You're Invited:Meet the Socket Team at RSAC and BSidesSF 2026, March 23–26.RSVP →
Socket
Book a DemoSign in
Socket
Book a DemoSign in

natasha

Package Overview
Dependencies
Maintainers
2
Versions
13
Alerts
File Explorer

Advanced tools

License
Socket logo

Install Socket

Detect and block malicious and high-risk dependencies

Install

natasha - pypi Package Compare versions

Comparing version
1.2.0
 to
1.3.0
+54
natasha/tests/test_span.py
import pytest
from natasha.span import (
Span,
envelop_spans
)
tests = [
[
[(0, 1)],
[],
[]
],
[
[],
[(0, 1)],
[[]]
],
[
[(0, 1), (1, 2)],
[(1, 2)],
[[(1, 2)]]
],
[
[(0, 1), (1, 2)],
[(0, 1)],
[[(0, 1)]]
],
[
[(0, 1), (1, 2)],
[(0, 1), (1, 2)],
[[(0, 1)], [(1, 2)]]
],
]
def adapt_spans(spans):
for start, stop in spans:
yield Span(start, stop, type=None)
@pytest.mark.parametrize('test', tests)
def test_envelope_spans(test):
spans, envelopes, target = test
spans = list(adapt_spans(spans))
envelopes = list(adapt_spans(envelopes))
target = [
list(adapt_spans(group))
for group in target
]
pred = list(envelop_spans(spans, envelopes))
assert pred == target
+7
-335
natasha.egg-info/PKG-INFO
Metadata-Version: 2.1
Name: natasha
Version: 1.2.0
Version: 1.3.0
Summary: Named-entity recognition for russian language

