This stuff explains how Mirai Translate is going differentiate the company from the others.
Please take a look.
Imagine a world in which people of different languages could communicate freely with each other, just as if they were speaking their own native language.
If language barriers are eliminated,businesses would accelerate globally and innovatively.
For example, everyday business processes could be streamlined and costs reduced, leading to higher growth.
Enhanced cross-border cooperation among companies could give birth to unprecedented business ideas and models.
We believe our “artificial intelligence approach to translation” will create a new standard in machine translation.
What could we do for your business?
This document discusses various topics in natural language processing including autocomplete, stemming, lemmatization, spelling correction, BLEU score, morphological analysis, transliteration, sentiment analysis, summarization, and confidence scores. It explains that autocomplete and auto-suggestion relate words based on prefix matching and popularity or occurrence ranking rather than being a simple function. It also discusses techniques for text normalization, stemming versus lemmatization, language modeling for spelling correction, n-gram modeling, and the use of edit distance and BLEU scores for evaluation.
This stuff explains how Mirai Translate is going differentiate the company from the others.
Please take a look.
Imagine a world in which people of different languages could communicate freely with each other, just as if they were speaking their own native language.
If language barriers are eliminated,businesses would accelerate globally and innovatively.
For example, everyday business processes could be streamlined and costs reduced, leading to higher growth.
Enhanced cross-border cooperation among companies could give birth to unprecedented business ideas and models.
We believe our “artificial intelligence approach to translation” will create a new standard in machine translation.
What could we do for your business?
This document discusses various topics in natural language processing including autocomplete, stemming, lemmatization, spelling correction, BLEU score, morphological analysis, transliteration, sentiment analysis, summarization, and confidence scores. It explains that autocomplete and auto-suggestion relate words based on prefix matching and popularity or occurrence ranking rather than being a simple function. It also discusses techniques for text normalization, stemming versus lemmatization, language modeling for spelling correction, n-gram modeling, and the use of edit distance and BLEU scores for evaluation.
EMNLP2014読み会 "Efficient Non-parametric Estimation of Multiple Embeddings per ...Yuya Unno
1. The document presents the Multi Sense Skip-gram (MSSG) model for learning multiple embeddings per word in vector space.
2. MSSG assigns a separate embedding to each sense of a word using a context vector. It extends the Skip-gram model by learning sense-specific embeddings.
3. The Non-Parametric MSSG (NP-MSSG) model extends MSSG by using a non-parametric approach to learn the context vectors instead of fixed vectors, allowing an unbounded number of senses per word.
Revised presentation slide for NLP-DL, 2016/6/22.
Recent Progress (from 2014) in Recurrent Neural Networks and Natural Language Processing.
Profile http://www.cl.ecei.tohoku.ac.jp/~sosuke.k/
Japanese ver. https://www.slideshare.net/hytae/rnn-63761483
Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
http://arxiv.org/abs/1609.08144
を読んでみたので、簡単にまとめました。間違い等は是非ご指摘ください。
Automatic Selection of Predicates for Common Sense Knowledge Expression長岡技術科学大学 自然言語処理研究室
Ai Makabi, Hiroshi Matsumoto and Kazuhide Yamamoto. Automatic Selection of Predicates for Common Sense Knowledge Expression. Proceedings of the Conference of the Pacific Association for Computational Linguistics (PACLING 2013), no page numbers (2013.9)
EMNLP2014読み会 "Efficient Non-parametric Estimation of Multiple Embeddings per ...Yuya Unno
1. The document presents the Multi Sense Skip-gram (MSSG) model for learning multiple embeddings per word in vector space.
2. MSSG assigns a separate embedding to each sense of a word using a context vector. It extends the Skip-gram model by learning sense-specific embeddings.
3. The Non-Parametric MSSG (NP-MSSG) model extends MSSG by using a non-parametric approach to learn the context vectors instead of fixed vectors, allowing an unbounded number of senses per word.
Revised presentation slide for NLP-DL, 2016/6/22.
Recent Progress (from 2014) in Recurrent Neural Networks and Natural Language Processing.
Profile http://www.cl.ecei.tohoku.ac.jp/~sosuke.k/
Japanese ver. https://www.slideshare.net/hytae/rnn-63761483
Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
http://arxiv.org/abs/1609.08144
を読んでみたので、簡単にまとめました。間違い等は是非ご指摘ください。
Automatic Selection of Predicates for Common Sense Knowledge Expression長岡技術科学大学 自然言語処理研究室
Ai Makabi, Hiroshi Matsumoto and Kazuhide Yamamoto. Automatic Selection of Predicates for Common Sense Knowledge Expression. Proceedings of the Conference of the Pacific Association for Computational Linguistics (PACLING 2013), no page numbers (2013.9)
Developing User-friendly and Customizable Text Analyzer長岡技術科学大学 自然言語処理研究室
Yuki Miyanishi and Kazuhide Yamamoto. Developing User-friendly and Customizable Text Analyzer. The International Conference on Practical Linguistics of Japanese (ICPLJ8), pp.172-173 (2014.3)
3. 関連研究
• Class-Dependent Modeling for Dialog
Translation [Finch et al., 2009]
• 入力からモダリティ(疑問)を推測
• 複数の翻訳モデルを用意し、入力のモダリティを考慮した翻訳
• Discriminative Reranking for SMT using Various
Global Features [Goh et al., 2010]
• N-best リランキングのために否定文や疑問文の情報を使用
否定・疑問の特徴語を考慮することで、
モダリティを保存した翻訳ができるか。
5. 人手による特徴語
not t
don Don
haven isn
No won
wasn doesn
didn cannot
hadn
? Why
Will What
Could Is
How Does
Can Do
Are Which
When Where
Have Does
Did Was
May
疑問 否定
語の位置情報を保存するため、大文字と小文字は区別しない
3
9. LLRによる特徴語例
can yet
any but
know worry
I anything
it so
afraid understand
what enough
do any
there have
this don
long it
isn did
your much
how time
疑問 否定
6
14. 翻訳例
入力 サーカスと動物園、どっちに行こうか。
ベースライン Let s go to the circus and, the zoo? (☓)
日英人手
Which one shall we go to the circus and
zoo? (○)
๏ 正しく翻訳できた例
入力 年に一度昇給を得る資格があります。
LLR30
Do you have any qualifications do you
get a raise once a year. (☓)
๏ 特徴語の抽出失敗による翻訳失敗
10
15. 翻訳例
入力 やさしく打ってくださいね。
人手(英語) Please go easy, isn t it? (☓)
人手(両方) Please go easy. (○)
๏ 英語のみの特徴語で翻訳に失敗した例
両言語の特徴語により選択するフレーズを制限している
๏ 英語のみの特徴語で翻訳に失敗した例
入力 キャンセルしてもかまいませんか。
ベースライン May I cancel? (○)
人手(両方) I don t mind if you cancel it? (☓)
人手特徴語でも常に否定、疑問のモダリティを表すとは限らない。
11