- Queries that are numbers. In general, numbers will not provide enough context to be able to determine if there is a correction and provide one. Most number queries are going to be ratedas is no correction needed
- Queries that are coordinates. For a query such as: “48.834670 2.455862”: It is reasonable to correct these to include the comma, since geographical coordinates are spelled with a comma to separate latitude and longitude
- Ambiguous/nonsensical queries/queries where no correction can be made because they are so nonsensical should be indicated as NOT having a spellings error
- Queries genitive cases should be reported as misspellings McDonald’s (McDonalds), Macy’s (Macys), l’arrabbiata ristorante (arrabiata ristorante). Missing apostrophes should be corrected when the business uses them in their official branding (as McDonald’s and Macy’s so) but we don’t need to worry about the overall grammar agreeing
- Queries that are half of the query unclear, but the other is clear, like this: “a partir de ayer de vue4182 n goddard rd lithonia” Second part is a real life address, first part is miscellaneous text, remember that the goal here is to avoid reformulating the query, including the deletion of misc or unnecesary text. If there are any spelling errors, the best thing to do would be to his “Yes” and fix them, even if they are within the misc text.
- Don’t include words that have been omitted by the original spelling. It is not important to include connecting punctuation for addresses but street names with missing diacritics should be corrected
- Abbreviations/initials are fine, the punctuation on these are not important/what we’re looking to capture with the rules. It’s fine to either include/not include them. So, for the example, “via g oberdan”, assuming oberdan is spelled correctly, we would mark it as not having a spelling
error. Please, rate the missing periods after numbers as errors, but not the missing periods after abbreviations
- Please disregard whether the address/business exist or not. This task is focusing on the spelling correction no on data accuracy
- If the search query is [Sha] or [orth], and based on the potential user locations, there could be numerous amounts of results, and corrections as it is vague. There is no need to take wild guesses on what queries might mean. Something like “walma” might have a somewhat likelywalmart” correction, but for cases where it could be one of dozens of things, you can just mark “No”
- Query with different intent for different locations. For example: the query is for iram, there are some options near some locations, however for other locations the only reasonable option is Irán.
Goal for this task is to imagine the lat/lngs as individual queries. If you think that it’s likely all lat/lngs have the same intent and some correction, they should say “Yes” in the second question. For a case like this, where “Iram” might have some meaning other than “Iran” for certain lat/lngs, the best answer is to select “No” on the second question. Use your best judgment considering prominence also
- For Mexican market: Query mix languages Query “carwhas cerca de mi” it has English components and Spanish components and all potential user locations are in the USA. Please, correct items to the spelling that the query seems to expect. You shouldn’t ever need to translate something from one language to another. Best correction will be “carwash cerca de mi”
- For Spanish market: Queries that are in Catalan, Euskera, Galician, etc. Shouldn’t be checked straight away as “Query in inappropriate language for locale “ “Other language” box because this is under the Spanish umbrella. However, if you are not familiar with that language feel free to select the option “Query in inappropriate language for locale “
- When a query is londen (dutch way of writing london) is this also a spell error since it’s not official spelling? when a query is partially in English for a Dutch address, for example kalverstraat 25 amsterdam netherlands, how should we rate it? Is this a spelling error since last word should be Nederland?
If “londen” is the accepted Dutch spelling of “London” and is appearing in the Dutch Spell Correction task, it should not be considered a spelling error. If the query using the English spelling of Netherlands is appearing in the Dutch task, it should be corrected. If it is appearing in the English task, it can be left alone.