@@ -332,336 +332,8 @@ Home-page: https://github.com/natasha/natasha

### Segmentation
* Segmentation — <a href="https://github.com/natasha/razdel#quality-performance">Razdel evalualtion section</a>
* Embedding — <a href="https://github.com/natasha/navec#evaluation">Navec evalualtion section</a>
* Morphology — <a href="https://github.com/natasha/slovnet#morphology-1">Slovnet Morph evaluation section</a>
* Syntax — <a href="https://github.com/natasha/slovnet#syntax-1">Slovnet Syntax evaluation section</a>
* NER — <a href="https://github.com/natasha/slovnet#ner-1">Slovnet NER evaluation section</a>
Natasha uses <a href="https://github.com/natasha/razdel">Razdel</a> for text segmentation.
`errors` — number of errors aggregated over 4 datasets, see <a href="https://github.com/natasha/razdel#quality-performance">Razdel evalualtion section</a> for more info.
<!--- token --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>errors</th>
<th>time</th>
</tr>
</thead>
<tbody>
<tr>
<th>razdel.tokenize</th>
<td>5439</td>
<td>9.898350</td>
</tr>
<tr>
<th>mystem</th>
<td>12192</td>
<td>17.210470</td>
</tr>
<tr>
<th>spacy</th>
<td>12288</td>
<td>19.920618</td>
</tr>
<tr>
<th>nltk.word_tokenize</th>
<td>130119</td>
<td>12.405366</td>
</tr>
</tbody>
</table>
<!--- token --->
<!--- sent --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>errors</th>
<th>time</th>
</tr>
</thead>
<tbody>
<tr>
<th>razdel.sentenize</th>
<td>32106</td>
<td>21.989045</td>
</tr>
<tr>
<th>deeppavlov/rusenttokenize</th>
<td>41722</td>
<td>32.535322</td>
</tr>
<tr>
<th>nltk.sent_tokenize</th>
<td>60378</td>
<td>29.916063</td>
</tr>
</tbody>
</table>
<!--- sent --->
### Embedding
Natasha uses <a href="https://github.com/natasha/navec">Navec pretrained embeddings</a>.
`precision` — Average precision over 4 datasets, see <a href="https://github.com/natasha/navec#evaluation">Navec evalualtion section</a> for more info.
<!--- emb1 --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>type</th>
<th>precision</th>
<th>init, s</th>
<th>disk, mb</th>
<th>ram, mb</th>
<th>vocab</th>
</tr>
</thead>
<tbody>
<tr>
<th>hudlit_12B_500K_300d_100q</th>
<td>navec</td>
<td>0.825</td>
<td>1.0</td>
<td>50.6</td>
<td>95.3</td>
<td>500K</td>
</tr>
<tr>
<th>news_1B_250K_300d_100q</th>
<td>navec</td>
<td>0.775</td>
<td>0.5</td>
<td>25.4</td>
<td>47.7</td>
<td>250K</td>
</tr>
<tr>
<th>ruscorpora_upos_cbow_300_20_2019</th>
<td>w2v</td>
<td>0.777</td>
<td>12.1</td>
<td>220.6</td>
<td>236.1</td>
<td>189K</td>
</tr>
<tr>
<th>ruwikiruscorpora_upos_skipgram_300_2_2019</th>
<td>w2v</td>
<td>0.776</td>
<td>15.7</td>
<td>290.0</td>
<td>309.4</td>
<td>248K</td>
</tr>
<tr>
<th>tayga_upos_skipgram_300_2_2019</th>
<td>w2v</td>
<td>0.795</td>
<td>15.7</td>
<td>290.7</td>
<td>310.9</td>
<td>249K</td>
</tr>
<tr>
<th>tayga_none_fasttextcbow_300_10_2019</th>
<td>fasttext</td>
<td>0.706</td>
<td>11.3</td>
<td>2741.9</td>
<td>2746.9</td>
<td>192K</td>
</tr>
<tr>
<th>araneum_none_fasttextcbow_300_5_2018</th>
<td>fasttext</td>
<td>0.720</td>
<td>7.8</td>
<td>2752.1</td>
<td>2754.7</td>
<td>195K</td>
</tr>
</tbody>
</table>
<!--- emb1 --->
### Morphology
Natasha uses <a href="https://github.com/natasha/slovnet#morphology">Slovnet morphology tagger</a>.
`accuracy` — accuracy on news dataset, see <a href="https://github.com/natasha/slovnet#morphology-1">Slovnet evaluation section</a> for more.
<!--- morph1 --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>accuracy</th>
<th>init, s</th>
<th>disk, mb</th>
<th>ram, mb</th>
<th>speed, sents/s</th>
</tr>
</thead>
<tbody>
<tr>
<th>slovnet</th>
<td>0.961</td>
<td>1.0</td>
<td>27</td>
<td>115</td>
<td>532.0</td>
</tr>
<tr>
<th>deeppavlov_bert</th>
<td>0.951</td>
<td>20.0</td>
<td>1393</td>
<td>8704</td>
<td>85.0 (gpu)</td>
</tr>
<tr>
<th>deeppavlov</th>
<td>0.940</td>
<td>4.0</td>
<td>32</td>
<td>10240</td>
<td>90.0 (gpu)</td>
</tr>
<tr>
<th>spacy</th>
<td>0.919</td>
<td>10.9</td>
<td>89</td>
<td>579</td>
<td>30.6</td>
</tr>
<tr>
<th>udpipe</th>
<td>0.918</td>
<td>6.9</td>
<td>45</td>
<td>242</td>
<td>56.2</td>
</tr>
</tbody>
</table>
<!--- morph1 --->
### Syntax
Natasha uses <a href="https://github.com/natasha/slovnet#syntax">Slovnet syntax parser</a>.
`uas`, `las` — accuracy on news dataset, see <a href="https://github.com/natasha/slovnet#syntax-1">Slovnet evaluation section</a> for more.
<!--- syntax1 --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>uas</th>
<th>las</th>
<th>init, s</th>
<th>disk, mb</th>
<th>ram, mb</th>
<th>speed, sents/s</th>
</tr>
</thead>
<tbody>
<tr>
<th>slovnet</th>
<td>0.907</td>
<td>0.880</td>
<td>1.0</td>
<td>27</td>
<td>125</td>
<td>450.0</td>
</tr>
<tr>
<th>deeppavlov_bert</th>
<td>0.962</td>
<td>0.910</td>
<td>34.0</td>
<td>1427</td>
<td>8704</td>
<td>75.0 (gpu)</td>
</tr>
<tr>
<th>spacy</th>
<td>0.876</td>
<td>0.818</td>
<td>10.9</td>
<td>89</td>
<td>579</td>
<td>31.6</td>
</tr>
<tr>
<th>udpipe</th>
<td>0.873</td>
<td>0.823</td>
<td>6.9</td>
<td>45</td>
<td>242</td>
<td>56.2</td>
</tr>
</tbody>
</table>
<!--- syntax1 --->
### NER
Natasha uses <a href="https://github.com/natasha/slovnet#ner">Slovnet NER tagger</a>.
`f1` — score aggregated over 4 datasets, see <a href="https://github.com/natasha/slovnet#ner-1">Slovnet evaluation section</a> for more.
<!--- ner1 --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>PER/LOC/ORG f1</th>
<th>init, s</th>
<th>disk, mb</th>
<th>ram, mb</th>
<th>speed, articles/s</th>
</tr>
</thead>
<tbody>
<tr>
<th>slovnet</th>
<td>0.97/0.91/0.85</td>
<td>1.0</td>
<td>27</td>
<td>205</td>
<td>25.3</td>
</tr>
<tr>
<th>deeppavlov_bert</th>
<td>0.98/0.92/0.86</td>
<td>34.5</td>
<td>2048</td>
<td>6144</td>
<td>13.1 (gpu)</td>
</tr>
<tr>
<th>deeppavlov</th>
<td>0.92/0.86/0.76</td>
<td>5.9</td>
<td>1024</td>
<td>3072</td>
<td>24.3 (gpu)</td>
</tr>
<tr>
<th>pullenti</th>
<td>0.92/0.82/0.64</td>
<td>2.9</td>
<td>16</td>
<td>253</td>
<td>6.0</td>
</tr>
</tbody>
</table>
<!--- ner1 --->
## Support

