File size: 6,368 Bytes
129a814
 
 
 
 
 
 
 
 
4da6a56
129a814
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
---
language: 
- ru

tags:
- sentiment analysis
- Russian
---

## XLM-RoBERTa-Base-ru-sentiment-RuReviews
XLM-RoBERTa-Base-ru-sentiment-RuReviews is a [XLM-RoBERTa-Base](https://huggingface.co/xlm-roberta-base) model fine-tuned on [RuReviews dataset](https://github.com/sismetanin/rureviews) of Russian-language reviews from the ”Women’s Clothes and Accessories” product category on the primary e-commerce site in Russia. 
<table>
<thead>
  <tr>
    <th rowspan="4">Model</th>
    <th rowspan="4">Score<br></th>
    <th rowspan="4">Rank</th>
    <th colspan="12">Dataset</th>
  </tr>
  <tr>
    <td colspan="6">SentiRuEval-2016<br></td>
    <td colspan="2" rowspan="2">RuSentiment</td>
    <td rowspan="2">KRND</td>
    <td rowspan="2">LINIS Crowd</td>
    <td rowspan="2">RuTweetCorp</td>
    <td rowspan="2">RuReviews</td>
  </tr>
  <tr>
    <td colspan="3">TC</td>
    <td colspan="3">Banks</td>
  </tr>
  <tr>
    <td>micro F1</td>
    <td>macro F1</td>
    <td>F1</td>
    <td>micro F1</td>
    <td>macro F1</td>
    <td>F1</td>
    <td>wighted</td>
    <td>F1</td>
    <td>F1</td>
    <td>F1</td>
    <td>F1</td>
    <td>F1</td>
  </tr>
</thead>
<tbody>
  <tr>
    <td>SOTA</td>
    <td>n/s</td>
    <td></td>
    <td>76.71</td>
    <td>66.40</td>
    <td>70.68</td>
    <td>67.51</td>
    <td>69.53</td>
    <td>74.06</td>
    <td>78.50</td>
    <td>n/s</td>
    <td>73.63</td>
    <td>60.51</td>
    <td>83.68</td>
    <td>77.44</td>
  </tr>
  <tr>
    <td>XLM-RoBERTa-Large</td>
    <td>76.37</td>
    <td>1</td>
    <td>82.26</td>
    <td>76.36</td>
    <td>79.42</td>
    <td>76.35</td>
    <td>76.08</td>
    <td>80.89</td>
    <td>78.31</td>
    <td>75.27</td>
    <td>75.17</td>
    <td>60.03</td>
    <td>88.91</td>
    <td>78.81</td>
  </tr>
  <tr>
    <td>SBERT-Large</td>
    <td>75.43</td>
    <td>2</td>
    <td>78.40</td>
    <td>71.36</td>
    <td>75.14</td>
    <td>72.39</td>
    <td>71.87</td>
    <td>77.72</td>
    <td>78.58</td>
    <td>75.85</td>
    <td>74.20</td>
    <td>60.64</td>
    <td>88.66</td>
    <td>77.41</td>
  </tr>
  <tr>
    <td>MBARTRuSumGazeta</td>
    <td>74.70</td>
    <td>3</td>
    <td>76.06</td>
    <td>68.95</td>
    <td>73.04</td>
    <td>72.34</td>
    <td>71.93</td>
    <td>77.83</td>
    <td>76.71</td>
    <td>73.56</td>
    <td>74.18</td>
    <td>60.54</td>
    <td>87.22</td>
    <td>77.51</td>
  </tr>
  <tr>
    <td>Conversational RuBERT</td>
    <td>74.44</td>
    <td>4</td>
    <td>76.69</td>
    <td>69.09</td>
    <td>73.11</td>
    <td>69.44</td>
    <td>68.68</td>
    <td>75.56</td>
    <td>77.31</td>
    <td>74.40</td>
    <td>73.10</td>
    <td>59.95</td>
    <td>87.86</td>
    <td>77.78</td>
  </tr>
  <tr>
    <td>LaBSE</td>
    <td>74.11</td>
    <td>5</td>
    <td>77.00</td>
    <td>69.19</td>
    <td>73.55</td>
    <td>70.34</td>
    <td>69.83</td>
    <td>76.38</td>
    <td>74.94</td>
    <td>70.84</td>
    <td>73.20</td>
    <td>59.52</td>
    <td>87.89</td>
    <td>78.47</td>
  </tr>
  <tr>
    <td>XLM-RoBERTa-Base</td>
    <td>73.60</td>
    <td>6</td>
    <td>76.35</td>
    <td>69.37</td>
    <td>73.42</td>
    <td>68.45</td>
