RFQ/BOM 0 Sign In / Register

Select Your Location

Machine reading comprehension: one of the important branches of artificial intelligence technology

July 03, 2021


Machine reading comprehension is one of the recent research hotspots in the field of natural language processing, and it is also a long-term goal of artificial intelligence in the process of processing and understanding human language. Thanks to the development of deep learning technology and large-scale annotated data sets, great progress has been made in using end-to-end neural networks to solve reading comprehension tasks.

Humans can easily read and understand their own mother tongue, but it is difficult for machines to read and understand natural language. If you want a machine to read and understand natural language, you need to convert the natural language into a numerical form that can be used to read, store, and calculate. After several natural languages are converted into numerical values, the machine determines the relationship between them through a series of operations on these numerical values, and then determines that an individual is in the whole according to the mutual relationship between individuals in a complete set ( Complete works).

Machine reading comprehension is a technology that uses algorithms to make computers understand the semantics of texts and answer related questions. Since the articles and questions are in the form of natural language, machine reading comprehension belongs to the category of natural language processing, and it is also one of the latest and hottest topics. In recent years, with the rapid development of machine learning, especially deep learning, the research of machine reading comprehension has made great progress, and it has emerged in practical applications.

With the development of machine reading comprehension technology, reading comprehension tasks are constantly escalating. From the early "cloze form" to the "single document reading comprehension" based on Wikipedia, such as the task of using the SQuAD designed by Stanford University as the data set; and further upgrading to the "multi-document" based on web (web page) data "Reading comprehension", the typical representative of this form is the task of using Microsoft's MS-MARCO and Baidu's DuReader as data sets.

At present, researchers have designed a variety of models for different reading comprehension tasks, and have achieved initial results. However, in the multi-document reading comprehension task, because there are many documents related to the question, there are more ambiguities, which may eventually lead to the wrong answer in the reading comprehension model. Faced with these questions, the human thinking mode usually is: first find multiple candidate answers, compare the content of multiple candidate answers, select the final answer, and find the answer with the highest accuracy.


Early reading comprehension models were mostly based on retrieval technology, that is, searching in articles based on questions and finding relevant sentences as answers. However, information retrieval mainly relies on keyword matching, and in many cases, the answers found solely on the text matching of questions and article fragments are not related to the question. With the development of deep learning, machine reading comprehension has entered the era of neural networks. The advancement of related technologies has greatly improved the efficiency and quality of the model, and the accuracy of the machine reading and comprehension model has been continuously improved.

Although the machine reading comprehension model based on deep learning has different structures, after years of practice and exploration, a stable frame structure has gradually formed. The input of the machine reading comprehension model is articles and questions. Therefore, the two parts must first be digitally coded to become an information unit that can be processed by a computer. In the coding process, the model needs to retain the semantics of the original sentence in the article. We call the coding module in the model the coding layer.

At the coding level, due to the correlation between the article and the question, the model needs to establish a connection between the article and the question. This can be solved by the attention mechanism in natural language processing. In this process, the reading comprehension model combines the semantics of the article and the question for consideration, and further deepens the model's understanding of the two. We call this module the interaction layer.

After the interaction layer, the model establishes a semantic connection between the article and the question, and can predict the answer to the question. The module that completes the prediction function is called the output layer; since there are many types of answers to the machine reading and comprehension task, the specific form of the output layer needs to be related to the answer type of the task. This can be solved through natural language processing technology.

Natural language processing is an important technical cornerstone to realize the vision of machine and human-computer interaction, and machine reading comprehension can be regarded as one of the crown jewels of natural language processing. Machine reading comprehension will allow knowledge acquisition not to be restricted by the human brain; but for the ultimate goal of machine reading comprehension of "understanding and thinking", this is just the beginning of a long march.

Experts believe that end-to-end deep neural networks can better discover some potential features in natural language processing, thereby improving the accuracy of machine reading comprehension. Deeper induction and summary of natural language, knowledge citation, reasoning attribution, knowledge graph and transfer learning will be the future development direction of machine reading comprehension.

As an important branch of artificial intelligence technology, machine reading comprehension will be increasingly used in various industries. As the internationally renowned scholar Professor Zhou Haizhong once predicted: "With the advancement of science and technology, the era of artificial intelligence is coming; by then, artificial intelligence technology will be widely used in various disciplines and will produce unexpected results."