Natural Language Processing

Scenario

After midterm elections have passed in the United States of America, politicians are already examining ways to appeal to the people. It is possible to gather this information by hiring large groups of interviewers, but it is a costly and timely process. Natural Language Processing (NLP) is a methodology used to analyze what people believe and support about politicians, especially on social media, and this is an effective way to sort through millions of comments. This is how politicians adapt their campaigns to gain more votes from citizens. Not only is NLP used in politics, but it is used in everyday life ranging from emails to educational systems. Although NL

What is it?

Natural Language Processing is the research on how natural language, or human language, can be interpreted and manipulated by computers. The purpose of this research is to make the interactions between humans and computers as easy and efficient as possible. The basis of NLP is not just limited to the field of computer science but even involves other fields such as mathematics, psychology, and linguistics. NLP is also used in a variety of fields, such as speech recognition, artificial intelligence, machine translation, user interfaces, and translation between human languages [1]. Early research in NLP focused on analyzing the structures of language and then developing technologies centered around it, such as speech recognition, speech synthesis, and machine translation. More recently, researchers have been focusing more on refining those technologies. These refinements include speech-to-speech translation engines and spoken dialogue systems [2].

How does it work?

There are many methods that NLP uses. One simple method for text uses part-of-speech sequences to decipher what the text is saying by figuring out whether each word is a noun, verb, adjective, and so on. This method also figures out names and dates, what certain prepositions refer to, and the general dependencies of each word [3]. Another method is called the User Preference Graph, which is what the autofill function on many phones is. This method looks at what the user types the most in the context of their sentences and then creates a flow chart of where possible sentences might lead. An example would be texting your friend your location and your phone offering the options of “home” or “work” when you type “I’m at” into your message. If NLP is used in the context of translating between languages, one method might be the Phrase-Based Machine Translation. This method is created by a computer learning from large, unstructured sets of texts that already have translations. When this method is being used, it breaks the original sentence into chunks. It then translates each chunk into every possible translation, and then combines these translations into all possible sentences. Finally, it chooses the best sentence based on the examples it had previously processed. There is also the processing of speech. Acoustic modeling is a method that creates phonemes, which are the individual sounds in a word, and then compares them to what the user says [2]. While these are still very advanced techniques and tend to work well, many people still find problems with these methods. For example, many people find the User Preference Graph method to be inaccurate to what they say in everyday conversations. Others have difficulty with the Acoustic Modeling method for a multitude of reasons. Due to the relatively small sample size the researchers test in, voice recognition softwares often have issues when figuring out what someone says if they have an accent or any sort of unusual speech patterns, such as stuttering.

What its significance?

Natural language processing is something everyone uses on a daily basis. It’s used in Google searches and autocorrect, which is something everyone utilizes. Even if you type in something wrong in a search bar, it’s corrected using NLP. [4] The computer is able to guess what it is you’re wanted based on millions of different similar searches. It’s also used a lot in social media, both for creators and just users in general. Creators can use it to see which posts do better than others, and it’s used for others to give them content similar to everything else they view. [4] Companies can use it to filter applications and look for certain keywords or requirements, instead of skimming through thousands of applications. [4]

Natural language processing fills in the gaps between human language and computer language. It makes it possible for computers to understand what we mean, and is a fantastic tool for us. It can collect much more data in a much faster and orderly way than humans ever could. [5] With that data, it can analyze it and decide what’s important, and filter it through in an unbiased way. [5] NLP is also a growing field. With the outlook of technology, and the thought that we’re heading in a more technologically based world, computers being able to understand human language is a big step. [5] It’s necessary if we’re to reach the future we dream of.

What are the Advantages and downsides?

