The Most Common and Interesting Applications of Natural Language Processing (NLP)
NLP (natural language processing) is a field of artificial intelligence where language, computer science, and machine learning come together. You could say this is one of the most challenging fields in machine learning. Text analysis and understanding of languages is not as easy as working with numbers and formulae are because of the simple reason that there really is no proper formula for expressions of thoughts, ideas, feelings, and opinions.
What is the goal of NLP?
Natural language processing is used in many fields and there are many different applications that result from the growth of NLP. We don’t need NLP to help humans who speak the same language when they can both speak and hear and are communicating in person or over telephone. Humans are smart, heck, humans are the smartest species we know as of now (which can be shocking and sad for some of us 😉). As long as we can udnerstand and communicate in a language, we can udnerstand so many things including the direct meaning, hidden meanings, and even sarcasm without any of us having to put in a lot of effort to help us understand these.
But what about other cases? That’s where natural language processing comes in to the rescue, to help us understand and communicate just as effectively when we don’t have the knowledge of language or the ability or the time needed. In this post, let us look at some of the most interesting applications of NLP.
Natural Language Understanding:
Helping the Differently-abled:
Speech-to-Text: In this process, the computer takes in the audio as input, works on changing the format from audio to text, and as we all know, computers understand only 0s and 1s, so the text would be broken down into 0s and 1s, and the output would again be in the form of text.
For those who can’t hear or are hard of hearing but can read well and have good eyesight, speech-to-text transformation can be of great help. Using this technique, they can use agents to help them translate speech to text, so that when someone asks them or tells them something, they can see the same content in the form of text, and they can choose to respond either by text or they can share their response by speech.
Text-to-Speech: This is the reverse process of speech-to-text conversion. For those who cannot see or read and can hear, this technique helps convert text messages or text content into audio content for them to read. Following this, the user can express their response in the form of voice or speech, and using the speech-to-text technique, their responses would be converted to text, which can be sent or shared by email or other applications.
Evidently, it’s not just those who are differently-abled who could benefit from these applications. Anyone who needs to use these apps just to make life easier or to multi-task can also benefit from these.
For example, devices that have the ability to read out text can help readers to listen to books in the form of audio (take PDF readers and Kindle for example), and applications like Google Assistant can take in voice input and share output with you in the form of either voice or text.
Helps break the Language Barrier:
Translation From One Language to Another:
This kind of natural language processing applications can be of great help because it helps break the barriers of language among people who speak and understand different languages. This also helps us really access content in other/foreign languages because having access but not knowing the language is as good as not having access to the content. This way, we wouldn’t need to pay someone to translate it for us either.
This application is extremely useful for students, professionals, business people, volunteers, travelers, and such. You can imagine.
But translation is still a pretty tricky business! Understanding context is one thing we human beings can naturally pick up, but that’s quite hard to get a machine to learn. For instance, see below:
Google Translate gets “merry Christmas” right in Spanish. It also correctly translates “I love nature” in Tamil. But it fails funnily when it translates “I am a queer ally.” Here, it understands the word “queer” to mean “strange” or “odd” and misses the context of gender! 😂 You get the point, the easier and more common a sentence is, Google can translate it easily. The rarer or more complicated the translation is, Google can still fail.
Scans Documents to Read Content:
Optical Character Recognition (OCR):
Well, AI can also scan images or documents and read the text. This is done using a technique called “optical character recognition”. Advanced, well-trained AIs can read content in several tens of languages and in many different fonts.
Tesseract, for example, can read documents in over 100 languages. On the other hand, Firebase ML, also a product of Google can have difficulties with some languages. But which of these two would be the best choice for OCR? That is something that you can’t say; you have to experiment and find out as it can vary on a case-to-case basis.
Languages supported in different versions of Tesseract
Tesseract documentation
tesseract-ocr.github.io
Natural Language Generation:
Helps Save Time:
Beyond natural language understanding, there is also natural language generation, where AI generates its own text for users. ChatGPT is the latest text generation tool that boasts of great advancements in NLP.
Text Summarization:
If you ask Google assistant some questions, you’d get responses in which Google assistant would quote answers that the AI thinks matches closely with yours form its own search results. That is, the AI would only read for you what is already written by someone else. But text summarization is where an AI would summarize texts for you. This could save a lot of time for people who don’t want to or don’t have the time to read several pages. Depending on how you program the AI, it can summarize articles or pages from a book or chapters of books or even entire books. But you should be wise enough to know that you’d not want an entire book summarized in a few paragraphs. Even if humans try to do the same, it can still fail in capturing the essence of the book.
Text summarization is just one of the applications of natural language generation. But a more common form of text generation that we are all familiar with would be suggestions. When you type something on search bars, it is the AI in the back-end that gives you the most relevant suggestions to choose from instead of having to type it all out.
As of now, the most happening and interesting natural language processing application that has left hundreds of thousands of people surprised is ChatGPT by Open AI. This is a language model which uses reinforcement learning, one that can answer some really complicated questions! But again, this is still getting developed. (We will see about ChatGPT in a different post.)
As we can see, tools that use natural language processing techniques help the differently-abled, help save time for those in a hurry, help students and professionals with text editing and text summarization, help travelers and others with translation, and even help all of us learn a lot more in short periods of time. Since language is everywhere, there would be no “intelligence” in “artificial intelligence” without the use of NLP. This is a field that is continuously growing and has a lot of potential.
Which applications of NLP are your favourites?
