Reading Club · Beijing Stage 1 deepqa framework and Siri architecture venue notes (PPT download)

ArticleDirectory

• Reader Conference · Workshop notes (including PPT download) of deepqa framework and Siri architecture in Beijing 1st)
• Preface
• I. Basic deepqa framework of Zhang Dong
• Ii. Basic architecture of Siri by Zhang junlin
• Download PPT
• Postscript

Reader Conference · Workshop notes (including PPT download) of deepqa framework and Siri architecture in Beijing 1st)

Preface

The reader will"Academic innovation, strong conduct, joint discussion, and mutual learning"For its purpose, unlimited themes and forms, it focuses on creating or creating a good academic atmosphere, a relaxed atmosphere for discussion, and a persistent enthusiasm for continuous in-depth research. In addition, the general official activities on the market are basically flashy, and there are few truly valuable substantive activities. I hope that the readers can take into account both.

The first phase of the reading fair was successfully held in Tsinghua Park yesterday. I hope it will continue later. (You are welcome to contact me if you want to share your speech,Especially those who have just made a new book.). Next, let's give a very brief description of the holding of the reading conference yesterday afternoon.(If you are not familiar with the reading club, please refer to my previous article:Http://blog.csdn.net/v_july_v/article/details/7237351).

First of all, I would like to thank all my friends who attended the reading club yesterday at Tsinghua fit building. I was surprised by the enthusiasm of everyone. (According to the statistics on the attendance list, the number of attendees is over 160, and the number of attendees will be controlled in the 2nd and later stages to increase the venue, although nearly half of our friends have no seats, they have no sense of confusion. In general, it is not bad. The figure shows my host (Thank you @ Jie _ Fu for providing the photo ):

Okay, let's talk about it in a few minutes. Let's take a look at the 1st topics of yesterday's Book Club: Zhang Dong's deepqa and 2nd topics: siri of Zhang junlin (the first topic in my speech was canceled due to insufficient time. At the end of this article, the download part of the PPT contains two topics ).

I. Basic deepqa framework of Zhang Dong

Before that, I must explain what deepqa is. Many friends, like me, may not know what deepqa is before going to the discussion yesterday. In fact, deepqa is very simple. Deep is the meaning of depth. Qa is Q (Question) & (Answer), One question and one answer.

To put it simply, deepqa is an artificial intelligence solution provided by IBM. In many Q & A competitions, deepqa won the competition, tried to understand human problems, and found answers from the existing knowledge base, score among the many answers to find the best answer.

I believe that our friends who are quick-thinking will answer this question? This is not the same as the search engine.

The search engine provides an answer based on the query results (of course, the answer does not satisfy you, or whether you can find the right answer is another thing ). As shown in (Dr. Zhang's PPT is based on the following three figures ):

How search engine solve Q &?

This deepqa is a bit like a real Technical Support/customer service system, and its Automatic Speech service can solve a considerable number of problems, however, the goal of deepqa is not only to answer certain simple questions, but to utilize computing capabilities to integrate information retrieval, natural language processing, Knowledge Representation and reasoning, machine learning, human-computer interaction, and other related technologies, analyze existing data to find the best answer.

So how deepqasolve Q &? As shown in:

The original PPT text is a bit problematic. I corrected it as follows: deepqa = massively parallel probabilistic evidence-based architecture generates and scores using hypotheses using a combination of 1000's of natural language processing, information retrieval, machine learning and reasoning algorithms, these gather, evaluate, weigh and balance different types.

The deep Q & A system establishes a large-scale parallel architecture based on Probabilistic evidence. It comprehensively utilizes thousands of natural language processing, information retrieval, machine learning, and reasoning.Algorithm, Analysis generates a variety of assumptions, and then collects, evaluates, and balances all kinds of evidence, and finally gives the best answer. As shown in:

For the deepqa architecture diagram above, you can see the Chinese version provided by the Blue violin at a glance, as shown below:

Because Zhang Dong focuses more on the discussion with you on the site, the core details are not too much (I hope that the lectures will be further discussed in the future ).

Subsequent discussions on deepqa

In the official exchange group of the Beijing clubCyqcyq raised a series of questions: 1. What is evidence? What is hypotheses? 2. Based on which algorithm can the donor | action | the receiver act as an evises or a hypotheses? 3. What algorithms are used to screen out a large number of actors, actions, and combinations of recipients? 4. Who determines the donor | action | the recipient is evidence? 5. How can we minimize this amount of judgment with manual intervention ?..

For the Question Answering process of Watson, refer to this article from the author's violin:Http://www.guokr.com/watson/article/6643,Its summary is as follows: "See Watson in jeopardy! Are you curious about how Watson understands and answers to the amazing performance of the preliminaries? Can Watson really think about problems like humans? Now let's take the first question he answered as an example. Let's explain the whole process of Watson seeing the question and giving the answer ...".

Ii. Basic architecture of Siri by Zhang junlin

Mr. Zhang junlin has 43 PPT pages, which are basically a huge volume of details. He talks about Siri's overall architecture, its input system, and the topics from text to intent and from intent to service, finally, I talked about the output system and how Siri went.

Of course, the lecture still does not involve more technical details and principles such as semantic analysis.

Download PPT

In the first phase of yesterday's reading conference, we downloaded the deepqa PPT of 1st theme Zhang Dong. Download the basic architecture PPT of Siri of 2nd theme Zhang junlin :.Updated: All original micro-disk download links are invalid,Correct packaging, unified PPT, and point-free:Http://download.csdn.net/detail/v_july_v/4061067. 02.13.17: 20 points.

An article written by Zuo Qisheng, a friend of the attendees: some notes about a machine learning speech on Tsinghua campus. Other participants are welcome to continue to share their notes or photos. I will post them one by one, so that more friends can understand them.

Postscript

Because I was not allowed yesterday, I canceled my original plan.3rd topics: Basic algorithm learning. If you are interested in robots, you can follow @ monkey robot on Sina Weibo. Finally, I would like to thank all the friends who attended the reading conference, Zhang dangru from the two lecturers, Tsinghua tianlei from the venue, those who registered for volunteer work, and Yang from Turing education.

Here are the photos of the site provided by Chen jizu. Thank you very much (sorted by the order of appearance ):

For more information, please follow our microblog @ Zhou Lei July. Let's see you next time (welcome to the audience and register as the Speaker, especially those who have recently published a new book. Welcome to some free 100 + venues ). If you have any questions, feel free to contact me, such as Zhou Lei July and July February 13, 2012.

Updated: I am eager to build a website for the reading club for publishing, registration, and communication. My idea is that one or two people who know the frontend can get a simple prototype in a week, and then take another two or three months to build a new prototype. I am not eager to seek success, but I want to get it out first. It is convenient for you and me to communicate with the readers of the Reading Club. It is also useful for the readers. Hope you can understand the frontend contact me.

Finally, the author of the first lecture of the Reading Club preliminarily confirmed: the latest news ,@Liu weipeng pongbaWith Xiang Liang @ xlvector_hulu (author of Recommendation System Practice), it is initially set as the Speaker and sharer for the middle April issue of the Book Fair to be held on April 9, 2nd. @ Liang bin Penny may also be present for help, I believe that compared with issue 1st, it is wonderful and exciting to expect ,:-).2012.03.04.