Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
about judging how relevant advert pages are to the search term the user has
entered.
Users want adverts that are useful to them (are directly related to their
search) and they may consider adverts that make them aware of an
alternative - perhaps a rival item or something complementary. But, no-one
wants to see adverts that have nothing to do with their search term - that's
just annoying!
You are shown those pages and rate them for how relevant they are to the
search query.
In the UHRS tool you are shown the search query and the ads landing page;
you are not shown the search page or the ad.
The guidelines can be overwhelming at first but, this module will help you
approach them in a methodical manner and know how to continue to refer to
them.
The guidelines are in two main sections:
• HRS Judging Guidelines
• Ad Landing Page Judging Appendix
HRS Judging Guidelines
The first section is an in-depth overview of how web pages are rated for
relevance for a user's search. This section explains important aspects such
as query intent and scope, and landing page considerations such as
relevance, freshness, market, authority, and quality. These are the important
takeaways of this section.
Do not approach the guidelines by trying to memorise every possible
scenario of the examples given. as some ratings given in the HRS approach
are different in the ALP approach which you will be mastering. You will
always have access to the guidelines and be able to reference them whilst
you judge. Instead, it is important to understand the foundations of the
project and the judging approach.
Ad Landing Page Judging Appendix
This section highlights the specific approach you will use to rate pages in
ALP. You will see the foundation is the same as the previous HRS section, the
important considerations given to query intent and scope, and the relevance
of the page being rated in relation to that query term.
It s important to understand the ALP ratings and how they applied. Again, do
not attempt to memorise every possibility and example in the guidelines but,
aim to understand the approach to arriving at those rating decisions.
Reading the guidelines cover to cover can be daunting so, we recommend
you review them in smaller sections and use this module as an
accompanying guide.
💡 As we review aspects in this module, after each topic discussed here, go
back to the guidelines and review that topic in the guidelines. It may not
even be a specific section for example, when we discuss 'scope' in this
module, go back to the guidelines and review the sections on scope but also
search the guidelines for the word 'scope' and review the specific examples
where scope has been explained as an important consideration.
Queries
Queries are terms we type into search engines. Think about some of the
queries you have recently searched - what were you hoping to find? were
you looking for information? were you looking to find something very specific
or be directed to a particular website? Were you searching to buy a product?
for information about a product? Both?
When users search for a query they have these thoughts in mind - they have
an intent. A query may have several intents, some more important to the
user than others.
When we decide a user's intent it is not an exact mathematical science! We
must put ourselves into the user's shoes. What sort of person would most
likely me searching that query? What would they like to see returned?
Let's use some flashcards to think about intent.
For the query "Justin Bieber".
How likely is it that the user has the following intents?
(flip the flashcard to see the answer)
Most people searching for 'Justin Bieber' will be his fans or have an interest in
him. Think about a famous person you might like to search for. What would
be your intents? Most likely the person with that name, and their official
website if they have one. Their latest news or products. It's very likely if they
are a singer or a musician that you'd like to see their music channels such as
You Tube channel of all their videos. It's likely you'd like some background on
them, maybe older news. But, it is unlikely you are wanting a very specific
page like a single music video from a few years ago - if you really wanted to
see that, you'd have written it in your search. Which brings us to scope...
We must consider several aspects about the landing page and its
relationship to the user's intent to decide how relevant it is.
Let's think about some of those considerations.
Freshness
Sometimes, the user's query may be looking for something that is happening
at the moment, and we can conclude freshness will be important.
Imagine the following user queries:
If you are rating in a non-Olympic year, it is most likely they are looking for
comprehensive information about Olympic rowing overall, previous events
etc.
But, if you are rating and the Olympics are happening, or it is the year of the
Olympics it is more likely that the user is looking for current Olympic rowing
results.
How can you tell a landing page is fresh and whether that is important?
Look out for creation dates, date articles are posted, copyright date etc?
Then think, how important is the content of the page based on freshness, for
example:
• A page with information about a historical event, it isn't as an
important that the page was written/created recently as the actual
information is not time sensitive.
• A page with latest news or sports results, it is very important it is an
up-to-date, regularly maintained page.
• A page with a sale on it, perhaps a buy this dress at a special price
now, well it is no good if that offer is now out of date!
Scope
We've thought about the scope of the query, what about the scope of the
landing page?
Does the landing page have a narrow or wide scope, and does that match
the users scope?
