<$BlogRSDUrl$>

Sunday, July 31, 2005

B.A.S.E jumps wmv dangerous unauthorised

Saturday, July 30, 2005

shuttle backflip Discovery spacestation tiles wmv

Friday, July 29, 2005

synchronised miming funny dancing song guys backyard

Thursday, July 28, 2005

song musicvideo NY indie car

Monday, July 25, 2005

Rating Tags with "rev" not "rel"

I've been looking at the microformats.org website and from their RelFAQ section about the "rel" attribute in hyperlinks, it looks as if the "rev" attribute would be more appropriate for rating tags. At the bottom of the FAQ page the final example using their VoteLinks micro format says:

"<a href="http://www2005.org" rev="vote-for">WWW2005 is a great conference</a>" should be interpreted as,
"The current document, as a proxy for the person who wrote it (e.g. the blogger whose blog it is), is a vote-for the resource indicated by http://www2005.org."
This is quite close to a "rated tag" interpretation where...

"<a href="http://www2005.org" rev="rating:5outof5">conference</a>" would be interpreted as,
"The current document, as a proxy for the person who wrote it (e.g. the blogger whose blog it is), rated the resource indicated by http://www2005.org as 5 out of 5 against the criteria of 'conference'".
One difference is that what I am calling the "tag", which in this case is "conference", is used as the criterion against which the rating is made. Also, in the original example the text of the link "WWW2005 is a great conference" represents to readers a positive assessment, but in the second example there is no representation of the rating for readers. I would anticipate that if the concept of "rated tags" were adopted tag clouds would be presented away from the body of any text. That is why I am playing around with the idea of representing the rating associated with the tag through the text size (with limited success as can be seen below - I'm aiming for a single line of tags in blue the size of which varies in accordance with the rating against that tag).

Song

Indie

NY

MusicVideo

Tuesday, July 12, 2005

Ratings, Reviews & Tags

This post looks at some of the characteristics of ratings, reviews and tags before going on to consider the possibility of rating tags in order to enable the discovery of resources. It assumes raters share this objective, rather than the tagging of resources for other personal objectives.

So, in order to discuss the possibility of using tags as criteria against which to provide ratings information, I think I first need to outline what I see as some of the advantages and disadvantages of ratings, reviews and tags when compared to each other.

Ratings are relatively quick and easy to create. The information, whatever numerical scale is used (4-star, 5-star, Out of 10 etc.), is suitable for computational analysis. Ratings are often an assessment of "quality" against some ontological categorisation of the resource, particularly when an "overall" rating is provided. Alternatively some criteria may have been defined such as "price" or "ease of use" and once these criteria are understood by raters and users they provide more specific ratings information.

Reviews, on the other hand, tend to be more extensive and time consuming/difficult to write. They can provide a more comprehensive picture of a resource/product, but are far less suitable for computational analysis (e.g. collaborative filtering techniques). Users also tend to be less likely to provide a review than a rating due to the increased effort involved in writing them, which means there is less data about the resource available.

Tags, being relatively easy to create, look set to become popular. There is an implicit assumption that when a resource is tagged with a particular word, that word is considered to be appropriate for the resource - at least in the opinion of the tagger. Taggers often create a number of tags for a resource so it is probably more appropriate to consider a collection, or cloud, of tags here. In some respects this cloud of tags can be considered to be a highly stripped down review with all the grammatical stuff removed. Whilst this reduces (but does not eliminate) the suitability for human consumption it enhances the suitability for computational analysis. The free form nature of tags means there is little agreed ontological structure; this has its advantages (e.g. low creation costs) and disadvantages (e.g. no agreed definitions) as discussed by Clay Shirky in Ontology is Overrated: Categories, Links, and Tags.

So, having briefly considered the characteristics of ratings, reviews and tags, is there any benefit to using tags as rating criteria? Well, a couple of quick wins are that the minimal ontological categorisation overhead means they would be relatively easy to create as long as the associated rating format/code is also simple - see note below. The provision of an explicit rating also makes it clear that the rater/tagger is rating the resource against a criteria defined by the tag, as opposed to any other personal tagging objectives the tagger may have.

In terms of assessing rated tags against pure ratings, reviews or tags, I have been trying to find a way to visualise the different characteristics. One model is to consider the corners of a triangle as representing the characteristics I outlined above for a rating, a review and a tag. In this model a collection of tags, rather than a single tag, would be represented by a point along one side of the triangle away from the "tag" corner towards the "review" corner - as I think the collection of tags is moving towards becoming a stripped down type of review. If it is also assumed that when a resource is tagged an implicit rating is being provided then this point would move into the triangle - some way towards the rating corner. Finally, the provision of explicit ratings with each tag in the tag cloud would move the characteristics profile further towards the ratings corner of the triangle, as detailed ratings are being provided for each tag criteria.

