What Anna knows, and how it grows
For the beta launch we count openly what Anna knows. From now on we measure after every build and add the result to the knowledge page. Over time it shows how it grows. We only count, we read no content. The count runs on a copy of the founder's store.
What the graph does not show
After build 9 and five weeks of use there are 129,469 statements in the store. Only 4,747 of them (3.7%) are the founder's own. The rest is base knowledge that ships with the app:
129,469 statements in total
Visible in the graph: up to 12,172 statements (9.4%). Never in the graph: about 117,297 (90.6%).
- Grammar 72,912 · 56.3%
- Vocabulary 35,363 · 27.3%
- Other prior knowledge 7,209 · 5.6%
- Ontology 9,238 · 7.1%
- Own (founder) 4,747 · 3.7%
The graph in the app shows your own statements and what they connect to, such as classes or a well-known person you name in a relation. That is up to 9% of all statements. It does not show words, grammar and relation terms. Anna needs this part to understand your sentences.
Vocabulary
Anna knows about 2,650 German and 2,950 English base words. Measured against the whole vocabulary of a language that is little: 1.8% of the roughly 151,000 headwords in the Duden and 1.7% of the 171,476 words in current use in the Oxford English Dictionary.
Which words they are says more. So we compared Anna's words with the official word lists:
English · Oxford 3000 and 5000
German · Goethe-Institut
We count literally. If Anna knows “street” but not “road”, “road” is missing. Known words also do not yet mean that Anna understands sentences at that level. That is what the test bench measures.
What comes next
Every build adds a new row to the knowledge page, and from two measurements on a curve as well. That way anyone can check whether and where Anna learns.