forked from filipg/aitech-eks-pub
1197 lines
38 KiB
Plaintext
1197 lines
38 KiB
Plaintext
|
{
|
||
|
|
"cells": [
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"# Zagęszczamy wektory\n",
|
||
|
|
"\n",
|
||
|
|
"Podstawowy problem z wektorową reprezentacją typu tf-idf polega na tym, że wektory dokumentów (i macierz całej kolekcji dokumentów) są _rzadkie_, tzn. zawierają dużo zer. W praktyce potrzebujemy bardziej \"gęstej\" czy \"kompaktowej\" reprezentacji numerycznej dokumentów. \n",
|
||
|
|
"\n",
|
||
|
|
"## _Hashing trick_\n",
|
||
|
|
"\n",
|
||
|
|
"Powierzchownie problem możemy rozwiązać przez użycie tzw. _sztuczki z haszowaniem_ (_hashing trick_). Będziemy potrzebować funkcji mieszającej (haszującej) $H$, która rzutuje na napisy na liczby, których reprezentacja binarna składa się z $b$ bitów:\n",
|
||
|
|
"\n",
|
||
|
|
"$$H : V \\rightarrow \\{0,\\dots,2^b-1\\}$$\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"\n"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Jako funkcji $H$ możemy np. użyć funkcji MurmurHash3."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 1,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"Hash64 0x6c3a641663470e2c"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"Hash64 0xa714568917576314"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"Hash64 0x875d9e7e413747c8"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"Hash64 0x13ce831936ebc69e"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"import Data.Digest.Murmur64\n",
|
||
|
|
"\n",
|
||
|
|
"hash64 \"komputer\"\n",
|
||
|
|
"hash64 \"komputerze\"\n",
|
||
|
|
"hash64 \"komputerek\"\n",
|
||
|
|
"hash64 \"abrakadabra\"\n"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"**Pytanie:** podobne napisy mają zupełnie różne wartości funkcji haszującej, czy to dobrze, czy to źle?"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Musimy tylko sparametryzować naszą funkcję rozmiarem \"odcisku\" (parametr $b$)."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 2,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"3628"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"25364"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"2877"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"50846"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"12"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"{-# LANGUAGE OverloadedStrings #-}\n",
|
||
|
|
"\n",
|
||
|
|
"import Data.Text\n",
|
||
|
|
"\n",
|
||
|
|
"-- pomocnicza funkcja, która konwertuje wartość specjalnego\n",
|
||
|
|
"-- typu Hash64 do zwykłej liczby całkowitej\n",
|
||
|
|
"hashValueAsInteger :: Hash64 -> Integer\n",
|
||
|
|
"hashValueAsInteger = toInteger . asWord64\n",
|
||
|
|
"\n",
|
||
|
|
"-- unpack to funkcja, która wartość typu String konwertuje do Text\n",
|
||
|
|
"hash :: Integer -> Text -> Integer\n",
|
||
|
|
"hash b t = hashValueAsInteger (hash64 $ unpack t) `mod` (2 ^ b)\n",
|
||
|
|
"\n",
|
||
|
|
"hash 16 \"komputer\"\n",
|
||
|
|
"hash 16 \"komputerze\"\n",
|
||
|
|
"hash 16 \"komputerem\"\n",
|
||
|
|
"hash 16 \"abrakadabra\"\n",
|
||
|
|
"hash 4 \"komputer\""
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"**Pytanie:** Jakie wartości $b$ będą bezsensowne?"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Sztuczka z haszowaniem polega na tym, że zamiast numerować słowa korzystając ze słownika, po prostu używamy funkcji haszującej. W ten sposób wektor będzie _zawsze_ rozmiar $2^b$ - bez względu na rozmiar słownika."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Zacznijmy od przywołania wszystkich potrzebnych definicji."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 3,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [],
|
||
|
|
"source": [
|
||
|
|
"{-# LANGUAGE OverloadedStrings #-}\n",
|
||
|
|
"{-# LANGUAGE QuasiQuotes #-}\n",
|
||
|
|
"\n",
|
||
|
|
"import Data.Text hiding(map, filter, zip)\n",
|
||
|
|
"import Text.Regex.PCRE.Heavy\n",
|
||
|
|
"\n",
|
||
|
|
"isStopWord :: Text -> Bool\n",
|
||
|
|
"isStopWord \"w\" = True\n",
|
||
|
|
"isStopWord \"jest\" = True\n",
|
||
|
|
"isStopWord \"że\" = True\n",
|
||
|
|
"isStopWord w = w ≈ [re|^\\p{P}+$|]\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"removeStopWords :: [Text] -> [Text]\n",
