Snorkel
The Snorkel team is now focusing their efforts on Snorkel Flow, an end-to-end AI application development platform based on the core ideas behind Snorkel—check it out here! The Snorkel project started at Stanford in 2016 with a simple technical bet: that it would increasingly be the training data, not the models, algorithms, or infrastructure, that decided whether a machine learning project succeed
LLM時代ã®Xæƒ…å ±åŽé›†è¡“｜ã¹ã„ãˆã‚Šã‚
AI for Everyoneã«ã¤ã„ã¦ã¯æ—¥æœ¬èªžç‰ˆã‚‚ã‚ã‚‹ã®ã¨ã€ã©ã¡ã‚‰ã®ã‚³ãƒ¼ã‚¹ã‚‚日本語å—幕付ãã§è¦‹ã‚‰ã‚Œã‚‹ï¼ˆå¤šåˆ†æ©Ÿæ¢°ç¿»è¨³ã§ã®è‹±èªžå—幕ã‹ã‚‰ã®ç¿»è¨³ã ãŒã€ç¿»è¨³ã®è³ªã¯æ‚ªããªã„）ã®ã§ã€è‹±èªžåˆ†ã‹ã‚‰ãªãã¦ã‚る程度何ã¨ã‹ãªã‚‹ã‚“ã˜ã‚ƒãªã„ã‹ã¨æ€ã„ã¾ã™ã€‚ ã‚ã¨ã€ä½™åŠ›ã®ã‚る人ã€æœ€æ–°ã®NLPç ”ç©¶ã‚’ç†è§£ã—ãŸã„人ã¯ã“ã¡ã‚‰ã®æœ¬ã‚’èªã‚€ã“ã¨ã‚’オススメã—ã¾ã™ã€‚アルゴリズムã®è©³ç´°ã¯å¿…ãšã—ã‚‚ç†è§£ã—ãªãã¦ã‚‚良ã„ã§ã™ãŒã€ã©ã‚“ãªã‚¿ã‚¹ã‚¯ãŒã‚ã‚‹ã®ã‹ã¯ç†è§£ã—ã¦ãŠã„ãŸæ–¹ãŒè‰¯ã„ã‹ã¨æ€ã„ã¾ã™ã€‚ NLPã®çŸ¥è˜ãŒLLMを応用ã™ã‚‹ä¸Šã§å®Ÿéš›ã«ã©ã†å½¹ã«ç«‹ã¤ã‹ã§ã™ãŒã€ä¾‹ãˆã°ã§è¨€ã†ã¨NLPçš„ã«ã¯å¯¾è©±ã®ä¸ã«ã¯ã€Œã‚¿ã‚¹ã‚¯æŒ‡å‘型対話（task-oriented dialogue）ã€ã¨ã€Œé›‘談（chit-chat dialogue）ã€ãŒã‚ã‚Šã¾ã™ã€‚ãã‚Œãžã‚Œå¯¾è©±ã®ä¸ã§é‡è¦è¦–ã•ã‚Œã‚‹ã‚‚ã®ã‹ã‚‰è©•ä¾¡ã®ä»•æ–¹ã¾ã§å…¨ç„¶é•ã†ã®ã§ã™ãŒã€NLPをやらãšã«LLMã‚’ã‚„ã£ã¦ã„る人ã¨è©±ã—ã¦ã„ã‚‹ã¨ã“れらをã”ã£ã¡ã‚ƒ
Speech and Language Processing
Speech and Language Processing (3rd ed. draft) Dan Jurafsky and James H. Martin Here's our August 20, 2024 release! Individual chapters and updated slides are below; Here is a single pdf of Aug 20, 2024 book! Feel free to use the draft chapters and slides in your classes, print it out, whatever, the resulting feedback we get from you makes the book better! Typos and comments are very welcome (just
作ã£ã¦ç†è§£ã™ã‚‹ Transformer / Attention - Qiita