Although NLP has numerous advantages, it has limitations because it must learn the writing patterns of humans. These downsides consist of domain specific languages, sarcasm, and ambiguity. For the effective use of NLP, it has to understand the domain-specifics of what it is analyzing. For example, schools teach different topics and classes, so NLP systems have to comprehend the different subjects students are learning to offer feedback on what level of understanding the students have for a class [6]. It becomes a time-consuming process for a computer to learn the subjects, books, and expected comprehension levels of students because large sums of data have to be taken into consideration such as topics being well-covered in class, the difficulty of a subject, and differentiating one subjects feedback from another for the same student. In an opinion analysis of students’ opinions on educational infrastructure, a research gap was evident as the NLP system did not understand the sarcasm detected in a student’s post. However, statistical classifiers were embedded into the NLP to detect sarcasm, punctuation, unigrams, word embedding similarity, and sentiment flips. For deep learning algorithms, recurrent neural networks (RNNs), long short-term memory (LSTM), and convolutional neural networks (CNNs) were used to detect sarcasm [6]. Sarcasm comes in different forms and emphasizes certain words to make the sarcasm obvious. Since the NLP systems do not easily recognize sarcasm without additional models, it creates a large gap of data. This creates inaccurate data and can skew the results and solutions for feedback, which creates setbacks for educational systems and grades. The ambiguity of a word is common in natural languages, but an NLP system will overcomplicate it. It may not recognize structural, syntactic, or lexical natures of a sentence, and it requires models such as BERT and WordNet 3.0 to distinguish these types of ambiguities. Scientists took positive and negative sentiments and examined them through these models, and the BERT model outperformed any other existing model with an F1-score of 77% [6]. Overall, NLP is inaccurate when examining large groups of data that can have different connotations, and it is an expensive process to use the models to decode the meaning and data behind comments and opinions. Even when the NLP’s models comprehend comments and give feedback, there is still a 23% of the data and feedback being incorrect and misleading for companies and schools to implement necessary actions for positive or negative feedback. [6]

What are the implications for higher education?

NLP is being used frequently in everyday life. It holds a place in medical fields, educational systems, and socialization among people. Computer scientists are combining deep learning and NLP to discuss with humans or even outcompete them in intelligence. As previously discussed, NLP is used for analyzing feedback for students, so the future of NLP consists of providing more accurate data and suggestions for students to better understand material [6]. After Covid-19 NLP should be more prevalent in schools to determine if students are improving or declining in test scores and what the education system needs to do to help students. On the other hand, NLP is now considered in the medical field, and doctors are hoping that NLP will correctly diagnose a patient based on other articles and similar symptoms posted online. NLP will be able to prescribe a patient with a drug depending on what their symptoms relate to, and it will differentiate between structured and unstructured data to provide accurate diagnoses [7]. Hospitals and healthcare networks like Covera Health are hiring AI scientists for deep learning and natural language processing. Their goal with NLP is to eliminate misdiagnoses and improve the methods for assessing and predicting the quality care provided by health workers [8]. NLP will be a breakthrough for patients who have been misdiagnosed by doctors, and it will aid medical businesses in diagnosing rare diseases based on what an NLP system’s feedback offers. Although medical and education systems plan to implement more NLP, it will take some time for NLP systems to understand sarcasm, ambiguity, and large specific domains, or it will provide inaccurate data that can harm people. Since doctors are planning to analyze structured and unstructured data with NLP, it will require additional finances to add models that can dictate whether a source or domain of information is accurate and up to date.

Natural Language Processing is frequently used in education systems, and teachers are improving the qualities of NLP to provide more accurate feedback of students’ grades. NLP can be applied to academic writing, assessments, and formatting exams. Using online resources, teachers gain data on how well a student is understanding a topic, and the teachers assist students to better comprehend information. NLP gathers information from websites and online databases that generates information for students to learn depending on the level of understanding they are at in the class. Before learning objectives are made for students, NLP will analyze the text and determine if it is appropriate for a certain grade level to learn. When students write responses, NLP will examine mistakes in their writing and show teachers the errors students made [9]. This will make the grading process much faster and easier for teachers, especially when they have large groups of students’ work to grade. It will also assist teachers in giving improved feedback as they will not be too exhausted to notice simple mistakes in other students’ writings. It is evident that NLP will enhance its suggestions to students and teachers in the future, so it will provide effective and efficient responses to difficult prompts and subjects. For example, NLP is improving to match human dialog for students to grasp a concept. A study was conducted between human and computer tutors, and it found that students learned more with a human tutor because they had unconstrained natural language dialogues with students. Teachers are trying to construct this unconstrained discussion into computers, so students will better understand the material and feel like they are talking to another person [10]. Although NLP is not advanced enough to be used as another person to talk to students about subjects, it is being used every day to provide feedback for grammatical errors. Once NLP technology improves, grading time for teachers will decrease and test scores will most likely improve as well.