|
||
|
|
"removeStopWords = filter (not . isStopWord)"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 4,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [],
|
||
|
|
"source": [
|
||
|
|
"{-# LANGUAGE OverloadedStrings #-}\n",
|
||
|
|
"{-# LANGUAGE QuasiQuotes #-}\n",
|
||
|
|
"{-# LANGUAGE FlexibleContexts #-}\n",
|
||
|
|
"\n",
|
||
|
|
"import Data.Text hiding(map, filter, zip)\n",
|
||
|
|
"import Prelude hiding(words, take)\n",
|
||
|
|
"import Text.Regex.PCRE.Heavy\n",
|
||
|
|
"import Data.Map as Map hiding(take, map, filter)\n",
|
||
|
|
"import Data.Set as Set hiding(map)\n",
|
||
|
|
"\n",
|
||
|
|
"tokenize :: Text -> [Text]\n",
|
||
|
|
"tokenize = map fst . scan [re|C\\+\\+|[\\p{L}0-9]+|\\p{P}|]\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"mockInflectionDictionary :: Map Text Text\n",
|
||
|
|
"mockInflectionDictionary = Map.fromList [\n",
|
||
|
|
" (\"kota\", \"kot\"),\n",
|
||
|
|
" (\"butach\", \"but\"),\n",
|
||
|
|
" (\"masz\", \"mieć\"),\n",
|
||
|
|
" (\"ma\", \"mieć\"),\n",
|
||
|
|
" (\"buta\", \"but\"),\n",
|
||
|
|
" (\"zgubiłem\", \"zgubić\")]\n",
|
||
|
|
"\n",
|
||
|
|
"lemmatizeWord :: Map Text Text -> Text -> Text\n",
|
||
|
|
"lemmatizeWord dict w = findWithDefault w w dict\n",
|
||
|
|
"\n",
|
||
|
|
"lemmatize :: Map Text Text -> [Text] -> [Text]\n",
|
||
|
|
"lemmatize dict = map (lemmatizeWord dict)\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"poorMansStemming = Data.Text.take 6\n",
|
||
|
|
"\n",
|
||
|
|
"normalize :: Text -> [Text]\n",
|
||
|
|
"normalize = map poorMansStemming . removeStopWords . map toLower . lemmatize mockInflectionDictionary . tokenize\n",
|
||
|
|
"\n",
|
||
|
|
"getVocabulary :: [Text] -> Set Text \n",
|
||
|
|
"getVocabulary = Set.unions . map (Set.fromList . normalize) \n",
|
||
|
|
" \n",
|
||
|
|
"idf :: [[Text]] -> Text -> Double\n",
|
||
|
|
"idf coll t = log (fromIntegral n / fromIntegral df)\n",
|
||
|
|
" where df = Prelude.length $ Prelude.filter (\\d -> t `elem` d) coll\n",
|
||
|
|
" n = Prelude.length coll\n",
|
||
|
|
" \n",
|
||
|
|
"vectorizeTfIdf :: Int -> [[Text]] -> Map Int Text -> [Text] -> [Double]\n",
|
||
|
|
"vectorizeTfIdf vecSize coll v doc = map (\\i -> count (v ! i) doc * idf coll (v ! i)) [0..(vecSize-1)]\n",
|
||
|
|
" where count t doc = fromIntegral $ (Prelude.length . Prelude.filter (== t)) doc "
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 5,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [],
|
||
|
|
"source": [
|
||
|
|
"import System.IO\n",
|
||
|
|
"import Data.List.Split as SP\n",
|
||
|
|
"\n",
|
||
|
|
"legendsh <- openFile \"legendy.txt\" ReadMode\n",
|
||
|
|
"hSetEncoding legendsh utf8\n",
|
||
|
|
"contents <- hGetContents legendsh\n",
|
||
|
|
"ls = Prelude.lines contents\n",
|
||
|
|
"items = map (map pack . SP.splitOn \"\\t\") ls\n",
|
||
|
|
"\n",
|
||
|
|
"labelsL = map Prelude.head items\n",
|
||
|
|
"collectionL = map (!!1) items\n",
|
||
|
|
"\n",
|
||
|
|
"collectionLNormalized = map normalize collectionL\n",
|
||
|
|
"voc' = getVocabulary collectionL\n",
|
||
|
|
"\n",
|
||
|
|
"vocLSize = Prelude.length voc'\n",
|
||
|
|
"\n",
|
||
|
|
"vocL :: Map Int Text\n",
|
||
|
|
"vocL = Map.fromList $ zip [0..] $ Set.toList voc'\n",
|
||
|
|
"\n",
|
||
|
|
"invvocL :: Map Text Int\n",
|
||
|
|
"invvocL = Map.fromList $ zip (Set.toList voc') [0..]\n",
|
||
|
|
"\n",
|
||
|
|
"lVectorized = map (vectorizeTfIdf vocLSize collectionLNormalized vocL) collectionLNormalized\n"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 6,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/html": [
|
||
|
|
"<style>/* Styles used for the Hoogle display in the pager */\n",
|
||
|
|
".hoogle-doc {\n",
|
||
|
|
"display: block;\n",
|
||
|
|
"padding-bottom: 1.3em;\n",
|
||
|
|
"padding-left: 0.4em;\n",
|
||
|
|
"}\n",
|
||
|
|
".hoogle-code {\n",
|
||
|
|
"display: block;\n",
|
||
|
|
"font-family: monospace;\n",
|
||
|
|
"white-space: pre;\n",
|
||
|
|
"}\n",
|
||
|
|
".hoogle-text {\n",
|
||
|
|
"display: block;\n",
|
||
|
|
"}\n",
|
||
|
|
".hoogle-name {\n",
|
||
|
|
"color: green;\n",
|
||
|
|
"font-weight: bold;\n",
|
||
|
|
"}\n",
|
||
|
|
".hoogle-head {\n",
|
||
|
|
"font-weight: bold;\n",
|
||
|
|
"}\n",
|
||
|
|
".hoogle-sub {\n",
|
||
|
|
"display: block;\n",
|
||
|
|
"margin-left: 0.4em;\n",
|
||
|
|
"}\n",
|
||
|
|
".hoogle-package {\n",
|
||
|
|
"font-weight: bold;\n",
|
||
|
|
"font-style: italic;\n",
|
||
|
|
"}\n",
|
||
|
|
".hoogle-module {\n",
|
||
|
|
"font-weight: bold;\n",
|
||
|
|
"}\n",
|
||
|
|
".hoogle-class {\n",
|
||
|
|
"font-weight: bold;\n",
|
||
|
|
"}\n",
|
||
|
|
".get-type {\n",
|
||
|
|
"color: green;\n",
|
||
|
|
"font-weight: bold;\n",
|
||
|
|
"font-family: monospace;\n",
|
||
|
|
"display: block;\n",
|
||
|
|
"white-space: pre-wrap;\n",
|
||
|
|
"}\n",
|
||
|
|
".show-type {\n",
|
||
|
|
"color: green;\n",
|
||
|
|
"font-weight: bold;\n",
|
||
|
|
"font-family: monospace;\n",
|
||
|
|
"margin-left: 1em;\n",
|
||
|
|
"}\n",
|
||
|
|
".mono {\n",
|
||
|
|
"font-family: monospace;\n",
|
||
|
|
"display: block;\n",
|
||
|
|
"}\n",
|
||
|
|
".err-msg {\n",
|
||
|
|
"color: red;\n",
|
||
|
|
"font-style: italic;\n",
|
||
|
|
"font-family: monospace;\n",
|
||
|
|
"white-space: pre;\n",
|
||
|
|
"display: block;\n",
|
||
|
|
"}\n",
|
||
|
|
"#unshowable {\n",
|
||
|
|
"color: red;\n",
|
||
|
|
"font-weight: bold;\n",
|
||
|
|
"}\n",
|
||
|
|
".err-msg.in.collapse {\n",
|
||
|
|
"padding-top: 0.7em;\n",
|
||
|
|
"}\n",
|
||
|
|
".highlight-code {\n",
|
||
|
|
"white-space: pre;\n",
|
||
|
|
"font-family: monospace;\n",
|
||
|
|
"}\n",
|
||
|
|
".suggestion-warning { \n",
|
||
|
|
"font-weight: bold;\n",
|
||
|
|
"color: rgb(200, 130, 0);\n",
|
||
|
|
"}\n",
|
||
|
|
".suggestion-error { \n",
|
||
|
|
"font-weight: bold;\n",
|
||
|
|
"color: red;\n",
|
||
|
|
"}\n",
|
||
|
|
".suggestion-name {\n",
|
||
|
|
"font-weight: bold;\n",
|
||
|
|
"}\n",
|
||
|
|
"</style><div class=\"suggestion-name\" style=\"clear:both;\">Eta reduce</div><div class=\"suggestion-row\" style=\"float: left;\"><div class=\"suggestion-warning\">Found:</div><div class=\"highlight-code\" id=\"haskell\">formatNumber x = printf \"% 7.2f\" x</div></div><div class=\"suggestion-row\" style=\"float: left;\"><div class=\"suggestion-warning\">Why Not:</div><div class=\"highlight-code\" id=\"haskell\">formatNumber = printf \"% 7.2f\"</div></div><div class=\"suggestion-name\" style=\"clear:both;\">Use zipWith</div><div class=\"suggestion-row\" style=\"float: left;\"><div class=\"suggestion-warning\">Found:</div><div class=\"highlight-code\" id=\"haskell\">map (\\ (lab, ix) -> lab <> \" \" <> similarTo simFun vs ix)\n",
|
||
|
|
" $ zip labels [0 .. (Prelude.length vs - 1)]</div></div><div class=\"suggestion-row\" style=\"float: left;\"><div class=\"suggestion-warning\">Why Not:</div><div class=\"highlight-code\" id=\"haskell\">zipWith\n",
|
||
|
|
" (curry (\\ (lab, ix) -> lab <> \" \" <> similarTo simFun vs ix))\n",
|
||
|
|
" labels [0 .. (Prelude.length vs - 1)]</div></div><div class=\"suggestion-name\" style=\"clear:both;\">Avoid lambda</div><div class=\"suggestion-row\" style=\"float: left;\"><div class=\"suggestion-warning\">Found:</div><div class=\"highlight-code\" id=\"haskell\">\\ l -> pack $ printf \"% 7s\" l</div></div><div class=\"suggestion-row\" style=\"float: left;\"><div class=\"suggestion-warning\">Why Not:</div><div class=\"highlight-code\" id=\"haskell\">pack . printf \"% 7s\"</div></div>"
|
||
|
|
],
|
||
|
|
"text/plain": [
|
||
|
|
"Line 5: Eta reduce\n",
|
||
|
|
"Found:\n",
|
||
|
|
"formatNumber x = printf \"% 7.2f\" x\n",
|
||
|
|
"Why not:\n",
|
||
|
|
"formatNumber = printf \"% 7.2f\"Line 11: Use zipWith\n",
|
||
|
|
"Found:\n",
|
||
|
|
"map (\\ (lab, ix) -> lab <> \" \" <> similarTo simFun vs ix)\n",
|
||
|
|
" $ zip labels [0 .. (Prelude.length vs - 1)]\n",
|
||
|
|
"Why not:\n",
|
||
|
|
"zipWith\n",
|
||
|
|
" (curry (\\ (lab, ix) -> lab <> \" \" <> similarTo simFun vs ix))\n",
|
||
|
|
" labels [0 .. (Prelude.length vs - 1)]Line 12: Avoid lambda\n",
|
||
|
|
"Found:\n",
|
||
|
|
"\\ l -> pack $ printf \"% 7s\" l\n",
|
||
|
|
"Why not:\n",
|
||
|
|
"pack . printf \"% 7s\""
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"import Text.Printf\n",
|
||
|
|
"import Data.List (take)\n",
|
||
|
|
"\n",
|
||
|
|
"formatNumber :: Double -> String\n",
|
||
|
|
"formatNumber x = printf \"% 7.2f\" x\n",
|
||
|
|
"\n",
|
||
|
|
"similarTo :: ([Double] -> [Double] -> Double) -> [[Double]] -> Int -> Text\n",
|
||
|
|
"similarTo simFun vs ix = pack $ Prelude.unwords $ map (formatNumber . ((vs !! ix) `simFun`)) vs\n",
|
||
|
|
"\n",
|
||
|
|
"paintMatrix :: ([Double] -> [Double] -> Double) -> [Text] -> [[Double]] -> Text\n",
|
||
|
|
"paintMatrix simFun labels vs = header <> \"\\n\" <> Data.Text.unlines (map (\\(lab, ix) -> lab <> \" \" <> similarTo simFun vs ix) $ zip labels [0..(Prelude.length vs - 1)])\n",
|
||
|
|
" where header = \" \" <> Data.Text.unwords (map (\\l -> pack $ printf \"% 7s\" l) labels)"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 7,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
" na_ak w_lud ba_hy w_lap ne_dz be_wy zw_oz mo_zu be_wy ba_hy mo_zu be_wy w_lud\n",
|
||
|
|
"na_ak 1.00 0.02 0.01 0.01 0.03 0.02 0.02 0.04 0.03 0.02 0.01 0.02 0.03\n",
|
||
|
|
"w_lud 0.02 1.00 0.02 0.05 0.04 0.01 0.03 0.04 0.06 0.01 0.02 0.03 0.06\n",
|
||
|
|
"ba_hy 0.01 0.02 1.00 0.01 0.02 0.03 0.03 0.04 0.08 0.22 0.01 0.04 0.01\n",
|
||
|
|
"w_lap 0.01 0.05 0.01 1.00 0.01 0.01 0.00 0.01 0.02 0.00 0.00 0.00 0.00\n",
|
||
|
|
"ne_dz 0.03 0.04 0.02 0.01 1.00 0.04 0.03 0.07 0.08 0.06 0.03 0.03 0.05\n",
|
||
|
|
"be_wy 0.02 0.01 0.03 0.01 0.04 1.00 0.01 0.03 0.21 0.01 0.02 0.25 0.01\n",
|
||
|
|
"zw_oz 0.02 0.03 0.03 0.00 0.03 0.01 1.00 0.04 0.03 0.00 0.01 0.02 0.02\n",
|
||
|
|
"mo_zu 0.04 0.04 0.04 0.01 0.07 0.03 0.04 1.00 0.10 0.02 0.09 0.05 0.04\n",
|
||
|
|
"be_wy 0.03 0.06 0.08 0.02 0.08 0.21 0.03 0.10 1.00 0.05 0.03 0.24 0.04\n",
|
||
|
|
"ba_hy 0.02 0.01 0.22 0.00 0.06 0.01 0.00 0.02 0.05 1.00 0.01 0.02 0.00\n",
|
||
|
|
"mo_zu 0.01 0.02 0.01 0.00 0.03 0.02 0.01 0.09 0.03 0.01 1.00 0.01 0.02\n",
|
||
|
|
"be_wy 0.02 0.03 0.04 0.00 0.03 0.25 0.02 0.05 0.24 0.02 0.01 1.00 0.02\n",
|
||
|
|
"w_lud 0.03 0.06 0.01 0.00 0.05 0.01 0.02 0.04 0.04 0.00 0.02 0.02 1.00"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"limit = 13\n",
|
||
|
|
"labelsLimited = Data.List.take limit labelsL\n",
|
||
|
|
"limitedL = Data.List.take limit lVectorized\n",
|
||
|
|
"\n",
|
||
|
|
"vectorNorm :: [Double] -> Double\n",
|
||
|
|
"vectorNorm vs = sqrt $ sum $ map (\\x -> x * x) vs\n",
|
||
|
|
"\n",
|
||
|
|
"toUnitVector :: [Double] -> [Double]\n",
|
||
|
|
"toUnitVector vs = map (/ n) vs\n",
|
||
|
|
" where n = vectorNorm vs\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"(✕) :: [Double] -> [Double] -> Double\n",
|
||
|
|
"(✕) v1 v2 = sum $ Prelude.zipWith (*) v1 v2\n",
|
||
|
|
"\n",
|
||
|
|
"cosineSim v1 v2 = toUnitVector v1 ✕ toUnitVector v2\n",
|
||
|
|
"\n",
|
||
|
|
"paintMatrix cosineSim labelsLimited limitedL"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Powyższa macierz reprezentuje porównanie przy użyciu podobieństwa kosinusowego. Spróbujmy teraz użyć gęstszych wektorów przy użyciu hashing trick. Jako wartość $b$ przyjmijmy 6.\n",
|
||
|
|
"\n",
|
||
|
|
"Zobaczmy najpierw, w które \"przegródki\" będą wpadały poszczególne wyrazy słownika.\n",
|
||
|
|
"\n",
|
||
|
|
"\n"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 8,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"[(\"0\",32),(\"00\",4),(\"01\",4),(\"07\",40),(\"09\",44),(\"1\",1),(\"10\",61),(\"100\",27),(\"12\",58),(\"13\",51),(\"131\",37),(\"15\",30),(\"16\",21),(\"17\",58),(\"18\",55),(\"19\",35),(\"1997r\",61),(\"2\",62),(\"20\",28),(\"2006\",44),(\"2008\",19),(\"2009\",4),(\"2010\",3),(\"22\",27),(\"23\",34),(\"24\",7),(\"25\",29),(\"26\",35),(\"27\",44),(\"28\",61),(\"29\",30),(\"3\",56),(\"30\",55),(\"300\",38),(\"31\",45),(\"4\",53),(\"40\",39),(\"42\",43),(\"48\",53),(\"49\",13),(\"5\",31),(\"50\",32),(\"56\",38),(\"57\",55),(\"6\",59),(\"7\",27),(\"8\",34),(\"a\",27),(\"aaa\",33),(\"absolu\",11),(\"absurd\",18),(\"aby\",12),(\"adnym\",10),(\"adres\",15),(\"adrese\",62),(\"afroam\",3),(\"afryce\",46),(\"agresy\",57),(\"ah\",37),(\"aha\",42),(\"aig\",56),(\"akadem\",18),(\"akcja\",0),(\"akcje\",21),(\"akompa\",13),(\"aktor\",26),(\"akurat\",7),(\"albino\",27),(\"albo\",44),(\"ale\",7),(\"alfa\",58),(\"alkoho\",56),(\"altern\",38),(\"ameryk\",11),(\"amp\",62),(\"anakon\",34),(\"analiz\",62),(\"andrze\",63),(\"anegdo\",43),(\"ang\",37),(\"anga\\380o\",27),(\"anglii\",33),(\"ani\",22),(\"anonsu\",36),(\"antono\",3),(\"antykr\",41),(\"apetyt\",16),(\"apolit\",39),(\"apropo\",54),(\"apteki\",20),(\"aqua\",59),(\"archit\",61),(\"aromat\",44),(\"artyku\",31),(\"asami\",22),(\"astron\",59),(\"asy\\347ci\",60),(\"atmosf\",37),(\"audycj\",50),(\"auta\",38)]"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"map (\\t -> (t, hash 6 t)) $ Data.List.take 100 $ Set.toList voc'"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"**Pytanie:** Czy jakieś dwa termy wpadły do jednej przegródki?"