[B! nlp] yassã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯

yass id:yass

nlpã«é–¢ã™ã‚‹yassã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (111)

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

${{author_name}}$
{{author_name}}{{created}}
{{ #comment }}{{ comment }}{{ /comment }}
- {{ label }}

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

æ‰‹å…ƒã«ç½®ã„ã¦ãŠãã¨å®‰å¿ƒã§ãã‚‹ã€æƒ…å ±ç³»ã®äººå‘ã‘ãªæ—¥æœ¬èªžã®æœ¬ã®ãƒªã‚¹ãƒˆ - EchizenBlog-Zwei
æœ€è¿‘ã€äººã«æœ¬ã‚’è–¦ã‚ã‚‹äº‹ãŒå¤šããªã£ãŸã€‚ã¨ã‚Šã‚ãˆãšã“ã®è¾ºã‚’èªã‚€ã¨ã„ã„ã§ã™ã‚ˆçš„ãªãƒªã‚¹ãƒˆã‚’ä½œã£ã¦ãŠãã¨ä¾¿åˆ©ã ã¨æ€ã£ãŸã®ã§ä½œã‚‹ã“ã¨ã«ã—ãŸã€‚ ä»¥ä¸‹ã€ã€Œäº‹å‰çŸ¥è˜ã®ã„ã‚‰ãªã„å…¥é–€æœ¬ã€ã€Œäº‹å‰çŸ¥è˜ã¯ã„ã‚‰ãªã„ã‘ã©æœ¬æ ¼çš„ãªæœ¬ã€ã€Œäº‹å‰çŸ¥è˜ãŒãªã„ã¨ä½•è¨€ã£ã¦ã‚‹ã‹ã‚ã‹ã‚‰ãªã„ã‘ã©æœ‰ç›Šãªæƒ…å ±ãŒæº€è¼‰ãªæœ¬ã€ã®3ã¤ã«ã‚ã‘ã¦åˆ—æŒ™ã™ã‚‹ã€‚ äº‹å‰çŸ¥è˜ã®ã„ã‚‰ãªã„å…¥é–€æœ¬ æ•°å¼å°‘ãªã‚ã€è„³è² è·ã®å°ã•ã‚ãªæœ¬ã‚’ã„ãã¤ã‹ã€‚ä½•ã‚’ã‚„ã‚‹ã«ã—ã¦ã‚‚ãƒ‡ãƒ¼ã‚¿æ§‹é€ ã€ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã€æ•°å¦ã¯ã‚„ã£ã¦ãŠãã¨å¹¸ã›ã«ãªã‚Œã‚‹ã‚ˆã€‚ æƒ…å ±æ¤œç´¢ã¨è¨€èªžå‡¦ç† ãƒ‡ãƒ¼ã‚¿ãƒžã‚¤ãƒ‹ãƒ³ã‚°ã¨ã‹è‡ªç„¶è¨€èªžå‡¦ç†ã¨ã‹ã‚„ã‚ŠãŸã„äººã«ã¯ã¨ã‚Šã‚ãˆãšã“ã‚Œã€‚ã•ã™ãŒã«å¤ã„è©±ãŒå¤šããªã£ã¦ããŸã®ã§ãã‚ãã‚æ–°ã—ã„å…¥é–€ç”¨æƒ…å ±æ¤œç´¢æœ¬ãŒã§ãªã„ã‹ãªã‚ã¨æ€ã£ã¦ã„ã‚‹ã€‚ å›³è§£ãƒ»ãƒ™ã‚¤ã‚ºçµ±è¨ˆã€Œè¶…ã€å…¥é–€ ä¼èª¬ã®ãƒ™ã‚¤ã‚¸ã‚¢ãƒ³å…ˆç”ŸãŒãƒ™ã‚¤ã‚ºã®åŸºç¤Žã‚’æ•™ãˆã¦ãã‚Œã‚‹æœ¬ã€‚ãƒ™ã‚¤ã‚ºã‚„ã‚ŠãŸã„äººã¯ã“ã‚Œã€‚ ç çŽ‰ã®ãƒ—ãƒã‚°ãƒ©ãƒŸãƒ³ã‚° ãƒ‡ãƒ¼ã‚¿æ§‹é€ ã¨ã‹ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã¨ã‹ã®è€ƒãˆæ–¹ã®åŸºç¤Žã‚’æ•™ãˆã¦ã
yass 2014/02/24
book

algorithm

data structure

machinelearning

nlp
ãƒªãƒ³ã‚¯
Statistical Semanticå…¥é–€~åˆ†å¸ƒä»®èª¬ã‹ã‚‰word2vecã¾ã§ #1
Products Video Streaming Enterprise Video Streaming Enterprise Content Delivery Network Video Distribution and Workflow OTT Video Management Video Monetization Plans & pricing Apps & services Live streaming iPhone & iPad app Android app Certified devices Production services Solutions Live Events Corporate Communications Marketing Media and Entertainment Resources Webinars Product sheets White pape
yass 2014/02/11
nlp

