This paper will propose an objective measure based on the information theoretic notion of mutual information, for estimating word association norms from computer readable corpora. (The standard method of obtaining word association norms, testing a few thousand subjects on a few hundred words, is both costly and unreliable.) The proposed measure, the association ratio, estimates word association norms directly from computer readable corpora, making it possible to estimate norms for tens of thousands of words

It is difficult for pure statistics-based machine translation systems to process long sentences. In addition, the domain dependent problem is a key issue under such a framework. Pure rule-based machine translation systems have many human costs in formulating rules and introduce inconsistencies when the number of rules increases. Integration of these two approaches reduces the difficulties associated with both. In this paper, an integrated model for machine translation system is proposed. A partial parsing method is adopted, and the translation process is performed chunk by chunk. In the synthesis module, the word order is locally rearranged within chunks via the Markov model. Since the length of a chunk is much shorter than that of a sentence, the disadvantage of the Markov model in dealing with long distance phenomena is greatly reduced. Structural transfer is fulfilled using a set of rules; in contrast, lexical transfer is resolved using bilingual constraints. Qualitative and quantitative knowledge is employed interleavingly and cooperatively, so that the advantages of these two approaches can be retained.

is to certify that we have read this thesis and that in our opinion it is fully

This paper discusses the problem of classification in component libraries when there is no standard terminology. The discussion is focused on libraries for software reuse [2] and the classification work done in the REBOOT

This study reports results from an empirical study of 109 software maintenance tasks in the software maintenance department of a Norwegian company. Findings were, amongst others, that: . While there was a reduction in the frequency of major unexpected problems from tasks solved by very inexperienced to medium experienced maintainers, additional years of experience did not lead to further reduction. . Increased experience did not lead to higher accuracy when predicting maintenance problems. . The prediction performance seems to be poor. A simple one-variable model outperformed the maintainer's predictions of maintenance problems. Our findings are supported by other studies reporting similar poor learning from experience in other domains and can be explained by learning hindrances in the maintenance process. To improve the learning from experience we recommend a stronger focus on creating better learning situations, higher quality of feedback, and training in the principles of probab...

description of the course

� Recommends ways to make the guidelines more comprehensible to those who seek to understand and implement them � Proposes a new vision for the WCAG 2.0

ectures using different languages. Moreover the systems studied differed in their size by up to two orders of magnitude and in the number of persons involved in their evolution by even more. Since the day to day control of the evolution process was in the hands of humans, differences between the several systems in their short term evolutionary behaviour were to be expected. The similarity of their long term behaviour, however, would have come as a surprise had not the 70s and 80 interpretation of the initial OS/360 observations, their subsequent phenomenological interpretation and the encapsulation of the observations and their interpretations in a set of laws of software evolution [leh74,78,80,96] prepared the investigators for such commonality. Thus the FEAST/1 results were seen as further support for six of the eight laws and supported many of the other conclusions that had been reached. The new evidence did, however, suggest some minor changes to the wording of the laws [leh98b].

day to day control of the evolution process was in the hands of humans, differences between the several systems in their short term evolutionary behaviour was to be expected. The similarity of their long term behaviour, however, would have come as a surprise had it not been anticipated by the 70s and 80 interpretation of the initial OS/360 observations, their subsequent phenomenological interpretation and the encapsulation of the observations and their interpretations in a set of laws of software evolution [3,4,5,6,13] prepared the investigators for such commonality. Thus the FEAST/1 results were seen as further support for six of the eight laws and supported many of the other conclusions that had been reached. The new evidence did, however, suggest some minor changes to the wording of the laws [15]. While, over a period of fifteen years, the laws were being developed no thought was given to any relationship between them. Each aspect of the observed behaviour was seen as characterisi

Three FEAST workshops were held at Imperial College during 1994/5 [fea94/5] to explore the FEAST hypothesis, itself formulated in 1993 [leh94]. The FEAST/1 project (1996- 8) [leh95] funded by EPSRC followed and led, in turn, to FEAST/2 (1999- 2001) [leh98]. Many of the results of these studies have been published over the past few years. They may be found on the FEAST web site at