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Stwórzmy najpierw funkcję, która będzie wektoryzowała pojedynczy term $t$. Po prostu stworzymy wektor, które będzie miał rozmiar $2^b$, wszędzie będzie miał 0 z wyjątkiem pozycji o numerze $H_b(t)$ - tam wpiszmy odwrotną częstość dokumentową."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 9,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,3.7727609380946383,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0]"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"wordVector :: Integer -> [[Text]] -> Text -> [Double]\n",
|
||
|
|
"wordVector b coll term = map selector [0..vecSize]\n",
|
||
|
|
" where vecSize = 2^b - 1\n",
|
||
|
|
" wordFingerprint = hash b term\n",
|
||
|
|
" selector i \n",
|
||
|
|
" | i == wordFingerprint = idf coll term\n",
|
||
|
|
" | otherwise = 0.0\n",
|
||
|
|
"\n",
|
||
|
|
"wordVector 6 collectionLNormalized \"ameryk\""
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Teraz wystarczy zsumować wektory dla poszczególnych słów, żeby otrzymać wektor dokumentu. Najpierw zdefiniujmy sobie sumę wektorową."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 10,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"[1.2,4.0]"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"(+++) :: [Double] -> [Double] -> [Double]\n",
|
||
|
|
"(+++) = Prelude.zipWith (+)\n",
|
||
|
|
"\n",
|
||
|
|
"[0.2, 0.5] +++ [1.0, 3.5]"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Przydatna będzie jeszcze funkcja, która tworzy wektor z samymi zerami o zadanej długości:"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 11,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0]"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"zero :: Int -> [Double]\n",
|
||
|
|
"zero s = Prelude.replicate s 0.0\n",
|
||
|
|
"\n",
|
||
|
|
"zero (2^6)"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 12,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"[5.242936783195232,0.0,0.0,0.0,0.0,0.0,0.0,0.0,2.856470206220483,0.0,0.0,1.1700712526502546,0.5947071077466928,0.0,5.712940412440966,3.0708470981669183,0.0,0.0,4.465908118654584,0.0,3.7727609380946383,0.0,0.0,0.0,0.0,4.788681510917635,0.0,3.7727609380946383,0.0,1.575536360758419,0.0,3.079613757534693,0.0,4.465908118654584,0.0,4.588010815455483,4.465908118654584,0.0,1.5214691394881432,0.0,0.0,0.0,0.0,4.465908118654584,2.5199979695992702,0.0,1.5214691394881432,8.388148398070203e-2,0.0,4.465908118654584,0.0,0.0,3.367295829986474,0.0,3.7727609380946383,0.0,1.5214691394881432,0.0,3.7727609380946383,0.0,0.0,0.0,3.367295829986474,0.0]"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"\n",
|
||
|
|
"vectorizeWithHashingTrick :: Integer -> [[Text]] -> [Text] -> [Double]\n",
|
||
|
|
"vectorizeWithHashingTrick b coll = Prelude.foldr ((+++) . wordVector b coll) (zero $ 2^b)\n",
|
||
|
|
"\n",
|
||
|
|
"vectorizeWithHashingTrick 6 collectionLNormalized $ collectionLNormalized !! 3\n"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Zobaczmy, jak zagęszczenie wpływa na macierz podobieństwa."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 13,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
" na_ak w_lud ba_hy w_lap ne_dz be_wy zw_oz mo_zu be_wy ba_hy mo_zu be_wy w_lud\n",
|
||
|
|
"na_ak 1.00 0.66 0.40 0.58 0.65 0.52 0.66 0.79 0.76 0.41 0.59 0.44 0.72\n",
|
||
|
|
"w_lud 0.66 1.00 0.51 0.53 0.66 0.43 0.57 0.76 0.68 0.38 0.47 0.42 0.62\n",
|
||
|
|
"ba_hy 0.40 0.51 1.00 0.42 0.55 0.29 0.41 0.54 0.58 0.54 0.47 0.24 0.50\n",
|
||
|
|
"w_lap 0.58 0.53 0.42 1.00 0.41 0.35 0.54 0.59 0.53 0.19 0.47 0.34 0.53\n",
|
||
|
|
"ne_dz 0.65 0.66 0.55 0.41 1.00 0.56 0.56 0.79 0.74 0.55 0.68 0.57 0.69\n",
