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

Update README.md

Browse Source
This commit is contained in:
bfsujason
2021-06-01 09:51:33 +08:00
committed by GitHub
parent beebd74ba2
commit e928835b8e
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
README.md
View File

@@ -21,7 +21,7 @@ We use [Intertext](https://wanthalf.saga.cz/intertext) to create the manual alig
In order to facilitate system evaluations, we delete the XML tags and save the clean gold alignment file with only sentence IDs in the [gold](./eval/mac/test/gold) directory
### Bible
The **Bible** corpus is located in [corpus/bible](./corpus/bible)
The **Bible** corpus is saved in [corpus/bible](./corpus/bible)
The directory makeup is similar to MAC-Test, except that there is no *intertext* directory for manual alignments.
@@ -34,7 +34,7 @@ All the experiments reported in the paper are conducted using [Google Colab](htt
### Job File
Before performing the automatic alignment, a job file is created for each aligner for batch processing. Each row in the job file represents an alignment task, which is made of three tab-separated file names for source, target and output text.
The job files for MAC-Test and Bible are located in *eval/mac/test/job* and *eval/bible/job*
The job files for MAC-Test and Bible are located in *eval/mac/test/job* and *eval/bible/job* respectively.
### Sentence Embeddings
Before embedding the source and target sentences, we use the following Python script to create the combinations of consecutive sentences: