ALEG

What I've done

• Another week spent performing formatting changes and tarting the system up ready for the public beta period. With the increased testing, quite a few actual errors (and not just formatting issues) were detected and fixed.

  Some of the major non-formatting changes are:

  • Oracle's interMedia text engine knobs were twiddled to treat apostrophes as punctuation (rather than part of a word), so searches for "Drover's Dog", "Drovers' Dog" and "Drovers Dog" will all return the same results. A parameter to map diacricitcs to their "base" characters was also enabled, although this is problematic for diacritics without base characters (oh well, probably better than nothing).

  • First known dates for expressions are now calculated and used by the stylesheets.

  • All the stylesheet work has been focussed on IE5, but later in the week I spent some time improving things for Netscape 6 users. We have to decide how much effort we want to spend support the now ancient Netscape 4 browser. Netscape loyalists often have a strong distaste for Microsoft's Internet Explorer, but at least if they download Netscape 6.1, AustLit should look OK (mostly)!

  • A major change was made to subject searches containing booleans which hopefully has made it more useful and intuitive. Previously, a search for "apple not orange" found works which had at least one topic which itself contained the word "apple" but not the word "orange". Now, a a search for "apple not orange" finds works which had at least one topic which itself contained the word "apple" but no topics which contain the word "orange".

  • Alternative titles stored at the expression and manifestation level are now being searched again. At some previous point, we changed the search algorithm to only search titles at the work level were searched.

  • The sort order for works is now always date by title (was just title).

  • Lists of work summaries now have checkboxes and individual works or the whole list can be selected for detailed display.

  • Dropped execution priority of long-running queries so that new short queries are given preference when the machine is very busy.

• The system has been stable this week, so maybe the undocumented parameter on the Java Virtual Machine (JVM) has done the trick. There was a press report of Sun about to release a new JVM specifically to boost performance on our type of Ultra Sparc III processor, but nothing else has surfaced yet.

  The system was quite busy on Thursday and Friday with the increased traffic. I've been running the Solaris "sar" utility to monitor CPU, I/O and memory usage, and memory and I/O are fine, but the CPU is sometimes very busy and facing a queue of work. Some of this is undoubtably due to the haste with which first-known-publication details were added to the work summary displays (including reviewing and critical works), as the volume of XML generated more than doubled. This should be addressed when we have a moment to tune XML production.

  The free Analog log analysis tool by Stephen Turner was installed and usage statistics are now being generated. (WARNING: you must read What the results mean before attempting to interpret anything in these usage statistics!)

• With the system being launched, many people have said kind words about the project. I'd like to add a special thanks to everyone on the team. I've worked on IT projects large and small, applications and infrastructure, in the public and private sector for over 20 years, and it has largely been an unedifying exprience observing waste, mismanagement and a general lack of accountability, enthusiasm, clear goals and results more often than not led by hardware and software vendors with their often ludicrous "value propositions" rather than the needs of the systems' users.

  But the people working on AustLit have been without exception focussed, enthusiastic, skilful and above all, a team dedicated to the best possible outcome using the available resources.

  Given the diversity of people and their physical distribution, this has been an amazing achievement. As with any large project, lots of things went wrong - the data conversion and merging for example was a disaster, and such a setback could have scuppered the project. But rather than panic and finger-pointing, the problem was addressed and solved.

  Much of the credit must go to the incredible vision and leadership of Marie-Louise and the dedication, innovation, planning and operational abilities of Kerry and Annette, but the hard work and great ideas of every member of the team from every institution was instrumental in bringing the project together.

  I'd like to think that we'd have 'noticed' the FRBR model a few days after Judith Pearce subtlely brought it to our intention, but just in case her intervention wasn't unnecessary, a special "thankyou" to her for this, and for marshalling the support of the NLA.

Next Week

• Email summary and detail results.

• Implement Advanced and keyword anywhere searching.

• Revise EAD production for Lu Rees Finding Aids (following changes devised by Megan and Marlene).

• William noticed that some of the place-of-publication data for loaded records is wrong where the name of the place of publication (town or city) occurs in more than one state (eg Richmond, Glebe). I'll investigate when we map the spatial thesaurus. Dan, Chris and Terry have also noticed similar incidents, so I think a very careful look at all place assignments to places occuring in multiple states/countries is warranted. (And Joan noticed Aberdeen NSW -v- Aberdeen Scotland!)

• Multiple creation events for a work as a mechanism for allowing date ranges to be associated with agents responsible for works, eg editors of a periodical.

Links of the week

• In the Beginning was the Command Line - Neal Stephenson

  A great weekend read if you are curious about computer operating system anthropology in the wake of the hoopla surrounding the Windows XP launch.

• The Death Knell of Noisy Searches - IT-director.com.

  Topic maps are able to declare a set of labels for topics and then to point to places where those topics are discussed and addressed. Essentially they provide, under the umbrella of a designated standard topic area, a way to provide information seekers with information across a variety documents. Whereas HTML is a metatag bound to the very document it is describing, topic maps exist independently of the document, allowing users and applications to understand the relationships between those documents.

• ALICE victorious in AI challenge - Rupert Goodwins, ZDNet (UK)

  ALICE got the highest score at this year's contest, held at the Science Museum in London on Saturday, although the silver and gold medals remain unawarded. The silver medal -- and $25,000 -- will go to any program able to convince half the judges that it is human ... To date, the behaviour of the humans involved has been considerably more entertaining than that of the robots

  Here's Alice and here's anti-Alice