Look at the table above and think, what kind of queries would those pages
satisfy?
Think about your own searches. How many times have you been
disappointed because a page returned for your query just didn't match what
you hoped for? How many times have you been disappointed because you
were looking for something specific and the landing page suggested was full
of other stuff?! Or you wanted wider information, and something too specific
was returned and you had to start your search again.
Authority
Authority on the internet is a big consideration today. How many times have
you fallen for 'fake news' or been wary of information on the internet?
We all want landing pages we can trust and for different subjects we all have
different sources we trust.
For entertainment news and celebrity gossip we know E! Online is a well-
respected and very popular but we wouldn't go there for in-depth political
news!
If we are looking for health information we want a site where we trust the
advice is safe and has a sound medical background, not unsound advice
written by an anonymous person., this could be really dangerous!
How do you know a site has authority in the subject in question, and can be
trusted? Many times you will be familiar with trusted sites in your market but
here are 8 top tips to consider when assessing a page:
Quality
Perfect
The perfect rating is only given in very specific circumstances.
A Perfect rating means the ads landing page is EXACTLY the page on all of
the internet the user is looking for without any doubt!
We can tell from the user's query that they are looking looking for an exact
place. This may be because they have told us in their query:
"walmart.com" the user has told us they are looking for a specific website,
so if the ALP was www.walmart.com it would be Perfect.
"walmart.com/toys" the user has told us they are looking for a specific
section on a named website so we can be absolutely certain they want to go
to www.walmart.com/toys and so if that was the page we were asked to rate
for their query - Perfect!
"walmart toys" the user's query tells us they specifically want to go to the
Walmart website AND they are looking for the toys section
so www.walmart.com/toys would be Perfect.
"Amazon" whilst the word Amazon may have different meanings, it is most
likely the user means Amazon the very famous online retailer. Imagine
saying the word 'Amazon' to lots of people, and their first thought would
be Amazon.com and if that was the alp returned it would be Perfect.
Think of the following scenarios are they Perfect or not? Flip the card to find
the answer.
Excellent
The Excellent ratings means the ALP strongly satisfies a user's most of very
likely intent.
In other words, this is a page that would make the user very happy 😃 The
page is very relevant to their query, it is useful, good quality, trustworthy,
just the sort of page they were hoping for when they did their search.
Imagine the user is searching "Saturn'"
The most likely intent is the user is looking for information on the planet,
Saturn.
Excellent results would be pages that had trustworthy and authoritative
information about the planet Saturn. Some excellent pages would include:
https://solarsystem.nasa.gov/planets/saturn/overview/ is the introduction
page to a whole section on Saturn from NASA who are very well respected
and an authoritative resource on astronomical and space information.
https://en.wikipedia.org/wiki/Saturn Wikipedia pages can vary drastically and
each must be rated on its own merits. This entry on Saturn is very robust,
well laid out, informative and is properly referenced.
https://www.britannica.com/place/Saturn-planet This is an in-depth article on
Saturn written by a well established and trustworthy encyclopedia. The
authors are listed, references and article versions available. The information
is well laid out, there is an interactive menu to the different sections. There
are some adverts on the page but they are not too intrusive.
Good
A Good alp moderately satisfies a most or very likely intent, or strong
satisfies a likely intent. So what does this really mean?!
A Good page is one which pleases the user. It's a page which answers their
most or very likely intent but there's something about the page which could
be better to make it Excellent. Maybe it could be better by having more
choice, or being a more popular and trustworthy website.
Maybe it's a really great page but it's not the kind of information or action
that the user's may goal was. A Good page may be a step closer to an
Excellent page. When considering if a page is Good or Excellent, ask yourself
'could it be better?', if yes, this is a strong indicator the page is Good.
Think of the following scenarios are they Good or not? Flip the card to find
the answer.
Fair
A fair rating is used when a page only weakly satisfies the user's most or
very likely intent, or strongly satisfies an unlikely intent.
In other words, Fair is when a page will make most people think 'this is just
to say okay but wasn't what I was hoping for' or, it is a page that will make
very few people very happy because it was there intent. Imagine most
people looking for Katy Perry would be disappointed with a profile for a non-
famous person sharing that name with a famous singer but, there would be
a very small amount of people perhaps looking for that particular Katy Perry
- strongly satisfy an unlikely intent, Fair!
