Updated
Definition
Entity resolution is the process of deciding which real-world brand, if any, a piece of text refers to, so that mentions of "Apple" the company are separated from apples the fruit and attached to a stable brand entity.
Brand names are hostile input. They collide with common words (Apple, Shell, Tide), with each other (Dove soap and Dove chocolate), and with their own variants: nicknames, ticker symbols, product lines, misspellings, and sub-brands that sometimes should roll up to the parent and sometimes should not. Naive keyword matching either drowns in false positives or, tuned tighter, silently drops the casual references where real conversation lives.
The working answer is to model brands as entities rather than strings: a stable identifier that aggregates every surface form, with context used to disambiguate collisions before a mention is attached. In Adveron, tracked brands are entities in this sense, which is what makes downstream data comparable: a mention count, a sentiment trend, or a share of voice figure means the same thing for every brand because the resolution step happened before the metric was computed.
They are distinct entities with a relationship, not one keyword bucket. That lets you query the sub-brand on its own or roll conversation up to the parent, and keeps a product launch from silently inflating the corporate brand's numbers when you did not want it to.
Ambiguity is handled before metrics, not after: context that does not support a confident match is excluded rather than guessed. Slightly conservative volume is a better foundation than volume inflated by fruit, weather, and unrelated companies.
Adveron is opening to a first wave of teams. Ask for a key and we will open your workspace with one credential for REST and MCP. Usage-priced credits, no seat licenses.
Get early access