ã“ã‚“ã«ã¡ã¯ã€‚ミクシィ AI ãƒãƒœãƒƒãƒˆäº‹æ¥éƒ¨ã§ã—ゃã¹ã‚‹ãƒãƒœãƒƒãƒˆã‚’作ã£ã¦ã„るインコã§ã™ã€‚ ã“ã®è¨˜äº‹ã¯ ミクシィグループ Advent Calendar 2018 ã®5日目ã®è¨˜äº‹ã§ã™ã€‚ ã“ã®è¨˜äº‹ã®ç›®çš„ ã“ã®è¨˜äº‹ã§ã¯2018å¹´ç¾åœ¨ DeepLearning ã«ãŠã‘る自然言語処ç†ã®ãƒ‡ãƒ•ã‚¡ã‚¯ãƒˆã‚¹ã‚¿ãƒ³ãƒ€ãƒ¼ãƒ‰ã¨ãªã‚Šã¤ã¤ã‚ã‚‹ Transformer を作るã“ã¨ã§ã€ Attention ベースã®ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã‚’ç†è§£ã™ã‚‹ã“ã¨ã‚’目的ã¨ã—ã¾ã™ã€‚ 機械翻訳ãªã©ã® Transformer, 自然言語ç†è§£ã® BERT ã‚„ãã®ä»–多ãã®ç¾åœ¨ SoTA ã¨ãªã£ã¦ã„る自然言語処ç†ã®ãƒ¢ãƒ‡ãƒ«ã¯ Attention ベースã®ãƒ¢ãƒ‡ãƒ«ã§ã™ã€‚ Attention ã‚’ç†è§£ã™ã‚‹ã“ã¨ãŒä»Šå¾Œã®è‡ªç„¶è¨€èªžå‡¦ç† x Deep Learning ã®å¿…é ˆã«ãªã£ã¦ãã‚‹ã®ã§ã¯ãªã„ã§ã—ょã†ã‹ã€‚ ã“ã®è¨˜äº‹ã«ã‹ã‹ã‚Œã¦ã„るコード (Github) æ´å² - RNN ã‹ã‚‰
最先端自然言語処ç†ãƒ©ã‚¤ãƒ–ラリã®æœ€é©ãªé¸æŠžã¨æœ‰ç”¨ãªåˆ©ç”¨æ–¹æ³• / pycon-jp-2020
spaCyを使ã£ãŸå…ˆé€²çš„ãªè‡ªç„¶è¨€èªžå‡¦ç† · ç„¡æ–™ã®ã‚ªãƒ³ãƒ©ã‚¤ãƒ³ã‚³ãƒ¼ã‚¹
ã“ã®ã‚³ãƒ¼ã‚¹ã«ã¤ã„ã¦spaCyã¯ç”£æ¥å¿œç”¨å‘ãã®è‡ªç„¶è¨€èªžå‡¦ç†ç”¨Pythonライブラリã§ã™ã€‚ã“ã®ç„¡æ–™ã®ã‚ªãƒ³ãƒ©ã‚¤ãƒ³ã‚³ãƒ¼ã‚¹ã§ã¯ã€ãƒ«ãƒ¼ãƒ«ãƒ™ãƒ¼ã‚¹ã¨æ©Ÿæ¢°å¦ç¿’を用ã„ãŸå…ˆé€²çš„ãªè‡ªç„¶è¨€èªžå‡¦ç†ã‚·ã‚¹ãƒ†ãƒ ã‚’spaCyã§ä½œã‚‹æ–¹æ³•ã‚’インタラクティブã«å¦ã¶ã“ã¨ãŒã§ãã¾ã™ã€‚ ç§ã«ã¤ã„ã¦ç§ã¯spaCyã®ã‚³ã‚¢é–‹ç™ºè€…ã§ã€Explosionã®å…±åŒå‰µæ¥è€…ã®ä¸€äººã®Inesã§ã™ã€‚AIや機械å¦ç¿’ã€è‡ªç„¶è¨€èªžå‡¦ç†ã®æœ€æ–°ã®é–‹ç™ºãƒ„ールを専門ã¨ã—ã¦ãŠã‚Šã€Web関連ã®ã‚‚ã®ã‚’作るã®ã‚‚大好ãã§ã™ã€‚ spaCyウェブサイトソースファイルInesã®Twitter
Topic Modeling in Python with Gensim
Topic Modeling is a technique to extract the hidden topics from large volumes of text. Latent Dirichlet Allocation(LDA) is a popular algorithm for topic modeling with excellent implementations in the Python’s Gensim package. The challenge, however, is how to extract good quality of topics that are clear, segregated and meaningful. This depends heavily on the quality of text preprocessing and the s
