By clicking “Accept”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.

NLP: Talk to an AI!

April 24, 2020

Newsletter April 2020

We're talking more and more with robots. Whether it is with conversational agents such as Google Home or Alexa, or with chatbots that pop up from the homepages of merchant sites, we interact with them more and more fluidly. Behind these exchanges, which aim to reproduce human conversation as faithfully as possible, lies a form of application of artificial intelligence: natural language processing (NLP). It is in short everything that allows the chatbot to read, decipher, understand and give meaning to human language in order to respond as pertinently and as "humanely" as possible. It is a technology that is also found in translation applications, word processor correctors, or Interactive Voice Responses (IVR) that call centers use to automatically process requests.

Unprecedented progress

This type of interaction between man and machine is not entirely new:

  • Turing's test, based on a machine's ability to mimic human conversation, dates from 1950,
  • the first chatbot, Eliza, dates back to the mid-1960s,
  • Those over 30 probably still remember Clippy, the personal assistant for Microsoft Office.

But in recent years, relying on the power of deep learning and the development of artificial intelligence, chatbots have become increasingly autonomous and efficient, able to adapt to the infinite nuances and possibilities opened up by human language, to spelling or grammatical mistakes, and even to detect certain emotions.

A key business interest

For companies, knowing how to analyze what their customers say about them, and being able to communicate with each of them in a fluid and coherent way, is an increasingly pressing business issue. And natural language processing plays a key role:

  • it allows to identify and analyze the opinions expressed in publications on the web or social networks, and to determine their tone (neutral, positive or negative). When you, like Starbucks, are mentioned more than 3 million times a month, the human eye is no longer enough. Based on the information thus gathered, marketing and communication teams can detect any change in brand perception, identify what's wrong or could be improved, or push a product or type of campaign in line with customer expectations;
  • it establishes an individualized mode of communication with the client, which contributes to establishing a special relationship between the company and the client. This can be particularly important in times of crisis. A good example is RATP's chatbot in times of strike, which accompanies its users on their daily transport journeys, while wiping away their bad moods. A good chatbot can help to alleviate this initial feeling, by responding in an effective, relevant and sometimes funny way when users have fun pushing him into his corner.  

But in order to take full advantage of the promises offered by this technology, AI must be able to learn from fine, reliable, cleanly labeled data.

Isahit: AI spokesman

At isahit, thanks to our community of women trained to the specificities of each project, we accompany our clients on the issues related to this technology:

  • We train the chatbots, from the client platform, to label phrases/questions asked by real clients to make them more accurate and intelligent. For example: arrival and departure locations, schedules or dates for airlines and airport groups ;
  • For a site selling transport and construction equipment for professionals, we label very technical data to improve research performance (axle configuration, cargo dimensions, engine power, brand of concrete pump, etc.).
  • We are also analyzing social network posts for one of our leading cosmetics/care clients, so that they can develop an algorithm to see and understand the trends that emerge. All this in 5 languages (French, English, German, Arabic, Spanish).
  • Finally, we are working on Optical Character Recognition (OCR), a technology linked to NLP algorithms that allows us to convert images into text. For example, a sales receipt by retranscribing the text printed on it or by validating the retranscription made by the algorithm.

In short, we adapt to the specific needs and challenges of each client to offer them reliable data, a sine qua non for the eloquence of their AI... and the relevance of their message.

A word from Isabelle: "With our new Data Residency offer, we will be able to pursue our training mission here in France. We are already in discussion with SSE players to get people who are far from the digital world on the train. An inclusive technology: it's possible! »

"The NLP allows you to get information faster, so you can make better decisions and be more productive for both designers and users. This technology will revolutionize communication around the world. At isahit, we are working on it every day, with the diversity of our community being one of the keys to building a relevant, sharp and contextualized message. »
Isabelle Mashola, CEO of isahit

View the newsletter in your browser

You might also like
this new related posts

Want to improve your knowledge
about AI and our data labeling tools & solutions?

Isahit has a wide range of solutions and tools that will help you train your algorithms. Click below to learn more!