Morphology, gemination, Ge'ez orthography and low-resource NLP.
Empirical acoustic measurements of Amharic gemination, minimal pair closure durations, and why neural TTS models require explicit phonemic duration modeling.
How to engineer deterministic grammar rules for Semitic agreement, Ethiopic sentence delimiters, and asymmetric honorifics using LanguageTool.
Empirical analysis of the 372-pair Amharic ErrorCorpus: non-word vs real-word errors, agglutinative clitic false alarms, and designing accurate spellcheckers.
Why static lemma lists fail in agglutinative Semitic languages, and how 39 Hunspell SFX rules recover 95% of Amharic noun plurals without dictionary bloat.
Neural Amharic Text-to-Speech: glottalic ejectives, the seven vowel orders, G2P for unwritten gemination, and VITS pipelines.
Automatic Speech Recognition for Amharic: proclitic attachment, code-switching, and CTC versus encoder-decoder models.
How Amharic's script and syntax affect text classification, named entity recognition and machine translation.
Ethiopic Optical Character Recognition: syllabary geometry and stroke ambiguity, degraded documents, TrOCR, and a Tesseract pipeline.
Unwritten gemination in Amharic: minimal pairs, verb types, G2P disambiguation in neural TTS, and duration modeling in ASR.
When to normalize Amharic homophones and when to preserve them, by NLP task, with a configurable Python normalizer.
Amharic's features that affect NLP: its 300+ character script, unwritten gemination, honorifics, morphology and word order.
Why Amharic is a low-resource language in NLP, and how its script, morphology and syntax affect computational models.
