PageRenderTime 58ms CodeModel.GetById 19ms RepoModel.GetById 1ms app.codeStats 0ms

/scripts/ems/web/analysis.php

https://bitbucket.org/ufal/mosesdecoder
PHP | 1447 lines | 1255 code | 119 blank | 73 comment | 350 complexity | 943906f8c9e3be0eb8f9f0e5c4d0f98b MD5 | raw file
Possible License(s): BSD-3-Clause, GPL-3.0, LGPL-3.0

Large files files are truncated, but you can click here to view the full file

    

  1. <?php 
    2. 

  2. 

  3. # main page frame, triggers the loading of parts
    4. 

  4. function show_analysis() {
    5. 

  5.   global $task,$user,$setup,$id,$set;
    6. 

  6.   global $dir;
    7. 

  7. 

  8.   head("Analysis: $task ($user), Set $set, Run $id");
    9. 

  9. 

  10. ?><script>
    11. 

  11. function show(field,sort,count,filter) {
    12. 

  12.   var url = '?analysis=' + field + '_show'
    13. 

  13.       + '&setup=<?php print $setup ?>'
    14. 

  14.       + '&id=<?php print $id ?>'
    15. 

  15.       + '&set=<?php print $set ?>'
    16. 

  16.       + '&sort=' + sort
    17. 

  17.       + '&count=' + count
    18. 

  18.       + '&filter=' + filter;
    19. 

  19.   new Ajax.Updater(field, url, { method: 'get', evalScripts: true });
    20. 

  20. }
    21. 

  21. function ngram_show(type,order,count,sort,smooth) {
    22. 

  22.   var url = '?analysis=ngram_' + type + '_show'
    23. 

  23.       + '&setup=<?php print $setup ?>'
    24. 

  24.       + '&id=<?php print $id ?>'
    25. 

  25.       + '&set=<?php print $set ?>'
    26. 

  26.       + '&order=' + order
    27. 

  27.       + '&smooth=' + smooth
    28. 

  28.       + '&sort=' + sort
    29. 

  29.       + '&count=' + count;
    30. 

  30.   var field = (type == "precision" ? "nGramPrecision" : "nGramRecall") + order;
    31. 

  31.   new Ajax.Updater(field, url, { method: 'get', evalScripts: true });
    32. 

  32. }
    33. 

  33. function generic_show(field,parameters) {
    34. 

  34.   var url = '?analysis=' + field + '_show'
    35. 

  35.       + '&setup=<?php print $setup ?>'
    36. 

  36.       + '&id=<?php print $id ?>'
    37. 

  37.       + '&set=<?php print $set ?>'
    38. 

  38.       + '&' + parameters;
    39. 

  39.   new Ajax.Updater(field, url, { method: 'get', evalScripts: true });
    40. 

  40. }
    41. 

  41. function highlight_phrase(sentence,phrase) {
    42. 

  42.   var input = "input-"+sentence+"-"+phrase;
    43. 

  43.   $(input).setStyle({ borderColor: 'red' });
    44. 

  44.   var output = "output-"+sentence+"-"+phrase;
    45. 

  45.   $(output).setStyle({ borderColor: 'red' });
    46. 

  46. }
    47. 

  47. function show_word_info(sentence,cc,tc,te) {
    48. 

  48.   var info = "info-"+sentence;
    49. 

  49.   document.getElementById(info).innerHTML = ''+cc+' occurrences in corpus, '+tc+' distinct translations, translation entropy: '+te;
    50. 

  50.   $(info).setStyle({ opacity: 1 });
    51. 

  51. }
    52. 

  52. function lowlight_phrase(sentence,phrase) {
    53. 

  53.   var input = "input-"+sentence+"-"+phrase;
    54. 

  54.   $(input).setStyle({ borderColor: 'black' });
    55. 

  55.   var output = "output-"+sentence+"-"+phrase;
    56. 

  56.   $(output).setStyle({ borderColor: 'black' });
    57. 

  57. }
    58. 

  58. function hide_word_info(sentence) {
    59. 

  59.   var info = "info-"+sentence;
    60. 

  60.   $(info).setStyle({ opacity: 0 });
    61. 

  61. }
    62. 

  62. function show_biconcor(sentence,phrase) {
    63. 

  63.   var div = "biconcor-"+sentence;
    64. 

  64.   var url = '?analysis=biconcor'
    65. 

  65.             + '&setup=<?php print $setup ?>&id=<?php print get_biconcor_version($dir,$set,$id); ?>&set=<?php print $set ?>'
    66. 

  66. 	    + '&sentence=' + sentence
    67. 

  67.             + '&phrase=' + encodeURIComponent(phrase);
    68. 

  68.   document.getElementById(div).innerHTML = "<center><img src=\"spinner.gif\" width=48 height=48></center>";
    69. 

  69.   $(div).setStyle({ borderStyle: 'solid', 'border-width': '3px', borderColor: 'black' });
    70. 

  70.   new Ajax.Updater(div, url, { method: 'get', evalScripts: true });
    71. 

  71. }
    72. 

  72. function close_biconcor(sentence) {
    73. 

  73.   var div = "biconcor-"+sentence;
    74. 

  74.   document.getElementById(div).innerHTML = "";
    75. 

  75.   $(div).setStyle({ borderStyle: 'none', 'border-width': '0px', borderColor: 'white' });
    76. 

  76. }
    77. 

  77. 

  78. </script>
    79. 

  79. </head>
    80. 

  80. <body>
    81. 

  81. <div id="nGramSummary"><?php ngram_summary()  ?></div>
    82. 

  82. <div id="CoverageDetails"></div>
    83. 

  83. <div id="PrecisionByCoverage"></div>
    84. 

  84. <div id="PrecisionRecallDetails"></div>
    85. 

  85. <div id="bleu">(loading...)</div>
    86. 

  86. <script language="javascript">
    87. 

  87. show('bleu','',5,'');
    88. 

  88. </script>
    89. 

  89. </body></html>
    90. 

  90. <?php
    91. 

  91. }
    92. 

  92. 

  93. function precision_by_coverage() {
    94. 

  94.   global $experiment,$evalset,$dir,$set,$id;
    95. 

  95.   $img_width = 1000;
    96. 

  96. 

  97.   print "<h3>Precision of Input Words by Coverage</h3>";
    98. 

  98.   print "The graphs display what ratio of words of a specific type are translated correctly (yellow), and what ratio is deleted (blue).";
    99. 

  99.   print " The extend of the boxes is scaled on the x-axis by the number of tokens of the displayed type.";
    100. 

  100. 

  101.   // load data
    102. 

  102.   $data = file(get_current_analysis_filename("precision","precision-by-corpus-coverage"));
    103. 

  103.   $total = 0;
    104. 

  104.   $log_info = array();
    105. 

  105.   for($i=0;$i<count($data);$i++) {
    106. 

  106.     $item = split("\t",$data[$i]);
    107. 

  107.     $info[$item[0]]["precision"] = $item[1];
    108. 

  108.     $info[$item[0]]["delete"] = $item[2];
    109. 

  109.     $info[$item[0]]["length"] = $item[3];
    110. 

  110.     $info[$item[0]]["total"] = $item[4];
    111. 

  111.     $total += $item[4];
    112. 

  112.     $log_count = -1;
    113. 

  113.     if ($item[0]>0) {
    114. 

  114. 	$log_count = (int) (log($item[0])/log(2));
    115. 

  115.     }
    116. 

  116.     if (!array_key_exists($log_count,$log_info)) {
    117. 

  117. 	$log_info[$log_count]["precision"] = 0;
    118. 

  118. 	$log_info[$log_count]["delete"] = 0;
    119. 

  119. 	$log_info[$log_count]["length"] = 0;
    120. 

  120. 	$log_info[$log_count]["total"] = 0;
    121. 

  121.     }
    122. 

  122.     $log_info[$log_count]["precision"] += $item[1];
    123. 

  123.     $log_info[$log_count]["delete"] += $item[2];
    124. 

  124.     $log_info[$log_count]["length"] += $item[3];
    125. 

  125.     $log_info[$log_count]["total"] += $item[4];    
    126. 

  126.   }
    127. 

  127.   print "<h4>By log<sub>2</sub>-count in the training corpus</h4>";
    128. 

  128.   precision_by_coverage_graph("byCoverage",$log_info,$total,$img_width,SORT_NUMERIC);
    129. 

  129. 

  130.   # load factored data
    131. 

  131.   $d = dir("$dir/evaluation/$set.analysis.".get_precision_analysis_version($dir,$set,$id));
    132. 

  132.   while (false !== ($file = $d->read())) {
    133. 

  133.     if (preg_match('/precision-by-corpus-coverage.(.+)$/',$file, $match)) {
    134. 

  134.       precision_by_coverage_factored($img_width,$total,$file,$match[1]);
    135. 

  135.     }
    136. 

  136.   }
    137. 

  137. }
    138. 

  138. 

  139. function precision_by_coverage_factored($img_width,$total,$file,$factor_id) {
    140. 

  140.   global $dir,$set,$id;
    141. 

  141.   $data = file(get_current_analysis_filename("precision",$file));
    142. 

  142.   for($i=0;$i<count($data);$i++) {
    143. 

  143.     $item = split("\t",$data[$i]);
    144. 

