Natural language processing (NLP) may include all language, it is simply getting a computer to understand and talk natural languages. These are also known as natural interfaces being the kind of interface humans use when communicating. Therefore, a computer can use a natural interface .
In this video, through Google duplex one can speak through a mobile phone and instruct it to perform specific tasks such as inserting an event in a calendar, booking a restaurant, set a timer ect. and it is carried out. It is worth noting that the restaurant may not even need to be on the web. The system calls the restaurant and makes a reservation .
Google Duplex: https://www.youtube.com/watch?v=-qCanuYrR0g
Source Alexei Dingli
An important sub-task of natural language processing is NLP machine translation which aims to get a piece of text in one language and translate it into another in a different way. This application was crucial in the sixties due to the cold war. The Americans had a lot of Russian documents requiring fast translation into English, and they did so using parallel text .
Another example of parallel text is the Bible, since it was translated into several languages, there are indeed 670 translations with each sentence in the Bible being numbered, so you can compare one sentence with another. However, when using literal translation, they ran into some problems, for instance, sentence 1438 of the Bible reads “the spirit is willing, but the flesh is weak” when given to the computer system, the result was “The vodka is good, but the meat is rotten” .
As a result, people lost interest in machine translation, as they thought it was not possible back then, so the research was abandoned – this is also referred to as the first AI winter .
However, today machine translation has advanced significantly, and when using Google translate, the text is easily translated from one language to another albeit with a couple of errors that may need tweaking. Natural language processing is extremely important when dealing with the processing of large amounts of text .
The following are some key examples:
- An example is parliamentary questions, generating large amounts of documents that need to be seen to by members of parliament on a daily basis. However, to carry out parliamentary research, AI can help - a system was tested in Japan, whereby they used AI to generate automatic replies for queries. Therefore, when a member of the public, submitted a query to a member of parliament, it is sent to the system to generate the reply which is then sent to the member of parliament for vetting, before sending it to the member of the public.
- Another use is for the generation of news, Bloomberg and the associated press are using AI to mine data, create text from datasets and draft pieces of information of circa 2000 words. The stories are submitted online with all the facts provided by AI.
- A similar system was developed by the University of Malta, known as the Adaptable Storybook. If a child is reading a book and is finding it difficult to read that particular page, AI may realise that the child is experiencing some difficulty, and when the child turns to the next page, the text on the next page is adjusted for the child to continue reading. Therefore, AI is adjusting the text to the child’s capabilities.
- A subtask of NLP is question answering, setting the task of submitting a question to an AI system and receiving the correct answer from that system. And in fact, the Government of North Carolina uses such a system to answer questions put forward by citizens. So when a citizen calls the call centre, the call is transferred to a chatbot. The chatbot analyses the question sent and provides a reply. The chatbot may reply with confidence, and if the reply is below what is expected then the query is submitted to a human operator. This allows the human operator to deal with the most complex of queries. So the AI system is capable of answering up to 90% of all queries. It's interesting to note that IBM Watson learnt over time and studied over 3000 documents about 16 city services and today he can answer up to 10,000 questions.
- NLP may be used to create intelligent chatbots, at this link you may find the first robot lawyer https://donotpay.com/. So if one receives a fine, the robot lawyer at this website will help you create and submit a petition to the petition board, and may possibly get waivered. To date, the robot lawyer has carried out 250,000 cases winning 160,000 cases, having a success rate of 64%, saving people up to $4 million in parking fines[1] .
- The same chatbot was now helps refugees seeking asylum in the US to apply for asylum status, the application is filled out and informs them whether they are eligible for protection or not[2] . The chatbot has the added value of enabling applicants to populate the required fields and translate the applications making it easier for applicants to fill in.
- For the legal profession, AI is also being used to scan advanced legal documents. By doing so AI can identify related case laws enabling lawyers to save time and focus when defending a particular case.
The following are some definitions:
Chatbots: NLP is used to analyse the input language of the customers to the chatbots. Based on the analysis, they provide appropriate responses to the customers or reroute the query to the relevant pages.
Talent Acquisition: sector in which accuracy related to the talent pool is an essential factor.
NLP in big data: means analysing data and NLP helps harness the pattern of such huge unstructured data.Source: https://mytechdecisions.com/it-infrastructure/choosing-natural-language-processing-business/