DolphinGemma
AI, bioacoustics and the science of animal communication.
DolphinGemma.com is an independent research platform exploring how AI models can analyze patterns in animal acoustic signals. We catalog research papers, preprints, datasets and scholarly resources across computational bioacoustics and animal communication.
What is DolphinGemma?
DolphinGemma is an exploratory research concept investigating how modern neural audio representations and sequence modeling techniques might assist in studying complex animal vocalizations, beginning with odontocete acoustic recordings.
Acoustic Structure Detection
Exploring how machine learning models can help detect recurring acoustic patterns, whistle contours, and pulse repetitions in bioacoustic datasets.
Neural Audio Representations
Examining how self-supervised audio codecs and embeddings can model high-frequency signals without relying solely on manual feature engineering.
Pattern vs. Translation
Maintaining a clear distinction between identifying statistical regularities in sound and unsubstantiated claims of semantic translation or language understanding.
What is Bioacoustics?
Bioacoustics is the scientific study of sound production, dispersion, and reception in animals. Modern bioacoustics combines ecology, acoustics, and signal processing to monitor wildlife and analyze communication non-invasively.
Passive Acoustic Monitoring
Deploying autonomous hydrophones and terrestrial acoustic sensors to record soundscapes over months or years without disrupting natural habitats.
Spectrogram & Signal Analysis
Transforming audio waveforms into time-frequency spectrograms to visualize frequency modulation, harmonics, and transient pulses across species.
Automated Classification
Using computational models to detect vocalizations, classify candidate signals, and process large-scale environmental audio archives.
Animal Communication Research
Scientists investigate how animals produce and use vocal signals in social interactions, navigation, and group coordination, examining repertoire diversity across cetaceans, birds, bats, and other vocal taxa.
Vocal Repertoire Categorization
Cataloging distinct signal classes such as frequency-modulated whistles, pulsed calls, and broadband echolocation clicks observed in odontocetes.
Context & Identity
Documenting how specific signals associate with behavioral states, individual identity (such as signature whistles), and social cohesion.
Communication vs. Language
Distinguishing acoustic pattern correlation from human linguistic equivalence. Research investigates communication dynamics without assuming human-like grammar.
Models & Research Projects
Documenting active initiatives across academic institutions, non-profit foundations, and AI laboratories.
DolphinGemma
Exploratory model concept investigating neural audio representations for dolphin acoustic signals.
Project CETI
Interdisciplinary scientific initiative applying machine learning to study sperm whale communication codas.
Earth Species Project
Non-profit organization developing open-source machine learning models for non-human bioacoustics.
BirdNET
Deep neural network developed for automated bird and wildlife sound identification from audio recordings.
The Research Hub
A curated directory connecting research papers, preprints, datasets and scholarly resources, along with species profiles and research organizations.
Papers & Preprints
Research papers, preprints, and scholarly publications with DOI links.
View papers →Datasets
Hydrophone recordings, annotated calls, and acoustic benchmarks.
View datasets →Species
Acoustic repertoires and vocal profiles of studied animals.
View species profiles →Organizations
Academic institutions, field projects, and bioacoustic laboratories.
View organizations →Committed to Open Scientific Evidence
We believe scientific progress requires accurate context. We prioritize primary literature, verifiable datasets, and transparent methodology while resisting sensationalized claims.