lsi

lda

nmf

word2vec

video

PFI
ãƒªãƒ³ã‚¯
NIPS2013èªã¿ä¼šã§word2vecè«–æ–‡ã®ç´¹ä»‹ã‚’ã—ã¾ã—ãŸ
å…ˆé€±ã€ @sla ã•ã‚“ä¸»å‚¬ã®NIPS2013èªã¿ä¼šã§ã€word2vecè«–æ–‡ï¼ˆæ£ç¢ºã«ã¯ç¶šå ±ï¼‰ã®ç´¹ä»‹ã‚’ã—ã¾ã—ãŸã€‚ ã¡ã‚‡ã£ã¨è§£èª¬ã‚’æ›¸ãã¾ã™ã€‚ ã“ã®ã¨ã“ã‚ã®æ·±å±¤å¦ç¿’ãƒ–ãƒ¼ãƒ ã¯è‡ªç„¶è¨€èªžå‡¦ç†ã«ã‚‚æ¥ã¦ã„ã¦ã€ãã‚Œã‚‰ã®ã‚¦ãƒã®1ã¤ã¨è¨€ã‚ã‚Œã¦ã„ã¾ã™ï¼ˆãŒã€å…¨ç„¶deepã£ã½ã•ã¯ãªã„ï¼‰ã€‚ æœ€åˆã®ãƒ¢ãƒãƒ™ãƒ¼ã‚·ãƒ§ãƒ³ãŒã©ã†ã„ã†ã¨ã“ã‚ã«ã‚ã£ãŸã‹ã¨ã„ã†ã®ã¯ã€ã¡ã‚‡ã£ã¨è‰²ã€…ã ã¨æ€ã„ã¾ã™ãŒï¼ˆãŠãã‚‰ãæœ€åˆã¯è¨€èªžãƒ¢ãƒ‡ãƒ«ã«ãŠã‘ã‚‹ä½Žé »åº¦èªžã®ç¢ºçŽ‡ã‚’ã‚¦ãƒžã‚¤ã“ã¨ãƒ¢ãƒ‡ãƒ«åŒ–ãƒ»æŽ¨å®šã—ãŸã‹ã£ãŸã‚“ã§ã¯ãªã„ã‹ãªï¼‰ã€ä½•ã¯ã¨ã‚‚ã‚ã‚Œå˜èªžã®æ„å‘³çš„ãªã‚ã‚‹ã„ã¯çµ±èªžçš„ãªæŒ¯ã‚‹èˆžã„ã‚’ãƒ™ã‚¯ãƒˆãƒ«è¡¨ç¾ã§è¡¨ã™ã¨ã„ã†ç ”ç©¶ãŒæµè¡Œã£ã¦ãŠã‚Šã¾ã™ã€‚ ãƒ™ã‚¯ãƒˆãƒ«è¡¨ç¾ã¨ã„ã†ã®ã¯ã€1ã¤ã®å˜èªžwã«å¯¾ã—ã¦ã€ãã®å˜èªžã‚’ã€Œè¡¨ç¾ã€ã™ã‚‹ã‚ˆã†ãªãƒ™ã‚¯ãƒˆãƒ« v(w) ã‚’ä½œã‚Šã¾ã™ã€‚ ãã‚“ãªã“ã¨ã„ã‚ã‚Œã¦ã‚‚ã€ä½œã‚Œã°ï¼Ÿã¨ã„ã†ã“ã¨ãªã‚“ã§ã™ãŒã€ã§ããŸãƒ™ã‚¯ãƒˆãƒ«ã«å¯¾ã—ã¦ä½•ã‹ã—ã‚‰ã€Œéƒ½åˆã®ã„ã„ã€æ€§è³ªãŒã§ãã‚‹ã“ã¨ãŒçœŸã®ç›®æ¨™ã§ã™ã€‚ ã€Œéƒ½åˆã®ã„ã„ã€
yass 2014/01/25
word2vec

nlp

machinelearning
ãƒªãƒ³ã‚¯
RNNLM
ã†ã¡ã®ä¼šç¤¾ã«ã¯æ–°æŠ€è¡“ç ”ç©¶ä¼šã¨ã„ã†ç ”ç©¶å“¡ãŒãŠäº’ã„ã«æ–°ã—ã„æŠ€è¡“ã‚’èª¿æŸ»ã—ã¦ç™ºè¡¨ã—ã‚ã†ä¼šè°ãŒã‚ã‚‹ã®ã§ã™ãŒï¼Œãã“ã§æœ€è¿‘è©±é¡Œã®word2vecã«ã¤ã„ã¦èª¿ã¹ãŸã“ã¨ã‚’å ±å‘Šã—ã¾ã—ãŸï¼Ž ãƒ‡ãƒ¢ã¯è¦‹ã›ã‚‰ã‚Œãªã„ã®ã§ã™ãŒï¼Œgoogle code ã® word2vec ã®ãƒšãƒ¼ã‚¸ã‹ã‚‰è½ã¨ã›ã‚‹ text8 ã¨ã„ã†ãƒ‡ãƒ¼ã‚¿ã‚’ä½¿ã£ãŸâ€man king -> woman queenâ€ã®ä¾‹ã¨ï¼Œæ¯Žæ—¥æ–°èžã‚³ãƒ¼ãƒ‘ã‚¹2000å¹´ç‰ˆã‚’MeCabã§å½¢æ…‹ç´ è§£æžã—ãŸä¸Šã§ï¼Œword2vecã§å¦ç¿’ã—ãŸãƒ‡ãƒ¼ã‚¿ã‚’ä½¿ã£ãŸï¼Œâ€ä¼šç¤¾ ä»•äº‹ -> å¦æ ¡ å‹‰å¼·â€ã¨ã„ã†é–¢ä¿‚ã‚’ãƒ‡ãƒ¢ã—ã¦ã¾ã™ã€‚ æ–‡ç« ã ã‘ã‹ã‚‰ã“ã“ã¾ã§å¦ç¿’ã§ãã‚‹ã®ã‹ã¨æ£ç›´é©šãã¾ã—ãŸï¼Ž
yass 2013/11/11
word2vec

