Goal:
> To determine how well the suggestion satisfy the user’s query, and
> To check the data accuracy (name and address) of the suggestion
There are two types of location intent:
> Explicit – the query includes a specific location where suggestions are
expected.
> Implicit – the location intent is not given, so you must use context clues in the form of “User Location” and “Viewport” to determine the area in which the user is expecting suggestions.
Query: Manila Jollibee
Viewport: Stale, near Mandaluyong City
User Location: (14.51719, 121.05321) near Brgy. Pinagsama, Taguig City
Location Intent: Explicit
Description: Query includes an explicit locality location intent “Manila” and a prominent fast food brand “Jollibee”. We can ignore the stale viewport location as well as the user location since the user has a specific location intent –
Manila Query: SM Mall Viewport: Fresh, near Tagaytay City
User Location: (14.0953, 120.93771) near Sky Ranch Tagaytay and located within the fresh
viewport
Location Intent: Implicit Description: Query includes a prominent POI/business shopping mall brand named “SM”. Since the user did not indicate a specific location, we should consider the user location within the fresh viewport in judging. Thus, all possible SM Malls might receive a starting rate of Excellent to Bad by distance from user location.
Case 1: with clear POI/Business intent and complete location intent
If the query string has a complete POI/Business name intent and the location indicated refers to a unique one, then judge them based on the exact location intent. Most of the time, user looks for a specific chain business in a specific area but research shows that no such branch in that area. Please refer to the following samples to calibrate your understanding:
Query: Ayala Malls Alfonso Cavite
Query Intent: The query string includes a chain business brand – “Ayala Malls” and a complete
explicit location “Alfonso, Cavite”
Real World Results: Research shows that there is no any Ayala Malls in the locality of Alfonso in the province of Cavite. Thus, we should promote the relevance rating of Ayala Malls outside the requested locality from Excellent to Bad by distance from Alfonso.
Case 2: with clear POI/Business intent & incomplete location intent
If the query has a complete POI/Business name intent and the location is an incomplete query string, then further research of the possible incomplete explicit location intent is required.
Query: Ayala Malls Serin Alfon
Query Intent: The query string includes a specific Ayala Malls branch and an incomplete location intent.
Real World Results: Research shows that the only intent for this query is to look for Ayala Malls Serin in Tagaytay City (near Alfonso, Cavite) – Excellent. Other Results:
Ayala Malls Cloverleaf and Ayala Malls Solenad – These results won’t match the query as user already indicated “Serin”, a narrow intent for specific mall branch. Rating: Bad.
Case 3: with incomplete POI/Business query string and no nearby results in real world
If the query has an incomplete POI/Business query string, then further research is required. Query string should always match the main intent or partial parts of the result listing. Please consider prominent features/POI/ businesses in your locale.
Query: MANG inaS
Query Intent: The primary intent is to look for a branch of a prominent fast food chain brand “Mang Inasal”.
User Location (within fresh viewport): near Siargao Airport in Siargao Island (9.8576, 126.01637)
Real World Results: Research shows that there is no any Mang Inasal near user location.
Leniency on distance is acceptable when there are only a few or no possible suggestions for the query in the requested area.
Sometimes, the query string may be misspelled or contains parts that may
have been entered incorrectly, and the suggestion corrects that issue. There could be minor or major misspellings that would lead to an auto correction of the query string. Research and locale knowledge are required to obtain the possible intent of the user.
Note: If the result listing matches the query string or matches by auto
correct, then the starting rate of Excellent to Bad will be applied.
If the result listing does not match the query string or spell correction is not considerable, then this is a No Match that would lead to a rating of Bad.