  144.     $factor = $item[0];
    145. 

  145.     $count = $item[1];
    146. 

  146.     $info_factored[$factor][$count]["precision"] = $item[2];
    147. 

  147.     $info_factored[$factor][$count]["delete"] = $item[3];
    148. 

  148.     $info_factored[$factor][$count]["length"] = $item[4];
    149. 

  149.     $info_factored[$factor][$count]["total"] = $item[5];
    150. 

  150.     $info_factored_sum[$factor]["precision"] += $item[2];
    151. 

  151.     $info_factored_sum[$factor]["delete"] += $item[3];
    152. 

  152.     $info_factored_sum[$factor]["length"] += $item[4];
    153. 

  153.     $info_factored_sum[$factor]["total"] += $item[5];
    154. 

  154.     $total_factored[$factor] += $item[5];
    155. 

  155.     $log_count = -1;
    156. 

  156.     if ($count>0) {
    157. 

  157. 	$log_count = (int) (log($count)/log(2));
    158. 

  158.     }
    159. 

  159.     $log_info_factored[$factor][$log_count]["precision"] += $item[2];
    160. 

  160.     $log_info_factored[$factor][$log_count]["delete"] += $item[3];
    161. 

  161.     $log_info_factored[$factor][$log_count]["length"] += $item[4];
    162. 

  162.     $log_info_factored[$factor][$log_count]["total"] += $item[5];    
    163. 

  163.   }
    164. 

  164.   print "<h4>By factor ".factor_name("input",$factor_id)."</h4>";
    165. 

  165.   precision_by_coverage_graph("byFactor",$info_factored_sum,$total,$img_width,SORT_STRING);
    166. 

  166. 

  167.   print "<h4>For each factor, by log<sub>2</sub>-count in the corpus</h4>";
    168. 

  168.   foreach ($log_info_factored as $factor => $info) {
    169. 

  169.     if ($total_factored[$factor]/$total > 0.01) {
    170. 

  170.       print "<table style=\"display:inline;\"><tr><td align=center><font size=-2><b>$factor</b></font></td></tr><tr><td align=center>";
    171. 

  171.       precision_by_coverage_graph("byCoverageFactor$factor",$info,$total_factored[$factor],10+2*$img_width*$total_factored[$factor]/$total,SORT_NUMERIC);
    172. 

  172.       print "</td></tr></table>";
    173. 

  173.     }
    174. 

  174.   }
    175. 

  175. }
    176. 

  176. 

  177. function precision_by_word($type) {
    178. 

  178.   global $dir,$set,$id;
    179. 

  179.   $byCoverage = -2;
    180. 

  180.   $byFactor = "false";
    181. 

  181.   if ($type == "byCoverage") {
    182. 

  182.       $byCoverage = (int) $_GET["type"];
    183. 

  183.   }
    184. 

  184.   else if ($type == "byFactor") {
    185. 

  185.       $byFactor = $_GET["type"];
    186. 

  186.   }
    187. 

  187.   else if (preg_match("/byCoverageFactor(.+)/",$type,$match)) {
    188. 

  188.       $byCoverage = (int) $_GET["type"];
    189. 

  189.       $byFactor = $match[1];
    190. 

  190.   }
    191. 

  191. 

  192.   $data = file(get_current_analysis_filename("precision","precision-by-input-word"));
    193. 

  193.   for($i=0;$i<count($data);$i++) {
    194. 

  194.     $line = rtrim($data[$i]);
    195. 

  195.     $item = split("\t",$line);
    196. 

  196. 

  197.     //# filter for count
    198. 

  198.     $count = $item[4];
    199. 

  199.     $log_count = -1;
    200. 

  200.     if ($count>0) {
    201. 

  201. 	$log_count = (int) (log($count)/log(2));
    202. 

  202.     }
    203. 

  203.     if ($byCoverage != -2 && $byCoverage != $log_count) {
    204. 

  204. 	continue;
    205. 

  205.     }
    206. 

  206.    
    207. 

  207.     //# filter for factor
    208. 

  208.     $word = $item[5];
    209. 

  209.     if ($byFactor != "false" && $byFactor != $item[6]) {
    210. 

  210. 	continue;
    211. 

  211.     }
    212. 

  212. 

  213.     $info[$word]["precision"] = $item[0];
    214. 

  214.     $info[$word]["delete"] = $item[1];
    215. 

  215.     $info[$word]["length"] = $item[2];
    216. 

  216.     $info[$word]["total"] = $item[3];
    217. 

  217.     $total += $item[3];
    218. 

  218.   }
    219. 

  219. 

  220.   print "<table border=1><tr><td align=center>Count</td><td align=center colspan=2>Precision</td><td align=center colspan=2>Delete</td><td align=center>Length</td></tr>\n";
    221. 

  221.   foreach ($info as $word => $wordinfo) {
    222. 

  222.       print "<tr><td align=center><a href=\"javascript:show('bleu','order',5,'".base64_encode($word)."')\">$word</a></td>";
    223. 

  223.       printf("<td align=right>%.1f%s</td><td align=right><font size=-1>%.1f/%d</font></td>",$wordinfo["precision"]/$wordinfo["total"]*100,"%",$wordinfo["precision"],$wordinfo["total"]);
    224. 

  224.       printf("<td align=right>%.1f%s</td><td align=right><font size=-1>%d/%d</font></td>",$wordinfo["delete"]/$wordinfo["total"]*100,"%",$wordinfo["delete"],$wordinfo["total"]);
    225. 

  225.       printf("<td align=right>%.3f</td>",$wordinfo["length"]/$wordinfo["total"]);
    226. 

  226.       print "</tr>";
    227. 

  227.   }
    228. 

  228.   print "</table>\n";
    229. 

  229. }
    230. 

  230. 

  231. function precision_by_coverage_latex($name,$log_info,$total,$img_width,$sort_type) {
    232. 

  232.   $keys = array_keys($log_info);
    233. 

  233.   sort($keys,$sort_type);
    234. 

  234.   
    235. 

  235.   $img_width /= 100;
    236. 

  236.   print "<div id=\"LatexToggle$name\" onClick=\"document.getElementById('Latex$name').style.display = 'block'; this.style.display = 'none';\" style=\"display:none;\"><font size=-2>(show LaTeX)</font></div>\n";
    237. 

  237.   print "<div id=\"Latex$name\" style=\"display:none;\">\n";
    238. 

  238.   print "<code>\\begin{tikzpicture}<br>";
    239. 

  239. 

  240.   print "% co-ordinates for precision<br>";
    241. 

  241.   for($line=0;$line<=9;$line++) {
    242. 

  242.       $height = 1.8-$line/10*1.8;
    243. 

  243.       print "\\draw[thin,lightgray] (0.2,-$height) ";
    244. 

  244.       print "node[anchor=east,black] {".$line."0\\%} -- ";
    245. 

  245.       print "($img_width,-$height) ;<br>\n"; 
    246. 

  246.   }
    247. 

  247.   print "% co-ordinates for deletion<br>\n";
    248. 

  248.   for($line=0;$line<=3;$line++) {
    249. 

  249.       $height = 2+$line/10*1.80;
    250. 

  250.       print "\\draw[thin,lightgray] (0.2,-$height) ";
    251. 

  251.       if ($line != 0) {
    252. 

  252. 	  print "node[anchor=east,black] {".$line."0\\%} ";
    253. 

  253.       }
    254. 

  254.       print "-- ($img_width,-$height) ;<br>\n"; 
    255. 

  255.   }
    256. 

  256. 

  257.   print "% boxes<br>\n";
    258. 

  258.   $total_so_far = 0;
    259. 

  259.   foreach ($keys as $i) {
    260. 

  260.     $prec_ratio = $log_info[$i]["precision"]/$log_info[$i]["total"];
    261. 

  261.     $x = .2+($img_width-.2) * $total_so_far/$total;
    262. 

  262.     $y = 1.80-($prec_ratio*1.80);
    263. 

  263.     $width = $img_width * $log_info[$i]["total"]/$total;
    264. 

  264.     $height = $prec_ratio*1.80;
    265. 

  265. 

  266.     $width += $x;
    267. 

  267.     $height += $y;
    268. 

  268.     
    269. 

  269.     print "\\filldraw[very thin,gray] ($x,-$y) rectangle($width,-$height) ;<br>";
    270. 

  270.     print "\\draw[very thin,black] ($x,-$y) rectangle($width,-$height);<br>";
    271. 

  271.     if ($width-$x>.1) {
    272. 

  272. 	print "\\draw (".(($x+$width)/2).",-1.8) node[anchor=north,black] {".$i."};<br>";
    273. 

  273.     }
    274. 

  274.  
    275. 

  275. 

  276.     $del_ratio = $log_info[$i]["delete"]/$log_info[$i]["total"];
    277. 

  277.     $height = $del_ratio*1.80;
    278. 

  278. 

  279.     $height += 2;
    280. 

  280. 

  281.     print "\\filldraw[very thin,lightgray] ($x,-2) rectangle($width,-$height);<br>\n";
    282. 

  282.     print "\\draw[very thin,black] ($x,-2) rectangle($width,-$height);<br>\n";
    283. 

  283. 

  284.     $total_so_far += $log_info[$i]["total"];   
    285. 

  285.   }
    286. 

