Based on the philosophy of the Semantic Web, Ontotext exploits text processing and automatic reasoning technologies to extract knowledge from texts and organise it conceptually in an ontology. Unlike common search engines, the Ontotext Portal directly accesses the concepts and entities of the ontology and presents the user with structured information instead of mere portions of texts. For each entity, the Ontotext Portal offers four different views: Articles (lists all the documents in which it is mentioned), Citografo (shows how often it is mentioned), Opinions (shows how often opinions are expressed about it), and Record (provides extra information about it).
ποΈ Yesterday at #LREC2026, Palma de Mallorca!
@lina_conti presented "Voice, Bias, and Coreference: An Interpretability Study of Gender in Speech Translation" at the poster session.
πPaper:
π»Code: https://github.com/lina-conti/voice-bias-coreference
#SpeechTranslation #NLProc
How does the granularity of speech-text pairs impact SpeechLLM performance, and what is the optimal way to interleave tokens? Furthermore, what are the best practices for generating synthetic data to boost training?π§
ποΈ Our paper on connecting Speech Foundation Models with LLMs is featured in the SpeechLMM Training Journal on Weights & Biases.
Read it π https://bit.ly/4svG7ll
SpeechLMM 2.0 coming this summer. π
#Meetween #SpeechLMM #AI #NLP