A heteronym is a word spelt the same as another but pronounced differently and with a different meaning – for example ‘lead’ (to guide) and ‘lead’ (the metal), or ‘wind’ (air) and ‘wind’ (to turn). Heteronyms are a subset of homographs. They pose a challenge for text-to-speech systems, because the correct pronunciation can only be inferred from context, often with the help of part-of-speech analysis.
Text-to-speech converts written text into spoken language, making it the counterpart of speech recognition (speech-to-text). Modern systems use neural networks to produce natural-sounding voices with appropriate intonation. TTS is used in telephone announcements, navigation devices, read-aloud functions for accessible websites and voice assistants.
Part-of-speech tagging automatically assigns each word in a sentence its grammatical category, such as noun, verb, adjective or preposition. Because many words are ambiguous (“book” as a noun or a verb, “light” as an adjective or a noun), the methods take the sentence context into account. POS tagging is a classic pre-processing step for parsing, information extraction or spell checking and is offered by libraries such as spaCy and NLTK.
Natural language processing is the branch of AI that enables computers to analyse, understand and generate human language in text or speech. It covers tasks such as translation, classification, summarisation, sentiment analysis and question answering. Rules and statistical methods once dominated; today the field is shaped by large language models.
Coreference resolution is a natural language processing task that identifies which expressions in a text refer to the same person, thing or matter. In “Ms Weber called. She would like a call back”, the system must understand that “she” means Ms Weber. This matters for summaries, information extraction and chatbots that need to keep track of references across several sentences or conversation turns.
