Terminology
Technical language is useful, but can also get in the way of conversation between fields. The same word can carry very different weight to a linguist, an AI researcher, an archivist, and a community language teacher. The word "AI" especially has become too broad a term to talk about meaningfully.
Terms
| Name | Category | Summary |
|---|---|---|
| Agent | AI & NLP | An AI system, usually built around a language model, that can take actions and use tools to pursue a goal over several steps. |
| Closed-Weight Models | AI & NLP | A model whose parameters are kept private, usable only through the owner's API. |
| Constrained Decoding | AI & NLP | Forcing a language model's output to fit a fixed structure or set of allowed choices instead of letting it generate free text. |
| Content Management System | AI & NLP | Software for storing, organizing, and publishing digital content (text, images, audio, video) through a structured interface, without hand-coding each page. |
| Data Scraping | AI & NLP | Collecting language material automatically from across the web instead of by hand. |
| Deterministic | AI & NLP | A system that always produces the same output for the same input, with no randomness. |
| Fine-Tuning | AI & NLP | Taking an already-trained model and training it a bit more on a smaller, specific set of examples to adapt it to a particular task or language. |
| Generative AI | AI & NLP | AI that produces new content (text, images, audio, code) rather than just analyzing existing material. |
| Language Model | AI & NLP | A model trained on text whose basic skill is predicting likely words. |
| Large Language Model | AI & NLP | A language model at the large end of the size scale, trained on enormous amounts of text. |
| Local Model | AI & NLP | A model run on infrastructure that a known, local entity owns or controls. |
| Low-Resource Language | AI & NLP | A language with little digital data for AI methods. |
| Machine Learning | AI & NLP | Building software that learns patterns from examples instead of being given explicit rules. |
| Machine Translation | AI & NLP | Automatically translating text from one language into another. |
| Model | AI & NLP | The trained artifact a machine learning process produces, stored as a list of numbers that turns input into output. |
| Natural Language Processing | AI & NLP | Building systems that work with human language. |
| Open-Weight Models | AI & NLP | A model whose parameters are released publicly, free to download and run. |
| Optical Character Recognition | AI & NLP | Converting a scan of printed or typed text into editable, searchable characters. |
| Probabilistic | AI & NLP | A system whose output involves chance, so the same input can produce different results. |
| Prompt Engineering | AI & NLP | Carefully wording the input you give a model to steer its output. |
| Semantic Search | AI & NLP | Searching by meaning rather than exact words. |
| Small Language Model | AI & NLP | A language model with few enough parameters to run on consumer hardware. |
| Summarization | AI & NLP | Using a language model to condense a longer text into a shorter one. |
| Text Recognition | AI & NLP | Turning an image of a document into machine-readable text. |
| Training | AI & NLP | The process of teaching a machine learning model by showing it many examples and adjusting its weights until it performs a task well. |
| Vision Language Model | AI & NLP | A model that takes images as input as well as text. |
| CARE Principles | Ethics & governance | Indigenous data governance principles: Collective benefit, Authority to control, Responsibility, Ethics. |
| Cognitive Offloading | Ethics & governance | Handing mental work over to a tool, and the worry that habitually doing so with AI may erode our own reasoning and skills. |
| Consent | Ethics & governance | Ongoing, revocable say over how language data is used. |
| Cultural Homogenization | Ethics & governance | The tendency of AI systems to flatten diverse cultures toward the dominant (largely Western, English-language) patterns in their training data. |
| Data Sovereignty | Ethics & governance | A people's right to govern data about their community. |
| Digital Redface | Ethics & governance | Using technology, especially AI, to imitate Indigenous languages or identities without the community. |
| Environmental Impact | Ethics & governance | The energy, water, and carbon cost of training and running AI systems. |
| Extractive Research | Ethics & governance | Taking from a community without returning value or control. |
| Misinformation | Ethics & governance | False information presented as true. |
| Misrepresentation | Ethics & governance | When a tool depicts a language, culture, or person inaccurately. |
| Ownership | Ethics & governance | Who holds rights over language materials and what is built from them. |
| Privacy | Ethics & governance | Control over who can access information about a person or community. |
| Traditional Knowledge | Ethics & governance | Knowledge, practices, and cultural expressions developed and held collectively within a community and passed down across generations. |
| Communicative Competence | Linguistics | The full ability to use a language appropriately in real social situations, not just to produce grammatically correct sentences. |
| Language Documentation | Linguistics | Recording and preserving a language's words, sounds, and texts. |
| Language Reclamation | Linguistics | A community claiming its right to speak its language and setting its own goals for it, in response to its own needs and perspectives. |
| Language Revitalization | Linguistics | Bringing a language back into active, intergenerational use. |