nlp
ãƒªãƒ³ã‚¯
word2vecã«è‹±è¾žéƒŽãƒ‡ãƒ¼ã‚¿ã‚’æ”¾ã‚Šè¾¼ã‚“ã§ã¿ãŸ - naoya_t@hatenablog
è‹±è¾žéƒŽã‚’word2vecã«æ”¾ã‚Šè¾¼ã‚“ã§ã¿ãŸã‚‰ã¡ã‚‡ã£ã¨é¢ç™½ã‹ã£ãŸã®ã§ãƒ¡ãƒ¢ã‚’ã€‚word2vecã«ã¤ã„ã¦ã¯å‰å›žã®è¨˜äº‹ã‚’å‚ç…§ã€‚ ä½¿ã£ãŸã®ã¯EIJI-138.TXTï¼ˆæœ€æ–°ã‚ˆã‚Šï¼‘ã¤å¤ã„ãƒãƒ¼ã‚¸ãƒ§ãƒ³ã§ã™ï¼‰ EDPã•ã‚“ã‹ã‚‰1980å††ãã‚‰ã„ã§è²·ãˆã¾ã™ã€‚ â– semantically-motivated {å½¢} : æ„å‘³è«–çš„ï½›ã„ã¿ã‚ã‚“ ã¦ãï½ã«å‹•æ©Ÿä»˜ã‘ã‚‰ã‚ŒãŸ â– semantically-restricted {å½¢} : æ„å‘³çš„ï½›ã„ã¿ ã¦ãï½ã«åˆ¶é™ï½›ã›ã„ã’ã‚“ï½ã•ã‚ŒãŸ â– semantics {å-1} : æ„å‘³è«–ï½›ã„ã¿ã‚ã‚“ï½ã€è¨˜å·è«–ï½›ãã”ã†ã‚ã‚“ï½ â– semantics {å-2} : ã€Šã‚³ã€‹ã€”ãƒ—ãƒã‚°ãƒ©ãƒ ã®ã€•å‹•ä½œ â– semantics : ã€ï¼ ã€‘ã‚»ãƒžãƒ³ãƒ†ã‚£ãƒƒã‚¯ã‚¹ã€ã€åˆ†ç¯€ã€‘seãƒ»manãƒ»tics â– semantics course : æ„å‘³è«–ï½›ã„ã¿ã‚ã‚“ï½ã®ã‚³ãƒ¼ã‚¹ â– semaphore {å-1} : æ‰‹æ——ä¿¡å·ï½›ã¦ã°ãŸ ã—ã‚“
yass 2013/10/02
word2vec

nlp
ãƒªãƒ³ã‚¯
Deep-learningã¯ãƒ©ãƒ†ãƒ³èªžã®å‹•è©žæ´»ç”¨ã‚’å¦ç¿’ã§ãã‚‹ã‹ï¼Ÿ Can deep-learning learn latin conjugation? - naoya_t@hatenablog
ãƒ©ãƒ†ãƒ³èªžãƒã‚¿ãŒç¶šãã¾ã™ãŒ å·¥è—¤ã•ã‚“ãŒãããŸã™ã§ç´¹ä»‹ã—ã¦ãŸ word2vec ãŒé¢ç™½ãã†ã ã£ãŸã®ã§ã€‚ https://code.google.com/p/word2vec/ ã§å°‘ã—éŠã‚“ã§ã¿ãŸã€‚ã„ã‚ã‚†ã‚‹ deep learning ã§ å˜èªžã®ãƒ™ã‚¯ãƒˆãƒ«è¡¨ç¾ã‚’å¦ç¿’ã—ã¦ãã‚Œã‚‹ã€‚ é¢ç™½ã„ã®ã¯ã€2ã¤ã®ãƒ™ã‚¯ãƒˆãƒ«ã®å·®ãŒã€2ã¤ã®å˜èªžã®é–¢ä¿‚ã‚’ã‚ˆãè¿‘ä¼¼ã—ã¦ãã‚Œã‚‹ã“ã¨ã€‚ It was recently shown that the word vectors capture many linguistic regularities, for example vector operations vector('Paris') - vector('France') + vector('Italy') results in a vector that is very close to vector('Rome'), and
yass 2013/10/02
word2vec

nlp
ãƒªãƒ³ã‚¯
word2vec in yhat: Word vector similarity | Daniel Rodriguez
A few weeks ago Google released some code to convert words to vectors called word2vec. The company I am currently working on does something similar and I was quite amazed by the performance and accuracy of Google's algorithm so I created a simple python wrapper to call the C code for training and read the training vectors into numpy arrays, you can check it out on pypi (word2vec). At the same time
yass 2013/10/02
word2vec

