Deep Learning for Search.pdf下载

2019年 Manning出版, 全英文 ,下面是preface: The field of natural language processing bewitched me as soon as I came to know about it nearly 10 years ago, while studying for my master’s degree. The promise that computers could help us understand the (already, even then) vast amount of textual documents in existence sounded like magic. I still remember how exciting it was to see my first NLP programs extract even vaguely correct and useful information from a few text documents. About the same time, at work, I was asked to do some consulting for a customer on their new open source search architecture. My colleague, who was an expert in the field, was busy on another project, so I was given a copy of Lucene in Action,1 which I studied for a couple of weeks; then I was sent out on the consulting job. A couple of years after I worked on that Lucene/Solr-based project, the new search engine went live (and, as far as I know, it’s still used). I can’t tell you how many times the search engine algorithms needed to be adjusted because of this or that query or this or that fragment of indexed text, but we made it work. I could see users’ queries, and I could see the data that was there to be retrieved, but a minimal difference in spelling or omitting a certain word could cause very relevant information to not show up in the search results. So while I was very proud of my work, I kept wondering how I could have done better to avoid the many manual interventions the product managers asked me to perform in order to provide the best possible user experience. Right after this, I quite by chance found myself involved in machine learning thanks to Andrew Ng’s first machine learning online class (which originated the Coursera MOOC series). I was so fascinated with the concepts behind the neural networks shown in the class that I decided to try to implement a small library for neural networks in Java myself, just for fun (http://svn.apache.org/repos/asf/labs/yay/). I started hunting for other online courses like Andrej Karpathy’s course on convolutional neural networks for visual recognition and Richard Socher’s course on deep neural networks for natural language processing. Since then, I have kept working on search engines, natural language processing, and deep learning, mostly in open source. A couple of years ago (!), Manning reached out to me to review a book on NLP, and I was naive enough to write at the bottom of my review that I would be interested in writing a book on search engines and neural networks. When Manning came back to me, expressing interest, I was kind of surprised, and wondered, do I really want to write a book on that? I realized that, yes, I was interested. While deep learning has revolutionized computer vision and natural language processing, there’s still a lot to uncover for its applications in search. I’m sure we can’t (yet?) rely on deep learning to automatically set up and tune search engines on our behalf, but it can help a lot in making the search engine user’s experience smoother. With deep learning, we can do things in search engines that we can’t do with other existing techniques so far, and we can use deep learning to enhance the techniques we already use in search engines. The journey toward making search engines more effective through deep neural networks has just started. I hope you enjoy it.