@@ -671,3 +343,3 @@

- Issues — https://github.com/natasha/natasha/issues
- Commercial support — http://lab.alexkuk.ru/natasha
- Commercial support — https://lab.alexkuk.ru

@@ -674,0 +346,0 @@ ## Development

+2
-1
natasha.egg-info/SOURCES.txt

@@ -45,2 +45,3 @@ README.md

natasha/tests/test_money.py
natasha/tests/test_name.py
natasha/tests/test_name.py
natasha/tests/test_span.py
+1
-1
natasha/__init__.py

@@ -20,2 +20,2 @@

__version__ = '1.2.0'
__version__ = '1.3.0'
+14
-32
natasha/span.py

@@ -23,34 +23,16 @@

def append_sentinel(items, sentinel=None):
for item in items:
yield item
yield sentinel
def envelop_spans(spans, envelopes):
if not spans or not envelopes:
return
spans = append_sentinel(spans)
span = next(spans)
envelopes = append_sentinel(envelopes)
envelope = next(envelopes)
buffer = []
while span and envelope:
if span.start < envelope.start:
span = next(spans)
elif span.stop <= envelope.stop:
buffer.append(span)
span = next(spans)
else:
if buffer:
yield buffer
buffer = []
envelope = next(envelopes)
if buffer:
yield buffer
index = 0
for envelope in envelopes:
chunk = []
while index < len(spans):
span = spans[index]
index += 1
if span.start < envelope.start:
continue
elif span.stop <= envelope.stop:
chunk.append(span)
else:
index -= 1
break
yield chunk
+7
-335
PKG-INFO
Metadata-Version: 2.1
Name: natasha
Version: 1.2.0
Version: 1.3.0
Summary: Named-entity recognition for russian language