nlp
ãƒªãƒ³ã‚¯
è¨€è‘‰ã¨æ¦‚å¿µã‚’ç´ä»˜ã‘ã‚‹ã€œGoogleã«ã‚ˆã‚‹ã‚»ãƒžãƒ³ãƒ†ã‚£ãƒƒã‚¯Webå†æ§‹ç¯‰ - éšæ‰€ä½œä¸»éŒ²
ã‚»ãƒžãƒ³ãƒ†ã‚£ãƒƒã‚¯Webã¨ã„ã†è¨€è‘‰ã®æ„å‘³ã«è¸ã¿è¾¼ã‚“ã æ–‡æ›¸ãƒ‡ãƒ¼ã‚¿å‡¦ç†ãŒè©±é¡Œã«ãªã£ã¦10å¹´ãã‚‰ã„çµŒã¤ã¨æ€ã„ã¾ã™ãŒã€è¯ã€…ã—ãã‚µãƒ¼ãƒ“ã‚¹ã¨ã—ã¦ä½¿ã‚ã‚Œã¦ã„ã‚‹ã¨ã¯è€³ã«ã—ã¾ã›ã‚“ã—ã€ãƒ¦ãƒ¼ã‚¶ãƒ¼ã¨ã—ã¦ã‚‚ãã®æ©æµã«é ã‹ã£ã¦ã‚‹æ„Ÿè¦šã¯ã‚ã‚Šã¾ã›ã‚“ã€‚ã‚„ã¯ã‚Šè¡¨è¨˜ã‚†ã‚Œã‚„åŒç¾©èªžã€æ„å‘³éšŽå±¤ã«é–¢ã™ã‚‹è†¨å¤§ãªãƒ«ãƒ¼ãƒ«ä½“ç³»ï¼ˆã‚ªãƒ³ãƒˆãƒã‚¸ãƒ¼ï¼‰ã‚’å®Œç’§ã«ä½œã‚Šä¸Šã’ãŸã‚Šã€ã‚³ãƒ³ãƒ†ãƒ³ãƒ„ã«å¯¾ã™ã‚‹ãƒ¡ã‚¿ãƒ‡ãƒ¼ã‚¿ä»˜ã‘ãŸã‚Šã¨ã„ã†äººæ‰‹ã®ãƒœãƒˆãƒ«ãƒãƒƒã‚¯ãŒè§£æ¶ˆã•ã‚Œãªã‹ã£ãŸã“ã¨ãŒåŽŸå› ãªã®ã§ã—ã‚‡ã†ã‹ã€‚ãã‚“ãªåœæ»žã—ãŸã‚¨ãƒªã‚¢ã«ã€GoogleãŒã„ã¤ã‚‚é€šã‚Šå°‘ã€…æš´åŠ›çš„ï¼ˆï¼Ÿï¼‰ãªæ–¹æ³•ã§ä¸€çŸ³ã‚’æŠ•ã˜ã‚‹æˆæžœã‚’å…¬é–‹ã—ãŸã‚ˆã†ã§ã™ã€‚ From Words to Concepts and Back: Dictionaries for Linking Text, Entities and IdeasÂ [Google Research Blog] ä¸èº«ã¯ã„ãŸã£ã¦ã‚·ãƒ³ãƒ—ãƒ«ã§ã€è‹±èªžç‰ˆWikipediaè¨˜äº‹1ãƒšãƒ¼ã‚¸ã‚’ã
yass 2013/09/01
" è‹±èªžç‰ˆWikipediaè¨˜äº‹1ãƒšãƒ¼ã‚¸ã‚’ãã‚Œãžã‚Œ1ã¤ã®æ¦‚å¿µã ã¨å®šç¾©ã—ã€ãã“ã¸ãƒã‚¤ãƒ‘ãƒ¼ãƒªãƒ³ã‚¯ãŒè²¼ã‚‰ã‚Œã‚‹æ™‚ã®ã‚¢ãƒ³ã‚«ãƒ¼ãƒ†ã‚ã‚¹ãƒˆã‚’å„è¨€èªžã®Wikipediaã‚„ãã®ä»–Webãƒšãƒ¼ã‚¸ã‹ã‚‰ã‚«ã‚¦ãƒ³ãƒˆ / Wikipediaè¨˜äº‹ãŒç´„750ä¸‡ã€æ•°ãˆãŸãƒªãƒ³ã‚¯ã¯ç´„1å„„7500ä¸‡ä»¶ "

google

wikipedia

nlp
ãƒªãƒ³ã‚¯
New community features for Google Chat and an update on Currents
Join the official community for Google Workspace administrators In the Google Cloud Community, connect with Googlers and other Google Workspace admins like yourself. Participate in product discussions, check out the Community Articles, and learn tips and tricks that will make your work and life easier. Be the first to know what's happening with Google Workspace. ______________ Learn about more Goo
yass 2013/08/29
"ã„ã‚ã‚†ã‚‹ deep learning ã§ å˜èªžã®ãƒ™ã‚¯ãƒˆãƒ«è¡¨ç¾ã‚’å¦ç¿’ã—ã¦ãã‚Œã‚‹ã€‚ é¢ç™½ã„ã®ã¯ã€2ã¤ã®ãƒ™ã‚¯ãƒˆãƒ«ã®å·®ãŒã€2ã¤ã®å˜èªžã®é–¢ä¿‚ã‚’ã‚ˆãè¿‘ä¼¼ã—ã¦ãã‚Œã‚‹ã“ã¨ã€‚"

nlp

google

deep learning

word2vec

vector
ãƒªãƒ³ã‚¯
Google Code Archive - Long-term storage for Google Code Project Hosting.
Code Archive Skip to content Google About Google Privacy Terms
yass 2013/08/17
" The word vectors can be also used for deriving word classes from huge data sets. This is achieved by performing K-means clustering on top of the word vectors. "

google

similarity

dimension

nlp

vector

deep learning

word2vec
ãƒªãƒ³ã‚¯
é«˜é€Ÿæ–‡å—åˆ—è§£æžã®"åˆ¥"ä¸–ç•Œ - æ°—ã¾ã¾ãªãƒ–ãƒã‚°
1æœˆã«ã€Œé«˜é€Ÿæ–‡å—åˆ—è§£æžã®ä¸–ç•Œã€ã‚’è³¼å…¥ã—ã¦ã‹ã‚‰åŠå¹´ãŒçµŒã¡ã¾ã—ãŸã€‚ä»¥ä¸‹ã€æ–‡å—åˆ—æœ¬ã¨å‘¼ã³ã¾ã™ã€‚ é«˜é€Ÿæ–‡å—åˆ—è§£æžã®ä¸–ç•Œâ€•â€•ãƒ‡ãƒ¼ã‚¿åœ§ç¸®ãƒ»å…¨æ–‡æ¤œç´¢ãƒ»ãƒ†ã‚ã‚¹ãƒˆãƒžã‚¤ãƒ‹ãƒ³ã‚° (ç¢ºçŽ‡ã¨æƒ…å ±ã®ç§‘å¦) ä½œè€…: å²¡é‡ŽåŽŸå¤§è¼”å‡ºç‰ˆç¤¾/ãƒ¡ãƒ¼ã‚«ãƒ¼: å²©æ³¢æ›¸åº—ç™ºå£²æ—¥: 2012/12/27ãƒ¡ãƒ‡ã‚£ã‚¢: å˜è¡Œæœ¬è³¼å…¥: 15äºº ã‚¯ãƒªãƒƒã‚¯: 324å›žã“ã®å•†å“ã‚’å«ã‚€ãƒ–ãƒã‚° (4ä»¶) ã‚’è¦‹ã‚‹ å…¨æ–‡æ¤œç´¢ã¨ã—ã¦ã€ã€ŒCSAã€ã‚„ã€ŒFM-Indexã€ãŒç´¹ä»‹ã•ã‚Œã¦ã„ã¾ã™ãŒã€ã€Œå…¨æ–‡æ¤œç´¢ã‚·ã‚¹ãƒ†ãƒ ã€ã‚’ä½œã‚‹ã«ã¯ã€ã“ã‚Œã‚‰ã ã‘ã§ã¯ä¸ååˆ†ã§ã™ã€‚ãªãœãªã‚‰ã€ä»¥ä¸‹ã®ã‚ˆã†ãªç‰¹å¾´ãŒã‚ã‚‹ã‹ã‚‰ã§ã™ã€‚ æ–‡æ›¸IDã®è˜åˆ¥ãŒé…ã„ã€‚ å„æ–‡æ›¸IDã«å‡ºç¾ã™ã‚‹é »åº¦ã‚’æ±‚ã‚ã‚‹ã®ãŒé…ã„ã€‚ ã¡ãªã¿ã«ã€è»¢ç½®ã‚¤ãƒ³ãƒ‡ãƒƒã‚¯ã‚¹(or N-gramã‚¤ãƒ³ãƒ‡ãƒƒã‚¯ã‚¹)ã‚’ä½¿ã£ãŸå ´åˆã€ã“ã‚Œã‚‰ã®å‡¦ç†ã¯é«˜é€Ÿã§ã™ãã€‚ ã‚¤ãƒ³ãƒ‡ãƒƒã‚¯ã‚¹ã‚’åœ§ç¸®ã—ã¦ã„ã‚‹ã®ã ã‹ã‚‰ã—ã‚‡ã†ãŒãªã„ã¨ã‚‚è€ƒãˆã‚‰ã‚Œã¾ã™ãŒã€ä½œã‚ŠãŸã„ã§ã™ã‚ˆãã‡ã€ã€Œå…¨æ–‡æ¤œç´¢ã‚·ã‚¹ãƒ†ãƒ ã€ã€‚ã“
yass 2013/07/07
nlp

