Despite the recent popularity of ultrasonic time-of-flight diffraction (TOFD) as a reliable non-destructive testing technique for the inspection of weld defects in steel structures, the critical stages of data processing and interpretation are still performed manually. This is subject to inevitable human errors due to reduced alertness arising from operator fatigue and visual strain when processing large volumes of data. This paper presents techniques developed for accurate sizing and positioning of weld flaws in TOFD D-scan data as an essential stage in a comprehensive TOFD inspection and interpretation system to aid the operator by automating some aspects of the processing and interpretation. Data manipulation and post-processing techniques have been specifically developed for the sizing of weld defects in TOFD data, significantly reducing the sizing and positioning errors for off-axis flaws. The mode-converted waves are utilised to enhance positional accuracy of flaws. The results achieved so far have been promising in tern's of accuracy, consistency and reliability.