@@ -332,336 +332,8 @@ Home-page: https://github.com/natasha/natasha

### Segmentation
* Segmentation — <a href="https://github.com/natasha/razdel#quality-performance">Razdel evalualtion section</a>
* Embedding — <a href="https://github.com/natasha/navec#evaluation">Navec evalualtion section</a>
* Morphology — <a href="https://github.com/natasha/slovnet#morphology-1">Slovnet Morph evaluation section</a>
* Syntax — <a href="https://github.com/natasha/slovnet#syntax-1">Slovnet Syntax evaluation section</a>
* NER — <a href="https://github.com/natasha/slovnet#ner-1">Slovnet NER evaluation section</a>
Natasha uses <a href="https://github.com/natasha/razdel">Razdel</a> for text segmentation.
`errors` — number of errors aggregated over 4 datasets, see <a href="https://github.com/natasha/razdel#quality-performance">Razdel evalualtion section</a> for more info.
<!--- token --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>errors</th>
<th>time</th>
</tr>
</thead>
<tbody>
<tr>
<th>razdel.tokenize</th>
<td>5439</td>
<td>9.898350</td>
</tr>
<tr>
<th>mystem</th>
<td>12192</td>
<td>17.210470</td>
</tr>
<tr>
<th>spacy</th>
<td>12288</td>
<td>19.920618</td>
</tr>
<tr>
<th>nltk.word_tokenize</th>
<td>130119</td>
<td>12.405366</td>
</tr>
</tbody>
</table>
<!--- token --->
<!--- sent --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>errors</th>
<th>time</th>
</tr>
</thead>
<tbody>
<tr>
<th>razdel.sentenize</th>
<td>32106</td>
<td>21.989045</td>
</tr>
<tr>
<th>deeppavlov/rusenttokenize</th>
<td>41722</td>
<td>32.535322</td>
</tr>
<tr>
<th>nltk.sent_tokenize</th>
<td>60378</td>
<td>29.916063</td>
</tr>
</tbody>
</table>
<!--- sent --->
### Embedding
Natasha uses <a href="https://github.com/natasha/navec">Navec pretrained embeddings</a>.
`precision` — Average precision over 4 datasets, see <a href="https://github.com/natasha/navec#evaluation">Navec evalualtion section</a> for more info.
<!--- emb1 --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>type</th>
<th>precision</th>
<th>init, s</th>
<th>disk, mb</th>
<th>ram, mb</th>
<th>vocab</th>
</tr>
</thead>
<tbody>
<tr>
<th>hudlit_12B_500K_300d_100q</th>
<td>navec</td>
<td>0.825</td>
<td>1.0</td>
<td>50.6</td>
<td>95.3</td>
<td>500K</td>
</tr>
<tr>
<th>news_1B_250K_300d_100q</th>
<td>navec</td>
<td>0.775</td>
<td>0.5</td>
<td>25.4</td>
<td>47.7</td>
<td>250K</td>
</tr>
<tr>
<th>ruscorpora_upos_cbow_300_20_2019</th>
<td>w2v</td>
<td>0.777</td>
<td>12.1</td>
<td>220.6</td>
<td>236.1</td>
<td>189K</td>
</tr>
<tr>
<th>ruwikiruscorpora_upos_skipgram_300_2_2019</th>
<td>w2v</td>
<td>0.776</td>
<td>15.7</td>
<td>290.0</td>
<td>309.4</td>
<td>248K</td>
</tr>
<tr>
<th>tayga_upos_skipgram_300_2_2019</th>
<td>w2v</td>
<td>0.795</td>
<td>15.7</td>
<td>290.7</td>
<td>310.9</td>
<td>249K</td>
</tr>
<tr>
<th>tayga_none_fasttextcbow_300_10_2019</th>
<td>fasttext</td>
<td>0.706</td>
<td>11.3</td>
<td>2741.9</td>
<td>2746.9</td>
<td>192K</td>
</tr>
<tr>
<th>araneum_none_fasttextcbow_300_5_2018</th>
<td>fasttext</td>
<td>0.720</td>
<td>7.8</td>
<td>2752.1</td>
<td>2754.7</td>
<td>195K</td>
</tr>
</tbody>
</table>
<!--- emb1 --->
### Morphology
Natasha uses <a href="https://github.com/natasha/slovnet#morphology">Slovnet morphology tagger</a>.
`accuracy` — accuracy on news dataset, see <a href="https://github.com/natasha/slovnet#morphology-1">Slovnet evaluation section</a> for more.
<!--- morph1 --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>accuracy</th>
<th>init, s</th>
<th>disk, mb</th>
<th>ram, mb</th>
<th>speed, sents/s</th>
</tr>
</thead>
<tbody>
<tr>
<th>slovnet</th>
<td>0.961</td>
<td>1.0</td>
<td>27</td>
<td>115</td>
<td>532.0</td>
</tr>
<tr>
<th>deeppavlov_bert</th>
<td>0.951</td>
<td>20.0</td>
<td>1393</td>
<td>8704</td>
<td>85.0 (gpu)</td>
</tr>
<tr>
<th>deeppavlov</th>
<td>0.940</td>
<td>4.0</td>
<td>32</td>
<td>10240</td>
<td>90.0 (gpu)</td>
</tr>
<tr>
<th>spacy</th>
<td>0.919</td>
<td>10.9</td>
<td>89</td>
<td>579</td>
<td>30.6</td>
</tr>
<tr>
<th>udpipe</th>
<td>0.918</td>
<td>6.9</td>
<td>45</td>
<td>242</td>
<td>56.2</td>
</tr>
</tbody>
</table>
<!--- morph1 --->
### Syntax
Natasha uses <a href="https://github.com/natasha/slovnet#syntax">Slovnet syntax parser</a>.
`uas`, `las` — accuracy on news dataset, see <a href="https://github.com/natasha/slovnet#syntax-1">Slovnet evaluation section</a> for more.
<!--- syntax1 --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>uas</th>
<th>las</th>
<th>init, s</th>
<th>disk, mb</th>
<th>ram, mb</th>
<th>speed, sents/s</th>
</tr>
</thead>
<tbody>
<tr>
<th>slovnet</th>
<td>0.907</td>
<td>0.880</td>
<td>1.0</td>
<td>27</td>
<td>125</td>
<td>450.0</td>
</tr>
<tr>
<th>deeppavlov_bert</th>
<td>0.962</td>
<td>0.910</td>
<td>34.0</td>
<td>1427</td>
<td>8704</td>
<td>75.0 (gpu)</td>
</tr>
<tr>
<th>spacy</th>
<td>0.876</td>
<td>0.818</td>
<td>10.9</td>
<td>89</td>
<td>579</td>
<td>31.6</td>
</tr>
<tr>
<th>udpipe</th>
<td>0.873</td>
<td>0.823</td>
<td>6.9</td>
<td>45</td>
<td>242</td>
<td>56.2</td>
</tr>
</tbody>
</table>
<!--- syntax1 --->
### NER
Natasha uses <a href="https://github.com/natasha/slovnet#ner">Slovnet NER tagger</a>.
`f1` — score aggregated over 4 datasets, see <a href="https://github.com/natasha/slovnet#ner-1">Slovnet evaluation section</a> for more.
<!--- ner1 --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>PER/LOC/ORG f1</th>
<th>init, s</th>
<th>disk, mb</th>
<th>ram, mb</th>
<th>speed, articles/s</th>
</tr>
</thead>
<tbody>
<tr>
<th>slovnet</th>
<td>0.97/0.91/0.85</td>
<td>1.0</td>
<td>27</td>
<td>205</td>
<td>25.3</td>
</tr>
<tr>
<th>deeppavlov_bert</th>
<td>0.98/0.92/0.86</td>
<td>34.5</td>
<td>2048</td>
<td>6144</td>
<td>13.1 (gpu)</td>
</tr>
<tr>
<th>deeppavlov</th>
<td>0.92/0.86/0.76</td>
<td>5.9</td>
<td>1024</td>
<td>3072</td>
<td>24.3 (gpu)</td>
</tr>
<tr>
<th>pullenti</th>
<td>0.92/0.82/0.64</td>
<td>2.9</td>
<td>16</td>
<td>253</td>
<td>6.0</td>
</tr>
</tbody>
</table>
<!--- ner1 --->
## Support

@@ -671,3 +343,3 @@

- Issues — https://github.com/natasha/natasha/issues
- Commercial support — http://lab.alexkuk.ru/natasha
- Commercial support — https://lab.alexkuk.ru

@@ -674,0 +346,0 @@ ## Development

+6
-334
README.md

@@ -324,336 +324,8 @@

### Segmentation
* Segmentation — <a href="https://github.com/natasha/razdel#quality-performance">Razdel evalualtion section</a>
* Embedding — <a href="https://github.com/natasha/navec#evaluation">Navec evalualtion section</a>
* Morphology — <a href="https://github.com/natasha/slovnet#morphology-1">Slovnet Morph evaluation section</a>
* Syntax — <a href="https://github.com/natasha/slovnet#syntax-1">Slovnet Syntax evaluation section</a>
* NER — <a href="https://github.com/natasha/slovnet#ner-1">Slovnet NER evaluation section</a>
Natasha uses <a href="https://github.com/natasha/razdel">Razdel</a> for text segmentation.
`errors` — number of errors aggregated over 4 datasets, see <a href="https://github.com/natasha/razdel#quality-performance">Razdel evalualtion section</a> for more info.
<!--- token --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>errors</th>
<th>time</th>
</tr>
</thead>
<tbody>
<tr>
<th>razdel.tokenize</th>
<td>5439</td>
<td>9.898350</td>
</tr>
<tr>
<th>mystem</th>
<td>12192</td>
<td>17.210470</td>
</tr>
<tr>
<th>spacy</th>
<td>12288</td>
<td>19.920618</td>
</tr>
<tr>
<th>nltk.word_tokenize</th>
<td>130119</td>
<td>12.405366</td>
</tr>
</tbody>
</table>
<!--- token --->
<!--- sent --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>errors</th>
<th>time</th>
</tr>
</thead>
<tbody>
<tr>
<th>razdel.sentenize</th>
<td>32106</td>
<td>21.989045</td>
</tr>
<tr>
<th>deeppavlov/rusenttokenize</th>
<td>41722</td>
<td>32.535322</td>
</tr>
<tr>
<th>nltk.sent_tokenize</th>
<td>60378</td>
<td>29.916063</td>
</tr>
</tbody>
</table>
<!--- sent --->
### Embedding
Natasha uses <a href="https://github.com/natasha/navec">Navec pretrained embeddings</a>.
`precision` — Average precision over 4 datasets, see <a href="https://github.com/natasha/navec#evaluation">Navec evalualtion section</a> for more info.
<!--- emb1 --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>type</th>
<th>precision</th>
<th>init, s</th>
<th>disk, mb</th>
<th>ram, mb</th>
<th>vocab</th>
</tr>
</thead>
<tbody>
<tr>
<th>hudlit_12B_500K_300d_100q</th>
<td>navec</td>
<td>0.825</td>
<td>1.0</td>
<td>50.6</td>
<td>95.3</td>
<td>500K</td>
</tr>
<tr>
<th>news_1B_250K_300d_100q</th>
<td>navec</td>
<td>0.775</td>
<td>0.5</td>
<td>25.4</td>
<td>47.7</td>
<td>250K</td>
</tr>
<tr>
<th>ruscorpora_upos_cbow_300_20_2019</th>
<td>w2v</td>
<td>0.777</td>
<td>12.1</td>
<td>220.6</td>
<td>236.1</td>
<td>189K</td>
</tr>
<tr>
<th>ruwikiruscorpora_upos_skipgram_300_2_2019</th>
<td>w2v</td>
<td>0.776</td>
<td>15.7</td>
<td>290.0</td>
<td>309.4</td>
<td>248K</td>
</tr>
<tr>
<th>tayga_upos_skipgram_300_2_2019</th>
<td>w2v</td>
<td>0.795</td>
<td>15.7</td>
<td>290.7</td>
<td>310.9</td>
<td>249K</td>
</tr>
<tr>
<th>tayga_none_fasttextcbow_300_10_2019</th>
<td>fasttext</td>
<td>0.706</td>
<td>11.3</td>
<td>2741.9</td>
<td>2746.9</td>
<td>192K</td>
</tr>
<tr>
<th>araneum_none_fasttextcbow_300_5_2018</th>
<td>fasttext</td>
<td>0.720</td>
<td>7.8</td>
<td>2752.1</td>
<td>2754.7</td>
<td>195K</td>
</tr>
</tbody>
</table>
<!--- emb1 --->
### Morphology
Natasha uses <a href="https://github.com/natasha/slovnet#morphology">Slovnet morphology tagger</a>.
`accuracy` — accuracy on news dataset, see <a href="https://github.com/natasha/slovnet#morphology-1">Slovnet evaluation section</a> for more.
<!--- morph1 --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>accuracy</th>
<th>init, s</th>
<th>disk, mb</th>
<th>ram, mb</th>
<th>speed, sents/s</th>
</tr>
</thead>
<tbody>
<tr>
<th>slovnet</th>
<td>0.961</td>
<td>1.0</td>
<td>27</td>
<td>115</td>
<td>532.0</td>
</tr>
<tr>
<th>deeppavlov_bert</th>
<td>0.951</td>
<td>20.0</td>
<td>1393</td>
<td>8704</td>
<td>85.0 (gpu)</td>
</tr>
<tr>
<th>deeppavlov</th>
<td>0.940</td>
<td>4.0</td>
<td>32</td>
<td>10240</td>
<td>90.0 (gpu)</td>
</tr>
<tr>
<th>spacy</th>
<td>0.919</td>
<td>10.9</td>
<td>89</td>
<td>579</td>
<td>30.6</td>
</tr>
<tr>
<th>udpipe</th>
<td>0.918</td>
<td>6.9</td>
<td>45</td>
<td>242</td>
<td>56.2</td>
</tr>
</tbody>
</table>
<!--- morph1 --->
### Syntax
Natasha uses <a href="https://github.com/natasha/slovnet#syntax">Slovnet syntax parser</a>.
`uas`, `las` — accuracy on news dataset, see <a href="https://github.com/natasha/slovnet#syntax-1">Slovnet evaluation section</a> for more.
<!--- syntax1 --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>uas</th>
<th>las</th>
<th>init, s</th>
<th>disk, mb</th>
<th>ram, mb</th>
<th>speed, sents/s</th>
</tr>
</thead>
<tbody>
<tr>
<th>slovnet</th>
<td>0.907</td>
<td>0.880</td>
<td>1.0</td>
<td>27</td>
<td>125</td>
<td>450.0</td>
</tr>
<tr>
<th>deeppavlov_bert</th>
<td>0.962</td>
<td>0.910</td>
<td>34.0</td>
<td>1427</td>
<td>8704</td>
<td>75.0 (gpu)</td>
</tr>
<tr>
<th>spacy</th>
<td>0.876</td>
<td>0.818</td>
<td>10.9</td>
<td>89</td>
<td>579</td>
<td>31.6</td>
</tr>
<tr>
<th>udpipe</th>
<td>0.873</td>
<td>0.823</td>
<td>6.9</td>
<td>45</td>
<td>242</td>
<td>56.2</td>
</tr>
</tbody>
</table>
<!--- syntax1 --->
### NER
Natasha uses <a href="https://github.com/natasha/slovnet#ner">Slovnet NER tagger</a>.
`f1` — score aggregated over 4 datasets, see <a href="https://github.com/natasha/slovnet#ner-1">Slovnet evaluation section</a> for more.
<!--- ner1 --->
<table border="0" class="dataframe">
<thead>
<tr style="text-align: right;">
<th></th>
<th>PER/LOC/ORG f1</th>
<th>init, s</th>
<th>disk, mb</th>
<th>ram, mb</th>
<th>speed, articles/s</th>
</tr>
</thead>
<tbody>
<tr>
<th>slovnet</th>
<td>0.97/0.91/0.85</td>
<td>1.0</td>
<td>27</td>
<td>205</td>
<td>25.3</td>
</tr>
<tr>
<th>deeppavlov_bert</th>
<td>0.98/0.92/0.86</td>
<td>34.5</td>
<td>2048</td>
<td>6144</td>
<td>13.1 (gpu)</td>
</tr>
<tr>
<th>deeppavlov</th>
<td>0.92/0.86/0.76</td>
<td>5.9</td>
<td>1024</td>
<td>3072</td>
<td>24.3 (gpu)</td>
</tr>
<tr>
<th>pullenti</th>
<td>0.92/0.82/0.64</td>
<td>2.9</td>
<td>16</td>
<td>253</td>
<td>6.0</td>
</tr>
</tbody>
</table>
<!--- ner1 --->
## Support

@@ -663,3 +335,3 @@

- Issues — https://github.com/natasha/natasha/issues
- Commercial support — http://lab.alexkuk.ru/natasha
- Commercial support — https://lab.alexkuk.ru

@@ -666,0 +338,0 @@ ## Development

+1
-1
setup.cfg
[bumpversion]
current_version = 1.2.0
current_version = 1.3.0
files = setup.py natasha/__init__.py

@@ -4,0 +4,0 @@ commit = True

+1
-1
setup.py

@@ -15,3 +15,3 @@

name='natasha',
version='1.2.0',
version='1.3.0',

@@ -18,0 +18,0 @@ description='Named-entity recognition for russian language',