full text search

algorithm
ãƒªãƒ³ã‚¯
é›†åˆã¨ã‹ãƒ™ã‚¯ãƒˆãƒ«ã®é¡žä¼¼åº¦ã®è¨ˆç®—ã®ãƒ¡ãƒ¢ - å”¯ç‰©æ˜¯çœŸ @Scaled_Wurm
Pythonã§å®Ÿè£…ã™ã‚‹é¡žä¼¼åº¦è¨ˆç®— - Screaming Loud #1283. å…±èµ·æ€§ã®è¨ˆç®—æ³• â†‘ã“ã®è¾ºã‚Šã®è¨˜äº‹ã‚’è¦‹ã¦ã€é›†åˆã¨ã‹ãƒ™ã‚¯ãƒˆãƒ«ã®é¡žä¼¼åº¦ã®è¨ˆç®—ã®è¨˜äº‹ã‚’ä¸‹æ›¸ãã®ã¾ã¾æ”¾ç½®ã—ã¦ã„ãŸã“ã¨ã‚’æ€ã„å‡ºã—ãŸã®ã§æ›¸ãä¸Šã’ãŸã€‚ é¡žä¼¼åº¦ã®è¨ˆç®—ã®ã‚³ãƒ¼ãƒ‰ã‚’æ›¸ã„ãŸã®ã§ãã‚Œã‚’è¼‰ã›ã‚‹ã ã‘ã«ã—ã‚ˆã†ã‹ã¨æ€ã£ãŸã®ã§ã™ãŒã€çŸ¥ã£ã¦ã„ã‚‹äººã«ã—ã‹ä¼ã‚ã‚Šãã†ã«ãªã„ã®ã§è‡ªç„¶è¨€èªžå‡¦ç†ã§ã‚ˆãä½¿ã†è©±ã®æ¦‚è¦ã ã‘ã§ã‚‚æ›¸ã„ã¨ãã¾ã™ã€‚ å°Žå…¥ è‡ªç„¶è¨€èªžå‡¦ç†ã®åˆ†é‡Žã§ã¯å˜èªžã®æ„å‘³ã‚’æ¯”è¼ƒã™ã‚‹ã¨ãã«ã€ã‚ã‚‹å˜èªžã®å‘¨ã‚Š(æ–‡è„ˆ)ã«å‡ºã¦ããŸå˜èªžã®ãƒ™ã‚¯ãƒˆãƒ«(æ–‡è„ˆãƒ™ã‚¯ãƒˆãƒ«)ã®é¡žä¼¼åº¦ã‚’è¨ˆç®—ã™ã‚‹ã“ã¨ãŒã‚ã‚‹ã€‚ ã“ã‚Œã¯ã€Œã‚ã‚‹å˜èªžã®æ„å‘³ã¯ãã®å‘¨å›²ã«å‡ºç¾ã™ã‚‹å˜èªžã«ã‚ˆã£ã¦ç‰¹å¾´ã¥ã‘ã‚‰ã‚Œã¦ã„ã‚‹ã€ã¨ã„ã†ä»®èª¬ã«åŸºã¥ã„ã¦ã„ã¦ã€æ–‡è„ˆãƒ™ã‚¯ãƒˆãƒ«ãŒä¼¼ã¦ã„ã‚Œã°ä¼¼ãŸã‚ˆã†ãªæ„å‘³ã€ä¼¼ãŸã‚ˆã†ãªçŠ¶æ³ã§ä½¿ã‚ã‚Œã‚‹å˜èªžãŒå¤šã„ã¨ã„ã†ã“ã¨ãŒè¨€ãˆã‚‹ã‹ã‚‰ã§ã‚ã‚‹ã€‚ Distributional semantics - Wi
yass 2013/05/12
similarity

vector

nlp
ãƒªãƒ³ã‚¯
Google Code Archive - Long-term storage for Google Code Project Hosting.
Code Archive Skip to content Google About Google Privacy Terms
yass 2013/04/29
LSI

