Pie Model for Lemmatization, POS Tagging, and Morphological Analysis of Eastern Armenian
- 1. École nationale des chartes-PSL / Calfa / LIPN, CNRS UMR 7030
- 2. LIPN, CNRS UMR 7030
- 3. SeDyL, UMR8202, INALCO, CNRS, IRD
Contributors
Data collectors:
- 1. Yerevan State University
Description
The models were trained for lemmatization, POS-tagging, and morphological analysis of Eastern Armenian. The training dataset used was the Eastern Armenian corpus from Universal Dependencies (09/2024 release), comprising 52,950 wordforms (42,337 for training / 5,370 for validation / 5,243 for testing). Sentences cover a large set of documents: blog, fiction, grammar-examples, legal, news, nonfiction. Note that the input data should be pre-tokenized.
The model development was part of the ANR project ANR-21-CE38-0006 "DALiH - Digitizing Armenian Linguistic Heritage", led by Victoria Khurshudyan (Inalco, SeDyL, CNRS, IRD), with initial contributions from Calfa. Models have been developed for the EMNLP 2024 conference (NLP4DH workshop), and rely on the PIE framework.
Data :
- Yavrumyan Marat and ArmTDP team : Github Repository
Results :
For detailed experiments and results, please refer to the linked publication. The following table displays accuracy (f1-score).
|
task_name |
all |
ambiguous-tokens |
known-tokens |
unknown-tokens |
|
abbr |
0.997 (0.8622) |
0.8864 (0.7991) |
0.997 (0.8631) |
0.9986 (0.7497) |
|
adptype |
0.9916 (0.6512) |
0.8246 (0.7106) |
0.9915 (0.6526) |
0.9945 (0.3324) |
|
animacy |
0.9588 (0.92) |
0.8949 (0.8311) |
0.966 (0.9327) |
0.733 (0.6866) |
|
aspect |
0.9909 (0.538) |
0.9727 (0.7321) |
0.993 (0.6421) |
0.9245 (0.4644) |
|
case |
0.9714 (0.9624) |
0.9167 (0.8801) |
0.977 (0.9672) |
0.792 (0.7241) |
|
definite |
0.9738 (0.9664) |
0.9076 (0.8619) |
0.9782 (0.9713) |
0.8346 (0.8481) |
|
degree |
0.9491 (0.2435) |
0.2378 (0.0961) |
0.9482 (0.2434) |
0.9794 (0.4948) |
|
lemma |
0.9909 (0.9502) |
0.9434 (0.7085) |
0.996 (0.9917) |
0.8298 (0.6842) |
|
mood |
0.9942 (0.8632) |
0.9762 (0.847) |
0.9959 (0.8962) |
0.9382 (0.6227) |
|
nametype |
0.9809 (0.359) |
0.8 (0.1778) |
0.9823 (0.3639) |
0.9369 (0.2956) |
|
number |
0.9619 (0.7737) |
0.9301 (0.8898) |
0.9664 (0.782) |
0.8181 (0.6069) |
|
number[psor] |
0.9954 (0.3326) |
0.2 (0.1667) |
0.996 (0.3327) |
0.9753 (0.3292) |
|
numform |
0.991 (0.3975) |
0.7115 (0.4157) |
0.9909 (0.3974) |
0.9952 (0.4994) |
|
numtype |
0.9968 (0.6147) |
0.8571 (0.8381) |
0.997 (0.623) |
0.9904 (0.6035) |
|
person |
0.9936 (0.9532) |
0.986 (0.9641) |
0.9954 (0.9702) |
0.9362 (0.66) |
|
person[psor] |
0.9949 (0.2494) |
0.2 (0.1667) |
0.9955 (0.2494) |
0.9753 (0.2469) |
|
polarity |
0.9785 (0.9558) |
0.9318 (0.9015) |
0.9799 (0.9588) |
0.9348 (0.8632) |
|
pos |
0.9911 (0.9881) |
0.9885 (0.9794) |
0.9959 (0.9928) |
0.8387 (0.5319) |
|
poss |
0.9959 (0.9283) |
0.8213 (0.7389) |
0.9958 (0.9289) |
0.9986 (0.4997) |
|
prontype |
0.995 (0.9162) |
0.9101 (0.8508) |
0.9951 (0.917) |
0.9918 (0.363) |
|
subcat |
0.9805 (0.9232) |
0.8349 (0.836) |
0.984 (0.9351) |
0.8696 (0.7021) |
|
tense |
0.9942 (0.9726) |
0.9767 (0.4651) |
0.995 (0.9776) |
0.9671 (0.7785) |
|
verbform |
0.985 (0.7933) |
0.9565 (0.7614) |
0.9869 (0.7964) |
0.9266 (0.7046) |
|
voice |
0.9753 (0.5995) |
0.8213 (0.807) |
0.9787 (0.6107) |
0.8655 (0.5098) |
Models can be used on Deucalion, the lemmatization service from École nationale des chartes-PSL.
