Reading Comprehension Systems
While browsing articles related to Machine Learning, I have found something quite interesting from the Computer Science Department of Stanford University called "Evaluating Reading Comprehension Systems". The following information compiled in this post is information taken from reliable sources which I will provide links to if you are interested in reading the full version. This post is simply meant to explain what "Reading Comprehension Systems" are in a simple matter.
What is Reading Comprehension Systems?
Example: (Source: https://nlp.stanford.edu/pubs/jia2017adversarial.pdf)
When you give the system this Article
The system should give out the following response
Question: “What is the name of the quarterback who was 38 in Super Bowl XXXIII?”
Original Prediction: John Elway
What the system just did, Was produce a question by understanding the given Paragraph w/ a predicted answer to the question it just produced.
Reading comprehension systems has started since the 1990s, For this, to work it has become vital to get the information it needs in a short amount of time. Thanks to the development of Natural Language Processing and AI, this has lead to the research of what we call today, Machine reading comprehension (RC).
"Reading comprehension" is the ability to read the text, process it, and understand the meaning. To understand the text, the reader must recognize every word and retrieve its meaning while combining this information with the syntactic knowledge to create meaningful sentences.
Examples
The document found here can better explain how this process works with graphical examples, most of the information written below was taken from the article, Source: https://nlp.stanford.edu/pubs/jia2017adversarial.pdf
The example shown above is an example from the Stanford Question Answering Dataset (SQuAD). “Our method tests whether systems can answer questions about paragraphs that contain adversarially inserted sentences, which are automatically generated to distract computer systems without changing the correct answer or misleading humans.” using this example, The BiDAF Ensemble model gets the answer correct, but it was mistaken with a distracting sentence that was marked in blue (See example, above)
If you'd like to learn how this example was produced, You can find this very illustration and information in the source document I provided above, Since explaining this will require explaining a lot about of the methods they used that can easily be found with more details in the provided document
Conclusion
Even though I have just now discovered about RC, and after researching about it for quite a while, I have come to the conclusion that this is a pretty interesting use of machine learning and AI And might experiment with this to create a program that can produce questions and learn from answers using past data, I'd like to see other projects utilizing this system to create some pretty cool concepts that could be used in the real world, Perhaps the same method but in reverse?
Disclaimer: I am in no way trying to steal or take credit for the content provided here, I have provided the source from where I got the information from, This post is meant to explain what RC is a simple matter. Information MAY BE INCORRECT from typos or misunderstandings, if you spot any issues with this post, Please contact me or leave a comment so I can make the correct changes. This will allow me to better understand from my errors and make the post more accurate for everyone else seeking for information like this, Thanks.
What is Reading Comprehension Systems?
Reading Comprehension, also known as RC are systems designed to understand a given text and return answers as a response to questions about the text.
Example: (Source: https://nlp.stanford.edu/pubs/jia2017adversarial.pdf)
When you give the system this Article
Article: Super Bowl 50
Paragraph: “Peyton Manning became the first quarterback ever to lead two different teams to multiple Super Bowls. He is also the oldest quarterback ever to play in a Super Bowl at age 39. The past record was held by John Elway, who led the Broncos to victory in Super Bowl XXXIII at age 38 and is currently Denver’s Executive Vice President of Football Operations and General Manager. Quarterback Jeff Dean had jersey number 37 in Champ Bowl XXXIV.”
The system should give out the following response
Question: “What is the name of the quarterback who was 38 in Super Bowl XXXIII?”
Original Prediction: John Elway
What the system just did, Was produce a question by understanding the given Paragraph w/ a predicted answer to the question it just produced.
Reading comprehension systems has started since the 1990s, For this, to work it has become vital to get the information it needs in a short amount of time. Thanks to the development of Natural Language Processing and AI, this has lead to the research of what we call today, Machine reading comprehension (RC).
"Reading comprehension" is the ability to read the text, process it, and understand the meaning. To understand the text, the reader must recognize every word and retrieve its meaning while combining this information with the syntactic knowledge to create meaningful sentences.
Examples
The document found here can better explain how this process works with graphical examples, most of the information written below was taken from the article, Source: https://nlp.stanford.edu/pubs/jia2017adversarial.pdf
The example shown above is an example from the Stanford Question Answering Dataset (SQuAD). “Our method tests whether systems can answer questions about paragraphs that contain adversarially inserted sentences, which are automatically generated to distract computer systems without changing the correct answer or misleading humans.” using this example, The BiDAF Ensemble model gets the answer correct, but it was mistaken with a distracting sentence that was marked in blue (See example, above)
 |
| Figure 2: An illustration of the AddSent and AddAny adversaries (From source) |
If you'd like to learn how this example was produced, You can find this very illustration and information in the source document I provided above, Since explaining this will require explaining a lot about of the methods they used that can easily be found with more details in the provided document
Conclusion
Even though I have just now discovered about RC, and after researching about it for quite a while, I have come to the conclusion that this is a pretty interesting use of machine learning and AI And might experiment with this to create a program that can produce questions and learn from answers using past data, I'd like to see other projects utilizing this system to create some pretty cool concepts that could be used in the real world, Perhaps the same method but in reverse?
Disclaimer: I am in no way trying to steal or take credit for the content provided here, I have provided the source from where I got the information from, This post is meant to explain what RC is a simple matter. Information MAY BE INCORRECT from typos or misunderstandings, if you spot any issues with this post, Please contact me or leave a comment so I can make the correct changes. This will allow me to better understand from my errors and make the post more accurate for everyone else seeking for information like this, Thanks.
Comments
Post a Comment