  286.   print "\\end{tikzpicture}</code>";
    287. 

  287.   print "</div>"; 
    288. 

  288. }
    289. 

  289. 

  290. function precision_by_coverage_graph($name,$log_info,$total,$img_width,$sort_type) {
    291. 

  291. 

  292.   $keys = array_keys($log_info);
    293. 

  293.   sort($keys,$sort_type);
    294. 

  294. 

  295.   print "<div id=\"Toggle$name\" onClick=\"document.getElementById('Table$name').style.display = 'none'; document.getElementById('LatexToggle$name').style.display = 'none'; document.getElementById('Latex$name').style.display = 'none'; this.style.display = 'none';\" style=\"display:none;\"><font size=-2>(hide table)</font></div>\n";
    296. 

  296.   precision_by_coverage_latex($name,$log_info,$total,$img_width,$sort_type);
    297. 

  297. 

  298.   print "<div id=\"Table$name\" style=\"display:none;\">\n";
    299. 

  299.   print "<table border=1><tr><td align=center>Count</td><td align=center colspan=2>Precision</td><td align=center colspan=2>Delete</td><td align=center>Length</td></tr>\n";
    300. 

  300.   foreach ($keys as $i) {
    301. 

  301.     if (array_key_exists($i,$log_info)) {
    302. 

  302.       print "<tr><td align=center>$i</td>";
    303. 

  303.       printf("<td align=right>%.1f%s</td><td align=right><font size=-1>%.1f/%d</font></td>",$log_info[$i]["precision"]/$log_info[$i]["total"]*100,"%",$log_info[$i]["precision"],$log_info[$i]["total"]);
    304. 

  304.       printf("<td align=right>%.1f%s</td><td align=right><font size=-1>%d/%d</font></td>",$log_info[$i]["delete"]/$log_info[$i]["total"]*100,"%",$log_info[$i]["delete"],$log_info[$i]["total"]);
    305. 

  305.       printf("<td align=right>%.3f</td>",$log_info[$i]["length"]/$log_info[$i]["total"]);
    306. 

  306.       print "<td><A HREF=\"javascript:generic_show('PrecisionByWord$name','type=$i')\">&#x24BE;</A></td>";
    307. 

  307.       print "</tr>";
    308. 

  308.    }
    309. 

  309.   }
    310. 

  310.   print "</table><div id=\"PrecisionByWord$name\"></div></div>";
    311. 

  311. 

  312.   print "<div id=\"Graph$name\" onClick=\"document.getElementById('Table$name').style.display = 'block'; document.getElementById('LatexToggle$name').style.display = 'block'; document.getElementById('Toggle$name').style.display = 'block';\">";
    313. 

  313.   print "<canvas id=\"$name\" width=$img_width height=300></canvas></div>";
    314. 

  314.   print "<script language=\"javascript\">
    315. 

  315. var canvas = document.getElementById(\"$name\");
    316. 

  316. var ctx = canvas.getContext(\"2d\");
    317. 

  317. ctx.lineWidth = 0.5;
    318. 

  318. ctx.font = '9px serif';
    319. 

  319. ";
    320. 

  320.   for($line=0;$line<=9;$line++) {
    321. 

  321.       $height = 180-$line/10*180;
    322. 

  322.       print "ctx.moveTo(20, $height);\n";
    323. 

  323.       print "ctx.lineTo($img_width, $height);\n";
    324. 

  324.       if ($line != 0) {
    325. 

  325. 	  print "ctx.fillText(\"${line}0\%\", 0, $height+4);";
    326. 

  326.       }
    327. 

  327.   }
    328. 

  328.   for($line=0;$line<=3;$line++) {
    329. 

  329.       $height = 200+$line/10*180;
    330. 

  330.       print "ctx.moveTo(20, $height);\n";
    331. 

  331.       print "ctx.lineTo($img_width, $height);\n";
    332. 

  332.       if ($line != 0) {
    333. 

  333. 	  print "ctx.fillText(\"${line}0\%\", 0, $height+4);";
    334. 

  334.       }
    335. 

  335.   }
    336. 

  336.   print "ctx.strokeStyle = \"rgb(100,100,100)\"; ctx.stroke();\n";
    337. 

  337. 

  338.   $total_so_far = 0;
    339. 

  339.   foreach ($keys as $i) {
    340. 

  340.     $prec_ratio = $log_info[$i]["precision"]/$log_info[$i]["total"];
    341. 

  341.     $x = (int)(20+($img_width-20) * $total_so_far / $total);
    342. 

  342.     $y = (int)(180-($prec_ratio*180));
    343. 

  343.     $width = (int)($img_width * $log_info[$i]["total"]/$total);
    344. 

  344.     $height = (int)($prec_ratio*180);
    345. 

  345.     print "ctx.fillStyle = \"rgb(200,200,0)\";";
    346. 

  346.     print "ctx.fillRect ($x, $y, $width, $height);";
    347. 

  347. 

  348.     $del_ratio = $log_info[$i]["delete"]/$log_info[$i]["total"];
    349. 

  349.     $height = (int)($del_ratio*180);
    350. 

  350.     print "ctx.fillStyle = \"rgb(100,100,255)\";";
    351. 

  351.     print "ctx.fillRect ($x, 200, $width, $height);";
    352. 

  352. 

  353.     $total_so_far += $log_info[$i]["total"];
    354. 

  354.    
    355. 

  355.     if ($width>3) {
    356. 

  356.       print "ctx.fillStyle = \"rgb(0,0,0)\";";
    357. 

  357.       // print "ctx.rotate(-1.5707);";
    358. 

  358.       print "ctx.fillText(\"$i\", $x+$width/2-3, 190);";
    359. 

  359.       //print "ctx.rotate(1.5707);";
    360. 

  360.     }
    361. 

  361.   }
    362. 

  362.   print "</script>";
    363. 

  363. }
    364. 

  364. 

  365. //# stats on precision and recall
    366. 

  366. function precision_recall_details() {
    367. 

  367. ?>
    368. 

  368. <table width=100%>
    369. 

  369.   <tr>
    370. 

  370.     <td width=25% valign=top><div id="nGramPrecision1">(loading...)</div></td>
    371. 

  371.     <td width=25% valign=top><div id="nGramPrecision2">(loading...)</div></td>
    372. 

  372.     <td width=25% valign=top><div id="nGramPrecision3">(loading...)</div></td>
    373. 

  373.     <td width=25% valign=top><div id="nGramPrecision4">(loading...)</div></td>
    374. 

  374.   </tr><tr>
    375. 

  375.     <td width=25% valign=top><div id="nGramRecall1">(loading...)</div></td>
    376. 

  376.     <td width=25% valign=top><div id="nGramRecall2">(loading...)</div></td>
    377. 

  377.     <td width=25% valign=top><div id="nGramRecall3">(loading...)</div></td>
    378. 

  378.     <td width=25% valign=top><div id="nGramRecall4">(loading...)</div></td>
    379. 

  379. </tr></table>
    380. 

  380. <script language="javascript">
    381. 

  381. ngram_show('precision',1,5,'',0);
    382. 

  382. ngram_show('precision',2,5,'',0);
    383. 

  383. ngram_show('precision',3,5,'',0);
    384. 

  384. ngram_show('precision',4,5,'',0);
    385. 

  385. ngram_show('recall',1,5,'',0);
    386. 

  386. ngram_show('recall',2,5,'',0);
    387. 

  387. ngram_show('recall',3,5,'',0);
    388. 

  388. ngram_show('recall',4,5,'',0);
    389. 

  389. </script>
    390. 

  390. <?php
    391. 

  391. }
    392. 

  392. 

  393. //# stats on ngram precision
    394. 

  394. function ngram_summary() {
    395. 

  395.   global $experiment,$evalset,$dir,$set,$id;
    396. 

  396. 

  397.   //# load data
    398. 

  398.   $data = file(get_current_analysis_filename("basic","summary"));
    399. 

  399.   for($i=0;$i<count($data);$i++) {
    400. 

  400.     $item = split(": ",$data[$i]);
    401. 

  401.     $info[$item[0]] = $item[1];
    402. 

  402.   }
    403. 

  403. 

  404.   print "<table cellspacing=5 width=100%><tr><td valign=top align=center bgcolor=#eeeeee>";
    405. 

  405.   //#foreach (array("precision","recall") as $type) {
    406. 

  406.   print "<b>Precision of Output</b>\n";
    407. 

  407.   $type = "precision";
    408. 

  408.   print "<table><tr><td>$type</td><td>1-gram</td><td>2-gram</td><td>3-gram</td><td>4-gram</td></tr>\n";
    409. 

  409.   printf("<tr><td>correct</td><td>%d</td><td>%d</td><td>%d</td><td>%d</td></tr>\n",
    410. 

  410. 	 $info["$type-1-correct"],
    411. 

  411. 	 $info["$type-2-correct"],
    412. 

  412. 	 $info["$type-3-correct"],
    413. 

  413. 	 $info["$type-4-correct"]);    
    414. 

  414.   printf("<tr><td>&nbsp;</td><td>%.1f%s</td><td>%.1f%s</td><td>%.1f%s</td><td>%.1f%s</td></tr>\n",
    415. 

  415. 	 $info["$type-1-correct"]/$info["$type-1-total"]*100,'%',
    416. 