「ã‚ã¾ã‚Šã«ã‚‚å±é™ºéŽãŽã‚‹ã€ã¨å±é™ºè¦–ã•ã‚ŒãŸæ–‡ç« 生æˆãƒ„ール「GPT-2ã€ã®æŠ€è¡“ã§ç”»åƒã‚’自動ã§ç”Ÿæˆã™ã‚‹ã“ã¨ã«æˆåŠŸ
äººå·¥çŸ¥èƒ½ã‚’ç ”ç©¶ã™ã‚‹éžå–¶åˆ©çµ„ç¹”ã®OpenAIãŒé–‹ç™ºã—ãŸAIã«ã‚ˆã‚‹æ–‡ç« 生æˆãƒ„ール「GPT-2ã€ã¯é«˜ç²¾åº¦ã®æ–‡ç« ãŒè‡ªå‹•ç”Ÿæˆã§ãã‚‹ãŸã‚ã€é–‹ç™ºé™£ãŒã€Œã‚ã¾ã‚Šã«ã‚‚å±é™ºéŽãŽã‚‹ã€ã¨å±æƒ§ã—ã¦è«–文公開を延期ã™ã‚‹äº‹æ…‹ã«ç™ºå±•ã—ã¾ã—ãŸã€‚ã“ã®GPT-2ã®ãƒˆãƒ¬ãƒ¼ãƒ‹ãƒ³ã‚°ã«ä½¿ç”¨ã•ã‚Œã‚‹ã‚¢ãƒ¼ã‚テクãƒãƒ£ã‚’直接é©ç”¨ã—ã€è‡ªå‹•ã§ç”»åƒã‚’生æˆã™ã‚‹æŠ€è¡“ãŒé–‹ç™ºã•ã‚Œã¦ã„ã¾ã™ã€‚ Image GPT https://openai.com/blog/image-gpt/ ã‚ã‚‹é ˜åŸŸã§å¦ç¿’ã•ã›ãŸãƒ¢ãƒ‡ãƒ«ã‚’ã€ã»ã‹ã®é ˜åŸŸã«è»¢ç”¨ã™ã‚‹æŠ€è¡“を転移å¦ç¿’(トランスファーラーニング)ã¨å‘¼ã³ã¾ã™ãŒã€GPT-2ã¯ã“ã®è»¢ç§»å¦ç¿’モデルã«ã‚ˆã£ã¦å¤§ããªæˆåŠŸã‚’å¾—ã¾ã—ãŸã€‚人間ãŒé–¢ã‚らãªã„教師ãªã—å¦ç¿’ã¯ã€GPT-2ã®ä»–ã«ã‚‚ã€Googleã®ã€ŒBERTã€ã‚„Facebookã®ã€ŒRoBERTaã€ãªã©ã§ã€ç›®è¦šã¾ã—ã„進æ©ã‚’é‚ã’ã¦ã„ã¾ã™ã€‚一方ã€è‡ªç„¶è¨€èªžã®åˆ†é‡Žã§ã¯è»¢ç§»å¦ç¿’モデルã®æˆåŠŸã¯ç›®è¦šã¾ã—ã„ã‚‚ã®ã®ã€ã“ã‚Œã¾
Recipes for building an open-domain chatbot
Building open-domain chatbots is a challenging area for machine learning research. While prior work has shown that scaling neural models in the number of parameters and the size of the data they are trained on gives improved results, we show that other ingredients are important for a high-performing chatbot. Good conversation requires a number of skills that an expert conversationalist blends in a
1
リリースã€éšœå®³æƒ…å ±ãªã©ã®ã‚µãƒ¼ãƒ“スã®ãŠçŸ¥ã‚‰ã›
最新ã®äººæ°—エントリーã®é…ä¿¡
j次ã®ãƒ–ックマーク
kå‰ã®ãƒ–ックマーク
lã‚ã¨ã§èªã‚€
eコメント一覧を開ã
oページを開ã
GitHub - mamorlis/nlpbook: 「自然言語処ç†ã®æ•™ç§‘書ã€ã‚µãƒãƒ¼ãƒˆã‚µã‚¤ãƒˆ
http://arxiv.org/pdf/2005.14165
Hugging Face
Recipes
GitHub - yoheikikuta/bert-japanese: BERT with SentencePiece for Japanese text.
{{#tags}}- {{label}}
{{/tags}}