Wenxin, who has reached the top of the glue list, is teaching again. One stop teaching is to understand information extraction

Time:2021-4-20

Recently, the authoritative ranking of natural language processing field — glue (general language understanding assessment benchmark) has been released. Ernie, a semantic understanding technology and platform developed by Baidu, topped the list again with a score of 90.9, leading Microsoft’s deberta / Turing nlrv4, Google’s T5 and other similar technologies developed by Alibaba and Huawei.

In the process of technology industry landing, Wenxin is also constantly making efforts to transform the latest world-class technological breakthroughs into easy-to-use product tools. Together with easydl platform, Wenxin provides a set of simple and efficient NLP development capabilities, effectively improving the efficiency of model development and application.

For the needs of user text information extraction, Wenxin is online“Text entity extraction”And“Text entity relation extraction”It supports extracting specific entities from massive information sources and defining the relationship between entities. For example, the information of the insured and the insured period in the insurance contract is extracted, and the relationship between them is established to assist the insurance manager to analyze and judge.

How to label data quickly?

How to achieve high precision model effect?

How to construct knowledge map from simple information extraction?

Can zero depth learning experience be used easily?

To solve the above problems, Baidu easydl-nlp special live class was held on March 25“From technical analysis to actual combat drill, text information extraction model”Attack, from data processing to model training, the whole process of 0 code visualization operation, teach you to quickly master the “information extraction” skills.

See the highlights of the course first!

Entity extraction and entity relationship extraction, efficient acquisition of knowledge

“Text entity extraction”, as the core task of text mining and information extraction, supports the extraction of specific fact information from massive information sources, and is an important basis for information retrieval, intelligent question answering, intelligent dialogue and other artificial intelligence applications; “text entity relationship extraction” can extract not only the predefined entity types, but also the relationship types between entities, and get the content Entity relation triples of semantic information can be used to construct and expand knowledge map. For example: “Wang Xuechun is the dubber of Qingwen in the 87 edition of a dream of Red Mansions.” We can extract the relationship of “Wang Xuechun – dubbing – dream of Red Mansions”.

Wenxin, who has reached the top of the glue list, is teaching again. One stop teaching is to understand information extraction

Online intelligent labeling, cost saving

In order to improve the ease of use of this ability, Wenxin also released a data annotation tool based on two tasks to solve the problem of data preparation, which supports marking directly in the text, bringing excellent annotation experience and higher annotation efficiency to the taggers. As shown in the figure below, we can mark the target information directly through visual operation, extract the enterprise entity and registered capital in financial contracts, and directly establish the relationship between them.

Wenxin, who has reached the top of the glue list, is teaching again. One stop teaching is to understand information extraction

In addition to the introduction of the above new functions, the course will also lead you through the whole process of practical operation, from understanding the principle to customizing the model, from following the operation to independent implementation, so as to quickly open up the two channels of text information mining.

Implementation of Baidu engineer hand in hand teaching case

In this open class, baidu engineer will practice the whole process of creating model, preparing data, training model, verifying model and publishing model, and lead you to customize an entity relationship extraction model based on sample data in three steps. What are you waiting for? Sign up quickly, let’s witness the magic charm of text intelligence!

time: 20:00-21:00, March 25

Registration method: scan the QR code in the poster, add a little assistant, wechat, notes“NLP”, get the exclusive registration channel. Participation in the course, andIqiyi VIP gold monthly card and Baidu brain customized MugWait for the surprise gift!

More “Course Introduction” and “course highlights” are shown in the figure below!

Wenxin, who has reached the top of the glue list, is teaching again. One stop teaching is to understand information extraction