  416. 	 $info["$type-2-correct"]/$info["$type-2-total"]*100,'%',
    417. 

  417. 	 $info["$type-3-correct"]/$info["$type-3-total"]*100,'%',
    418. 

  418. 	 $info["$type-4-correct"]/$info["$type-4-total"]*100,'%');
    419. 

  419.   printf("<tr><td>wrong</td><td>%d</td><td>%d</td><td>%d</td><td>%d</td></tr>\n",
    420. 

  420. 	 $info["$type-1-total"]-$info["$type-1-correct"],
    421. 

  421. 	 $info["$type-2-total"]-$info["$type-2-correct"],
    422. 

  422. 	 $info["$type-3-total"]-$info["$type-3-correct"],
    423. 

  423. 	 $info["$type-4-total"]-$info["$type-4-correct"]);
    424. 

  424.   print "</table>";
    425. 

  425.   //}
    426. 

  426. 

  427.   print "<A HREF=\"javascript:generic_show('PrecisionRecallDetails','')\">details</A> ";
    428. 

  428.   if (file_exists(get_current_analysis_filename("precision","precision-by-corpus-coverage"))) {
    429. 

  429.     print "| <A HREF=\"javascript:generic_show('PrecisionByCoverage','')\">precision of input by coverage</A> ";
    430. 

  430.   }
    431. 

  431. 

  432.  print "</td><td valign=top valign=top align=center bgcolor=#eeeeee>";
    433. 

  433. 

  434.   $each_score = explode(" ; ",$experiment[$id]->result[$set]);
    435. 

  435.   $header = "";
    436. 

  436.   $score_line = "";
    437. 

  437.   for($i=0;$i<count($each_score);$i++) {
    438. 

  438.     if (preg_match('/([\d\(\)\.\s]+) (BLEU[\-c]*)/',$each_score[$i],$match) ||
    439. 

  439.         preg_match('/([\d\(\)\.\s]+) (IBM[\-c]*)/',$each_score[$i],$match)) {
    440. 

  440.       $header .= "<td>$match[2]</td>";
    441. 

  441.       $score_line .= "<td>$match[1]</td>";
    442. 

  442.     }
    443. 

  443.   }
    444. 

  444.   print "<b>Metrics</b><table border=1><tr>".$header."</tr><tr>".$score_line."</tr></table>";
    445. 

  445. 

  446.   printf("<p>length-diff: %d (%.1f%s)",$info["precision-1-total"]-$info["recall-1-total"],($info["precision-1-total"]-$info["recall-1-total"])/$info["recall-1-total"]*100,"%");
    447. 

  447. 

  448.   // coverage
    449. 

  449.   if (file_exists(get_current_analysis_filename("coverage","corpus-coverage-summary"))) {
    450. 

  450.     print "</td><td valign=top align=center bgcolor=#eeeeee>";
    451. 

  451.     print "<div id=\"CoverageSummary\">";
    452. 

  452.     coverage_summary();
    453. 

  453.     print "</div>";
    454. 

  454.   }
    455. 

  455. 

  456.   // phrase segmentation
    457. 

  457.   if (file_exists(get_current_analysis_filename("basic","segmentation")) ||
    458. 

  458.       file_exists(get_current_analysis_filename("basic","rule"))) {
    459. 

  459.     print "</td><td valign=top align=center bgcolor=#eeeeee>";
    460. 

  460.     print "<div id=\"SegmentationSummary\">";
    461. 

  461.     segmentation_summary();
    462. 

  462.     print "</div>";
    463. 

  463.   }
    464. 

  464. 

  465.   // rules
    466. 

  466.   if (file_exists(get_current_analysis_filename("basic","rule"))) {
    467. 

  467.     print "</td><td valign=top align=center bgcolor=#eeeeee>";
    468. 

  468.     print "<div id=\"RuleSummary\">";
    469. 

  469.     rule_summary();
    470. 

  470.     print "</div>";
    471. 

  471.   }
    472. 

  472. 

  473.   print "</td></tr></table>";
    474. 

  474. }
    475. 

  475. 

  476. // details on ngram precision/recall
    477. 

  477. function ngram_show($type) {
    478. 

  478.   global $set,$id,$dir;
    479. 

  479. 

  480.   // load data
    481. 

  481.   $order = $_GET['order'];
    482. 

  482.   $data = file(get_current_analysis_filename("basic","n-gram-$type.$order"));
    483. 

  483.   for($i=0;$i<count($data);$i++) {
    484. 

  484.      $item = split("\t",$data[$i]);
    485. 

  485.      $line["total"] = $item[0]; 
    486. 

  486.      $line["correct"] = $item[1];
    487. 

  487.      $line["ngram"] = $item[2];
    488. 

  488.      $ngram[] = $line;
    489. 

  489.   }
    490. 

  490. 

  491.   // sort option
    492. 

  492.   $sort = $_GET['sort'];
    493. 

  493.   $smooth = $_GET['smooth'];
    494. 

  494.   if ($sort == '') {
    495. 

  495.     $sort = 'ratio_worst';
    496. 

  496.     $smooth = 1;
    497. 

  497.   }
    498. 

  498.   
    499. 

  499.   // sort index
    500. 

  500.   for($i=0;$i<count($ngram);$i++) {
    501. 

  501.     if ($sort == "abs_worst") {
    502. 

  502.       $ngram[$i]["index"] = $ngram[$i]["correct"] - $ngram[$i]["total"];
    503. 

  503.     }
    504. 

  504.     else if ($sort == "ratio_worst") {
    505. 

  505.       $ngram[$i]["index"] = ($ngram[$i]["correct"] + $smooth) / ($ngram[$i]["total"] + $smooth);
    506. 

  506.     }
    507. 

  507.   }
    508. 

  508. 

  509.   // sort
    510. 

  510.   function cmp($a, $b) {
    511. 

  511.     if ($a["index"] == $b["index"]) {
    512. 

  512.         return 0;
    513. 

  513.     }
    514. 

  514.     return ($a["index"] < $b["index"]) ? -1 : 1;
    515. 

  515.   }
    516. 

  516. 

  517.   usort($ngram, 'cmp');
    518. 

  518. 

  519.   // display
    520. 

  520.   $count = $_GET['count'];
    521. 

  521.   if ($count == 0) { $count = 5; }
    522. 

  522. 

  523.   print "<B>$order-gram $type</B><br><font size=-1>sorted by ";
    524. 

  524.   if ($sort == "ratio_worst") {
    525. 

  525.     print "ratio ";
    526. 

  526.     print "smooth-$smooth ";
    527. 

  527.     print "<A HREF=\"javascript:ngram_show('$type',$order,$count,'ratio_worst',$smooth+1)\">+</A> ";
    528. 

  528.     print "<A HREF=\"javascript:ngram_show('$type',$order,$count,'ratio_worst',$smooth-1)\">-</A> ";
    529. 

  529.   }
    530. 

  530.   else {
    531. 

  531.     print "<A HREF=\"javascript:ngram_show('$type',$order,$count,'ratio_worst',1)\">ratio</A> ";
    532. 

  532.   }
    533. 

  533.   if ($sort == "abs_worst") {
    534. 

  534.     print "absolute ";
    535. 

  535.   }
    536. 

  536.   else {
    537. 

  537.     print "<A HREF=\"javascript:ngram_show('$type',$order,$count,'abs_worst',0)\">absolute</A> ";
    538. 

  538.   }
    539. 

  539.   print "showing $count ";
    540. 

  540.   if ($count < 9999) {
    541. 

  541.     print "<A HREF=\"javascript:ngram_show('$type',$order,$count+5,'$sort',$smooth)\">more</A> ";
    542. 

  542.     print "<A HREF=\"javascript:ngram_show('$type',$order,9999,'$sort',$smooth)\">all</A> ";
    543. 

  543.   }
    544. 

  544.   else {
    545. 

  545.     print "<A HREF=\"javascript:ngram_show('$type',$order,5,'$sort',$smooth)\">top5</A> ";
    546. 

  546.   }
    547. 

  547.   print "</font><br>\n";
    548. 

  548. 

  549.   print "<table width=100%>\n";
    550. 

  550.   print "<tr><td>$order-gram</td><td>ok</td><td>x</td><td>ratio</td></tr>\n";
    551. 

  551.   for($i=0;$i<$count && $i<count($ngram);$i++) {
    552. 

  552.     $line = $ngram[$i];
    553. 

  553.     print "<tr><td>".$line["ngram"]."</td>";
    554. 

  554.     print "<td>".$line["correct"]."</td>";
    555. 

  555.     print "<td>".($line["total"]-$line["correct"])."</td>";
    556. 

  556.     printf("<td>%.3f</td></tr>",$line["correct"]/$line["total"]);
    557. 

  557.   }
    558. 

  558.   print "</table>\n";
    559. 

  559. }
    560. 

  560. 

  561. // details on ngram coverage
    562. 

  562. function coverage_details() {
    563. 

  563.   global $dir,$set,$id;
    564. 

  564. 

  565.   $count = array(); $token = array();
    566. 

  566.   foreach (array("ttable","corpus") as $corpus) {
    567. 

  567.     foreach (array("token","type") as $b) {
    568. 

  568.       for($i=0;$i<=7;$i++) {
    569. 

