Ever since the advent of Machines, humans had to interact with them to make them perform the assigned tasks. What started with pulling of some levers or humans bending over conveyor belts, has reached a place where we talk to machines in our own natural languages, and they understand! Not only in just communication, but in many other language processing scenarios, machines are taking over attracting the attention of Data Scientists the world over to the persistently unfolding field of Natural Language Processing (NLP). In this post we shall look at some of the state-of-the-art NLP techniques and their applications.
NLP may be interpreted as the exercise of imparting to machines the ability to comprehend Human Languages. In other words, NLP is the field of Artificial Intelligence that enables the machines to read or listen to human languages, understand them and even respond in the same. The two main components of NLP are Natural Language Understanding (NLU) and Natural Language Generation (NLG), the input and output of these could be speech or written text. In NLU, computer software is used to understand input sentences, as speech or text, enabling direct human-computer interaction in natural languages, without needing the formal syntaxes of computer languages. NLG, as the name suggests is the process of generating plain-language output from the input data. That is, given some data, the machine can write a story from it. The use cases of both NLU and NLG, in general NLP, are on the rise, as we shall see later in this post, with many enterprises turning to these technologies to engage with their customers.
Before diving into the applications of NLP, let’s have a quick glance at its history. Though the origin can be traced as far back as 1950s when Alan Turing, the father of Computer Science, came up with his famous Turing Test, notable success came to the fore only by mid 1960s with the contrivance named ELIZA, a simulated psychotherapist, which could sustain a reasonable human-like conversation. Up till the late 1980s, which saw the dawn of the so-called “statistical revolution” era, there was a slump in NLP with programmers working on some rule-based systems which were rarely robust, and the pace picked up dramatically with the introduction of Machine Learning Algorithms in NLP.
The year 2018 has been the most phenomenal so far in terms of NLP development with the entrance of the mighty BERT (Bidirectional Encoder Representations from Transformers) published by Google AI Language team, followed forthwith by a challenger, the OpenAI GTP-2 model, which was not released publicly for fear of misuse, like using it for continuously generating fake news, as the ‘deepfake’ text generated by this model is too good to the level where it cannot be differentiated from those written by humans. Prior to these heavyweights, the most sought after was the Word2Vec originally published in 2013, among other smart schemes like ELMo, ULMFiT, Semi-supervised Sequence Learning, to name a few.
The applications of NLP are briskly on the rise of which some are:
Healthcare sector has major potential for NLP applications and some are already in use, of which one of the most important is disease prediction using electronic clinical records and speech recordings of the patient, which can make predictions on conditions related to cardiovascular system, schizophrenia, depression among others. Amazon Comprehend Medical is one service that uses NLP to extract disease conditions from clinical reports.
This needs no further explanation as they are almost ubiquitous mostly as Amazon’s Alexa, Google Assistant or Apple’s Siri. These are intelligent voice driven interfaces responding to vocal commands aiding one to do tasks as rudimentary as setting the alarm timer to purchasing things online, or even control the appliances at home.
A lot of insights into customer choices and preferences can be obtained using sentiment analysis, by identifying and extracting information from social media platforms. This allows organizations not only to realize what all are being said about their products and services, but also to identify the aspects that drive a customer’s choices and decisions.
It is the process of automatic conversion of one natural language to another, preserving the sense and fluency. The quality of machine translation has improved drastically recently allowing it to spread to more languages.
This is the process of automatically extracting information about a particular topic from multiple texts to produce summaries, so that a user will have the summary enriched by multiple sources, without having to go through each individually.
These are only a few of the applications of NLP as an exhaustive list is impossible due to the sheer enormity of the fields in which it is being implemented. The Deep Learning team of Curvelogics, the parent establishment of Data Science Academy has been into NLP and has produced some exciting results in Automated Text Redaction, Information Extraction, Question-Answer Bots, Sentiment Analysis and so on. The horizon of NLP has suddenly been widened with the arrival of big shots like BERT and GTP-2, which are undeniably the next big things. With NLP getting refined to the level where machines communicate using natural languages with superior fluency, it’s just a matter of time before the traditional ways of communicating with the machines become obsolete.