Why should we care about Natural Language Disambiguation?

Natural language processing (NLP) is one of the latest applications of artificial intelligence (AI). NLP allows computers to process and understand the complexities of human language, and derive knowledge from it that can be used for a broad range of tasks.

Quite literally, the ability of computers to "hear" humans and make sense of what they say, NLP is used for translations, transcriptions, speech recognition, sentiment analysis, named entity recognition, pattern analysis, and much more.

This technology is going to become the new market standard soon enough, even more so as the planetary Covid-19 pandemic is accelerating the digital transition. In fact, unlike other industries, more than half of the digital leaders increased their budget by at least 10% compared to 2019.

In the upcoming years, NLP is going to make widespread use of BERT (Bidirectional Encoder Representations from Transformers) and ELMo (Embeddings from Language Models) models, which take full advantage of an extensive training on immense amounts of data. NLP tech is also going to become much more accessible even to less tech-savvy professional now that low-code/no-code tools are becoming commonplace.

Other than being used for practical applications, together with computer vision, this technology is a turning point in providing next-gen intelligent computers with the senses they need to perceive their surroundings and interact with the environment. (Read Computer Vision: Revolutionizing Research in 2020 and Beyond.)

Personal digital assistants (PDA) are probably one of the most immediate use of NLP. Alexa, Cortana, Siri, and OK Google couldn't "talk" or interact with us if they weren't able to understand what we say. A rather complex task, if we keep in mind that a computer cannot grasp all the subtle (but fundamental) non-verbal cues that so heavily characterize human language. Even a slight tilt of an eyebrow or the tone of our voice can be used to convey irony, humor or disappointment — and completely subvert the meaning of a sentence. NLP is therefore critical to make these assistant smarter and more reactive. The introduction of XLM-R and M2M-100 multilingual machine translation models has made NLP even more global instead of having to rely on English data only.

And intelligent virtual assistants are becoming part of a much broader evolution of the marketing industry — a revolution that has already begun and that is changing the way we purchase and consume our products.

NLP is used for topic extraction, relationship extraction, automatic text summarization, and ultimately sentiment analysis. Opinion mining is the latest trend for which NLP is used to monitor social media and obtain real-time insights on what customers think, want, and feel.

All these uses are part of the larger field of behavioral analytics, which allow marketers and corporations to understand our individual needs to customize their offer. Data coming from virtual assistants is integrated with search engine queries, social media interactions, email text, and... well, probably even the conversations recorded by our IoTs.

In a nutshell, the more we talk and write, the better the machines will know us, and the more personalized our products and shopping experience will be.

But shopping is just the tip of the iceberg. NLP can be used to extract information from everything we say and write, and determine, for example, if we're committing a fraud, or perpetrating an act of terrorism that could be therefore stopped or prevented.

NLP is being used to tackle issues such as cyberbullying, reducing offensive or racist language, and automatic detection of fake news. And while this is also opening a new debate on how social media should be regulated, NLP applications are currently being tested “live” on million of users every day.