Talk:Contextualization (computer science)
 | This article is rated Stub-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects: | ||||||||||
| |||||||||||
Circular logic in the lede
[edit]I read that first paragraph and wondered what sort of tongue twister I had stumbled across.
contextualization is the process of identifying the data relevant to an entity (e.g., a person or a city) based on the entity's contextual information
This is a tautological statement, and does not resolve in any way the intended interpretation of contextual, nor the means of doing so. Is this done manually, is it automated? It reads like it is a manual process, but as it's linked to by Transformer (deep learning), I expect this to be automated. There, there is no mention of what qualities are used to "contextualize" a given word-token, so I look to the article for clarification. Unfortunately, here I was presented with a vague depiction of a reason to classify data, but not how; the format is likewise indistinct, and it's not clear how the metadata is to be collected; is it a part of the original data?
Not all of that needs to be answered in the lede―but the lack of anything to resolve these fairly casual questions meant that I'm leaving more confused than I arrived. I'd appreciate this getting some enrichment, as it were.