Case 1: Incomplete query string with possible spell correction
Query String | Possible Suggestion | Explanation |
Sampagi | Sampaguita Subdivision, Caloocan City | Match. The query suggestion spell corrects the common term “sampaguita” |
Forbs Park Makti | Forbes Park, Makati City | Match. It corrects the words “Forbes” and “Makati”. It is obviously the only intent of the user. |
Pashig Public Ma | Pasig Public Market, Pasig City | Match. It corrects the misspelling of the locality name “Pasig” |
Marjet! Market! | Market! Market! Taguig City | Match. It corrects the name of the prominent shopping mall in Taguig City |
Estella Steet, Makati, Metro , | Estrella Street, Makati, Metro Manila | Match. It corrects a unique possible partial street address street address |
Case 2: Incomplete query string without spell correction
Query String | Possible Suggestion | Explanation |
Bacolod, Lib | Bacolod City, Negros Occidental | Not Match. The suggestion does not match the second word “Lib”. Possible match is the barangay Bacolod in Libon, Albay |
Puregold Kalen | Puregold Caloocan City | Not Match. The brand Puregold matches the query string but not the possible location modifier. Research shows that Puregold Kalentong is the best match. |
Mercury Drug, Manda | South Star Drug, Mandaluyong City | Not Match. The suggested location modifier matches the incomplete query string but not the pharmacy brand. |
Victory Chu | Victory Liner, Cubao | Not Match. The query matches the word “Victory”. Further research that user would look for a Victory Church location, not a bus terminal |
Case 3: Complete query string without spell correction
Query String | Possible Suggestion | Explanation |
Mati City | Makati City | Not Match. The user query is complete and refers to a unique locality while the suggestion refers to another locality. |
Nuvali | Novaliches, Quezon City | Not Match. The query is complete and refers to a prominent neighborhood in Santa Rosa, Laguna I \ |
Query String | Possible Suggestion | Explanation |
Robinsons Place Lipa | Robinsons Place Antipolo | Not Match. Both are Robinsons Malls but a location modifier mismatch. |
Cebu | Cebuana Lhuillier | Match. The primary intent is the city of Cebu. But the listed suggestion also matches the query string. This suggestion is a prominent pawnshop chain. |
Davao | Davao International Airport | Match. The primary intent is the city of Davao. The listed suggestion matches the query string and a prominent POI airport. |
Case 4: Complete query string with spell correction
Query String | Possible Suggestion | Explanation |
Siurgao Island | Siargao Island | Match. The suggestion matches the query string by minor spell correction. |
SM Lanag Premier | SM Lanang Premier | Match. The suggestion made a minor auto correction for the term “Lanang” |
Batanesss | Bataan | Mismatch. Although the query string has misspellings/extra letters, research shows that the prominent intent could be “Batanes” |
Little Baguio | Baguio City | Mismatch. Research shows that Little Baguio is a sublocality in San Juan while Baguio City is an unrelated locality address. addres |
The user could issue a query string without a possible relevant result in the real world. This means that the query might have some extreme misspellings, nonsense wordings, etc. In this case, further research is required. It is essential to have an extensive research in case we missed the possible user’s intent.
Users might have a possible non-prominent map intent.
Query String | Possible Suggestion | Explanation |
zzyzx | Club zzyzx Manila | Excellent. The query appears to be nonsense but further research shows that such business exists in Manila. |
zzzzz | Club zzyzx Manila | Bad. The suggestion does not match and has no connection with the query string. It is non-sense and has no real world relevant result. |
coffeeeeeeee | Starbucks | Excellent to Bad. Result suggestions satisfy the intent to look for a coffee shop. Although the query string seems harsh, the main intent is clear to look for a business that serves coffee. |
Depending on how ambiguous the query intent is, the rating could receive a starting grade of Excellent. It always depends on the distance from the location modifier and its prominence. Further research is required to know the prominence of the possible suggestions. There might be more than one
possible user’s intent in a single query string.
7.Business/POI is Closed or Does Not Exist
Mark the^business/POI as closed or does not exist if there is evidence to prove
the business indicated in the suggestion is no longer operational. When you can’t make a decision about business closure due to lack of resources, assume that the business exists and rate the Name and Address as “Can’t
Verify”. The following samples below focus on determining business closure when they are no longer visible via online street imageries.
Result Listing: Shakey’s
Coordinates: (14.53788, 121.01766)
2017 Street Imagery 2023 Street Imagery
Explanation: It is true that we don’t judge the coordinates or pin accuracy in AutoComplete tasks. But these coordinates would give you a huge tip to determine the closure of a POI/Business. In above images, the Shakey’s
branch is visible in 2017 street imagery while 2023 street imagery shows another business named Tenya. This is strong evidence of closure. Please include a shortened link of street imagery every time you rate a business/POI closed based on street imagery differences.
It is important to review the Guidelines and the Country Specific Guidelines while you review this document. We hope you will find this document helpful in reaching a high-quality standard for AutoComplete tasks.
AutoComplete Guidelines:
https://geoworkerz.telusinternational.ai/mapping/StartWorking/Documents/AC%20Guideline s.pdf
Country Specific Guidelines:
https://geoworkerz.telusinternational.ai/mapping/MappingPDF/CSG_EN_PH.pdf
Please follow the flowchart in GL. 4. Relevance when you make your decision on the relevance rating.
Tip: If you are unsure whether a suggestion would satisfy primary or secondary intent, ask yourself:
intent?
SAMPLE A. Please research and rate the relevance rating following the GL. 4.
Relevance Rating Relevance flowchart.
