model: function () fit: function (x, n_iter, convergence_tol = -1, ...) get_history: function () get_word_vectors: function () initialize: function (word_vectors_size, vocabulary, x_max, learning_rate = 0.15, shuffle: FALSE verbose: TRUE Private: alpha: 0.75 b_i: -0.088758796453476 -0.200479492545128 -0.276277631521225 ... b_j: 0.158077865839005 0.00269329198636115 -0.506954908370972 ... cost_history: 0.0582876185658562 0.0376007230450009 0.0264438356106707 ... fitted: TRUE glove_fitter: Rcpp_GloveFitter grain_size: 100000 initial: NULL internal_matrix_format: dgTMatrix lambda: 0 learning_rate: 0.15 max_cost: 10 vocab_terms: 3000 體型較呵護 m62 護膚新生兒 s12 級白金特極頂級 ... w_i: 0.194914728403091 -0.0265734232962132 -0.611702501773834 ... w_j: 0.273217976093292 -0.193755224347115 0.475706458091736 0 ... word_vectors_history: NULL word_vectors_size: 50 x_max: 10 [程式範例]: library(text2vec) keyword <- as.character(article_list$keyword[1]) keyword <- enc2utf8(keyword) #轉UTF8 keyword <- strsplit(keyword,',') ########計算不重複的詞 # iterator token <- itoken(keyword) # to create unique word matrix vocab <- create_vocabulary(token, ngram=c(1, 1)) #詞,頻率,文章佔比 ##只篩出現5次以上的詞 #vocab <- prune_vocabulary(vocab, term_count_min = 5L) ########向量化 # vectorization of words vectorizer <- vocab_vectorizer(vocab, grow_dtm= FALSE, #don't vectorize input skip_grams_window= 5L) #use window of 5 for context words # tcm= term co-occurrence matrix 字段共現矩陣 tcm <- create_tcm(token, vectorizer) # glove fitting model, 分解TCM矩陣 glove <- GlobalVectors$new(word_vectors_size = 50, vocabulary = vocab, x_max = 10) glove$fit(tcm, n_iter = 20) #詞向量 word.vec <- glove$word_vectors$w_i + #文字向量 glove$word_vectors$w_j #脈絡向量就是最後一行這出了問題不知道是不是因為text2vec裡面的glove()已經被刪除改成GlobalVectors()的關係所以這條就失敗了 [環境敘述]: R version 3.4.0 (2017-04-21) Platform: x86_64-w64-mingw32/x64 (64-bit) Running under: Windows 7 x64 (build 7601) Service Pack 1 Matrix products: default locale: [1] LC_COLLATE=Chinese (Traditional)_Taiwan.950 LC_CTYPE=Chinese (Traditional)_Taiwan.950 [3] LC_MONETARY=Chinese (Traditional)_Taiwan.950 LC_NUMERIC=C [5] LC_TIME=Chinese (Traditional)_Taiwan.950 attached base packages: [1] stats graphics grDevices utils datasets methods base other attached packages: [1] stringi_1.1.2 RODBC_1.3-14 text2vec_0.4.0 loaded via a namespace (and not attached): [1] compiler_3.4.0 magrittr_1.5 R6_2.2.0 Matrix_1.2-9 tools_3.4.0 [6] Rcpp_0.12.11 codetools_0.2-15 grid_3.4.0 iterators_1.0.8 foreach_1.4.3 [11] data.table_1.10.4 digest_0.6.10 RcppParallel_4.3.20 lattice_0.20-35 [關鍵字]: text2vec, environment, private -- ※ 發信站: 批踢踢實業坊(ptt.cc), 來自: 36.224.218.7 ※ 文章網址: https://www.ptt.cc/bbs/R_Language/M.1501063745.A.5E3.html ※ 編輯: augustana (36.224.218.7), 07/26/2017 18:12:22