|
||
|
|
"be_wy 0.52 0.43 0.29 0.35 0.56 1.00 0.51 0.54 0.64 0.28 0.59 0.61 0.49\n",
|
||
|
|
"zw_oz 0.66 0.57 0.41 0.54 0.56 0.51 1.00 0.72 0.61 0.29 0.55 0.48 0.63\n",
|
||
|
|
"mo_zu 0.79 0.76 0.54 0.59 0.79 0.54 0.72 1.00 0.79 0.49 0.73 0.58 0.79\n",
|
||
|
|
"be_wy 0.76 0.68 0.58 0.53 0.74 0.64 0.61 0.79 1.00 0.49 0.72 0.61 0.74\n",
|
||
|
|
"ba_hy 0.41 0.38 0.54 0.19 0.55 0.28 0.29 0.49 0.49 1.00 0.37 0.32 0.48\n",
|
||
|
|
"mo_zu 0.59 0.47 0.47 0.47 0.68 0.59 0.55 0.73 0.72 0.37 1.00 0.53 0.71\n",
|
||
|
|
"be_wy 0.44 0.42 0.24 0.34 0.57 0.61 0.48 0.58 0.61 0.32 0.53 1.00 0.54\n",
|
||
|
|
"w_lud 0.72 0.62 0.50 0.53 0.69 0.49 0.63 0.79 0.74 0.48 0.71 0.54 1.00"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"lVectorized' = map (vectorizeWithHashingTrick 6 collectionLNormalized) collectionLNormalized\n",
|
||
|
|
"limitedL' = Data.List.take limit lVectorized'\n",
|
||
|
|
"\n",
|
||
|
|
"paintMatrix cosineSim labelsLimited limitedL'"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"**Pytanie:** Co się stanie, gdy zwiększymy $b$, a co jeśli zmniejszymi?\n",
|
||
|
|
"\n",
|
||
|
|
"Zalety sztuczki z haszowaniem:\n",
|
||
|
|
"\n",
|
||
|
|
"* zagwarantowany stały rozmiar wektora\n",
|
||
|
|
"* szybsze obliczenia\n",
|
||
|
|
"* w naturalny sposób uwzględniamy termy, których nie było w początkowej kolekcji (ale uwaga na idf!)\n",
|
||
|
|
"* nie musimy pamiętać odzworowania rzutującego słowa na ich numery\n",
|
||
|
|
"\n",
|
||
|
|
"Wady:\n",
|
||
|
|
"\n",
|
||
|
|
"* dwa różne słowa mogą wpaść do jednej przegródki (szczególnie częste, jeśli $b$ jest za małe)\n",
|
||
|
|
"* jeśli $b$ ustawimy za duże, wektory mogą być nawet większe niż w przypadku standardowego podejścia\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"\n"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"## Word2vec\n",
|
||
|
|
"\n",
|
||
|
|
"A może istnieje dobra wróżka, która dałaby nam dobre wektory słów (z których będziemy składali proste wektory dokumentów przez sumowanie)?\n",
|
||
|
|
"\n",
|
||
|
|
"**Pytanie:** Jakie własności powinny mieć dobre wektory słów?\n",
|
||
|
|
"\n",
|
||
|
|
"Tak! Istnieją gotowe \"bazy danych\" wektorów. Jedną z najpopularniejszych (i najstarszych) metod uzyskiwania takich wektorów jest Word2vec. Jak dokładnie Word2vec, dowiemy się później, na dzisiaj po prostu użyjmy tych wektorów.\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"\n",
|
||
|
|
"\n"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Najpierw wprowadźmy alternatywną normalizację zgodną z tym, jak został wygenerowany model."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 14,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"ala"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"ma"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"kota"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"normalize' :: Text -> [Text]\n",
|
||
|
|
"normalize' = removeStopWords . map toLower . tokenize\n",
|
||
|
|
"\n",
|
||
|
|
"normalize' \"Ala ma kota.\""
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 15,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"mam"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"kumpla"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"ktory"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"zdawal"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"walentynki"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"i"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"polozyl"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"koperte"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"dla"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"laski"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"z"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"kartka"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"na"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"desce"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"rozdzielczej"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"egzaminator"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"wziol"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"ta"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