Reproduce “Automatic Ambiguity Detection” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Richard Sproat, Jan van Santen
Researchers at Bell Labs – Lucent Technologies developed an automated method to detect and quantify word ambiguity using only unlabeled text corpora. This approach, which computes a 'polysemy index' based on the angular distribution of contextual terms, successfully identified multiple distinct senses for words like 'bass' while confirming single dominant senses for others such as 'oxidation'.
Key insights
- Word ambiguity can be quantitatively measured and ranked using a polysemy index derived from contextual term distributions.
- Corpus-based methods can identify and quantify word senses without reliance on pre-existing, often incomplete, lexical resources like WordNet.
- Distinct contextual clusters and multiple peaks in angular distributions of surrounding terms indicate high polysemy, while concentrated distributions suggest a single dominant sense.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.