Natural Language Processing

By letting the computer read and understand texts, and then produce text from a given text, you open up the possibility of chatting with the computer. This is called a chatbot. Chatbots are most often used for automatic responses from customer service and similar tasks where there is a desire to have questions answered. This can reduce the cost of having customer service staff to answer FAQs.

Many texts have an overall purpose or topic. By predicting a topic, these documents can be placed in the same category. This deals with use cases such as classification of genre, classification of news articles and similar elements with text. Such topics are often used in connection with recommendation systems that can automatically suggest products of the same topic.

Clustering is a technique which allows you to categorise data without having defined categories in order to sort and create coherence in a large amount of data. This can be used to learn about the underlying, often unknown, clusters in data. This can also be used advantageously for text. The clustering of text data can identify trends and topics in data that are previously unknown. This can be anything from documents to feedback and comments or results from questionnaires or similar surveys.

By teaching the computer to read, it also enables the computer to skim through a text and produce a compressed version of it. This might be a brief summary of an online article, an automatic summary of documents or as a tool to screen the applications of potential candidates responding to a job ad.

A relatively simple NLP tool is subjective analysis (sentiment analysis), where an attempt is made to rank a given text based on the positive or negative charge of the text.

This may seem superfluous at first, but in many cases, it has a huge, positive impact on a system. Systems with built-in sentiment analysis can be used to rank important emails, bad/good reviews, detect spam or malicious messages, both as email and on social media. The technique can also be used to turn a subjective sentence into an objective, measurable unit, e.g. for satisfaction analysis.

It is normal for text data to contain a lot of filler words, so the sentence makes sense when addressed to other people. Computers, on the other hand, do not need sentences to make sense in the way we humans do. Keyword spotting/extraction is a technique whereby important or relevant keywords are extracted from texts. This technique can be used to extract names from documents, figures from invoices or adjectives and nouns from texts.