文件列表:

2017_dataset (0, 2019-07-30)
2017_dataset\1994_cumulative.tsv (1500, 2019-07-30)
2017_dataset\1994_single.tsv (1500, 2019-07-30)
2017_dataset\1995_cumulative.tsv (1751, 2019-07-30)
2017_dataset\1995_single.tsv (1378, 2019-07-30)
2017_dataset\1996_cumulative.tsv (2201, 2019-07-30)
2017_dataset\1996_single.tsv (1559, 2019-07-30)
2017_dataset\1997_cumulative.tsv (2534, 2019-07-30)
2017_dataset\1997_single.tsv (1531, 2019-07-30)
2017_dataset\1998_cumulative.tsv (2863, 2019-07-30)
2017_dataset\1998_single.tsv (1544, 2019-07-30)
2017_dataset\1999_cumulative.tsv (3059, 2019-07-30)
2017_dataset\1999_single.tsv (1463, 2019-07-30)
2017_dataset\2000_cumulative.tsv (3185, 2019-07-30)
2017_dataset\2000_single.tsv (1647, 2019-07-30)
2017_dataset\2001_cumulative.tsv (3438, 2019-07-30)
2017_dataset\2001_single.tsv (1618, 2019-07-30)
2017_dataset\2002_cumulative.tsv (3601, 2019-07-30)
2017_dataset\2002_single.tsv (1382, 2019-07-30)
2017_dataset\2003_cumulative.tsv (3707, 2019-07-30)
2017_dataset\2003_single.tsv (1295, 2019-07-30)
2017_dataset\2004_cumulative.tsv (3932, 2019-07-30)
2017_dataset\2004_single.tsv (1318, 2019-07-30)
2017_dataset\2005_cumulative.tsv (3978, 2019-07-30)
2017_dataset\2005_single.tsv (1059, 2019-07-30)
2017_dataset\2006_cumulative.tsv (4269, 2019-07-30)
2017_dataset\2006_single.tsv (1361, 2019-07-30)
2017_dataset\2007_cumulative.tsv (4329, 2019-07-30)
2017_dataset\2007_single.tsv (1256, 2019-07-30)
2017_dataset\2008_cumulative.tsv (4408, 2019-07-30)
2017_dataset\2008_single.tsv (1265, 2019-07-30)
2017_dataset\2009_cumulative.tsv (4556, 2019-07-30)
2017_dataset\2009_single.tsv (1220, 2019-07-30)
2017_dataset\2010_cumulative.tsv (4666, 2019-07-30)
2017_dataset\2010_single.tsv (1115, 2019-07-30)
2017_dataset\2017_fulltest.json (53960, 2019-07-30)
2019_dataset (0, 2019-07-30)
2019_dataset\2010_single.tsv (489, 2019-07-30)
... ...

# Diachronic Armed Conflicts Diachronic armed conflicts prediction with news texts and word embeddings Code and data for the paper: [*One-to-X analogical reasoning on word embeddings: a case for diachronic armed conflict prediction from news texts*](https://aclweb.org/anthology/papers/W/W19/W19-4724/) (in Proceedings of the 1st International Workshop on Computational Approaches to Historical Language Change, 2019) by Andrey Kutuzov, Erik Velldal and Lilja vrelid ## Embeddings models Word embeddings we trained can be found at the [NLPL Vectors repository](http://vectors.nlpl.eu/repository/): - [CBOW incremental embeddings for Gigaword (1995-2010)](http://vectors.nlpl.eu/repository/11/191.zip) - [CBOW incremental embeddings for News on the Web (2010-2017)](http://vectors.nlpl.eu/repository/11/192.zip) ## Running `python3 multanalogies_diachronic.py --testfile TESTFILE [--visual VISUAL] [--year YEAR] [--finyear FINYEAR] [--lmbd LMBD] [--candidates CANDIDATES] [--threshold] [--modeldir MODELDIR]` For example: `python3 multanalogies_diachronic.py --year 2013 --finyear 2015 --testfile 2019_dataset/fulltest.json --threshold --modeldir NoW/` will learn transformation matrices from locations to armed groups on the embeddings from years 2013 and 2014. These projections will be tested on the next years (2014 and 2015 correspondingly). The `fulltest.json` file will be used as the source of gold data, the system will employ the cosine threshold technique. The word embedding models for all the years must be available in the `NoW` directory. ## Replicating experiments from [Kutuzov et al. 2017] `python3 transform_diachronic.py --modelfile PATH_TO_FIRST_EMBEDDING --reference PATH_TO_FIRST_GOLD_DATA` For example: `python3 transform_diachronic.py --modelfile 2000.bin --reference 2000_single.tsv` will learn the transformation matrix on the embeddings and gold data from 2000 and test it on the next year (2001). `test_all_2017_way.sh` will test all the years and output the results to the `results.tsv` file.

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：