  569.         foreach (array("6+","2-5","1","0") as $range) {
    570. 

  570.           $count[$corpus][$b][$i][$range] = 0;
    571. 

  571.         }
    572. 

  572.         $total[$corpus][$b][$i] = 0;
    573. 

  573.       }
    574. 

  574.     }
    575. 

  575.     $data = file(filename_fallback_to_factored(get_current_analysis_filename("coverage","$corpus-coverage-summary")));
    576. 

  576.     for($i=0;$i<count($data);$i++) {
    577. 

  577.       $item = split("\t",$data[$i]);
    578. 

  578.       if ($item[1]>5) {
    579. 

  579.         $count[$corpus]["type"][$item[0]]["6+"] += $item[2];
    580. 

  580. 	$count[$corpus]["token"][$item[0]]["6+"] += $item[3];
    581. 

  581.       }
    582. 

  582.       else if ($item[1]>1) {
    583. 

  583.         $count[$corpus]["type"][$item[0]]["2-5"] += $item[2];
    584. 

  584. 	$count[$corpus]["token"][$item[0]]["2-5"] += $item[3];
    585. 

  585.       }
    586. 

  586.       else if ($item[1]==1) {
    587. 

  587.         $count[$corpus]["type"][$item[0]]["1"] += $item[2];
    588. 

  588. 	$count[$corpus]["token"][$item[0]]["1"] += $item[3];
    589. 

  589.       }
    590. 

  590.       else {
    591. 

  591.         $count[$corpus]["type"][$item[0]]["0"] += $item[2];
    592. 

  592. 	$count[$corpus]["token"][$item[0]]["0"] += $item[3];
    593. 

  593.       }
    594. 

  594.       $total[$corpus]["type"][$item[0]] += $item[2];
    595. 

  595.       $total[$corpus]["token"][$item[0]] += $item[3];
    596. 

  596.     }
    597. 

  597.   }
    598. 

  598.   print "<b>coverage</b><br>\n";
    599. 

  599.   print "<table width=100%><tr>";
    600. 

  600.   foreach (array("token","type") as $by) {
    601. 

  601.     for($i=1;$i<=4;$i++) {
    602. 

  602.       print "<td align=center><b>$i-gram ($by)</b><br>\n";
    603. 

  603.       print "<table><tr><td></td><td>model</td><td>corpus</td></tr>\n";
    604. 

  604.       foreach (array("0","1","2-5","6+") as $range) {
    605. 

  605.         print "<tr><td>$range</td>";
    606. 

  606.         foreach (array("ttable","corpus") as $corpus) {
    607. 

  607.           printf("<td align=right nowrap>%d (%.1f%s)</td>",$count[$corpus][$by][$i][$range],100*$count[$corpus][$by][$i][$range]/($total[$corpus][$by][$i]+0.0001),"%");
    608. 

  608.         }
    609. 

  609.         print "</tr>\n";
    610. 

  610.       }
    611. 

  611.       print "</table></td>\n";
    612. 

  612.     }
    613. 

  613.     print "</tr><tr>";
    614. 

  614.   }
    615. 

  615.   print "</tr></table>\n";
    616. 

  616. 

  617.   $data = file(filename_fallback_to_factored(get_current_analysis_filename("coverage","ttable-unknown")));
    618. 

  618.   for($i=0;$i<count($data);$i++) {
    619. 

  619.     list($word,$count) = split("\t",$data[$i]);
    620. 

  620.     $item["word"] = $word;
    621. 

  621.     $item["count"] = rtrim($count);
    622. 

  622.     $unknown[] = $item;
    623. 

  623.   }
    624. 

  624. 

  625.   function cmp($a,$b) {
    626. 

  626.     if ($a["count"] > $b["count"]) {
    627. 

  627.       return -1;
    628. 

  628.     }
    629. 

  629.     else if ($a["count"] < $b["count"]) {
    630. 

  630.       return 1;
    631. 

  631.     }
    632. 

  632.     else {
    633. 

  633.       return strcmp($a["word"],$b["word"]);
    634. 

  634.     }
    635. 

  635.   }
    636. 

  636. 

  637.   usort($unknown, 'cmp');
    638. 

  638. 

  639.   print "<b>unknown words (to model)</b><br>\n";
    640. 

  640.   print "<table><tr><td valign=top><table>";
    641. 

  641.   $state = 5;
    642. 

  642.   foreach ($unknown as $item) {
    643. 

  643.     if ($item["count"] < $state) {
    644. 

  644.       if ($state == 5) { print "</table>"; }
    645. 

  645.       print "</td><td valign=top><b>".$item["count"].":</b> ";
    646. 

  646.       $state = $item["count"];
    647. 

  647.       if ($state == 1) { print "<font size=-1>"; }
    648. 

  648.     }
    649. 

  649.     else if ($state<5) {
    650. 

  650.       print ", ";
    651. 

  651.     }
    652. 

  652.     if ($state == 5) {
    653. 

  653.       print "<tr><td>".$item["count"]."</td><td>".$item["word"]."</td></tr>";
    654. 

  654.     }
    655. 

  655.     else {
    656. 

  656.       print $item["word"];
    657. 

  657.     }
    658. 

  658.   }
    659. 

  659.   print "</font></td></tr></table>\n";
    660. 

  660. }
    661. 

  661. 

  662. function filename_fallback_to_factored($file) {
    663. 

  663.   if (file_exists($file)) {
    664. 

  664.     return $file;
    665. 

  665.   }
    666. 

  666.   $path = pathinfo($file);
    667. 

  667.   $dh = opendir($path['dirname']);
    668. 

  668.   while (($factored_file = readdir($dh)) !== false) {
    669. 

  669.     if (strlen($factored_file) > strlen($path['basename']) &&
    670. 

  670. 	substr($factored_file,0,strlen($path['basename'])) == $path['basename'] &&
    671. 

  671. 	preg_match("/0/",substr($factored_file,strlen($path['basename'])))) {
    672. 

  672.       return $path['dirname']."/".$factored_file;
    673. 

  673.     }
    674. 

  674.   }
    675. 

  675.   // found nothing...
    676. 

  676.   return $file;
    677. 

  677. }
    678. 

  678. 

  679. function factor_name($input_output,$factor_id) {
    680. 

  680.   global $dir,$set,$id;
    681. 

  681.   $file = get_current_analysis_filename("coverage","factor-names");
    682. 

  682.   if (!file_exists($file)) {
    683. 

  683.     return $factor_id;
    684. 

  684.   }
    685. 

  685.   $in_out_names = file($file);
    686. 

  686.   $names = explode(",",trim($in_out_names[($input_output == "input")?0:1]));
    687. 

  687.   return "'".$names[$factor_id]."' ($factor_id)";
    688. 

  688. }
    689. 

  689. 

  690. // stats on ngram coverage
    691. 

  691. function coverage_summary() {
    692. 

  692.   global $dir,$set,$id,$corpus;
    693. 

  693. 

  694.   if (array_key_exists("by",$_GET)) { $by = $_GET['by']; }
    695. 

  695.   else { $by = 'token'; }
    696. 

  696. 

  697.   $total = array(); $count = array();
    698. 

  698.   foreach (array("ttable","corpus") as $corpus) {
    699. 

  699.     foreach (array("token","type") as $b) {
    700. 

  700.       foreach (array("6+","2-5","1","0") as $c) {
    701. 

  701.         $count[$corpus][$b][$c] = 0;
    702. 

  702.       }
    703. 

  703.       $total[$corpus][$b] = 0;
    704. 

  704.     }
    705. 

  705.     $data = file(filename_fallback_to_factored(get_current_analysis_filename("coverage","$corpus-coverage-summary")));
    706. 

  706.     for($i=0;$i<count($data);$i++) {
    707. 

  707.       $item = split("\t",$data[$i]);
    708. 

  708.       if ($item[0] == 1) {
    709. 

  709.         if ($item[1]>5) {
    710. 

  710.           $count[$corpus]["type"]["6+"] += $item[2];
    711. 

  711. 	  $count[$corpus]["token"]["6+"] += $item[3];
    712. 

  712.         }
    713. 

  713.         else if ($item[1]>1) {
    714. 

  714.           $count[$corpus]["type"]["2-5"] += $item[2];
    715. 

  715.   	  $count[$corpus]["token"]["2-5"] += $item[3];
    716. 

  716.         }
    717. 

  717.         else if ($item[1]==1) {
    718. 

  718.           $count[$corpus]["type"]["1"] += $item[2];
    719. 

  719. 	  $count[$corpus]["token"]["1"] += $item[3];
    720. 

  720.         }
    721. 

  721.         else {
    722. 

  722.           $count[$corpus]["type"]["0"] += $item[2];
    723. 

  723. 	  $count[$corpus]["token"]["0"] += $item[3];
    724. 

  724.         }
    725. 

  725.         $total[$corpus]["type"] += $item[2];
    726. 

  726.         $total[$corpus]["token"] += $item[3];
    727. 

  727.       }
    728. 

  728.     }
    729. 

  729.   }
    730. 

  730. 

  731.   print "<b>Coverage</b>\n";
    732. 

  732.   print "<table><tr><td></td><td>model</td><td>corpus</td></tr>\n";
    733. 

  733.   foreach (array("0","1","2-5","6+") as $range) {
    734. 

  734.     print "<tr><td>$range</td>";
    735. 

  735.     foreach (array("ttable","corpus") as $corpus) {
    736. 

  736.       printf("<td align=right nowrap>%d (%.1f%s)</td>",$count[$corpus][$by][$range],100*$count[$corpus][$by][$range]/($total[$corpus][$by]+0.0001),"%");
    737. 

  737.     }
    738. 

  738.     print "</tr>\n";
    739. 

  739.   }
    740. 

  740.   print "</table>\n";
    741. 

  741.   if ($by == 'token') { print "by token"; } else {
    742. 

  742.     print "<A HREF=\"javascript:generic_show('CoverageSummary','by=token')\">by token</A> ";
    743. 

  743.   }
    744. 

  744.   print " / ";
    745. 

  745.   if ($by == 'type') { print "by type"; } else {
    746. 

  746.     print "<A HREF=\"javascript:generic_show('CoverageSummary','by=type')\">by type</A> ";
    747. 

  747.   }
    748. 

  748.   print " / ";
    749. 

  749.   print "<div id=\"CoverageDetailsLink\"><A HREF=\"javascript:generic_show('CoverageDetails','')\">details</A></div> ";
    750. 

  750. }
    751. 

  751. 

  752. 

  753. // stats on segmenation (phrase-based)
    754. 

  754. function segmentation_summary() {
    755. 

  755.   global $dir,$set,$id;
    756. 

  756. 

  757.   if (array_key_exists("by",$_GET)) { $by = $_GET['by']; }
    758. 

  758.   else { $by = 'word'; }
    759. 

  759. 

  760.   $count = array();
    761. 

  761.   for($i=0;$i<=4;$i++) {
    762. 

  762.     $count[$i] = array();
    763. 

  763.     for($j=0;$j<=4;$j++) {
    764. 

  764.       $count[$i][$j] = 0;
    765. 

  765.     }
    766. 

  766.   }
    767. 

  767. 

  768.   $total = 0;
    769. 

  769.   $file = get_current_analysis_filename("basic","segmentation");
    770. 

  770.   if (file_exists($file)) {
    771. 

  771.     $data = file($file);
    772. 

  772.     for($i=0;$i<count($data);$i++) {
    773. 

  773.       list($in,$out,$c) = split("\t",$data[$i]);
    774. 

  774.       if ($by == "word") { $c *= $in; }
    775. 

  775.       if ($in>4)  { $in  = 4; }
    776. 

  776.       if ($out>4) { $out = 4; }
    777. 

  777.       $total += $c;
    778. 

  778.       $count[$in][$out] += $c;
    779. 

  779.     }
    780. 

  780.   }
    781. 

  781.   else {
    782. 

  782.     $data = file(get_current_analysis_filename("basic","rule"));
    783. 

  783.     for($i=0;$i<count($data);$i++) {
    784. 

  784.       $field = split("\t",$data[$i]);
    785. 

  785.       $type = $field[0];
    786. 

  786.       $rule = $field[1];
    787. 

  787.       if (count($field) > 2) { $c = $field[2]; } else { $c = 0; }
    788. 

  788.       if ($type == "rule") {
    789. 

  789.         list($rule_in,$in,$nt,$rule_out,$out) = split(":",$rule);
    790. 

  790.         if ($by == "word") { $c *= $in; }
    791. 

  791.         if ($in>4)  { $in  = 4; }
    792. 

  792.         if ($out>4) { $out = 4; }
    793. 

  793.         $total += $c;
    794. 

  794.         $count[$in][$out] += $c;
    795. 

  795.       }
    796. 

  796.     }
    797. 

  797.   }
    798. 

  798. 

  799.   print "<b>Phrase Segmentation</b><br>\n";
    800. 

  800.   print "<table>";
    801. 

  801.   print "<tr><td></td><td align=center>1</td><td align=center>2</td><td align=center>3</td><td align=center>4+</td></tr>";
    802. 

  802.   for($in=1;$in<=4;$in++) {
    803. 

  803.     print "<tr><td nowrap>$in".($in==4?"+":"")." to</td>";
    804. 

  804.     for($out=1;$out<=4;$out++) {
    805. 

  805.       if (array_key_exists($in,$count) &&
    806. 

  806.           array_key_exists($out,$count[$in])) {
    807. 

  807. 	  $c = $count[$in][$out];
    808. 

  808.       } 
    809. 

  809.       else { $c = 0; }
    810. 

  810.       printf("<td align=right nowrap>%d (%.1f%s)</td>",$c,100*$c/$total,"%");
    811. 

  811.     }
    812. 

  812.     print "</tr>";
    813. 

  813.   }
    814. 

  814.   print "</table>\n";
    815. 

  815.   if ($by == 'word') { print "by word"; } else {
    816. 

  816.     print "<A HREF=\"javascript:generic_show('SegmentationSummary','by=word')\">by word</A> ";
    817. 

  817.   }
    818. 

  818.   print " / ";
    819. 

  819.   if ($by == 'phrase') { print "by phrase"; } else {
    820. 

  820.     print "<A HREF=\"javascript:generic_show('SegmentationSummary','by=phrase')\">by phrase</A> ";
    821. 

  821.   }
    822. 

  822. }
    823. 

  823. 

  824. // hierarchical rules used in translation
    825. 

  825. function rule_summary() {
    826. 

  826.   global $dir,$set,$id;
    827. 

  827.   $data = file(get_current_analysis_filename("basic","rule"));
    828. 

  828.   $rule = array(); $count = array(); $count_nt = array(); $count_w = array();
    829. 

  829.   $nt_count = 0; $total = 0;
    830. 

  830.   foreach ($data as $item) {
    831. 

  831.     $field = split("\t",$item);
    832. 

  832.     $type = $field[0];
    833. 

  833.     $d = $field[1];
    834. 

  834.     if (count($field) > 2) { $d2 = $field[2]; } else { $d2 = 0; }
    835. 

  835.     if ($type == "sentence-count") {
    836. 

  836. 	$sentence_count = $d;
    837. 

  837.     }
    838. 

  838.     else if ($type == "glue-rule") {
    839. 

  839. 	$glue_rule = $d / $sentence_count;
    840. 

  840.     }
    841. 

  841.     else if ($type == "depth") {
    842. 

  842. 	$depth = $d / $sentence_count;
    843. 

  843.     }
    844. 

  844.     else {
    845. 

  845. 	list($rule_in,$word_in,$nt,$rule_out,$word_out) = split(":",$d);
    846. 

  846. 	$rule_in = preg_replace("/a/","x",$rule_in);
    847. 

  847. 	$rule_in = preg_replace("/b/","y",$rule_in);
    848. 

  848. 	$rule_in = preg_replace("/c/","z",$rule_in);
    849. 

  849. 	$rule_out = preg_replace("/a/","x",$rule_out);
    850. 

  850. 	$rule_out = preg_replace("/b/","y",$rule_out);
    851. 

  851. 	$rule_out = preg_replace("/c/","z",$rule_out);
    852. 

  852. 	$nt_count += $d2 * $nt;
    853. 

  853. 	if (!array_key_exists($d,$rule)) { $rule[$d] = 0; }
    854. 

  854. 	$rule[$d] += $d2;
    855. 

  855. 	if (!array_key_exists($nt,$count)) { $count[$nt] = 0; }
    856. 

  856. 	$count[$nt] += $d2;
    857. 

  857. 	$just_nt = preg_replace("/\d/","",$rule_in)."-".preg_replace("/\d/","",$rule_out);
    858. 

  858. 	$no_wc = preg_replace("/\d/","W",$rule_in)."-".preg_replace("/\d/","",$rule_out);
    859. 

  859. 	if ($just_nt == "-") { $just_nt = "lexical"; }
    860. 

  860. 	if (!array_key_exists($just_nt,$count_nt)) { $count_nt[$just_nt] = 0; }
    861. 

  861. 	$count_nt[$just_nt] += $d2;
    862. 

  862. 	if (!array_key_exists($no_wc,$count_w)) { $count_w[$no_wc] = 0; }
    863. 

  863. 	$count_w[$no_wc] += $d2;
    864. 

  864. 	$total += $d2;
    865. 

  865.     }
    866. 

  866.   }
    867. 

  867.   print "<b>Rules</b><br>\n";
    868. 

  868.   printf("glue rule: %.2f<br>\n",$glue_rule);
    869. 

  869.   printf("tree depth:  %.2f<br>\n",$depth);
    870. 

  870.   printf("nt/rule: %.2f<br>\n",$nt_count/$total);
    871. 

  871.   print "<table>\n";
    872. 

  872.   foreach ($count_nt as $rule => $count) {
    873. 

  873.       printf("<tr><td>%s</td><td align=right>%d</td><td align=right>%.1f%s</td></tr>\n",$rule,$count,$count/$total*100,'%');
    874. 

  874.   }
    875. 

  875.   print "</table>\n";
    876. 

  876. }
    877. 

  877. 

  878. // annotated sentences, navigation
    879. 

  879. function bleu_show() {
    880. 

  880. 

  881.   $count = $_GET['count'];
    882. 

  882.   if ($count == 0) { $count = 5; }
    883. 

  883. 

  884.   $filter = ""; 
    885. 

  885.   if (array_key_exists("filter",$_GET)) { 
    886. 

  886.     $filter = base64_decode($_GET['filter']);
    887. 

  887.   }
    888. 

  888. 

  889.   print "<b>annotated sentences</b><br><font size=-1>sorted by: ";
    890. 

  890. 

  891.   if ($_GET['sort'] == "order" || $_GET['sort'] == "") { print "order "; }
    892. 

  892.   else {
    893. 

  893.     print "<A HREF=\"javascript:show('bleu','order',$count,'".base64_encode($filter)."')\">order</A> ";
    894. 

  894.   }
    895. 

  895.   if ($_GET['sort'] == "best") { print "best "; }
    896. 

  896.   else {
    897. 

  897.     print "<A HREF=\"javascript:show('bleu','best',$count,'".base64_encode($filter)."')\">best</A> ";
    898. 

  898.   }
    899. 

  899.   if ($_GET['sort'] == "25") { print "25% "; }
    900. 

  900.   else {
    901. 

  901.     print "<A HREF=\"javascript:show('bleu','25',$count,'".base64_encode($filter)."')\">25%</A> ";
    902. 

  902.   }
    903. 

  903.   if ($_GET['sort'] == "avg") { print "avg "; }
    904. 

  904.   else {
    905. 

  905.     print "<A HREF=\"javascript:show('bleu','avg',$count,'".base64_encode($filter)."')\">avg</A> ";
    906. 

  906.   }
    907. 

  907.   if ($_GET['sort'] == "75") { print "75% "; }
    908. 

  908.   else {
    909. 

  909.     print "<A HREF=\"javascript:show('bleu','75',$count,'".base64_encode($filter)."')\">75%</A> ";
    910. 

  910.   }
    911. 

  911.   if ($_GET['sort'] == "worst") { print "worst; "; }
    912. 

  912.   else {
    913. 

  913.     print "<A HREF=\"javascript:show('bleu','worst',$count,'".base64_encode($filter)."')\">worst</A>; ";
    914. 

  914.   }
    915. 

  915. 

  916.   print "showing: $count ";
    917. 

  917.   print "<A HREF=\"javascript:show('bleu','" . $_GET['sort'] . "',5+$count,'".base64_encode($filter)."')\">more</A> ";
    918. 

  918.   print "<A HREF=\"javascript:show('bleu','" . $_GET['sort'] . "',9999,'".base64_encode($filter)."')\">all</A>";
    919. 

  919. 

  920.   if ($filter != "") {
    921. 

  921.     print "; filter: '$filter'";
    922. 

  922.   }
    923. 

  923.   sentence_annotation($count,$filter);
    924. 

  924.   print "<p align=center><A HREF=\"javascript:show('bleu','" . $_GET['sort'] . "',5+$count,'".base64_encode($filter)."')\">5 more</A> | ";
    925. 

  925.   print "<A HREF=\"javascript:show('bleu','" . $_GET['sort'] . "',10+$count,'".base64_encode($filter)."')\">10 more</A> | ";
    926. 

  926.   print "<A HREF=\"javascript:show('bleu','" . $_GET['sort'] . "',20+$count,'".base64_encode($filter)."')\">20 more</A> | ";
    927. 

  927.   print "<A HREF=\"javascript:show('bleu','" . $_GET['sort'] . "',50+$count,'".base64_encode($filter)."')\">50 more</A> | ";
    928. 

  928.   print "<A HREF=\"javascript:show('bleu','" . $_GET['sort'] . "',100+$count,'".base64_encode($filter)."')\">100 more</A> | ";
    929. 

  929.   print "<A HREF=\"javascript:show('bleu','" . $_GET['sort'] . "',9999,'".base64_encode($filter)."')\">all</A> ";
    930. 

  930. }
    931. 

  931. 

  932. // annotated sentences core: reads data, sorts sentences, displays them
    933. 

  933. function sentence_annotation($count,$filter) {
    934. 

  934.   global $set,$id,$dir,$biconcor;
    935. 

  935. 

  936.   # get input
    937. 

  937.   $filtered = array();
    938. 

  938.   $file = get_current_analysis_filename("coverage","input-annotation");
    939. 

  939.   if (file_exists($file)) {
    940. 

  940.     $input = file($file);
    941. 

  941.     # filter is so specified
    942. 

  942.     if ($filter != "") {
    943. 

  943.       for($i=0;$i<count($input);$i++) {
    944. 

  944.         $item = explode("\t",$input[$i]);
    945. 

  945. 	$word = explode(" ",$item[0]);
    946. 

  946. 	$keep = 0;
    947. 

  947. 	for($j=0;$j<count($word);$j++) {
    948. 

  948. 	  if ($word[$j] == $filter) { 
    949. 

  949. 	    $keep = 1; 
    950. 

  950. 	  }
    951. 

  951. 	}
    952. 

  952. 	if (!$keep) { $filtered[$i] = 1; }	
    953. 

  953.       }
    954. 

  954.     }  
    955. 

  955.   }
    956. 

  956.   
    957. 

  957.   # load bleu scores 
    958. 

  958.   $data = file(get_current_analysis_filename("basic","bleu-annotation"));
    959. 

  959.   for($i=0;$i<count($data);$i++) {
    960. 

  960.     $item = split("\t",$data[$i]);
    961. 

  961.     if (! array_key_exists($item[1],$filtered)) {
    962. 

  962.       $line["bleu"] = $item[0]; 
    963. 

  963.       $line["id"] = $item[1]; 
    964. 

  964.       $line["system"] = $item[2];
    965. 

  965.       $line["reference"] = ""; 
    966. 

  966.       for($j=3;$j<count($item);$j++) {
    967. 

  967.         if ($j>3) { $line["reference"] .= "<br>"; };
    968. 

  968.         $line["reference"] .= $item[$j];
    969. 

  969.       } 
    970. 

  970.       $bleu[] = $line;
    971. 

  971.     }
    972. 

  972.   }
    973. 

  973. 

  974.   # sort and label additional sentences as filtered
    975. 

  975.   global $sort;
    976. 

  976.   function cmp($a, $b) {
    977. 

  977.     global $sort;
    978. 

  978.     if ($sort == "order") {
    979. 

  979.       $a_idx = $a["id"];
    980. 

  980.       $b_idx = $b["id"];
    981. 

  981.     }
    982. 

  982.     else if ($sort == "worst" || $sort == "75") {
    983. 

  983.       $a_idx = $a["bleu"];
    984. 

  984.       $b_idx = $b["bleu"];
    985. 

  985.       if ($a_idx == $b_idx) { 
    986. 

  986.         $a_idx = $b["id"];
    987. 

  987.         $b_idx = $a["id"];
    988. 

  988.       }
    989. 

  989.     }
    990. 

  990.     else if ($sort == "best" || $sort == "avg" || $sort == "25") {
    991. 

  991.       $a_idx = -$a["bleu"];
    992. 

  992.       $b_idx = -$b["bleu"];
    993. 

  993.       if ($a_idx == $b_idx) { 
    994. 

  994.         $a_idx = $a["id"];
    995. 

  995.         $b_idx = $b["id"];
    996. 

  996.       }
    997. 

  997.     }
    998. 

  998.     if ($a_idx == $b_idx) {
    999. 

  999.       return 0;
    1000. 

  1000.     }
    1001. 

  1001.     return ($a_idx < $b_idx) ? -1 : 1;
    1002. 

  1002.   }
    1003. 

  1003.   $sort = $_GET['sort'];
    1004. 

  1004.   if ($sort == '') {
    1005. 

  1005.     $sort = "order";
    1006. 

  1006.   }
    1007. 

  1007.   usort($bleu, 'cmp');
    1008. 

  1008. 

  1009.   $offset = 0;
    1010. 

  1010.   if ($sort == "25" || $sort == "75") {
    1011. 

  1011.     $offset = (int) (count($bleu)/4);
    1012. 

  1012.   }
    1013. 

  1013.   else if ($sort == "avg") {
    1014. 

  1014.     $offset = (int) (count($bleu)/2);
    1015. 

  1015.   }
    1016. 

  1016. 

  1017.   $retained = array();
    1018. 

  1018.   for($i=$offset;$i<$count+$offset && $i<count($bleu);$i++) {
    1019. 

  1019.     $line = $bleu[$i]; 
    1020. 

  1020.     $retained[$line["id"]] = 1;
    1021. 

  1021.   }
    1022. 

  1022. 

  1023.   # get segmentation (phrase alignment)
    1024. 

  1024.   $file = get_current_analysis_filename("basic","segmentation-annotation");
    1025. 

  1025.   if (file_exists($file)) {
    1026. 

  1026.     $data = file($file);
    1027. 

  1027.     for($i=0;$i<count($data);$i++) {
    1028. 

  1028.       if ($filter == "" || array_key_exists($i,$retained)) {
    1029. 

  1029. 	$segment = 0;
    1030. 

  1030. 	foreach (split(" ",$data[$i]) as $item) {
    1031. 

  1031. 	  list($in_start,$in_end,$out_start,$out_end) = split(":",$item);
    1032. 

  1032. 	  $segment++;
    1033. 

  1033. 	  $segmentation[$i]["input_start"][$in_start] = $segment;
    1034. 

  1034. 	  $segmentation[$i]["input_end"][$in_end] = $segment;
    1035. 