dimension

java

nlp

random projection

SVD
ãƒªãƒ³ã‚¯
è‡ªç„¶è¨€èªžå‡¦ç†ã«ãŠã‘ã‚‹é¡žä¼¼åº¦å¦ç¿’(æ©Ÿæ¢°å¦ç¿’ã«ãŠã‘ã‚‹è·é›¢å¦ç¿’)ã«ã¤ã„ã¦ - æ¦è”µé‡Žæ—¥è¨˜
Twitter ã§ã‚°ãƒ©ãƒ•ç†è«–ã«é–¢ã™ã‚‹è©±é¡ŒãŒä¸ŠãŒã£ã¦ã„ãŸã®ã§ã€æœ€è¿‘èª¿ã¹ã¦ã„ã‚‹è·é›¢å¦ç¿’(distance metric learning)ã«ã¤ã„ã¦å°‘ã—ã¾ã¨ã‚ã¦ã¿ã‚‹ã€‚ã‚«ãƒ¼ãƒãƒ«ã¨ã‹è·é›¢(é¡žä¼¼åº¦)ã¨ã‹ã‚’å¦ç¿’ã™ã‚‹ã¨ã„ã†è©±(ã‚«ãƒ¼ãƒãƒ«ã¨ã„ã†ã®ã¯2ç‚¹é–“ã®è¿‘ã•ã‚’æ¸¬ã‚‹é–¢æ•°ã ã¨æ€ã£ã¦ã‚‚ã‚‰ãˆã‚Œã°)ã€‚ ã“ã®åˆ†é‡Žã§ã¯ Liu Yang ã«ã‚ˆã‚‹A comprehensive survey on distance metric learning (2005) ãŒåŒ…æ‹¬çš„ãªã‚µãƒ¼ãƒ™ã‚¤è«–æ–‡ã¨ã—ã¦æœ‰åãªã‚ˆã†ã ãŒã€ãã‚Œã®ã‚¢ãƒƒãƒ—ãƒ‡ãƒ¼ãƒˆ(ã‹ã¤ç°¡ç•¥)ç‰ˆã¨ã—ã¦åŒã˜è‘—è€…ã«ã‚ˆã‚‹An overview of distance metric learning (2007) ãŒå‡ºã¦ã„ã‚‹ã®ã§ã€ãã‚Œã‚’ã•ã‚‰ã«ç°¡ç•¥åŒ–ã—ã¦ãŠå±Šã‘ã™ã‚‹(å…ƒè«–æ–‡è‡ªä½“æœ¬æ–‡ã¯3ãƒšãƒ¼ã‚¸ã—ã‹ãªã„ã—ã€å¼•ç”¨æ–‡çŒ®ã®ã‚ã¨ã«è¡¨ãŒ2ãƒšãƒ¼ã‚¸ã‚ã£ã¦ã€ãã‚Œãžã‚Œç›¸é•ç‚¹ã¨å…±é€šç‚¹ãŒã¾ã¨ã¾ã£ã¦ã„ã‚‹ã®ã§ã€ã“ã‚Œã‚’è¦‹ã‚‹ã¨éž
yass 2013/04/13
machinelearning

nlp

similarity
ãƒªãƒ³ã‚¯
http://blog.yuku-t.com/entry/20110623/1308810518
yass 2013/02/23
LSI

lda

nlp

gensim

dimension
ãƒªãƒ³ã‚¯
Tree Edit Distanceã¨è‡ªç„¶è¨€èªžå‡¦ç†ã¸ã®å¿œç”¨ - Preferred Networks Research & Development
æµ·é‡Žã§ã™ã€‚ã¡ã‚‡ã£ã¨æ™‚é–“ãŒã‚ã„ã¦ã—ã¾ã„ã¾ã—ãŸãŒã€æ˜¨å¹´ã®12æœˆã«é–‹å‚¬ã•ã‚ŒãŸNTCIR-9ã¨ã„ã†ä¼šè°ã®Recognizing Inference in TExt (RITE)ã¨ã„ã†ã‚¿ã‚¹ã‚¯ã«ã€å‰è·ã®æ–¹ã€…ã¨å…±è‘—ã§å‡ºå ´ã—ã¾ã—ãŸã€‚ Syntactic Difference Based Approach for NTCIR-9 RITE Task. Yuta Tsuboi, Hiroshi Kanayama, Masaki Ohno and Yuya Unno. NTCIR-9, 2011. [pdf] å«æ„é–¢ä¿‚èªè˜ã¨ã„ã‚ã‚Œã‚‹ã“ã®ã‚¿ã‚¹ã‚¯ã¯ã€å¤§é›‘æŠŠã«è¨€ã†ã¨ä¸Žãˆã‚‰ã‚ŒãŸ2ã¤ã®æ–‡ãŒåŒã˜æ„å‘³ã®ã“ã¨ã‚’è¨€ã£ã¦ã„ã‚‹ã‹ã©ã†ã‹åˆ¤å®šã—ãªã•ã„ã¨ã„ã†ã‚¿ã‚¹ã‚¯ã§ã™ï¼ˆåŽ³å¯†ã«ã¯ä¸€æ–¹ã‹ã‚‰ã‚‚ã†ä¸€æ–¹ãŒå¸°çµã§ãã‚‹ã‹ã®åˆ¤å®šã§ã™ï¼‰ã€‚ä»Šæ—¥ã¯ã€ãã®ä¸ã§ä½¿ã£ãŸTree Edit Distance (TED) ã«ã¤ã„ã¦è§£èª¬ã—ã¾ã™ã€‚ TEDã¯2ã¤ã®é †åºä»˜ãæœ¨ãŒ
yass 2013/02/23
tree

