A carregar...
Publicação
Classification problem in real estate corpora: Furniture detection in real estate listings
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.17 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
The real estate market has been a conservative industry, but it has recently seen an increased number of data-driven solutions and software products aimed at real estate professionals. Casafari is at the forefront of this development and is rapidly building a portfolio of data services. All of these services rely on clean data extracted from multiple websites. The current project contributed to several of the company’s products by developing a classifier that detects if a mention of furniture is present in the online property listings. As a result, a new filter was introduced on the company’s metasearch page, and a new furniture attribute was made available for an improved valuation of the property. In addition to the improvements in the website navigation and property valuation accuracy, the current project contributed to the improvement of theory. Several state-of-the-art NLP models were tested and evaluated against the developed classifier, with some showing very competitive results without any training and reliance on pre-labeled data. However, for these models to be considered for production, there needs to be a significant improvement in speed and GPU requirements.
Descrição
Internship Report presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Business Analytics
Palavras-chave
NLP Text Classification Zero-shot learning Keywords Supervised Learning Real Estate