iproskurina commited on
Commit
ee4df18
·
verified ·
1 Parent(s): 4c1ec95

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +133 -0
README.md CHANGED
@@ -1,3 +1,136 @@
1
  ---
 
 
 
2
  license: mit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ pretty_name: RILEC (Russian L1 Interference Learner English Corpus)
3
+ language:
4
+ - en
5
  license: mit
6
+ arxiv: 2603.07366
7
+ task_categories:
8
+ - token-classification
9
+ - text-classification
10
+ size_categories:
11
+ - 10K<n<100K
12
+ tags:
13
+ - l1-interference
14
+ - grammatical-error-detection
15
+ - learner-corpus
16
+ - learner-english
17
+ - russian-l1
18
+ - second-language-acquisition
19
+ - data-augmentation
20
+ - synthetic-data
21
+ - language-transfer
22
+ - english-as-a-foreign-language
23
+ - english-as-a-second-language
24
+ - ESL
25
+ - REALEC
26
+ - RILEC
27
+ - rule-based-generation
28
+ - large-language-models
29
+ configs:
30
+ - config_name: all_aug
31
+ default: true
32
+ data_files:
33
+ - split: full
34
+ path: 'All_aug - full.tsv'
35
+ sep: "\t"
36
+ header: null
37
+ column_names: [label, span, sentence]
38
+ - config_name: gpt_errorifier
39
+ data_files:
40
+ - split: full
41
+ path: 'GPT_errorifier - full.tsv'
42
+ sep: "\t"
43
+ header: null
44
+ column_names: [label, span, sentence]
45
+ - config_name: llm_errorifier
46
+ data_files:
47
+ - split: full
48
+ path: 'LLM_errorifier - full.tsv'
49
+ sep: "\t"
50
+ header: null
51
+ column_names: [label, span, sentence]
52
+ - config_name: rule_based_errorifier
53
+ data_files:
54
+ - split: full
55
+ path: 'Rule_based_errorifier - full.tsv'
56
+ sep: "\t"
57
+ header: null
58
+ column_names: [label, span, sentence]
59
  ---
60
+
61
+ # RILEC: Russian L1 Interference Learner English Corpus
62
+
63
+ We introduce **RILEC**, a corpus for the detection of L1 Russian interference errors in English learner texts. We use expert-annotated examples from the [Russian Error-Annotated Learner English Corpus (REALEC)](https://realec.org/) and augment them using neural and rule-based approaches. In our experiments, we work with **18,830 sentences** covering five types of L1 interference.
64
+
65
+ We investigate errors associated with the influence of Russian on English learner writing, including lexical choice, literal translation, tense semantics, transliteration, and word forms. Our aim is to improve automatic detection of these errors in learner texts.
66
+
67
+ **[Paper (LREC 2026)](https://aclanthology.org/2026.lrec-1.848/)** · **[arXiv](https://arxiv.org/abs/2603.07366)** · **[Code](https://github.com/harlamovads/RILEC)** · **[Interference error detector](https://huggingface.co/Zlovoblachko/en_stellar_butterfly)**
68
+
69
+ ## L1-interference annotation
70
+
71
+ | Error type | Description |
72
+ | --- | --- |
73
+ | **Copying Expression** | Word-for-word translation of Russian expressions and collocations into English. |
74
+ | **Synonyms** | Selection of an inappropriate English word when a Russian lexeme corresponds to several English equivalents. |
75
+ | **Tense Semantics** | Use of an English tense influenced by the meaning or grammatical form of the corresponding Russian construction. |
76
+ | **Transliteration** | Use of a Russian word written with the Latin alphabet in an English sentence. |
77
+ | **Word Form Transmission** | Transfer of a grammatical category or form from Russian into English. |
78
+
79
+ ## Usage
80
+
81
+ ```
82
+ import csv
83
+ import pandas as pd
84
+ from huggingface_hub import hf_hub_download
85
+
86
+ path = hf_hub_download(
87
+ repo_id="Zlovoblachko/RILEC",
88
+ repo_type="dataset",
89
+ filename="All_aug - full.tsv",
90
+ )
91
+
92
+ data = pd.read_csv(
93
+ path,
94
+ sep="\t",
95
+ header=None,
96
+ names=["label", "span", "sentence"],
97
+ usecols=[0, 1, 2],
98
+ quoting=csv.QUOTE_NONE,
99
+ )
100
+
101
+ labels = {
102
+ "Copying expression", "Synonyms", "Tense semantics",
103
+ "Transliteration", "Word form transmission",
104
+ }
105
+ data = data[data["label"].isin(labels)].reset_index(drop=True)
106
+ print(data.head())
107
+ ```
108
+
109
+ ## L1-interference error detection
110
+
111
+ We formulate L1-interference error detection as a **multi-span classification task** and train RoBERTa-based span classifiers on the augmented subsets.
112
+ We release the trained [RILEC interference detector](https://huggingface.co/Zlovoblachko/en_stellar_butterfly) and the [evaluation data](https://huggingface.co/datasets/Zlovoblachko/L1_true_errors) on Hugging Face.
113
+
114
+ ## Citation
115
+
116
+ ```
117
+ @inproceedings{kharlamova-proskurina-2026-rilec,
118
+ title = "{RILEC}: Detection and Generation of {L}1 {R}ussian Interference Errors in {E}nglish Learner Texts",
119
+ author = "Kharlamova, Darya and
120
+ Proskurina, Irina",
121
+ editor = "Piperidis, Stelios and
122
+ Bel, N{\'u}ria and
123
+ van den Heuvel, Henk and
124
+ Ide, Nancy and
125
+ Krek, Simon and
126
+ Toral, Antonio",
127
+ booktitle = "Proceedings of the Fifteenth Language Resources and Evaluation Conference",
128
+ month = may,
129
+ year = "2026",
130
+ address = "Palma de Mallorca, Spain",
131
+ publisher = "ELRA Language Resource Association",
132
+ url = "https://aclanthology.org/2026.lrec-1.848/",
133
+ doi = "10.63317/5bas35hntskv",
134
+ pages = "10826--10837"
135
+ }
136
+ ```