mira/bertalign
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge branch 'main' of https://github.com/bfsujason/bertalign into main

Browse Source
This commit is contained in:
nlpfun
2021-05-18 01:07:00 +08:00
parent 9771860e30 62e801b0bf
commit 717d235f2a
1 changed files with 17 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
README.md Normal file
View File

@@ -0,0 +1,17 @@
# Bertalign
word embedding-based bilingual sentence aligner
## Evaluation Corpus
This section describes the procedure of creating the evaluation corpora: the manually aligned corpus (MAC) of Chinese-English literary texts and the Bible corpus aligned at the verse level.
### MAC
Firstly, 5 chapters and their translations are sampled from each of the 6 novels included in MAC, obtaining a corpus of 30 bitexts. We then split the corpus into **MAC-Dev** and **MAC-Test** with the former containing 6 chapters and the latter 24 chapters.
The **MAC-Test** is saved in [corpus/mac/test](./corpus/mac/test)
The sampling schemes for building **MAC-Test** can be found at [corpus/mac/test/meta_data.tsv](./corpus/mac/test/meta_data.tsv)
There are 4 subdirectories in **MAC-Test**. The [split](/corpus/mac/test/split) directory contains the sentence-split source texts, target texts and the machine translations of source texts, which are required by **Bleualign** to perform automatic alignment.
The inputs to **Hunalign** are saved in the [tok](/corpus/mac/test/tok) directory.
The emb directory is made up of the overlapping sentences and their embeddings for Vecalign and BertAlign.