Eliciting Implicit Evocations Using Word Embeddings and Knowledge Representation

• Published in: Scalable Uncertainty Management pp. 78 - 92
• Main Authors: Harispe, Sébastien, Medjkoune, Massissilia, Montmain, Jacky
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing
• Series: Lecture Notes in Computer Science
• ISBN: 9783319675817; 3319675818
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-67582-4_6

Automatic elicitation of implicit evocations - i.e. indirect references to entities (e.g. objects, persons, locations) - is central for the development of intelligent agents able of understanding the meaning of written or spoken natural language. This paper focuses on the definition and evaluation of models that can be used to summarize a set of words into a unique unambiguous entity identifier selected from a given ontology; the ability to accurately perform this task being a prerequisite for the detection and elicitation of implicit evocations on spoken and written contents. Among the several strategies explored in this contribution, we propose to compare hybrid approaches taking advantages of knowledge bases (symbolic representations) and word embeddings defined from large text corpora analysis. The results we obtain highlight the relative benefits of mixing symbolic representations with classic word embeddings for this task.