1.Introduction

In this document, we will cover some frequent errors we have encountered when reviewing autocomplete tasks in the fr_FR market and provide examples to help illustrate.

2.Query suggestion matching the query but no real-world result near the VP/user.

Review guidelines section 4.13. Query and Category Suggestions :

Query and category suggestions are rated by considering the user intent and the possible suggestions within the viewport. For a query/category suggestion to be valid it needs to provide some use to the user. Simply matching the user query is not enough on its own.

If there are no real world results near the fresh viewport, user or location modifier, the query or category suggestion should be rated Bad.

Query String Suggestion  Rating and Explanation

[SU]

Fresh viewport in San Jose, CA

3 supermarket

Supermarket is a common store category and there are suggestions inside the viewport as well as around it.

4 supercuts

Relevance

Good (Distance/Prominence)

Supercuts is a popular salon chain but there are no results inside the viewport. The closest branch is just outside the viewport, so the query suggestion is demoted to Good.

5 super 8

Relevance

Bad (Distance / Prominence)

The Super 8 motel chain is fairly prominent, but there are no results within or around the viewport. The suggestion is therefore demoted to Bad.

Example

In this case the fresh viewport is on Riom-ès-Montagnes and the query is short : “Ap”. The suggestion “Apple store” is matching but there is no Apple Store in the viewport or near it. There is a locality matching instead :Apchen”.

It is also the case if the query is complete and the suggestion is exactly matching the query string, but no real world result within or around the fresh viewport (or any explicit / implicit location).

“Sherpa” is a chain of supermarkets only located in mountain areas, here the viewport is on the region “Bretagne” where they don’t exist. This suggestion is rated Bad.

.

3.Very prominent result promoted due to high prominence

Review guidelines section 4.3.3. Prominence :

The prominence of a feature refers to its popularity, including the number of people visiting and media sources referencing it. Prominence can vary based on the test locale and even local knowledge. Consider the following list to get a general idea of the concept of prominence, ordered from the most prominent to the least prominent:

The feature is known internationally
The feature is known in the country
The feature is known in the suggestion region
The feature is known locally
The feature may not even be known locally

Suggestions with high prominence may still be relevant even when far away from the user. Suggestions with low prominence, on the other hand, may not be as relevant even if they are close to the user.

Example

In this case, Madagascar is a country and it is considered prominent even if it is very far. Thus, because this result is promoted and there are no matching result near the fresh viewport centered near Bourg-en-Bresse in France, this result is rated Excellent.

 

4.Unexpected language/script

Review guidelines section 3.1. Suggestion Name/Title in Unexpected Language or Script :

A suggestion’s name or title is in an expected language or script when it is:

In the language/script of the test locale

In the language/script of the query

In a language/script of the suggestion’s region

Any combination of any of the languages or scripts described above

An official company, chain, or brand name commonly used in the market, even if this is not any of the languages or scripts described above

When all or part of a suggestion’s name or title is in an unexpected language or script, users will not understand it. In these cases, select the Result name/title is in unexpected language or script checkbox. No further rating is required.

Examples : Tronçais Forest, bakery, Markt, spiaggia etc

Example

In the case below, “Dam” is the English word for “barrage”, the Result name/ title is in unexpected language or script checkbox must be selected.

5.Query with location modifier

Review guidelines section 2.3.1 Explicit Location

When there is an explicit location stated in the query, the user and viewport locations can be ignored because the user has stated exactly where they want to see suggestions. Explicit locations can include specific addresses or location modifiers, such as neighborhoods and localities.

Rate explicit location intents differently depending on how many possible suggestions exist within the requested area.

 

5.1.How to demote results if they are not in the location requested

It depends if there are real world results in the location requested and how many.

One suggestion within the requested area

As there is only one Intermarché within Felletin, the next closest Intermarché outside the explicit location are still considered Good. The demotions are applied for distance gradually.

 

Many suggestions within the requested area

Because of the high number of restaurants within the explicit area, any suggestion outside is considered too far and is rated Bad.

 

.

5.2.How to rate if there is no results in the location requested

No suggestions within the requested area

Where no suggestions exists within the explicit area, like in the IKEA example, the closest one receives Excellent. Gradual demotions are applied for locations further out, like the one in Bordeaux.

.

6.Spell corrections

Review guidelines section 4.1.5. Spell Corrections

The user query may be misspelled or contains parts that may have been entered incorrectly, and the suggestion corrects that issue. This type of match should only be used for obvious issues, when no suggestion for the actual query exists at the user location or when the user intent clearly is satisfied by the autocorrected suggestion. > This kind of correction also happens when a query address does not exist but the suggestion address does, correcting the issue and creating a match.

6.1.Considered a spell correction

When the user query contains a possible typo and the result contains a string that is one-letter different and the key typed is close on the keyboard, the result has possible relevance.

Examples

 

In this case, there is no close result for something beginning with « Garde de Lyo… » but the letter D is near E and R and the query can be corrected to « Gare de Lyon ».

Since Gare de Lyon is relevant and prominent as it is the name of a train station in Paris, this spell correct result is considered the primary intent. Relevance for the result is Excellent

.6.2. Not a spell correction

In this case, the user is in fresh viewport, there is a matching locality near the user : “Nandy”. The result with a misspell correction is a more prominent locality far from the user, but here the intent was clear because of the implicit location intent, this misspell correction is not considered a match.

Leave a Reply

Your email address will not be published. Required fields are marked *

Related Post