Hi,
when I read something, i usually want to start from the beginning - i.e. the first story of a series. When i search for something, the results a ordered by relevance (or whatever metric is selected) on a per-story basis, which means all parts of a series are scattered around in a "random" order and most of the time, you find some of the later stories before the first part, which to me is really annoying.
Therefore i suggest to add an option to the advanced search to group the results by series somehow.
I know this sounds very simple at first but becomes increasingly complex quickly, so here are a couple of more detailed ideas/thoughts.
1. Presentation:
It would be nice to have a "block" for each series, containing a header with info like the series title, description, story count (and how many of them match the current search), dates of the first and latest stories
and then a list of stories below, which could be limited to all the chapters that actually match the search pattern (but i would list them in their normal order), maybe prepended with the first stories info for better orientation
a nice convenience feature would be if i could expand the list to see all chapters, basically as if the series page was embedded, with the relevant chapters highlighted in some way
2. search & ordering Metric:
the easy option would be to simply replace the first occurence of each series (a story from this series) with a result block for the entire series and remove/skip all subsequent results pointing to this series
a more advanced approach could be:
a story matches a search, whenever it contains the search term(s) and it's relevance is calculated on the current relevance algorithm.
each series group appears in the results when at least one story of the series matches the search
the groups are ordered by the sum, minimum, maximum or average value of all the stories, potentially only taking stories into account that actually match the current search (maybe even give the user the option to choose a merging strategy)
this value could be further refined by weighing in the percentage of stories in the series that match the search
alternatively, the current algorithm could be expanded to cover all stories in a series at once and directly generate a value for the entire series. (in which case i would expect the values to be similar to a averaging merge strategy from above)
depending on the details, algorithms that use an average or minimal score for the group are prone to putting longer stories at a disadvantage, which might not be desired,
whereas minimum or sum based algorithms would probably favour stories with more chapters, so some normalization on the chapter count and length would be a good idea
Another quick and dirty solution might be to simply add a link to the series page to the search results block.
I'd be happy to help testing and tweaking an algorithm, if desired.
What do you guys think?
when I read something, i usually want to start from the beginning - i.e. the first story of a series. When i search for something, the results a ordered by relevance (or whatever metric is selected) on a per-story basis, which means all parts of a series are scattered around in a "random" order and most of the time, you find some of the later stories before the first part, which to me is really annoying.
Therefore i suggest to add an option to the advanced search to group the results by series somehow.
I know this sounds very simple at first but becomes increasingly complex quickly, so here are a couple of more detailed ideas/thoughts.
1. Presentation:
It would be nice to have a "block" for each series, containing a header with info like the series title, description, story count (and how many of them match the current search), dates of the first and latest stories
and then a list of stories below, which could be limited to all the chapters that actually match the search pattern (but i would list them in their normal order), maybe prepended with the first stories info for better orientation
a nice convenience feature would be if i could expand the list to see all chapters, basically as if the series page was embedded, with the relevant chapters highlighted in some way
2. search & ordering Metric:
the easy option would be to simply replace the first occurence of each series (a story from this series) with a result block for the entire series and remove/skip all subsequent results pointing to this series
a more advanced approach could be:
a story matches a search, whenever it contains the search term(s) and it's relevance is calculated on the current relevance algorithm.
each series group appears in the results when at least one story of the series matches the search
the groups are ordered by the sum, minimum, maximum or average value of all the stories, potentially only taking stories into account that actually match the current search (maybe even give the user the option to choose a merging strategy)
this value could be further refined by weighing in the percentage of stories in the series that match the search
alternatively, the current algorithm could be expanded to cover all stories in a series at once and directly generate a value for the entire series. (in which case i would expect the values to be similar to a averaging merge strategy from above)
depending on the details, algorithms that use an average or minimal score for the group are prone to putting longer stories at a disadvantage, which might not be desired,
whereas minimum or sum based algorithms would probably favour stories with more chapters, so some normalization on the chapter count and length would be a good idea
Another quick and dirty solution might be to simply add a link to the series page to the search results block.
I'd be happy to help testing and tweaking an algorithm, if desired.
What do you guys think?