"karteke"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"i"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"powiedzial"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"ze"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"ma"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"znade"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"wypisal"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"mu"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"papierek"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"i"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"po"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"egzaminie"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"hehe"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"filmik"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"dobry"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"collectionLNormalized' = map normalize' collectionL\n",
|
||
|
|
"collectionLNormalized' !! 3"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 17,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"[-2.305081844329834,0.3418600857257843,4.44999361038208,0.9008448719978333,-2.1629886627197266,1.0206516981124878,4.157524108886719,2.5060904026031494,-0.17275184392929077,4.085052967071533,2.236677408218384,-2.3315281867980957,0.5224806070327759,0.15804219245910645,-1.5636622905731201,-1.2624900341033936,-0.3161393105983734,-1.971177101135254,1.4859644174575806,-0.1742715835571289,1.209444284439087,4.063786193728447e-2,-0.2808700501918793,-0.5895432233810425,-4.126195430755615,-2.690922260284424,1.4975452423095703,-0.25380706787109375,-4.5767364501953125,-1.7726246118545532,2.938936710357666,-0.7173141837120056,-2.4317402839660645,-4.206724643707275,0.6768773198127747,2.236821413040161,4.1044291108846664e-2,1.6991114616394043,1.2354476377367973e-2,-3.079916000366211,-1.7430219650268555,1.8969229459762573,-0.4897139072418213,1.1981141567230225,2.431124687194824,0.39453181624412537,1.9735784530639648,2.124225378036499,-4.338796138763428,-0.954145610332489,3.3927927017211914,0.8821511268615723,5.120451096445322e-3,2.917816638946533,-2.035374164581299,3.3221969604492188,-4.981880187988281,-1.105080008506775,-4.093905448913574,-1.5998111963272095,0.6372298002243042,-0.7565107345581055,0.4038744270801544,0.685226321220398,2.137610912322998,-0.4390018582344055,1.007287859916687,0.19681350886821747,-2.598611354827881,-1.8872140645980835,1.6989527940750122,1.6458508968353271,-5.091184616088867,1.4902764558792114,-0.4839307367801666,-2.840092420578003,1.0180696249008179,0.7615311741828918,1.8135554790496826,-0.30493396520614624,3.5879104137420654,1.4585649967193604,3.2775094509124756,-1.1610190868377686,-2.3159284591674805,4.1530327796936035,-4.67172384262085,-0.8594478964805603,-0.860812783241272,-0.31788957118988037,0.7260096669197083,0.1879102736711502,-0.15789580345153809,1.9434200525283813,-1.9945732355117798,1.8799400329589844,-0.5253798365592957,-0.2834266722202301,-0.8012301921844482,1.5093021392822266]"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"100"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"{-# LANGUAGE OverloadedStrings #-}\n",
|
||
|
|
"{-# LANGUAGE BangPatterns #-}\n",
|
||
|
|
"\n",
|
||
|
|
"import Data.Word2Vec.Model\n",
|
||
|
|
"import Data.Maybe (catMaybes, fromJust)\n",
|
||
|
|
"import qualified Data.Vector.Storable as V\n",
|
||
|
|
"\n",
|
||
|
|
"model <- readWord2VecModel \"tiny.bin\"\n",
|
||
|
|
"\n",
|
||
|
|
"toOurVector :: WVector -> [Double]\n",
|
||
|
|
"toOurVector (WVector v _) = map realToFrac $ V.toList v\n",
|
||
|
|
"\n",
|
||
|
|
"balwanV = toOurVector $ fromJust $ getVector model \"bałwan\"\n",
|
||
|
|
"balwanV\n",
|
||
|
|
"Prelude.length balwanV\n",
|
||
|
|
"\n",
|
||
|
|
"vectorizeWord2vec model d = Prelude.foldr (+++) (zero 100) $ map toOurVector $ catMaybes $ map (getVector model) d\n",
|
||
|
|
"\n",
|
||
|
|
"collectionLVectorized'' = map (vectorizeWord2vec model) collectionLNormalized'"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 18,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
