| auto-translation |
|
|
- |
| fmt.h |
|
312 |
- |
| linguistproject |
|
|
- |
| numerus.cpp |
|
15415 |
- |
| po.cpp |
|
33862 |
- |
| qm.cpp |
|
25794 |
- |
| qmake-parser |
|
|
- |
| qph.cpp |
|
5196 |
- |
| releasehelper.cpp |
|
4685 |
- |
| releasehelper.h |
|
1096 |
- |
| simtexth.cpp |
How similar are two texts? The approach used here relies on co-occurrence
matrices and is very efficient.
Let's see with an example: how similar are "here" and "hither"? The
co-occurrence matrix M for "here" is M[h,e] = 1, M[e,r] = 1, M[r,e] = 1, and 0
elsewhere; the matrix N for "hither" is N[h,i] = 1, N[i,t] = 1, ...,
N[h,e] = 1, N[e,r] = 1, and 0 elsewhere. The union U of both matrices is the
matrix U[i,j] = max { M[i,j], N[i,j] }, and the intersection V is
V[i,j] = min { M[i,j], N[i,j] }. The score for a pair of texts is
score = (sum of V[i,j] over all i, j) / (sum of U[i,j] over all i, j),
a formula suggested by Arnt Gulbrandsen. Here we have
score = 2 / 6,
or one third.
The implementation differs from this in a few details. Most importantly,
repetitions are ignored; for input "xxx", M[x,x] equals 1, not 2.
|
7266 |
- |
| simtexth.h |
The matrix has 20 * 20 = 400 entries. This requires 50 bytes, or 13
words. Some operations are performed on words for more efficiency.
|
2439 |
- |
| translator.cpp |
|
26221 |
- |
| translator.h |
= "auto" |
9152 |
- |
| translatormessage.cpp |
|
4020 |
- |
| translatormessage.h |
|
6388 |
- |
| trlib |
|
|
- |
| ts.cpp |
|
27462 |
- |
| ts.xsd |
|
7648 |
- |
| validator.cpp |
|
7943 |
- |
| validator.h |
|
2045 |
- |
| xliff.cpp |
Implementation of XLIFF file format for Linguist
|
29360 |
- |
| xmlparser.cpp |
|
2172 |
- |
| xmlparser.h |
|
1056 |
- |