Week 7

Monday (29th Sept)
Looked through Makino's latest code and commented out the polling code. Re-compiled and rebuilt the solution, then ran TextDisplay.exe again. Observed that the memory usage still seemed to be increasing, albeit alot slower than before. Commented out the code for advertising as well, then re-compiled and rebuilt the solution again. Left TextDisplay.exe running to test for any memory leak again.
Wednesday (1st Oct)
Ken informed me that TextDisplay.exe was still leaking memory, at a rate of about 128K/min. Modified the code and passed it on to Ken.
Thursday (2nd Oct)
Prepared for presentation to visitors from Philips, but none of them showed up in the end. Perhaps the visit got cancelled again. In any case, while I was awaiting their arrival, I worked more on the visualisation code. Apart from commenting out the polling and advertisement(s) code, I studied Makino's code in blogwallviz.cpp and textgl.cpp. Suspect that the memory leak may have arisen from some of the manually allocated memory to glyph pointers, in particular in the getGlyphs function. Added a couple of 'delete's to try and solve the leaks by returning the allocated memory to the memory store. Left TextDisplay.exe running again for testing.
Friday (3rd Oct)
The modifications from yesterday did not seem to help. TextDisplay.exe was still leaking memory. Talked to Ken and Nimesha and they suggested that I email Makino to see if he is able to solve the problem, since he is certainly more familiar with his own code. E-mailed Makino with the modified code attached. Studying "Handbook of Natural Language Processing" to find ways to utilise the PosTagged definitions as obtained last week. Parsing techniques and semantic analysis may prove to be useful subsequent steps. It may be possible to analyse the structure of discourse and thus recognise the intention of the SMS input.