I am going to listen to Laura Martín-Pérez, computational linguist at Dail Software, and I am going to see the movie ‘The Arrival’. Laura says that in this science fiction film the linguistic relativity or Sapir-Whorf hypothesis (Edward Sapir and Benjamin Whorf) is reflected, which shuffles the idea that our language determines our way of thinking. That is, there is a relationship between grammatical categories and the way in which people understand and conceptualize the world around us. Thus, the world will not be conceptualized the same by a Chinese as by a Francophone. Or by a Spaniard.
At one point in the film, Laura continues, it is observed how the linguist tries to decipher the language of aliens, which is basically the same thing that she (Laura) does in the company where she works: “I teach the machine to understand the language that must be processed / understood for different purposes ”. And this is what we call Natural Language Processing (NLP).
The PLN has existed since the 50s, but the change is that it is in a few years when solutions of machine learning and deep learning, machine and deep learning (ML and DL), which provide are used over much more accurate understanding of language. Natural language processing techniques allow the analysis of an interaction, the tone of a conversation, or the monitoring of social media, for example.
No one doubts that language is a capacity that only humans have, and that ultimately it also expresses the level of intelligence of each person. But through artificial intelligence, and more specifically from the PLN, linguistic capabilities are already being provided to machines, opening a new world that today, experts say, we are not able to imagine. Siri, Google Assistant or Amazon Echo are just the first of their applications, and the beginning of a journey to a place that no analyst or expert can predict.
Machines need to understand the complexities of language and how humans communicate to make use of it. Advances in sentiment analysis, question answering, and multi-tasking joint learning are making it possible for AI to truly understand human beings and the way we communicate. And, in addition, in each language the PLN is different. Any development with PLN and AI carried out in English, for example, is useless for Spanish.
Salesforce, which is one of the companies that have most developed solutions and tools based on artificial intelligence and natural language processing, uses its ‘Einstein AI’ platform so that, for example, brands can obtain real-time analysis of the sentiment in emails, social media, and chat text, in order to provide better experiences for those brands’ customers. Accurate sentiment analysis enables service agents to know which customers are dissatisfied, or to identify deficiencies in a product, measure overall satisfaction, and even monitor brand perception through social media. However, this type of intelligent system also requires context, so that it can differentiate between different meanings. And this is where the PLN has one of its greatest challenges, to build models that can learn from the context and share knowledge even while performing many tasks at the same time.
Dail Software, where Laura Martín-Pérez works, also develops artificial intelligence solutions using natural language processing, among other techniques. Such as Full Talk, a rich conversational platform or chatbot to improve customer service and automate business processes. Or Agora, a suite of AI solutions for the analysis and 360-degree management of online interactions. Or, for example multilingual email, an email service that allows interaction between people of different languages.
And we still haven’t talked about what artificial intelligence and the PLN can do for journalism, information and online media to help automate certain processes. We will do so in future articles, and we will see examples that already exist in, above all, the United States.
