By Martin Vongsuravatana (FootbAI) 
1.8 billion visits were recorded by ChatGPT in March 2024, making AI one of the most exciting and trending technological topics of recent years. ChatGPT is an artificial intelligence system that uses Natural Language Processing (NLP).
What is Natural Language Processing ?
NLP is defined as an AI’s ability to understand or express the natural language used by humans, either orally or in writing. By natural language we mean what a person will say or write in their own words without paying particular attention to syntax or spelling. This is in contrast to computer languages, which are strict in their writing and require advanced skills. NLP breaks down this barrier and brings AI closer to everyone.
Let’s take a first example. Let’s imagine that I have data on the rankings and calendar of English national championships. I’d like to be able to have a team’s ranking and fixture list available on request. Without NLP, I would have to code an algorithm with a precise syntax to meet this request. With NLP, I’d just enter a question like, “What is Arsenal’s ranking and their next 3 matches?” The AI would reply directly.
NLP is easy for everyone to use, which explains its popularity and the growth in its use in recent years. Chatbots and automated file classifiers are now common in many firms.
The two approaches to NLP
Two approaches can be used to create an NLP model: a symbolic approach and a statistical approach.
Historically, the symbolic approach has been used first. This involves teaching our model the clear rules of a language through examples illustrating these syntactic and grammatical rules. The aim of this approach was to reduce AI errors as much as possible, but it is extremely costly in terms of time and budget. What’s more, it can quickly be confronted with cases that it has never seen before and become irrelevant.
The statistical approach has taken over, especially with ChatGPT. It learned the rules from vast amounts of data. ChatGPT4 uses no less than 175 billion parameters, for example. Like the suggestion of the next words in a Google search, this approach is probabilistic. Although it may have its limitations, this approach, used on a large scale, has revolutionised the sector.
Let’s go back to our example. Gunners’ is the nickname of the Arsenal team. If we use the statistical approach, our AI model will correctly answer the question ‘What is the Gunners ranking?’ In terms of probability, this question refers to the football club 90% of the time. On the other hand, with the symbolic approach, the model could give the wrong answer by giving the ranking of any gunner in a field because it hasn’t seen enough examples and doesn’t identify the most likely answer.
Limits and hallucinations
NLP does have its limitations. Firstly, the responses it generates are often very standard. Even if this saves time, a human touch is often needed at the end of the process, particularly to add creativity. To obtain very precise results, the exchange with the AI needs to be well worked out, with what is known as prompt engineering, and there is a form of code here.
Secondly, NLP can give rise to quite grotesque errors, sometimes with a small margin of error. An AI can solve a very complicated mathematical problem but sometimes make a mistake on a question to which a 10-year-old child would know the answer. These are known as hallucinations. In many cases, these hallucinations are not serious because they are corrected by a human behind them. But they do pose a problem in sectors such as medicine, the legal sector or banking, where even a 1% error rate is too high.
What’s next?
NLP is a fascinating subject that is driving the AI sector. It has many practical applications, and the world of football is seizing on it. Sentients Sports is developing an NLP-based solution that will revolutionise the sector: ScoutGPT. This solution makes it easy to target players using simple natural language phrases. It has been trained on specialised football data to achieve the best performance.
For our explanation of machine learning, see our article.
If you have any questions about how to integrate natural language processing and artificial intelligence into your football club or sports organisation, don’t hesitate to contact our team.
Source :