Thanks to recent progress in the field of optical interferometry, instrument sensitivities have now reached the level achieved in the domain of new space missions dedicated to exoplanet and stellar studies. Combining interferometry with other observational approaches enables the determination of stellar parameters and helps improve our understanding of stellar physics. In this paper, we aim to demonstrate a new way of using stellar atmosphere models for a joint interpretation of spectroscopic and interferometric observations. Starting from a discrete grid of one-dimensional (1D) stellar atmosphere models, we developed a training algorithm, based on an artificial neural network, capable of estimating the spectrum and intensity profile of a star over a range of wavelengths and viewing angles. A minimisation algorithm based on the trained function allowed for the simultaneous fitting of the observational spectrum and interferometric complex visibilities. As a result, coherent and precise stellar parameters can be extracted. We show the ability of the trained function to match the modelled intensity profiles of stars in the effective temperature range of $4500$ K to $7000$ K and surface gravity range of $3$ to $5$ dex, with a relative precision to the model that is better than $0.05<!PCT!>$. Using simulated interferometric data and actual spectroscopic measurements, we demonstrated the performance of our algorithm on a sample of five benchmark stars. Using this method, we achieved an accuracy within $0.5<!PCT!>$ for the angular diameter, radius, and surface gravity, and within $20$ K for the effective temperature. This paper demonstrates a new method of using interferometric data combined with spectroscopic observations. This approach offers an improved determination of the radius, effective temperature, and surface gravity of stars.