similarity

nlp
ãƒªãƒ³ã‚¯
ä¹±æŠžã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã¨è‡ªç„¶è¨€èªžå‡¦ç†: Locality Sensitive Hashing ã‚’ç”¨ã„ãŸ é«˜é€Ÿãªåè©žã‚¯ãƒ©ã‚¹ã‚¿ãƒªãƒ³ã‚°
Company Logo ä¹±æŠžã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã¨è‡ªç„¶è¨€èªžå‡¦ç†: Locality Sensitive Hashing ã‚’ç”¨ã„ãŸ é«˜é€Ÿãªåè©žã‚¯ãƒ©ã‚¹ã‚¿ãƒªãƒ³ã‚° Randomized Algorithms and NLP: Using Locality Sensitive Hash Functions for High Speed Noun Clustering D. Ravichandran, P. Pantel, and E. Hovy (ACL 2005) èªã¿â¼¿: å²¡å´Žç›´è¦³ Company Logo ã“ã®è«–æ–‡ã«ã¤ã„ã¦ â¼¿æ³•ã¨ãã®èƒŒæ™¯ç†è«–ãŒã™ã”ãèˆˆå‘³æ·±ã„ â€“ ä»–ã«ã‚‚ã„ã‚ã„ã‚ä½¿ãˆãã† ä¹±æŠžã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã®é›°å›²æ°—ãŒå‘³ã‚ãˆã‚‹ â€“ ãƒ©ãƒ³ãƒ€ãƒ ãŒç¹”ã‚Šãªã™ä¸æ€è°ãªä¸–ç•Œ ã™ãã‚ŒãŸå®Ÿè£…â½…æ³•ã‚’â½°ã—ã¦ã„ã‚‹ â€“ èª°ã§ã‚‚ç°¡å˜ã«å®Ÿè£…ã§ãã‚‹ 2005å¹´ã®ç ”ç©¶ã ãŒï¼Œã„ã¾ã ã«â¾Šè¤ªã›ã¦ã¯ã„ãªã„ â€“ ç”»åƒé¡žä¼¼æ¤œç´¢ã§Lo
yass 2013/02/11
LSH

nlp

similarity
ãƒªãƒ³ã‚¯
Approximate Scalable Bounded Space Sketch for Large Data NLP
1. Approximate Scala ble Bounded Space Sketch for Large Data NLP Amit Goyal and Hal Daume III (EM NLP2011) ç´¹ä»‹è€… : æ¾ç”° 2011/10/13 1 2. Abstract â€¢ Count-Min sketch(ä»¥ä¸‹CM)ã¨ã„ã† sketch-based ã‚¢ãƒ« ã‚´ãƒªã‚ºãƒ ã‚’å˜èªžãƒšã‚¢ã®è¨ˆæ•°ã«ç”¨ã„ã‚‹ â€¢ CMã‚’ç”¨ã„ã‚‹ã“ã¨ã§ã€å˜èªžãƒšã‚¢ãã‚Œè‡ªä½“ã‚’ä¿æŒã™ã‚‹ ã“ã¨ãªãã€å…±èµ·é »åº¦ã®è¿‘ä¼¼å€¤ã‚’(ç©ºé–“,æ™‚é–“ã¨ã‚‚ã«) åŠ¹çŽ‡ã‚ˆãä¿æŒã™ã‚‹ã“ã¨ãŒã§ãã‚‹ â€¢ å…±èµ·é »åº¦ã®è¿‘ä¼¼å€¤ã‚’ä»¥ä¸‹ã®ä¸‰ã¤ã«ã‚¿ã‚¹ã‚¯ã«é©ç”¨ ã—ã€state-of-the-artãªæ‰‹æ³•ã€åŠã³Exact Countã¨æ€§èƒ½ ã‚’æ¯”è¼ƒ â€“ å˜èªžã®æ„Ÿæƒ…æ¥µæ€§åˆ¤åˆ¥ â€“ åˆ†å¸ƒã‚¢ãƒ—ãƒãƒ¼ãƒã«åŸºã¥ãå˜èªžã®é¡žä¼¼æ€§åˆ¤åˆ¥ â€“ æ•™å¸«ãªã—ä¿‚ã‚Šå—ã‘è§£æž â€¢ 49GBã®ãƒ†ã‚ã‚¹ãƒˆã‹ã‚‰ã®å…±èµ·æƒ…å ±ã‚’8
yass 2013/02/10
count

nlp

Count-Min Sketch
ãƒªãƒ³ã‚¯
è¶…å¤§è¦æ¨¡ãƒ†ã‚ã‚¹ãƒˆã«ãŠã‘ã‚‹N-gramçµ±è¨ˆ - Negative/Positive Thinking
ã¯ã˜ã‚ã« è¶…å¤§è¦æ¨¡ãªãƒ†ã‚ã‚¹ãƒˆãƒ‡ãƒ¼ã‚¿ã§ã®N-gramçµ±è¨ˆã‚’å–ã‚‹å ´åˆã€ãã‚‚ãã‚‚ãƒ‡ãƒ¼ã‚¿ãŒãƒ¡ãƒ¢ãƒªã«ã®ã‚‰ãªãã¦SuffixArrayã‚’ä½¿ã£ãŸã‚«ã‚¦ãƒ³ãƒˆã‚‚ç„¡ç†ã ã£ãŸã‚Šã™ã‚‹ã€‚è¿‘ä¼¼å€¤ã§ã‚ˆã„å ´åˆã€åŠ¹çŽ‡çš„ãªæ–¹æ³•ãŒã‚ã‚‹ã¨çŸ¥ã£ãŸã®ã§ã¡ã‚‡ã£ã¨ãƒ¡ãƒ¢ï¼†è©¦ã—ã¦ã¿ãŸã€‚ ä¸Žãˆã‚‰ã‚Œã‚‹ãƒ‡ãƒ¼ã‚¿ å¤§é‡ã®ãƒ‡ãƒ¼ã‚¿ãŒã‚¹ãƒˆãƒªãƒ¼ãƒ å½¢å¼ã§ä¸Žãˆã‚‰ã‚Œã‚‹ã¨ã™ã‚‹ é«˜é€Ÿã«ã©ã‚“ã©ã‚“ä¸Žãˆã‚‰ã‚Œã‚‹ ä¾‹ãˆã°ã€ãƒ‡ãƒ¼ã‚¿ãƒ‘ã‚±ãƒƒãƒˆç›£è¦–ã‚„twitterãªã© ã‚«ã‚¦ãƒ³ãƒˆã—ãŸã„ãƒ‡ãƒ¼ã‚¿ã®ç¨®é¡žãŒè†¨å¤§ ç¨®é¡žã‚’ãƒ¡ãƒ¢ãƒªã«ä¿æŒã™ã‚‹ã®ãŒç„¡ç† ã‚¹ãƒˆãƒªãƒ¼ãƒ ãƒ‡ãƒ¼ã‚¿ã«ãŠã‘ã‚‹é »åº¦ã‚«ã‚¦ãƒ³ãƒˆæ³• æ£ç¢ºãªã‚«ã‚¦ãƒ³ãƒˆã¯é›£ã—ã„ãŒã€è¿‘ä¼¼çš„ã«é »åº¦ã‚«ã‚¦ãƒ³ãƒˆã‚’è¡Œã†ã“ã¨ãŒã§ãã‚‹ Sticky Sampling Algorithmã¯è§£é‡ˆãŒé–“é•ã£ã¦ã„ã‚‹ã‹ã‚‚ã—ã‚Œãªã„ Sticky Sampling Algorithm ã‚«ã‚¦ãƒ³ãƒˆã™ã‚‹è¦ç´ ã‚’ã‚µãƒ³ãƒ—ãƒªãƒ³ã‚°ã§é¸ã¶æ–¹æ³• ä¿æŒã™ã‚‹ã®ã¯ä»¥ä¸‹ã®2ã¤ã®ãƒšã‚¢ã®é›†åˆ e : è¦ç´ (ä¾‹ãˆã°N-gram) f
yass 2013/02/10
count