That last paragraph may not have been so clear without a picture (sorry!), but my point is that a collection of tags with rating information has different characteristics to pure ratings, reviews or tag clouds. Such tag ratings are relatively easy to create, have some of the properties of a stripped down review, and are extremely suitable for computational analysis. They could also be highly distributed in nature if included in peoples' websites, or blogs, for example.

I would like to make it clear that I am not suggesting that all tags have an explicit rating provided with them. Tags without a rating can have a different meaning to the same tag with a rating - an "MP3" tag on its own indicates to me metadata about the resource, in this case that it is an MP3 file, while the tag "MP3" with a rating of 4 out of 5 indicates to me that against the qualities one may expect from an MP3 file (in terms of bitrate, lack of pops and distortions etc) in the opinion of the tagger/rater this resource scores 4 out of 5. I discussed this in my posting on 11 July 2005. Also, some other tags do not provide suitable rating criteria - again touched upon in that post.

In summary, the advantages (and disadvantages) of tagging (folksonomy), when compared to ontological categorisation schemes, can be transferred to rating mechanisms. The associated game for raters/taggers would be to capture the essence of a resource by selecting appropriate tags and rating the resource against each tag criteria. The objective would be to enable the discovery of such resources by others, more powerfully than ontological categorisation rating techniques, by combining rating, review and tag characteristics.

*Note: Representation of Rated Tag
One (completely unofficial and unauthorised) way to represent a rated tag would be to insert a rating into the rel="X" of a hyperlink as follows:
<a href="http://www.resource.net" rel="rating:4outof5">Tag</a>
It looks like a normal Tag and if the font size could be linked to the rating, higher ratings could be represented by larger tags which would result in a human readable representation as well. I'm still struggling to achieve this using the CSS style options!

Monday, July 11, 2005

Trial Illustration

Just a quick attempt to illustrate tags and ratings visually for blog readers...
RockBallad Song MP3 RockCountryDamn, I was hoping to get the tags on the same line. Anyway, the bigger the rating the bigger the tag font size and I've also changed the link behind the tags to the web page where the resource (song/mp3 in this case) can be downloaded. Experiment over.

UPDATE: This looked a lot different in "Preview" when compared to the "Published" version. I was aiming for a line of blue tags sized according to the rating I have given the resource against each tag.

Friday, July 08, 2005

Initial Thoughts on Rating Tags

Before posting more generally on how tags and ratings can possibly be combined I thought I would experiment with the technique first to see how it works. The basic format for each rating would consist of five elements as outlined below:
<rater>...</rater>
<resource>...</resource>
<time>...</time>
<tag>...</tag>
<rating>...</rating>
The <rater>...</rater> entry refers to the URL where the data is stored. So in this case it is the webpage of this blog (http://soundrater.blogspot.com). Normally I would expect it to refer to some form of feed as the data is more suitable for processing by computers rather than reading by individuals. The important point is that the person who created the metadata exercises ownership over the location found in the <rater>...</rater> entry so that others cannot spoof their ratings. It can be considered, by default, the ID of the rater for purposes of subsequent analysis.

The location where the item or content being rated can be accessed is identified in <resource>...</resource>. I prefer to think of items/content as resources that are of potential use to people/computers accessing them, although this may not be a widely held opinion as the examples below can all be considered to be "content". The <time>...</time> entry refers to when the rating was published and would be done in some appropriate format that is widely accepted - I have just made up an entry below.

The entry in the <tag>...</tag> element is chosen by the "rater" for the identified "resource" at the specified "time" as per other tagging systems - so nothing much new there. However, it (the tag) is then used as the basis, or criteria, against which a rating is given in the <rating>...</rating> element for the resource. So each tag enables a different rating, although the "resource" and "rater" are the same. Personally I prefer a rating scale out of 100, despite the popularity of the five-star scheme, as it gives scope for finer granularity.

Each rating would require these five elements to be valid. Also, in the examples below, I have repeated all five entries so that it is easy to determine if each rating has the necessary five elements to make it valid or not. I will discuss this structure in my next posting. So here are a few examples:
<rater>http://soundrater.blogspot.com</rater>
<resource>http://www.lisarein.com/lisarein-jamesandmarybeth.mp3</resource>
<time>20050616T143428</time>
<tag>song</tag>
<rating>85</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://www.lisarein.com/lisarein-jamesandmarybeth.mp3</resource>
<time>20050616T143428</time>
<tag>mp3</tag>
<rating>83</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://www.lisarein.com/lisarein-jamesandmarybeth.mp3</resource>
<time>20050616T143428</time>
<tag>RockCountry</tag>
<rating>65</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://www.lisarein.com/lisarein-jamesandmarybeth.mp3</resource>
<time>20050616T143428</time>
<tag>RockBallad</tag>
<rating>90</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://www.lisarein.com/lisarein-jamesandmarybeth.mp3</resource>
<time>20050616T143428</time>
<tag>attribution1.0</tag>
<rating>100</rating>
So, when tagged as a "song" I have given the resource a rating of 85 out of 100. It is a bit more interesting to rate it against the "mp3" tag as the file is an mp3 file. It could therefore be argued that as a resource either is or isn't an mp3 the rating should either be 0 or 100 (in this case 100). However, in my opinion, the "mp3" tag is a criteria which also encompasses issues such as bitrate, distortions/pops etc. In other words, how does the resource rate against the qualities I expect an mp3 (or more accurately a resource tagged "mp3") to possess? As the song takes about 11 seconds to get started and the bitrate is 160kbps I have given it a rating of 83 out of 100.

