- About crossMining
In the advanced options, you can first determine the number of iterations in the first and second phases of the lexicon creation. (It is difficult to make a general recommendation concerning the optimum number of iterations, as this differs from case to case. The value depends especially on the selected language pair, but also on the selected language direction. For example, the optimum values for the language direction German-English are usually different from those for the language direction English-German due to the different morphology of the two languages.)
Therefore, the preset values in the advanced crossMining settings must be considered as a starting point for test purposes. As you conduct your tests, you should optimize these values for your specific data.
For the second phase of the lexicon creation, you can also determine that it is to be skipped. For test purposes, this may be useful if the results of the first phase are to be checked.
In numerical mathematics, the repeated application of the same computing procedure is referred to as "iteration". The results of an iteration step are used as starting values for the next step in order to get closer and closer to a satisfactory final result.
In the first phase of the lexicon creation, the probability of word equivalents is analyzed. In the subsequent second phase, the results of the first phase are further processed. In this context, especially the position of the words is included in the probability calculation.
Moreover, you can specify the minimum probability and the minimum frequency. In this way, you can determine the probability and frequency threshold from which the equivalents are to be written to the statistical lexica.
If the option Save intermediate results is enabled, the results of the analyses will be saved after every iteration. This option is only relevant for tests or optimizations by the Across Professional Services team or for inquiries to be sent to the Across Support.