Selected Bibliography:
Vidal-Gorène, C., Khurshudyan, V., & Donabédian-Demopoulos, A. (2020, December). Recycling and comparing morphological annotation models for Armenian diachronic-variational corpus processing. In Proceedings of the 7th Workshop on NLP for Similar Languages, Varieties and Dialects (pp. 90-101).
Vidal-Gorène C., Tomeh N., and Khurshudyan V. (2024, November). Cross-Dialectal Transfer and Zero-Shot Learning for Armenian Varieties: A Comparative Analysis of RNNs, Transformers and LLMs. In Proceedings of the 4th International Conference on Natural Language Processing for Digital Humanities, pages 438–449, Miami, USA. Association for Computational Linguistics.
Files
Files
(637.6 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:8d6f9719ca811acd87fd4739aba182b2
|
24.3 MB | Download |
|
md5:cc4cc90c1007d31c03eef889412d9015
|
24.3 MB | Download |
|
md5:fb89e7745baa19f29160be22c342abe0
|
24.3 MB | Download |
|
md5:4b3698bd9e2e887a84d56acb27a5b8b5
|
24.3 MB | Download |
|
md5:dcf428431b68b0dbe75ac8201525f143
|
24.3 MB | Download |
|
md5:0e94932d78e8fc8d6027978221977567
|
24.3 MB | Download |
|
md5:afd4972ca3680b807c71d84963ffa3f9
|
24.3 MB | Download |
|
md5:743d351c8af0f085c1a185d0e3f42e5f
|
54.3 MB | Download |
|
md5:84b3dc67a4870283ac561774a948488b
|
24.3 MB | Download |
|
md5:f6bce3a5eeecd20573e6f64f46de9994
|
24.3 MB | Download |
|
md5:18fcc7d3e7c50bd49b4082b783dfa5be
|
24.3 MB | Download |
|
md5:678200f987b679aeac7bf257d9e59141
|
24.3 MB | Download |
|
md5:c54c5e83cc270781e37a17457054c395
|
24.3 MB | Download |
|
md5:455e452603af16307f08291de42a5762
|
24.3 MB | Download |
|
md5:adf797662ec835fe35449074675047d6
|
24.3 MB | Download |
|
md5:21e25f361679e025d69a3e8ecbd4dcf9
|
24.3 MB | Download |
|
md5:54ef7a2d2641014ba143a4c527c3b678
|
24.3 MB | Download |
|
md5:d0382cfc8cf7365efe325fb2ccb5a615
|
24.3 MB | Download |
|
md5:9f6011039c800559165429947a807beb
|
23.7 MB | Download |
|
md5:9a75b44607f2b5a09d0c2cb2a34e65df
|
24.3 MB | Download |
|
md5:68fb99cb74b4d8f8fb96d5eecc8557dc
|
24.4 MB | Download |
|
md5:2cd683d50e0ae0abaee381814caa4b54
|
24.3 MB | Download |
|
md5:3f60b4882c8b1ce1d3aa3ce5bc5d123c
|
24.3 MB | Download |
|
md5:4ec2420faaf664f47df04ceb4c69aad2
|
24.3 MB | Download |
|
md5:ead902ddc5ab9e10192fff46594fe531
|
24.3 MB | Download |
Additional details
Related works
- Is described by
- Publication: https://aclanthology.org/2024.nlp4dh-1.42/ (URL)
- Is source of
- Software: https://dh.chartes.psl.eu/deucalion/fr (URL)
Funding
- Agence Nationale de la Recherche
- DALiH – Digitizing Armenian Linguistic Heritage: Armenian Multivariational Corpus and Data Processing ANR-21-CE38-0006