A Comparative Study of Text Representations for French Real-Estate Classified Advertisements Information Extraction

Lucie Cadorel (1, 2) , Andrea G. B. Tettamanzi (1)

1 WIMMICS - Web-Instrumented Man-Machine Interactions, Communities and Semantics
2 Septeo Proptech

Text representations are widely used in NLP tasks such as text classification. Very powerful models have emerged and been trained on huge corpora for different languages. However, most of the pre-trained models are domain-agnostic and fail on domain-specific data. We perform a comparison of different text representations applied to French Real Estate classified advertisements through several text classification tasks to retrieve some key attributes of a property. Our results demonstrate the limitations of pre-trained models on domain-specific data and small corpora, but also the strength of text representation, in general, to capture underlying knowledge about language and stylistic specificities.

Mots clés: Text Representations; Information Extraction; Real-Estate Market

Domaines: Intelligence artificielle [cs.AI]

Lucie Cadorel, Andrea G. B. Tettamanzi. A Comparative Study of Text Representations for French Real-Estate Classified Advertisements Information Extraction. ENIGMA 2023 : 1st Workshop on AI-driven heterogeneous data management : Completing, merging, handling inconsistencies and query-answering, Sep 2023, Rhodes, Greece. pp.55-63. ⟨hal-04346759⟩