Why would a page be weakly satisfying for a user's search? Here are some
common culprits:
Bad
It is very important to understand how the Bad rating is used in ALP.
We strongly recommend study the Bad rating section in the ALP Appendix
before reviewing the Bad section in the HRS section where the Bad examples
would actually be marked Very Bad for ALP rating.
The Bad rating is used in ALP when it weakly satisfies a likely or unlikely
intent. In other words, it isn't a great result for an intent many people didn't
have anyway!!
What can be confusing is the word 'bad' is negative and so you may think
that everything labelled bad must be awful throw out of the results, no good
for anyone but in ALP we use Bad to label a result that has a very weak
relationship to the user's query, there is an outside chance the user may
actually visit the ad landing page.
Let's consider these examples:
Query (en-US): Bangkok
This query is very wide in scope, it is the name of a city and one which is a
popular tourist destination. Users will want a result matching in scope,
maybe an authoritative overview of Bangkok and what it has to offer, a wide
range of hotels and lodgings, reviews etc.
It is likely the user is looking for lodgings in this popular tourist destination
(to be more likely the user would have specified they were looking for
accommodation in their query, don't you think?)
The ALP on the right is a booking page for a single hotel in Bangkok.
It is a result that as an ad landing page a few people may visit or find useful.
It weakly satisfies a likely intent
Rating: Bad
Very Bad
The Very Bad rating is any ad landing page that is poorly satisfying,
regardless of what the user's intent was. It is also used if the ad landing page
satisfies in any way, a very unlikely query intent.
In other words, a Very Bad ad landing page is one no user wants to see!
In the guidelines, examples in the HRS section which are labelled Bad are
labelled Very Bad for the ALP project so remember this when searching for
examples in the guidelines always double check - is is an example of Bad in
the HRS section (if so it is ALP Very Bad) or is it an example of Bad in the ALP
Appendix and then it is Bad as we have just discussed in this module.
Very Bad is also used if the ad landing page is spam or a parked domain, is
misleadingly providing content from another site, tries to affect the user's
computer such as having them download unknown programs that may
contain malware, or is a link redirecting the user to the new location of the
url.
Let's look at some examples:
No Judgement Ratings
No Judgement ratings are made when you cannot assess the relevance of
the page because it is hidden behind a login wall, it is not currently available
or it is in a language you cannot understand.
NJ: Login
The NJ: Login rating is used when you cannot see the page as you would
need to login, such as a subscription newspaper article.
Do not use the NJ: Login if:
• the user's query is asking for the login page. Rate the page for
relevance. Is is the login page the user is looking for? Is there an easy way
for the user to login on the ad landing page or navigate to the login page?
• it is obvious that even you could see behind the login page it would not
be relevant. If the user's query is for "water lillies" but the ALP is for a
newspaper article about share prices but the main article is hidden unless
you could login in, you can confidently assess it wouldn't be relevant and you
can rate it Very Bad.
NJ: PDNL
This rating stands for No Judgement: Page Did Not Load. This rating is used
when a page is temporarily unavailable or displays an error message.
💡 Always refresh the ALP just in case it is a current loading issue.
If the page continues not to load, showing an error message etc then it is
NJ:PDNL.
In ALP we must consider if the PDNL information is the main focus, or if the
page also contains other information which should be rated for relevance.
Such as situation is often seen on commerce sites which says '0 results for
your search' but they provide alternative suggestions and so the ALP should
be rated for how useful this is. Note: 'recently viewed' items is not the same
as suggested items for a search term than has returned 0 results.
The HRS Guidelines section on NJ: PDNL has several examples of ALP pages
and how they would be rated for PDNL, please take a moment to review
them.
NJ: Foreign
The NJ: Foreign rating is only used when their is foreign intent in the user's
query which means the user has written something which suggests they
would welcome an ALP in a foreign language. This may be because they
have asked the query in a foreign language for their market (such as a US
user writing a query in French - sounds like they may be interested in a
French language result!).
If the ALP is in a foreign language, and it is possible it could be useful to the
user who has displayed foreign intent then we use the NJ:Foreign rating. We
are saying this ALP may be useful to the user but we can't tell as we don't
speak that language. For example, if a US user was looking for an entity
associated with a different language, a TV soap opera that only airs in
Spanish, it is reasonable to presume an ALP in Spanish could be of use to
them and if we did not speak Spanish and could not assess the page's
relevance we could rate it NJ:Foreign.