    <td>67.45</td>
    <td>74.05</td>
    <td>74.26</td>
    <td>70.44</td>
    <td>71.40</td>
    <td>60.19</td>
    <td>87.90</td>
    <td>78.28</td>
  </tr>
  <tr>
    <td>RuBERT</td>
    <td>73.45</td>
    <td>7</td>
    <td>74.03</td>
    <td>66.14</td>
    <td>70.75</td>
    <td>66.46</td>
    <td>66.40</td>
    <td>73.37</td>
    <td>75.49</td>
    <td>71.86</td>
    <td>72.15</td>
    <td>60.55</td>
    <td>86.99</td>
    <td>77.41</td>
  </tr>
  <tr>
    <td>MBART-50-Large-Many-to-Many</td>
    <td>73.15</td>
    <td>8</td>
    <td>75.38</td>
    <td>67.81</td>
    <td>72.26</td>
    <td>67.13</td>
    <td>66.97</td>
    <td>73.85</td>
    <td>74.78</td>
    <td>70.98</td>
    <td>71.98</td>
    <td>59.20</td>
    <td>87.05</td>
    <td>77.24</td>
  </tr>
  <tr>
    <td>SlavicBERT</td>
    <td>71.96</td>
    <td>9</td>
    <td>71.45</td>
    <td>63.03</td>
    <td>68.44</td>
    <td>64.32</td>
    <td>63.99</td>
    <td>71.31</td>
    <td>72.13</td>
    <td>67.57</td>
    <td>72.54</td>
    <td>58.70</td>
    <td>86.43</td>
    <td>77.16</td>
  </tr>
  <tr>
    <td>EnRuDR-BERT</td>
    <td>71.51</td>
    <td>10</td>
    <td>72.56</td>
    <td>64.74</td>
    <td>69.07</td>
    <td>61.44</td>
    <td>60.21</td>
    <td>68.34</td>
    <td>74.19</td>
    <td>69.94</td>
    <td>69.33</td>
    <td>56.55</td>
    <td>87.12</td>
    <td>77.95</td>
  </tr>
  <tr>
    <td>RuDR-BERT</td>
    <td>71.14</td>
    <td>11</td>
    <td>72.79</td>
    <td>64.23</td>
    <td>68.36</td>
    <td>61.86</td>
    <td>60.92</td>
    <td>68.48</td>
    <td>74.65</td>
    <td>70.63</td>
    <td>68.74</td>
    <td>54.45</td>
    <td>87.04</td>
    <td>77.91</td>
  </tr>
  <tr>
    <td>MBART-50-Large</td>
    <td>69.46</td>
    <td>12</td>
    <td>70.91</td>
    <td>62.67</td>
    <td>67.24</td>
    <td>61.12</td>
    <td>60.25</td>
    <td>68.41</td>
    <td>72.88</td>
    <td>68.63</td>
    <td>70.52</td>
    <td>46.39</td>
    <td>86.48</td>
    <td>77.52</td>
  </tr>
</tbody>
</table>

The table shows per-task scores and a macro-average of those scores to determine a models’s position on the leaderboard. For datasets with multiple evaluation metrics (e.g., macro F1 and weighted F1 for RuSentiment), we use an unweighted average of the metrics as the score for the task when computing the overall macro-average. The same strategy for comparing models’ results was applied in the GLUE benchmark.

## Citation
If you find this repository helpful, feel free to cite our publication:

```
@article{Smetanin2021Deep,
  author = {Sergey Smetanin and Mikhail Komarov},
  title = {Deep transfer learning baselines for sentiment analysis in Russian},
  journal = {Information Processing & Management},
  volume = {58},
  number = {3},
  pages = {102484},
  year = {2021},
  issn = {0306-4573},
  doi = {0.1016/j.ipm.2020.102484}
}
```

Dataset:
```
@INPROCEEDINGS{Smetanin2019Sentiment,
  author={Sergey Smetanin and Michail Komarov},
  booktitle={2019 IEEE 21st Conference on Business Informatics (CBI)},
  title={Sentiment Analysis of Product Reviews in Russian using Convolutional Neural Networks},
  year={2019},
  volume={01},
  pages={482-486},
  doi={10.1109/CBI.2019.00062},
  ISSN={2378-1963},
  month={July}
}
```