Terrier Users :  Terrier Forum terrier.org
General discussion about using/developing applications using Terrier 
[SOLVED] Difference between Single-Pass and Two-Pass Indexing
Posted by: Abigot ()
Date: March 26, 2012 02:30PM

Good afternoon to all Terrier users,

I read the documentation of Terrier 3.5 about the two types of indexing but I couldn't get the main differences of the two methods.
What I understood is that the Two-Pass indexing create the direct file to store some temporary results while Single-Pass does everything with the memory.

Thus, Two-Pass allows to use directly some retrieval models that use the direct file, whereas you have to create the direct file "manually" if you used Single-Pass and want to use these retrieval models.

Here comes my question (finally) : are these differences the only differences between the two methods?
or does the two methods are also different in terms of performance for the quality of the information retrieval process (meaning do the two methods have different impacts on measures such as precision, recall...)?

Thanks for your time, your answer may save me some usefull time smiling smiley

Anthony

______________________________
Anthony



Edited 1 time(s). Last edit at 03/26/2012 04:50PM by Abigot.

Options: ReplyQuote
Re: Difference between Single-Pass and Two-Pass Indexing
Posted by: ounis ()
Date: March 26, 2012 04:33PM

Hi,

The documentation provides details about the differences between single-pass indexing and the classical two-pass indexing: [terrier.org]

In a nuthshell, the single-pass indexing method is much more efficient, but does not indeed produce a direct file: so no query expansion is supported a priori. However, you can generate the direct file using the Inverted2DirectIndexBuilder - The classical two-pass indexing method is slower but produces both an inverted file and a direct file.

Both the single-pass and classical two-pass indexing methods produce exactly the same inverted file.

Thanks
Iadh

Options: ReplyQuote
Re: Difference between Single-Pass and Two-Pass Indexing
Posted by: Abigot ()
Date: March 26, 2012 04:42PM

Thank you very much Iadh, I must have missed "Both the single-pass and classical two-pass indexing methods produce exactly the same inverted file."

Thank you again for your time,

______________________________
Anthony

Options: ReplyQuote
Re: [SOLVED] Difference between Single-Pass and Two-Pass Indexing
Posted by: craigm ()
Date: March 26, 2012 07:25PM

This is missing from the documentation - I have added an issue to address this in the next release: [terrier.org]

Craig

Options: ReplyQuote


Sorry, only registered users may post in this forum.
This forum powered by Phorum.