We would not use NJ: Foreign rating:
• if their was no foreign intent in the user query, the foreign language
result would be Very Bad.
• if the user's foreign language intent does not match the foreign
language of the ALP, this would be Very Bad. If a US user asked a query in
Spanish but the ALP was in Chinese, it is very unlikely.
• if the user is looking for someone or an entity with a foreign name. If a
US user was searching for the famous, Serbian tennis player, Novak Djokovic
it is very unlikely they would expect to see a result in the Serbian language.
Similarly, if the US user was looking for the film Les Miserables, this is the
name of a famous English language film and the user would expect English
language results and ALPs in French would be Very Bad.
As well as wanting to visit the site when writing a URL query, as an ALP
project we need to consider whether complimentary and competitive results
would also be useful to the user. Because the intent of wanting to visit the
query site is so strong we must take this into consideration when making our
judgement.
Query ALP Rating Reasoning
Adidas Adidas.com Perfect The ALP would
official website strongly satisfies
the user most
likely intent to
visit the Adidas
homepage,
being interested
in Adidas
products and
brand
Adidas Nike.com official Fair The ALP would
website strongly satisfy
an unlikely
intent to check
out an Adidas
competitor.
Their query
suggest an
overall interest
in the Adidas
brand and
products so is
wide in scope in
that respect.
Seeing a
competitor could
be of interest to
them
Adidas.com Adidas.com Perfect The ALP strongly
official website satisfies the
users most
likely, and very
clear
navigational
intent to visit
adidas.com
Adidas.com Nike.com official Bad The user's intent
website has such a
strong
navigational
intent (they
typed in a url,
they really want
to go there!)
that looking at a
competitor
won't be as
satisfying so it is
Bad.
Even if it is not a URL query, if it has a very strong navigational intent use a
similar approach and considerations. Look out for user's making clear they
have a specific website/page in mind with their wording. For example,
"Manchester United website" has a very strong navigational intent - the
user is telling us they want Manchester United's website. But, "Manchester
United" while the most likely intent is to visit the official website can also
have other strong intents such as wanting latest news, scores etc.
💡 Check out the ALP Appendix in the guidelines for examples of Navigational
queries, including URL queries.
Personal Names
Personal names (and sometimes brand names or sports teams) can sound
very regular to some of us but if we do a side search we discover they are
names very important to other people - they are famous. Maybe you don't
follow the latest reality TV show so have missed out on knowing who the
latest celebrity is! This is why it can be important to do a side search to
double check.
If the name is associated with a famous person or entity, they will be the
most likely intent.
For non-famous names, the likelihood the person on the ALP is who the
person is looking for will depend on several factors. Just how many details
did the user give you, more than just a name?
If a user is looking "John Smith, dentist in Annaheim" this gives you some
more details to be a detective with. An ALP returning a John Smith who is a
baker in Chicago will not be satisfying.
If the user is just looking for a name - how common is that name? A side
search or using an online directory or searching social media can give you a
clue as to home common the name is and the likelihood of how useful an ALP
may be.
Aggregator Sites
An aggregator site brings together content from other sites to show a
collection to the user. If the ALP is from an aggregator site it should still be
rated on whether it is relevant and useful to the user.
Let's consider some popular type of aggregator sites you may see landing
pages from.
Shopping Aggregators
These are a popular and common type of aggregator site. If you are looking
to buy an item, a shopping aggregator site may show you lots of of versions
of that item and different places they are available from. How satisfying it is
to the user will depend on the pages content and how it matches the user's
query intent. Rate the landing page not the fact it is an aggregator page
Search Aggregators
A search aggregator is when a result is presented like a search result, or a
search result from another search provider is presented. This is poorly
satisfying for a user has already put their query in a search engine - who
wants to go from search to search to search!
Unless the ALP includes some viable and useful information actually on the
page that could be rated for relevance, it will be a Very Bad result.
Similarly, if it is an aggregator that is nothing more than lots of adverts, this
is poorly satisfying.
Job Aggregators
If the user is searching for a job, an aggregator can be a useful result. In fact,
job hunters will want to see a landing page with many relevant job postings.
A user who is looking for a job and is searching for a particular job website
(which are often aggregator sites) will consider seeing competitive results
from another job site.
Pay close attention to the user's query and level of detail they have
specified.
Let's look at an example: