Update README.md
Browse files
README.md
CHANGED
|
@@ -72,7 +72,7 @@
|
|
| 72 |
</head>
|
| 73 |
<body>
|
| 74 |
<h1>COMBO vs Trankit vs Stanza — UD Evaluation Report</h1>
|
| 75 |
-
<p class="subtitle">
|
| 76 |
COMBO on UD 2.17 · Trankit on UD 2.5 · Stanza on UD 2.12
|
| 77 |
·
|
| 78 |
<span style="background:#d4edda;padding:1px 4px;font-size:0.8em">highest</span>
|
|
@@ -132,10 +132,6 @@
|
|
| 132 |
<tr><td class="system">Trankit[large]</td><td class="highest">99.99</td><td class="highest">100.00</td><td class="highest">99.98</td><td class="highest">99.64</td><td class="highest">98.17</td><td class="highest">98.01</td><td class="highest">97.65</td><td>98.30</td><td class="highest">95.48</td><td class="highest">93.97</td><td class="highest">93.06</td><td class="highest">90.13</td><td class="highest">90.92</td></tr>
|
| 133 |
<tr><td class="system">Trankit[base]</td><td>99.96</td><td class="highest">100.00</td><td>99.95</td><td>99.42</td><td>97.40</td><td>97.12</td><td>96.68</td><td>98.13</td><td>94.97</td><td>93.27</td><td>92.08</td><td>88.13</td><td>89.85</td></tr>
|
| 134 |
<tr><td class="system">Stanza</td><td>99.82</td><td class="highest">100.00</td><td class="lowest">99.57</td><td class="lowest">98.52</td><td class="lowest">94.59</td><td class="lowest">93.64</td><td class="lowest">92.92</td><td class="lowest">97.46</td><td class="lowest">91.09</td><td class="lowest">88.71</td><td class="lowest">86.78</td><td class="lowest">80.27</td><td class="lowest">84.87</td></tr>
|
| 135 |
-
<tr><td class="treebank" rowspan="4">Czech<br><small>CLTT</small></td><td class="system">COMBO[base]</td><td class="lowest">88.05</td><td class="lowest">60.17</td><td class="lowest">88.06</td><td class="lowest">86.98</td><td class="lowest">77.81</td><td class="lowest">77.75</td><td class="lowest">75.73</td><td class="lowest">83.31</td><td class="lowest">72.70</td><td class="lowest">70.76</td><td class="lowest">72.89</td><td class="lowest">61.07</td><td class="lowest">68.61</td></tr>
|
| 136 |
-
<tr><td class="system">Trankit[large]</td><td class="highest">99.89</td><td>98.89</td><td class="highest">99.85</td><td class="highest">99.17</td><td class="highest">93.80</td><td class="highest">94.02</td><td class="highest">93.53</td><td class="highest">97.81</td><td class="highest">92.53</td><td class="highest">90.76</td><td class="highest">88.90</td><td class="highest">81.07</td><td class="highest">86.35</td></tr>
|
| 137 |
-
<tr><td class="system">Trankit[base]</td><td>99.82</td><td class="highest">100.00</td><td>99.76</td><td>98.93</td><td>93.36</td><td>93.68</td><td>93.21</td><td>97.66</td><td>90.20</td><td>88.01</td><td>85.13</td><td>77.66</td><td>82.80</td></tr>
|
| 138 |
-
<tr><td class="system">Stanza</td><td>99.23</td><td>98.23</td><td>99.23</td><td>97.38</td><td>88.25</td><td>88.22</td><td>87.42</td><td>95.78</td><td>83.53</td><td>79.71</td><td>76.52</td><td>66.32</td><td>74.12</td></tr>
|
| 139 |
<tr><td class="treebank" rowspan="4">Czech<br><small>FICTREE</small></td><td class="system">COMBO[base]</td><td class="lowest">99.68</td><td class="lowest">94.11</td><td class="lowest">99.59</td><td>98.65</td><td>96.63</td><td>97.10</td><td>96.03</td><td class="highest">98.69</td><td>93.86</td><td>92.45</td><td>91.34</td><td>87.09</td><td>90.12</td></tr>
|
| 140 |
<tr><td class="system">Trankit[large]</td><td class="highest">99.98</td><td class="highest">99.50</td><td class="highest">99.98</td><td class="highest">99.11</td><td class="highest">97.26</td><td class="highest">97.99</td><td class="highest">97.06</td><td>98.64</td><td class="highest">96.50</td><td class="highest">94.85</td><td class="highest">93.44</td><td class="highest">89.79</td><td class="highest">91.46</td></tr>
|
| 141 |
<tr><td class="system">Trankit[base]</td><td>99.97</td><td>99.38</td><td>99.97</td><td>98.94</td><td>96.47</td><td>97.09</td><td>96.12</td><td>98.61</td><td>95.85</td><td>93.86</td><td>92.10</td><td>87.13</td><td>90.16</td></tr>
|
|
@@ -153,8 +149,6 @@
|
|
| 153 |
<tr><td class="system">Trankit[large]</td><td>99.36</td><td class="highest">92.60</td><td>99.36</td><td>96.52</td><td>95.57</td><td>96.63</td><td>94.99</td><td class="highest">97.37</td><td>92.25</td><td>89.68</td><td>86.51</td><td>82.56</td><td class="highest">84.19</td></tr>
|
| 154 |
<tr><td class="system">Trankit[base]</td><td class="lowest">99.21</td><td>91.09</td><td class="lowest">99.21</td><td class="lowest">96.20</td><td class="lowest">95.18</td><td class="lowest">96.25</td><td class="lowest">94.53</td><td>97.19</td><td>91.09</td><td>88.18</td><td>84.63</td><td>80.47</td><td>82.44</td></tr>
|
| 155 |
<tr><td class="system">Stanza</td><td class="highest">99.87</td><td>78.61</td><td class="highest">99.87</td><td>96.67</td><td>95.92</td><td>96.83</td><td>95.20</td><td class="lowest">95.86</td><td class="lowest">88.25</td><td class="lowest">85.32</td><td class="lowest">80.00</td><td class="lowest">76.72</td><td class="lowest">75.15</td></tr>
|
| 156 |
-
<tr><td class="treebank" rowspan="2">English<br><small>ATIS</small></td><td class="system">COMBO[base]</td><td class="lowest">99.48</td><td class="lowest">7.90</td><td class="lowest">99.48</td><td class="lowest">98.26</td><td class="lowest">99.48</td><td class="lowest">97.34</td><td class="lowest">96.79</td><td class="lowest">99.11</td><td class="lowest">64.87</td><td class="lowest">63.57</td><td class="lowest">55.72</td><td class="lowest">48.78</td><td class="lowest">55.46</td></tr>
|
| 157 |
-
<tr><td class="system">Stanza</td><td class="highest">100.00</td><td class="highest">87.55</td><td class="highest">100.00</td><td class="highest">99.15</td><td class="highest">100.00</td><td class="highest">98.74</td><td class="highest">98.19</td><td class="highest">99.65</td><td class="highest">94.45</td><td class="highest">92.61</td><td class="highest">90.08</td><td class="highest">87.49</td><td class="highest">89.72</td></tr>
|
| 158 |
<tr><td class="treebank" rowspan="1">English<br><small>CHILDES</small></td><td class="system">COMBO[base]</td><td>99.86</td><td>98.97</td><td>99.13</td><td>96.22</td><td>94.49</td><td>99.13</td><td>93.65</td><td>97.15</td><td>91.78</td><td>89.11</td><td>86.93</td><td>82.49</td><td>84.90</td></tr>
|
| 159 |
<tr><td class="treebank" rowspan="2">English<br><small>ESLSPOK</small></td><td class="system">COMBO[base]</td><td class="lowest">99.80</td><td class="lowest">91.23</td><td class="lowest">99.80</td><td class="highest">98.52</td><td class="highest">98.43</td><td class="lowest">99.80</td><td class="highest">97.99</td><td class="lowest">99.80</td><td class="lowest">92.75</td><td class="highest">91.03</td><td class="lowest">88.97</td><td class="lowest">86.91</td><td class="lowest">88.97</td></tr>
|
| 160 |
<tr><td class="system">Stanza</td><td class="highest">100.00</td><td class="highest">92.81</td><td class="highest">100.00</td><td class="lowest">98.28</td><td class="lowest">98.32</td><td class="highest">100.00</td><td class="lowest">97.88</td><td class="highest">100.00</td><td class="highest">93.12</td><td class="lowest">90.73</td><td class="highest">89.43</td><td class="highest">87.45</td><td class="highest">89.43</td></tr>
|
|
@@ -222,12 +216,6 @@
|
|
| 222 |
<tr><td class="system">Trankit[large]</td><td>99.85</td><td>93.50</td><td>99.85</td><td class="highest">98.41</td><td class="highest">98.41</td><td class="highest">96.34</td><td class="highest">95.84</td><td>96.73</td><td class="highest">95.25</td><td class="highest">93.87</td><td class="highest">91.74</td><td class="highest">85.40</td><td class="highest">86.54</td></tr>
|
| 223 |
<tr><td class="system">Trankit[base]</td><td class="lowest">99.75</td><td>93.57</td><td class="lowest">99.75</td><td>98.04</td><td>98.04</td><td>95.41</td><td>94.73</td><td>96.55</td><td>94.16</td><td>92.80</td><td>89.84</td><td>82.39</td><td>84.83</td></tr>
|
| 224 |
<tr><td class="system">Stanza</td><td class="highest">99.90</td><td class="highest">93.66</td><td class="highest">99.90</td><td class="lowest">97.71</td><td class="lowest">97.71</td><td class="lowest">94.72</td><td class="lowest">94.01</td><td class="lowest">96.05</td><td class="lowest">91.22</td><td class="lowest">88.87</td><td class="lowest">84.19</td><td class="lowest">77.99</td><td class="lowest">79.08</td></tr>
|
| 225 |
-
<tr><td class="treebank" rowspan="4">Hebrew<br><small>HTB</small></td><td class="system">COMBO[base]</td><td class="lowest">99.65</td><td class="lowest">99.59</td><td class="lowest">56.55</td><td class="lowest">55.11</td><td class="lowest">55.09</td><td class="lowest">53.23</td><td class="lowest">52.94</td><td class="lowest">54.94</td><td class="lowest">39.57</td><td class="lowest">37.97</td><td class="lowest">37.82</td><td class="lowest">30.34</td><td class="lowest">36.34</td></tr>
|
| 226 |
-
<tr><td class="system">Trankit[large]</td><td>99.81</td><td>99.69</td><td class="highest">96.31</td><td class="highest">94.32</td><td class="highest">94.32</td><td class="highest">93.03</td><td class="highest">92.39</td><td class="highest">93.48</td><td class="highest">88.41</td><td class="highest">86.04</td><td class="highest">82.23</td><td class="highest">74.92</td><td class="highest">78.35</td></tr>
|
| 227 |
-
<tr><td class="system">Trankit[base]</td><td>99.79</td><td class="highest">100.00</td><td>96.03</td><td>93.75</td><td>93.75</td><td>91.96</td><td>91.24</td><td>93.21</td><td>87.02</td><td>84.55</td><td>80.34</td><td>72.38</td><td>76.52</td></tr>
|
| 228 |
-
<tr><td class="system">Stanza</td><td class="highest">99.98</td><td class="highest">100.00</td><td>92.29</td><td>89.99</td><td>89.99</td><td>88.68</td><td>88.03</td><td>89.71</td><td>79.38</td><td>77.18</td><td>71.95</td><td>65.35</td><td>68.81</td></tr>
|
| 229 |
-
<tr><td class="treebank" rowspan="2">Hebrew<br><small>IAHLTWIKI</small></td><td class="system">COMBO[base]</td><td class="lowest">99.31</td><td class="lowest">88.21</td><td class="lowest">91.44</td><td class="lowest">88.75</td><td class="lowest">88.76</td><td class="lowest">83.92</td><td class="lowest">82.85</td><td class="lowest">89.77</td><td class="lowest">77.98</td><td class="lowest">76.44</td><td class="lowest">71.64</td><td class="lowest">60.09</td><td class="lowest">69.51</td></tr>
|
| 230 |
-
<tr><td class="system">Stanza</td><td class="highest">99.64</td><td class="highest">96.78</td><td class="highest">94.61</td><td class="highest">91.69</td><td class="highest">91.69</td><td class="highest">86.88</td><td class="highest">86.14</td><td class="highest">92.51</td><td class="highest">83.23</td><td class="highest">81.19</td><td class="highest">75.58</td><td class="highest">64.88</td><td class="highest">73.14</td></tr>
|
| 231 |
<tr><td class="treebank" rowspan="4">Hindi<br><small>HDTB</small></td><td class="system">COMBO[base]</td><td class="lowest">99.83</td><td class="lowest">99.20</td><td class="lowest">99.83</td><td>97.99</td><td>97.67</td><td>94.05</td><td>92.50</td><td class="highest">98.81</td><td>95.30</td><td class="lowest">92.31</td><td class="lowest">89.24</td><td class="lowest">79.38</td><td class="lowest">87.99</td></tr>
|
| 232 |
<tr><td class="system">Trankit[large]</td><td>99.88</td><td class="highest">99.91</td><td>99.88</td><td class="highest">98.01</td><td class="highest">97.70</td><td class="lowest">93.91</td><td>92.38</td><td class="lowest">96.54</td><td class="highest">95.95</td><td class="highest">92.96</td><td class="highest">89.79</td><td>79.69</td><td class="highest">88.47</td></tr>
|
| 233 |
<tr><td class="system">Trankit[base]</td><td>99.89</td><td>99.64</td><td>99.89</td><td class="lowest">97.77</td><td class="lowest">97.38</td><td>94.03</td><td class="lowest">92.33</td><td class="lowest">96.54</td><td>95.68</td><td>92.70</td><td>89.59</td><td>79.60</td><td>88.28</td></tr>
|
|
|
|
| 72 |
</head>
|
| 73 |
<body>
|
| 74 |
<h1>COMBO vs Trankit vs Stanza — UD Evaluation Report</h1>
|
| 75 |
+
<p class="subtitle">106 treebanks · F1 scores ·
|
| 76 |
COMBO on UD 2.17 · Trankit on UD 2.5 · Stanza on UD 2.12
|
| 77 |
·
|
| 78 |
<span style="background:#d4edda;padding:1px 4px;font-size:0.8em">highest</span>
|
|
|
|
| 132 |
<tr><td class="system">Trankit[large]</td><td class="highest">99.99</td><td class="highest">100.00</td><td class="highest">99.98</td><td class="highest">99.64</td><td class="highest">98.17</td><td class="highest">98.01</td><td class="highest">97.65</td><td>98.30</td><td class="highest">95.48</td><td class="highest">93.97</td><td class="highest">93.06</td><td class="highest">90.13</td><td class="highest">90.92</td></tr>
|
| 133 |
<tr><td class="system">Trankit[base]</td><td>99.96</td><td class="highest">100.00</td><td>99.95</td><td>99.42</td><td>97.40</td><td>97.12</td><td>96.68</td><td>98.13</td><td>94.97</td><td>93.27</td><td>92.08</td><td>88.13</td><td>89.85</td></tr>
|
| 134 |
<tr><td class="system">Stanza</td><td>99.82</td><td class="highest">100.00</td><td class="lowest">99.57</td><td class="lowest">98.52</td><td class="lowest">94.59</td><td class="lowest">93.64</td><td class="lowest">92.92</td><td class="lowest">97.46</td><td class="lowest">91.09</td><td class="lowest">88.71</td><td class="lowest">86.78</td><td class="lowest">80.27</td><td class="lowest">84.87</td></tr>
|
|
|
|
|
|
|
|
|
|
|
|
|
| 135 |
<tr><td class="treebank" rowspan="4">Czech<br><small>FICTREE</small></td><td class="system">COMBO[base]</td><td class="lowest">99.68</td><td class="lowest">94.11</td><td class="lowest">99.59</td><td>98.65</td><td>96.63</td><td>97.10</td><td>96.03</td><td class="highest">98.69</td><td>93.86</td><td>92.45</td><td>91.34</td><td>87.09</td><td>90.12</td></tr>
|
| 136 |
<tr><td class="system">Trankit[large]</td><td class="highest">99.98</td><td class="highest">99.50</td><td class="highest">99.98</td><td class="highest">99.11</td><td class="highest">97.26</td><td class="highest">97.99</td><td class="highest">97.06</td><td>98.64</td><td class="highest">96.50</td><td class="highest">94.85</td><td class="highest">93.44</td><td class="highest">89.79</td><td class="highest">91.46</td></tr>
|
| 137 |
<tr><td class="system">Trankit[base]</td><td>99.97</td><td>99.38</td><td>99.97</td><td>98.94</td><td>96.47</td><td>97.09</td><td>96.12</td><td>98.61</td><td>95.85</td><td>93.86</td><td>92.10</td><td>87.13</td><td>90.16</td></tr>
|
|
|
|
| 149 |
<tr><td class="system">Trankit[large]</td><td>99.36</td><td class="highest">92.60</td><td>99.36</td><td>96.52</td><td>95.57</td><td>96.63</td><td>94.99</td><td class="highest">97.37</td><td>92.25</td><td>89.68</td><td>86.51</td><td>82.56</td><td class="highest">84.19</td></tr>
|
| 150 |
<tr><td class="system">Trankit[base]</td><td class="lowest">99.21</td><td>91.09</td><td class="lowest">99.21</td><td class="lowest">96.20</td><td class="lowest">95.18</td><td class="lowest">96.25</td><td class="lowest">94.53</td><td>97.19</td><td>91.09</td><td>88.18</td><td>84.63</td><td>80.47</td><td>82.44</td></tr>
|
| 151 |
<tr><td class="system">Stanza</td><td class="highest">99.87</td><td>78.61</td><td class="highest">99.87</td><td>96.67</td><td>95.92</td><td>96.83</td><td>95.20</td><td class="lowest">95.86</td><td class="lowest">88.25</td><td class="lowest">85.32</td><td class="lowest">80.00</td><td class="lowest">76.72</td><td class="lowest">75.15</td></tr>
|
|
|
|
|
|
|
| 152 |
<tr><td class="treebank" rowspan="1">English<br><small>CHILDES</small></td><td class="system">COMBO[base]</td><td>99.86</td><td>98.97</td><td>99.13</td><td>96.22</td><td>94.49</td><td>99.13</td><td>93.65</td><td>97.15</td><td>91.78</td><td>89.11</td><td>86.93</td><td>82.49</td><td>84.90</td></tr>
|
| 153 |
<tr><td class="treebank" rowspan="2">English<br><small>ESLSPOK</small></td><td class="system">COMBO[base]</td><td class="lowest">99.80</td><td class="lowest">91.23</td><td class="lowest">99.80</td><td class="highest">98.52</td><td class="highest">98.43</td><td class="lowest">99.80</td><td class="highest">97.99</td><td class="lowest">99.80</td><td class="lowest">92.75</td><td class="highest">91.03</td><td class="lowest">88.97</td><td class="lowest">86.91</td><td class="lowest">88.97</td></tr>
|
| 154 |
<tr><td class="system">Stanza</td><td class="highest">100.00</td><td class="highest">92.81</td><td class="highest">100.00</td><td class="lowest">98.28</td><td class="lowest">98.32</td><td class="highest">100.00</td><td class="lowest">97.88</td><td class="highest">100.00</td><td class="highest">93.12</td><td class="lowest">90.73</td><td class="highest">89.43</td><td class="highest">87.45</td><td class="highest">89.43</td></tr>
|
|
|
|
| 216 |
<tr><td class="system">Trankit[large]</td><td>99.85</td><td>93.50</td><td>99.85</td><td class="highest">98.41</td><td class="highest">98.41</td><td class="highest">96.34</td><td class="highest">95.84</td><td>96.73</td><td class="highest">95.25</td><td class="highest">93.87</td><td class="highest">91.74</td><td class="highest">85.40</td><td class="highest">86.54</td></tr>
|
| 217 |
<tr><td class="system">Trankit[base]</td><td class="lowest">99.75</td><td>93.57</td><td class="lowest">99.75</td><td>98.04</td><td>98.04</td><td>95.41</td><td>94.73</td><td>96.55</td><td>94.16</td><td>92.80</td><td>89.84</td><td>82.39</td><td>84.83</td></tr>
|
| 218 |
<tr><td class="system">Stanza</td><td class="highest">99.90</td><td class="highest">93.66</td><td class="highest">99.90</td><td class="lowest">97.71</td><td class="lowest">97.71</td><td class="lowest">94.72</td><td class="lowest">94.01</td><td class="lowest">96.05</td><td class="lowest">91.22</td><td class="lowest">88.87</td><td class="lowest">84.19</td><td class="lowest">77.99</td><td class="lowest">79.08</td></tr>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 219 |
<tr><td class="treebank" rowspan="4">Hindi<br><small>HDTB</small></td><td class="system">COMBO[base]</td><td class="lowest">99.83</td><td class="lowest">99.20</td><td class="lowest">99.83</td><td>97.99</td><td>97.67</td><td>94.05</td><td>92.50</td><td class="highest">98.81</td><td>95.30</td><td class="lowest">92.31</td><td class="lowest">89.24</td><td class="lowest">79.38</td><td class="lowest">87.99</td></tr>
|
| 220 |
<tr><td class="system">Trankit[large]</td><td>99.88</td><td class="highest">99.91</td><td>99.88</td><td class="highest">98.01</td><td class="highest">97.70</td><td class="lowest">93.91</td><td>92.38</td><td class="lowest">96.54</td><td class="highest">95.95</td><td class="highest">92.96</td><td class="highest">89.79</td><td>79.69</td><td class="highest">88.47</td></tr>
|
| 221 |
<tr><td class="system">Trankit[base]</td><td>99.89</td><td>99.64</td><td>99.89</td><td class="lowest">97.77</td><td class="lowest">97.38</td><td>94.03</td><td class="lowest">92.33</td><td class="lowest">96.54</td><td>95.68</td><td>92.70</td><td>89.59</td><td>79.60</td><td>88.28</td></tr>
|