In the last years, Cognitive Systems are increasingly appearing, offering new ways for developing Question Answering solutions able to autonomously extract an answer for a question formulated in natural language. Currently, to the best of our knowledge, most of the available Question Answering solutions are designed for the English language and use SQL-like knowledge bases to provide factual answers to a natural language question. Starting from these considerations, this work presents a preliminary Question Answering framework for closed-domains, like Cultural Heritage. It has been expressly thought to extract factual answers from collections of documents by operating with the Italian language. Such a framework exploits a variety of NLP methods for the Italian language to help the understanding of user's questions and the extraction of precise answers from textual passages contained into documents. Moreover, Deep Learning techniques have been used to proficiently understand the topic of a question, whereas a rule-based approach relying on dictionaries has been applied for the annotation and indexing of collections of documents in Italian, enabling their usage into a state-of-the-art Information Retrieval engine. An experimental session has also been arranged, showing very promising preliminary results.
Towards a Framework for Closed-Domain Question Answering in Italian
Emanuele Damiano;Massimo Esposito;Giuseppe De Pietro
2016
Abstract
In the last years, Cognitive Systems are increasingly appearing, offering new ways for developing Question Answering solutions able to autonomously extract an answer for a question formulated in natural language. Currently, to the best of our knowledge, most of the available Question Answering solutions are designed for the English language and use SQL-like knowledge bases to provide factual answers to a natural language question. Starting from these considerations, this work presents a preliminary Question Answering framework for closed-domains, like Cultural Heritage. It has been expressly thought to extract factual answers from collections of documents by operating with the Italian language. Such a framework exploits a variety of NLP methods for the Italian language to help the understanding of user's questions and the extraction of precise answers from textual passages contained into documents. Moreover, Deep Learning techniques have been used to proficiently understand the topic of a question, whereas a rule-based approach relying on dictionaries has been applied for the annotation and indexing of collections of documents in Italian, enabling their usage into a state-of-the-art Information Retrieval engine. An experimental session has also been arranged, showing very promising preliminary results.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.