lossy counting

probabilistic data structure

top-k

n-gram

nlp
ãƒªãƒ³ã‚¯
æ©Ÿæ¢°å¦ç¿’ã¨è‡ªç„¶è¨€èªžå‡¦ç†ã¨ãƒ“ãƒƒã‚°ãƒ‡ãƒ¼ã‚¿ - Preferred Networks Research & Development
å²¡é‡ŽåŽŸã§ã™ã€‚ æƒ…å ±å‡¦ç†å¦ä¼šä¸»å‚¬ã®é€£ç¶šã‚»ãƒŸãƒŠãƒ¼ã€Œãƒ“ãƒƒã‚°ãƒ‡ãƒ¼ã‚¿ã¨ã‚¹ãƒžãƒ¼ãƒˆãªç¤¾ä¼šã€ã§ã®æ©Ÿæ¢°å¦ç¿’ã®å›žã€è‡ªç„¶è¨€èªžå‡¦ç†ã®å›žã§ã®è¬›æ¼”è³‡æ–™ã‚’å…¬é–‹ã—ã¾ã—ãŸã€‚ ä»Šå¹´ã¯ãƒ“ãƒƒã‚°ãƒ‡ãƒ¼ã‚¿ã¨ã„ã†è¨€è‘‰ãŒåºƒã¾ã£ãŸã¨ã„ã†ã“ã¨ã§ã€ã“ã®ãƒ†ãƒ¼ãƒžã§è©±ã™æ©Ÿä¼šãŒå¤šã‹ã£ãŸã§ã™ã€‚ä»Šã¯ãƒ“ãƒƒã‚°ãƒ‡ãƒ¼ã‚¿ã¨ã„ã†ã¨ãã‚Œã‚’æ”¯ãˆã‚‹ã‚¤ãƒ³ãƒ•ãƒ©ã€ã‚¯ãƒ©ã‚¦ãƒ‰ã€DBãªã©ãŒã¾ãšæ³¨ç›®ã•ã‚Œã¦ã„ã¾ã™ãŒã€æˆ‘ã€…ã¨ã—ã¦ã¯å®Ÿéš›ãã‚Œã‚’ä½¿ã£ã¦ä½•ã‚’ã™ã‚‹ã®ã‹ã€ä½•ãŒå®Ÿç¾ã§ãã‚‹ã®ã‹ã¨ã„ã†ã¨ã“ã‚ã‚’æ³¨ç›®ã—ã¦ã„ã¾ã™ã€‚ PFIã¯å…ƒã€…ã“ã†ã—ãŸãƒ‡ãƒ¼ã‚¿ã‚’åˆ†æžã—ã¦ä¾¡å€¤ã‚’æä¾›ã™ã‚‹ï¼ˆæ¤œç´¢ã‚¨ãƒ³ã‚¸ãƒ³ã¨ã‹ã‚‚ãã®ç¯„ç–‡ã«å…¥ã‚‹ã¨æ€ã„ã¾ã™ï¼‰ã“ã¨ã‚’ãšã£ã¨ç¶šã‘ã¦ããŸã‚ã‘ã§ã™ãŒã€ãƒ“ãƒƒã‚°ãƒ‡ãƒ¼ã‚¿ã¨ã„ã†è¨€è‘‰ãŒåºƒã¾ã£ã¦ãã‚ŒãŸãŠã‹ã’ã§ã“ã®è€ƒãˆãŒã‚ˆã‚Šå—ã‘å…¥ã‚Œã‚‰ã‚Œæ§˜ã€…ãªæ¥ç•Œã®æ–¹ã¨éšåˆ†ã¨è©±ãŒã—ã‚„ã™ããªã£ãŸã¨æ€ã„ã¾ã™ã€‚ ä»¥ä¸‹ã®è¬›æ¼”è³‡æ–™ã§ã¯ã€ä»Šãƒ“ãƒƒã‚°ãƒ‡ãƒ¼ã‚¿ã®ä¸ã§ã‚‚æ©Ÿæ¢°å¦ç¿’ã¨è‡ªç„¶è¨€èªžå‡¦ç†ã®åˆ†é‡Žã«ãŠã„ã¦æˆ‘ã€…ãŒã©ã“ã«æ³¨ç›®ã—ã¦ã„ã‚‹ã®ã‹ã‚’è©±ã‚’ã—ã¾ã—ãŸã€‚
yass 2012/12/26
NLP

pfi

bigdata

machinelearning

toread
ãƒªãƒ³ã‚¯
1 2 3 4 5 6 æ¬¡ã®ãƒšãƒ¼ã‚¸