Gravs answers make sense if the number is the slope of the trend line of the dataset. In that case, the greater the magnitude of the slope, the more trend there is in the data; if the data were random, you'd expect the slope to be about 0 (flat line).
Well, if as I stated, the correlation is the slope of the trend line, then only the magnitude matters in the strength of the correlation. A strong negative correlation is still a strong correlation; for example, average rainfall and days of draught would have a negative slope.