U ovom radu dan je pregled metoda za klasifikaciju korisničkih upita i obradu prirodnog jezika korištenjem strojnog učenja. Dohvaćeni su podaci sa društvenih mreža koji su zatim ručno klasificirani, obrađeni te korišteni za izgradnju i evaluaciju klasifikatora. Korišteni algoritmi za klasifikaciju su Naivni Bayesov algoritam i strojevi sa potpornim vektorima. Rezultati evaluacije dobiveni su u obliku matrice zabune i općenitog postotka ispravnog klasificiranja upita u testnom skupu. Upiti se uspješno klasificiraju te sustav pokazuje dobre rezultate i na zahtjevnijim upitima korisnika. Model je implementiran unutar web aplikacije u koju korisnik unosi upit i osnovne podatke, aplikacija klasificira upit kroz implementirane modele te konačan rezultat zajedno sa upitom i podacima sprema u bazu podataka i ispisuje u web aplikaciji.

This paper reviews the methods for classifying user queries and natural language processing using machine learning. Data from social networks has been fetched, which are then manually classified, processed and used to construct and evaluate the classifier. Used classification algorithms are Naive Bayes algorithm and support vector machines. The test set of data was submitted to classifier, and an evaluation of the obtained classes was conducted for each query. Evaluation results were obtained in the form of confusion matrix and the overall percentage of correct classification of queries in the test set. Queries are successfully classified, and the system shows good results even on more demanding user queries. Model is implemented within a web application where user inputs the query and personal data, application classifies query through the implemented models, and final result along with query and data saves it in database and prints it in the web application.