"[-26.834667675197124,2.568521626293659,37.66925026476383,9.381511189043522,-32.04328362643719,-19.734033070504665,55.21128339320421,14.215368987061083,23.60182836651802,38.74189975857735,0.16257449332624674,-47.983866568654776,-36.917382495012134,36.08420217037201,13.996580198407173,-30.473296120762825,21.28328724205494,30.601420499384403,-40.5945385559462,16.043263137340546,-8.694086126983166,-41.90418399870396,-10.448782376945019,-0.21028679609298706,9.586350612342358,-46.172676257789135,46.27567541599274,11.25023115798831,9.00947591662407,-43.525397814810276,22.09978771582246,56.93886440992355,-23.428963833488524,-1.4649565666913986,21.969609811902046,-21.504647210240364,24.955158293247223,-8.328911297023296,-31.118815276771784,0.22846409678459167,12.212224327027798,-28.337586268782616,-24.105730276554823,3.36764569953084,8.270942151546478,33.71851025521755,30.665825616568327,-24.134687054902315,-31.72916578501463,35.20022106170654,71.15121555328369,-15.448215141892433,-41.27439119666815,3.0322337672114372,9.768462024629116,38.911416467279196,-9.848581969738007,-20.030757322907448,6.734442539513111,-84.9070791369304,38.147536396980286,4.3607237339019775,-25.426255017518997,5.240264508873224,-32.71464269608259,2.095752328634262,2.4292337521910667,32.93906496465206,-51.44473773613572,0.5551527962088585,-6.1982685178518295,20.187213011085987,-52.809339098632336,-10.458874322474003,13.979218572378159,-38.16066548228264,27.336308609694242,5.3437707126140594,-32.01269288826734,-38.117460787296295,-9.337415304034948,38.90077601373196,-2.158842660486698,-44.878454223275185,23.69188129901886,-54.10413733869791,-41.30505630373955,-37.28948371112347,-65.8488347530365,32.51569982431829,3.781733974814415,72.77320172637701,6.847739472985268,63.77478001266718,24.26227615773678,7.260737741366029,10.931276574730873,-17.388786104973406,9.978045962750912,5.968699499964714]"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"collectionLVectorized'' !! 3"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": 19,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"data": {
|
||
|
|
"text/plain": [
|
||
|
|
" na_ak w_lud ba_hy w_lap ne_dz be_wy zw_oz mo_zu be_wy ba_hy mo_zu be_wy w_lud\n",
|
||
|
|
"na_ak 1.00 0.92 0.85 0.77 0.87 0.90 0.92 0.88 0.87 0.87 0.89 0.89 0.89\n",
|
||
|
|
"w_lud 0.92 1.00 0.92 0.72 0.93 0.93 0.91 0.94 0.95 0.86 0.94 0.94 0.96\n",
|
||
|
|
"ba_hy 0.85 0.92 1.00 0.69 0.89 0.91 0.83 0.89 0.95 0.86 0.87 0.94 0.90\n",
|
||
|
|
"w_lap 0.77 0.72 0.69 1.00 0.60 0.74 0.67 0.65 0.68 0.58 0.68 0.73 0.66\n",
|
||
|
|
"ne_dz 0.87 0.93 0.89 0.60 1.00 0.90 0.87 0.95 0.94 0.86 0.93 0.90 0.95\n",
|
||
|
|
"be_wy 0.90 0.93 0.91 0.74 0.90 1.00 0.89 0.89 0.91 0.85 0.91 0.96 0.94\n",
|
||
|
|
"zw_oz 0.92 0.91 0.83 0.67 0.87 0.89 1.00 0.89 0.86 0.86 0.91 0.85 0.90\n",
|
||
|
|
"mo_zu 0.88 0.94 0.89 0.65 0.95 0.89 0.89 1.00 0.97 0.85 0.95 0.91 0.96\n",
|
||
|
|
"be_wy 0.87 0.95 0.95 0.68 0.94 0.91 0.86 0.97 1.00 0.84 0.93 0.95 0.95\n",
|
||
|
|
"ba_hy 0.87 0.86 0.86 0.58 0.86 0.85 0.86 0.85 0.84 1.00 0.83 0.85 0.84\n",
|
||
|
|
"mo_zu 0.89 0.94 0.87 0.68 0.93 0.91 0.91 0.95 0.93 0.83 1.00 0.91 0.96\n",
|
||
|
|
"be_wy 0.89 0.94 0.94 0.73 0.90 0.96 0.85 0.91 0.95 0.85 0.91 1.00 0.94\n",
|
||
|
|
"w_lud 0.89 0.96 0.90 0.66 0.95 0.94 0.90 0.96 0.95 0.84 0.96 0.94 1.00"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
"metadata": {},
|
||
|
|
"output_type": "display_data"
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"source": [
|
||
|
|
"limitedL'' = Data.List.take limit collectionLVectorized''\n",
|
||
|
|
"\n",
|
||
|
|
"paintMatrix cosineSim labelsLimited limitedL''"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"execution_count": null,
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [],
|
||
|
|
"source": []
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"metadata": {
|
||
|
|
"kernelspec": {
|
||
|
|
"display_name": "Haskell",
|
||
|
|
"language": "haskell",
|
||
|
|
"name": "haskell"
|
||
|
|
},
|
||
|
|
"language_info": {
|
||
|
|
"codemirror_mode": "ihaskell",
|
||
|
|
"file_extension": ".hs",
|
||
|
|
"mimetype": "text/x-haskell",
|
||
|
|
"name": "haskell",
|
||
|
|
"pygments_lexer": "Haskell",
|
||
|
|
"version": "8.10.4"
|
||
|
|
}
|
||
|
|
},
|
||
|
|
"nbformat": 4,
|
||
|
|
"nbformat_minor": 4
|
||
|
|
}
|