We annotated RuSERRC with 2 more kinds of information:
There is information about nested entities for the terms at different levels. We used BIO tagging as well.
Here an example of nested entities:
| token | tag | nested_1 | nested_2 |
|---|---|---|---|
| автоматизированное | B-TAG | ||
| извлечение | I-TAG | B-TAG | |
| речевых | I-TAG | I-TAG | B-TAG |
| действий | I-TAG | I-TAG | I-TAG |
We linked the terms with the entities from Wikidata.
In column "wiki_id" you can find an id to the Wikidata entity and ids of the term (including nested ones) which is linked.
For example,
| id | token | tag | nested_1 | nested_2 | wiki_id |
|---|---|---|---|---|---|
| 0 | автоматизированное | B-TAG | |||
| 1 | извлечение | I-TAG | B-TAG | ||
| 2 | речевых | I-TAG | I-TAG | B-TAG | Q25481968:2,3 |
| 3 | действий | I-TAG | I-TAG | I-TAG |
means that the tokens with ids 2 and 3 ("речевых действий") are linked with an entity Q25481968.
Also we extracted information about entities from Wikidata (version from 27.01.2021) such as name, description and aliases in Russian (removing claims and sitelinks) and made it publicly available. You can download it from here.