Query:
Viewport Center: [14.58446, 120.99125]
User Location: [14.601410358038308, 121.00820051204715]
Suggestion: ZEN ROOMS MABINI ERMITA
What is the correct relevance rating for this suggestion?
A. Excellent
0
Relevance Rating Steps
Step 1. Review the Query StringStep
If you research, you will find that ‘Ermit
is not a complete word. The user might have not completed typing or mistyped another word.
Step 2. Suggestion Matches?
No. This hotel’s name is ‘Zen Rooms’. “Mabini Ermita’ is merely a location modifier, nonessential to the POI brand name (https://goo.gl/maps/BVhU2KNo6uVH1kQw5). This suggestion should not be considered a match even though the query string’s location modifier matches the suggestion. The user is unlikely to be looking for this particular hotel by typing ‘Ermit’. Following the rating relevance flow chart, apply Bad for the rating. Please review GL 4.1.1. General Match [san die] and 4.2 User Intent [Hollywoo] Arclight Hollywood example.
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. Review GL. 4.1.5. Spell Corrections.
SAMPLE B. Please research and rate the relevance rating following the GL. 4.
Relevance Rating Relevance flowchart.
Query: clememte
Viewport Age: FRESH
Viewport Center: [14.52815, 120.99208]
User Location: [14.52714, 120.98135]
Suggestion: Clemente Jose Bridge
What is the correct relevance rating for this suggestion?
A. Excellent
Step 1. Review the Query String
If you research, you will find that ‘clememte’ is not a complete word. The user might have mistyped since n and m are right next to each other on the keyboard. This is a case when the user query may be misspelled or contains parts that may have been entered incorrectly, and the suggestion corrects that issue. Please review GL.4.1.5. Spell Corrections list of query string examples.
Considering that ‘clememte’ was misspelled and the user was meant to type ‘clemente’, the suggestion ‘Clemente Jose Bridge’ matches the query by spell correction.
The initial rating can be Excellent to Acceptable depending on the user intent. Clemente Jose Bridge is a small bridge that the user is unlikely to be particularly looking for this bridge. It is reasonable to say that the initial User Intent is Good.
Step 4: Research Real World suggestions and apply demotion based on
Once you’ve made your initial rating by establishing that there is a relationship between the query and the suggestion and determining the user intent, you will need to consider the suggestion’s distance and prominence. Most important for this judgment is the concept of “all possible suggestions”. Review GL. 4.3. Distance and Prominence. Clemente Jose Bridge is located outside the viewport however, research shows that there are not many results with ‘Clemente’ near the viewport, thus it is reasonable to rate it Excellent. Excellent due to relevant distance from the viewport and Good due to demotion by prominence. Both ratings are accepted in this case.
Do not demote a suggestion for distance when:
Demote a suggestion for distance when:
Review GL.4.3.2. Distance and 4.6 Few Possible Suggestions and Greater Distance.
SAMPLE C. Please research and rate the relevance rating following the GL. 4.
Relevance Rating Relevance flowchart
Query: Tea &
Viewport Age: FRESH
Viewport Center: [14.37712, 121.32751]
User Location: [14.350043106688338, 121.34125718590937]
Suggestion: ELLY Tea & Kafei Main Branch
What is the correct relevance rating for this suggestion?
A. Excellent
Step 1. Review the Query String
From research, ‘Tea &’ seems not to exist near the user location. The user might have not completed typing after ‘&’.
The suggestion ‘Tea &’ matches the query string. Depending on how ambiguous this purpose is, the rating can range from Excellent to Acceptable. In cases where the suggestion is gibberish, rate Bad. Review GL.4.13. Query and Category Suggestions, Incomplete and Ambiguous Query Suggestions.
ELLY Tea & Kafei Main Branch is not a prominent POI or part of the chain business. However, there is a possibility that the user is looking for a place to drink. It is safe to say that the initial User Intent is Excellent-Good.
Step 4. Research Real World suggestions and apply demotion based on Distance/Prominence
Suggestions are generally expected in or near the viewport and suggestions inside the area cannot be rated Bad for distance alone. Review GL. 2.3.2. Implicit Location. Although the suggestion is significantly far from the user location, from research, this suggestion matches the query, thus rated Excellent. Leniency on distance is acceptable when there are only a few possible suggestions for the query in the requested area. You might notice that there ^ytare cafes such as Kkopi.tea, Thirst.EA and The Tea Gang near the user. Per GL. 4.1.1.
General Match [bristol,] example, these cafes are not considered a match as they don’t take the “&” into consideration. Since the user has made the effort to type it, we expect suggestions to address the “&”.