You can't detect allele-specific expression in RNA-seq unless you know the allelic status of the DNA. For example, if you see entirely As expressed at one location, does it mean that the DNA is entirely A, or does it mean that it's a het A/C with allele-specific expression?

You could, however, detect a subset of biased expression patterns. If you see 90% As and 10% Cs, you can be fairly confident that the DNA is A/C and that you're seeing biased expression of the A allele. The caveat here is that you'll miss any site that is 100% expression of one allele or the other (for the reasons outlined above).

Really, to do this type of analysis, you're going to want to have RNA and DNA sequence from the same sample.

Roughly, the process for calling variants in RNA-seq is going to be:

1. Align the reads using an RNA-seq aligner (gsnap, STAR, tophat2)
2. Call variants (multiple options here, also)
3. For each variant, look at the proportion of reads containing the variant

This process will likely have a high false-positive rate unless you use some heuristic filtering.

I use Partek Genomics Suite to estimate allele-specific expression from RNA-Seq data alone.