tromso-2006-01-notes

Opened at 13.xx

• project updates
• project evaluation

Project Evaluation

At least three points for each category:

Project management

Improvements:

• Ilona:
  • the lack of information has been my problem beacuse I haven't always read the cvs logs. One thing is that I don't always understand if some problem affects my work.
• Børre:
  • nothing
• Saara:
  • monday meeting memos: clumsy, difficult to find where the decision was made, because the same text is present in the memo from meeting to meeting.
• Tomi:
  • better monitoring or progress report tool
  • similar comments about the memos as Saara as well
• Linda:
  • more pattern oriented generalized documentation also for linguistics
  • having more specific goals for linguistics and revising them frequently as unexpected things can come up.
• Trond:
  • The project management tools are not functioning the way we want them to do.
  • The weekly meetings are not able to balance between what to include there and what to include elsewhere.
  • Recruiting new people is cumbersome.
• Sjur:
  • better timetables and plans (more detailed)
  • better feedback on time spent on each task (= more bureaucracy, hm…)

Satisfactory:

• Ilona:
  • Most problems small, and get fixed easily
• Saara:
  • good management overall: project members are equal and the team is motivated.
  • communication: newsgroup is working fine and enough in most cases, if you use it.
• Linda:
  • feels very comfortable in the project(s) and with her work
• Trond:
  • the cooperation between divvun and disamb is good
• Sjur:
  • overall progress is good

Programmer work

Improvements:

• Ilona:
  • It's going on very fine, the list for propernames works nicely and is rather easy to use. Some definitions should be maybe added to the lexicon rules, but it's mostly my problem to do it.
• Børre:
  • short meetings on what to prioritize, and how to schedule the tasks
• Saara:
  • better prioritization
  • more comments, more documents and perhaps short specs of the bigger tasks before the coding starts.
• Tomi:
  • process guidelines would be nice
    • specification, implementation, coding, testing & documentation;
    • review each step (at least briefly)
• Linda:
  • it could be interesting to know a bit more about what actually happens on programming level
• Trond:
  • there has been some double work
  • the linguists do not know what is going on (solution: have a session on the common corpus during this week).
  • the xsl procedure is not working as intended.

Satisfactory:

• Saara:
  • reviews are a good idea.
• Linda:
  • technical problems were solved quickly even over long physical distance, good communication with not-programmers
• Trond:
  • there is steady progress
  • new solutions are constantly developed
• Sjur:
  • progress
  • twolc testing

Linguistic work

Improvements:

• Ilona:
  • There is lexc lexicon documentation missing in the source code
  • Capital letters. The program counts words written in capitals to unknown-list, still. Does it mean that the analysator cannot analyse them? If so, it may be bad. Solution: Use the (very slow) all-caps script
  • Checking how the adjectives actually work, right now there are problems with smávis, which doesn't work properly. Well, the unknown-list is a big help in this because it usually shows the failures, unless the word can be analyzed differently and thus gets past the list.
• Linda:
  • regular expressions for date etc.
  • work more with semantics, also in the lexicon
  • more discussion on particular issues
  • we need patterns, stereotype examples respectively "receipes" for writing particular rules
• Trond:
  • We have not been able to create a forum for linguistic exchange. The programmers' goals are the same across the projects (except the clearly different parts), whereas the linguists have slightly different goals.
  • Thomas has been left alone on the 3-part compound issue
  • There are open issues (numerals, G3)
• Sjur:
  • standardised testing for lexc
    • utilising and improving the tools of the testing catalogue
    • including tests in the code?
  • standardised testing for vislcg (Correct! mode should be working)
  • better lexicalisation tools
  • separate newsgroup for linguistic discussions, incl. normative ones
  • code reviews
  • consistent coding style
    • coding style guidelines

Satisfactory:

• Tomi, Børre & Saara:
  • it's excellent!
• Trond:
  • We have good linguistic rule sets.

Other improvements?

• Børre:
  • higher speed on gathering texts (prioritize, discuss if there are problems)
• Linda:
  • more mother tongue speakers for checking (maybe a list of who can be asked about what)
• Trond:
• Sjur:

Notes for the discussion: Testing the morphological transducers? The generation issue