Internal Search Engine and Recommender System with Natural Language Processing in PaaS

Nowadays, Natural Language Processing con-tributes to the advancement of different areas and applications such as voice recognition, web search engines, social mining, etc. This scientific article applies natural language processing techniques to develop an internal search engine and recommender system in e-commerce. The system is deployed in a PaaS with a NoSQL database and a dataset with 100 documents on literary works. Fifteen fields structure each document, and for the system implementation, only six document fields were considered. The search engine compares a query with the corpus of documents using the Jaccard coefficient, the Sorensen-Dice coefficient, and the Salton cosine. Moreover, the recommender system applies similarity metrics to find books similar to the last ones viewed by the user. The tools were validated through functional and non-functional tests. The functional tests used human validation through a satisfaction survey in 25 people between 23 and 25 years old with a university education. The results showed that at least 65 % of the users rated the tools with the highest satisfaction score. Stress and load tests were carried out for the non-functional tests, obtaining a high latency in some cases because free services were used.