I had some difficulty coming up with a tag to represent genre. RockCountry isn't quite right so I only gave it a rating of 65, although I still decided to include this information as it may be useful. Having contacted Lisa Rein (the artist) it appears that RockBallad is appropriate which is why it scores 90/100.

Finally, as this song is licensed under a Creative Commons Attribution 1.0 license I have represented this using <tag>attribution1.0</tag> and giving it a rating entry of 100. This is because the song either is, or isn't, licensed in this way and I can see no reasonable grading scale. It's an issue I need to think about. There is also a question of whether or not this type of tag rating complies with the attribution license requirements as the only reference to Lisa Rein occurs in the <resource>...</resource> entry, and this would not be the case if the MP3 file were hosted on sites such as Archive.org or ourmedia.org. However, as the rating is of a file hosted on someone else's system I tend to think attribution is not required, although I would welcome clarification on this point.

The next example is relatively straightforward:
<rater>http://soundrater.blogspot.com</rater>
<resource>http://www.archive.org/download/Miracle_Car_video/sam_bisbee_miracle_car.mov</resource>
<time>20050616T143428</time>
<tag>song</tag>
<rating>95</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://www.archive.org/download/Miracle_Car_video/sam_bisbee_miracle_car.mov</resource>
<time>20050616T143428</time>
<tag>musicvideo</tag>
<rating>85</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://www.archive.org/download/Miracle_Car_video/sam_bisbee_miracle_car.mov</resource>
<time>20050616T143428</time>
<tag>indie</tag>
<rating>90</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://www.archive.org/download/Miracle_Car_video/sam_bisbee_miracle_car.mov</resource>
<time>20050616T143428</time>
<tag>NY</tag>
<rating>87</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://www.archive.org/download/Miracle_Car_video/sam_bisbee_miracle_car.mov</resource>
<time>20050616T143428</time>
<tag>Attribution-NonCommercial-ShareAlike2.0</tag>
<rating>100</rating>
The major difference between this and the last example is that it is a music video rather than just an audio file - hence the "musicvideo" tag. I included the "NY" tag as it seemed to me to be an integral part of the video, and I guess would help people looking for music videos set in NY if that is what they desired.

Finally, an item I have found quite difficult to tag and rate, possibly because I have been trying to anticipate who would be looking for such content and what they would be using it for (see above). On reflection, I think it is important not to make assumptions about matters such as usage, but instead to focus on capturing the essence of the item when selecting tags with appropriate ratings. By accurately representing a resource in this way the technique could be used as a basis to facilitate its discovery. So in this example I have chosen "song" "apple" "girl" "fun" "squeaky" "wmv" and "copyright" with ratings as detailed below.
<rater>http://soundrater.blogspot.com</rater>
<resource>http://http://www.archive.org/download/Apple_Song/thevideo.wmv</resource>
<time>20050616T143428</time>
<tag>song</tag>
<rating>40</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://http://www.archive.org/download/Apple_Song/thevideo.wmv</resource>
<time>20050616T143428</time>
<tag>apple</tag>
<rating>85</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://http://www.archive.org/download/Apple_Song/thevideo.wmv</resource>
<time>20050616T143428</time>
<tag>girl</tag>
<rating>85</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://http://www.archive.org/download/Apple_Song/thevideo.wmv</resource>
<time>20050616T143428</time>
<tag>fun</tag>
<rating>90</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://http://www.archive.org/download/Apple_Song/thevideo.wmv</resource>
<time>20050616T143428</time>
<tag>squeaky</tag>
<rating>90</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://http://www.archive.org/download/Apple_Song/thevideo.wmv</resource>
<time>20050616T143428</time>
<tag>wmv</tag>
<rating>70</rating>

<rater>http://soundrater.blogspot.com</rater>
<resource>http://http://www.archive.org/download/Apple_Song/thevideo.wmv</resource>
<time>20050616T143428</time>
<tag>copyright</tag>
<rating>100</rating>

It would be interesting to hear how others think this (and the other two items) should be tagged and rated...

This page is powered by Blogger. Isn't yours?