Natural language processing is a vital key to the future of technology. It will help to make using computers and other technologies easier for most people. There are many methods of NLP, but all of them have the basis of analyzing human language and putting it in terms the computer knows. Once the computer understands the original message, it can do almost anything to it, whether it translates it, sorts it, or applies it to other applications. There are many kinds of people that use NLP: doctors, politicians, businesses, and even the average person. Political figures can use it to decipher how they’re being perceived by their audience, and then use that information to appeal to them and try to attract more people to their campaign. Teachers can use it to grade papers, analyze students’ scores, and learn how to improve students’ understanding of topics. Doctors use it to analyze patients’ symptoms and give them swift and accurate diagnoses. It is used every day when people text their friends or ask Siri or Alexa a question. NLP is significant because it offers adaptable feedback on large scales, offers smart responses, can easily find trends in large groups of data, and can even change how people respond in their everyday lives. There are disadvantages to NLP, such as its difficulty in understanding domain-specific language, sarcasm, and ambiguity. It also takes time for computers to learn all the necessary information for each specific topic. NLP progresses further every day, and it seems that its focus is now on providing more accurate data and suggestions to users. Education systems are also planning to use NLP more for academic writing, assessments, and formatting exams. It will also be used by teachers for things like grading exams and analyzing test scores. Overall, Natural Language Processing is a useful tool that continues to grow to accommodate more people in more fields.

References

[1] G. G. Chowdhury, “Natural language processing,” Strathprints, 31-Jan-2003. [Online]. Available: https://strathprints.strath.ac.uk/2611/. [Accessed: 14-Nov-2022].

[2] V. Shah, A. Jain, and G. Kulkarni, “Natural language processing - ecole Enti Confindustriali Lombardi per l ...,” ResearchGate, Jan-2018. [Online]. Available: https://www.myecole.it/biblio/wp-content/uploads/2020/11/7semt_2018-Natural_Language_Processing.pdf. [Accessed: 14-Nov-2022].

[3] J. Hirschberg and C. Manning, “Advances in natural language processing | science,” Science, 17-Jul-2015. [Online]. Available: https://www.science.org/doi/10.1126/science.aaa8685. [Accessed: 14-Nov-2022].

[4] A. Sharma, “Applications of natural language processing (NLP),” Analytics Vidhya, 23-Jul-2021. [Online]. Available: https://www.analyticsvidhya.com/blog/2020/07/top-10-applications-of-natural-language-processing-nlp/. [Accessed: 19-Nov-2022].

[5] “Natural language processing (NLP): What it is and why it matters,” SAS. [Online]. Available: https://www.sas.com/en_us/insights/analytics/what-is-natural-language-processing-nlp.html#:~:text=NLP%20is%20important%20because%20it,speech%20recognition%20or%20text%20analytics. [Accessed: 19-Nov-2022].

[6] T. Shaik, X. Tao, Y. Li, C. Dann, J. McDonald, P. Redmond, and L. Galligan, “A review of the trends and challenges in adopting natural language processing methods for education feedback analysis,” IEEE Access, vol. 10, pp. 56720–56739, May 2022 [Online]. Available: https://ieeexplore.ieee.org/document/9781308/authors#authors. [Accessed: 13-Nov-2022]

[7] P. Johri, S. K. Khatri, A. T. Al-Taani, M. Sabharwal, S. Suvanov, and A. Kumar, “Natural language processing: History, evolution, application, and future work,” pp. 365–375, Jan. 2021 [Online]. Available: file:///C:/Users/Brealynn%20Harper/Downloads/NaturalLanguageProcessing-paper-Al-Taani-2021.pdf. [Accessed: 13-Nov-2022]

[8] “AI scientist, Deep Learning for Natural Language Processing,” 2021. [Online]. Available: https://ai-jobs.net/job/5273-ai-scientist-deep-learning-for-natural-language-processing/. [Accessed: 14-Nov-2022]

[9] K. M. Alhawiti, “Natural language processing and its use in education,” International Journal of Advanced Computer Science and Applications, vol. 5, no. 12, pp. 72–76, 2014 [Online]. Available: https://thesai.org/Downloads/Volume5No12/Paper_10-Natural_Language_Processing.pdf. [Accessed: 13-Nov-2022].

[10] D. Litman, “Natural Language Processing for Enhancing Teaching and Learning”, AAAI, vol. 30, no. 1, Mar. 2016. Available: file:///C:/Users/Brealynn%20Harper/Downloads/12310-56401-1-PB.pdf. [Accessed 13-Nov-2022].