  1035. 	  $segmentation[$i]["output_start"][$out_start] = $segment;
    1036. 

  1036. 	  $segmentation[$i]["output_end"][$out_end+0] = $segment;
    1037. 

  1037. 	}
    1038. 

  1038.       }
    1039. 

  1039.     }
    1040. 

  1040.   }
    1041. 

  1041. 

  1042.   # get hierarchical data
    1043. 

  1043.   $hierarchical = 0;
    1044. 

  1044.   $file = get_current_analysis_filename("basic","input-tree");
    1045. 

  1045.   if (file_exists($file)) {
    1046. 

  1046.     $data = file($file);
    1047. 

  1047.     $span = 0;
    1048. 

  1048.     $last_sentence = -1;
    1049. 

  1049.     $nt_count = array();
    1050. 

  1050.     for($i=0;$i<count($data);$i++) {
    1051. 

  1051.       list($sentence,$brackets,$nt,$words) = split("\t",$data[$i]);
    1052. 

  1052.       if ($sentence != $last_sentence) { $span = 0; }
    1053. 

  1053.       $last_sentence = $sentence;
    1054. 

  1054.       if ($filter == "" || array_key_exists($sentence,$retained)) {
    1055. 

  1055. 	$segmentation[$sentence][$span]["brackets"] = $brackets;
    1056. 

  1056. #	  $segmentation[$sentence][$span]["nt"] = $nt;
    1057. 

  1057. 	$segmentation[$sentence][$span]["words"] = rtrim($words);
    1058. 

  1058. 	if ($nt != "") { $nt_count[$nt]=1; }
    1059. 

  1059. 	$span++;
    1060. 

  1060.       }
    1061. 

  1061.     }
    1062. 

  1062.     $hierarchical = 1;
    1063. 

  1063. #      if (count($nt_count) <= 2) {
    1064. 

  1064. #	  foreach ($segmentation as $sentence => $segmentation_span) {
    1065. 

  1065. #	      foreach ($segmentation_span as $span => $type) {
    1066. 

  1066. #		  $segmentation[$sentence][$span]["nt"]="";
    1067. 

  1067. #	      }
    1068. 

  1068. #	  }
    1069. 

  1069. #     }
    1070. 

  1070.   }
    1071. 

  1071.   $file = get_current_analysis_filename("basic","output-tree");
    1072. 

  1072.   if (file_exists($file)) {
    1073. 

  1073.     $data = file($file);
    1074. 

  1074.     $span = 0;
    1075. 

  1075.     $last_sentence = -1;
    1076. 

  1076.     $nt_count = array();
    1077. 

  1077.     for($i=0;$i<count($data);$i++) {
    1078. 

  1078.       list($sentence,$brackets,$nt,$words) = split("\t",$data[$i]);
    1079. 

  1079.       if ($sentence != $last_sentence) { $span = 0; }
    1080. 

  1080.       $last_sentence = $sentence;
    1081. 

  1081.       if ($filter == "" || array_key_exists($sentence,$retained)) {      
    1082. 

  1082. 	$segmentation_out[$sentence][$span]["brackets"] = $brackets;
    1083. 

  1083. 	$segmentation_out[$sentence][$span]["nt"] = $nt;
    1084. 

  1084. 	$segmentation_out[$sentence][$span]["words"] = rtrim($words);
    1085. 

  1085. 	if ($nt != "") { $nt_count[$nt]=1; }
    1086. 

  1086. 	$span++;
    1087. 

  1087.       }
    1088. 

  1088.     }
    1089. 

  1089.     # no non-terminal markup, if there are two or less non-terminals (X,S)
    1090. 

  1090.     if (count($nt_count) <= 2) {
    1091. 

  1091.       foreach ($segmentation_out as $sentence => $segmentation_span) {
    1092. 

  1092. 	foreach ($segmentation_span as $span => $type) {
    1093. 

  1093. 	  $segmentation_out[$sentence][$span]["nt"]="";
    1094. 

  1094. 	}
    1095. 

  1095.       }
    1096. 

  1096.     }
    1097. 

  1097.   }
    1098. 

  1098.   $file = get_current_analysis_filename("basic","node");
    1099. 

  1099.   if (file_exists($file)) {
    1100. 

  1100.     $data = file($file);
    1101. 

  1101.     $n = 0;
    1102. 

  1102.     $last_sentence = -1;
    1103. 

  1103.     for($i=0;$i<count($data);$i++) {
    1104. 

  1104.       list($sentence,$depth,$start_div,$end_div,$start_div_in,$end_div_in,$children) = split(" ",$data[$i]);
    1105. 

  1105.       if ($sentence != $last_sentence) { $n = 0; }
    1106. 

  1106.       $last_sentence = $sentence;
    1107. 

  1107.       if ($filter == "" || array_key_exists($sentence,$retained)) {
    1108. 

  1108. 	$node[$sentence][$n]['depth'] = $depth;
    1109. 

  1109. 	$node[$sentence][$n]['start_div'] = $start_div;
    1110. 

  1110. 	$node[$sentence][$n]['end_div'] = $end_div;
    1111. 

  1111. 	$node[$sentence][$n]['start_div_in'] = $start_div_in;
    1112. 

  1112. 	$node[$sentence][$n]['end_div_in'] = $end_div_in;
    1113. 

  1113. 	$node[$sentence][$n]['children'] = rtrim($children);
    1114. 

  1114. 	$n++;
    1115. 

  1115.       }
    1116. 

  1116.     }
    1117. 

  1117.   } 
    1118. 

  1118. 

  1119.   # display
    1120. 

  1120.   if ($filter != "") { 
    1121. 

  1121.       print " (".(count($input)-count($filtered))." retaining)";
    1122. 

  1122.   }
    1123. 

  1123.   print "</font><BR>\n";
    1124. 

  1124. 

  1125.   $biconcor = get_biconcor_version($dir,$set,$id);
    1126. 

  1126.   //print "<div id=\"debug\">$sort / $offset</div>";
    1127. 

  1127.   for($i=$offset;$i<$count+$offset && $i<count($bleu);$i++) {
    1128. 

  1128.      $line = $bleu[$i]; 
    1129. 

  1129.      if ($hierarchical) {
    1130. 

  1130. 	 annotation_hierarchical($line["id"],$segmentation[$line["id"]],$segmentation_out[$line["id"]],$node[$line["id"]]);
    1131. 

  1131.      }
    1132. 

  1132.      if ($input) {
    1133. 

  1133.        print "<div id=\"info-".$line["id"]."\" style=\"border-color:black; background:#ffff80; opacity:0; width:100%; border:1px;\">0 occ. in corpus, 0 translations, entropy: 0.00</div>\n";
    1134. 

  1134.        if ($biconcor) {
    1135. 

  1135. 	   print "<div id=\"biconcor-".$line["id"]."\" class=\"biconcor\"><font size=-2>(click on input phrase for bilingual concordancer)</font></div>";
    1136. 

  1136.        }
    1137. 

  1137.        if ($hierarchical) {
    1138. 

  1138.          sentence_annotation_hierarchical("#".$line["id"],$line["id"],$input[$line["id"]],$segmentation[$line["id"]],"in");
    1139. 

  1139.        }
    1140. 

  1140.        else {
    1141. 

  1141. 	 print "<font size=-2>[#".$line["id"]."]</font> ";
    1142. 

  1142.          input_annotation($line["id"],$input[$line["id"]],$segmentation[$line["id"]],$filter);
    1143. 

  1143.        }
    1144. 

  1144.      }
    1145. 

  1145.      //else {
    1146. 

  1146.        // print "<font size=-2>[".$line["id"].":".$line["bleu"]."]</font> ";
    1147. 

  1147.      //}
    1148. 

  1148.      if ($hierarchical) {
    1149. 

  1149.        sentence_annotation_hierarchical($line["bleu"],$line["id"],$line["system"],$segmentation_out[$line["id"]],"out");
    1150. 

  1150.      }
    1151. 

  1151.      else {
    1152. 

  1152.        print "<font size=-2>[".$line["bleu"]."]</font> ";
    1153. 

  1153.        output_annotation($line["id"],$line["system"],$segmentation[$line["id"]]);
    1154. 

  1154.      }
    1155. 

  1155.      print "<br><font size=-2>[ref]</font> ".$line["reference"]."<hr>";
    1156. 

  1156.   }
    1157. 

  1157. }
    1158. 

  1158. 

  1159. function coverage($coverage_vector) {
    1160. 

  1160.   # get information from line in input annotation file
    1161. 

  1161.   $coverage = array();
    1162. 

  1162.   foreach (split(" ",$coverage_vector) as $item) {
    1163. 

  1163.     if (preg_match("/[\-:]/",$item)) {
    1164. 

  1164.       $field = preg_split("/[\-:]/",$item);
    1165. 

  1165.       $from = $field[0];
    1166. 

  1166.       $to = $field[1];
    1167. 

  1167.       if (count($field)>2){ $coverage[$from][$to]["corpus_count"]=$field[2]; }
    1168. 

  1168.       if (count($field)>3){ $coverage[$from][$to]["ttable_count"]=$field[3]; }
    1169. 

  1169.       if (count($field)>4){ $coverage[$from][$to]["ttabel_entropy"]=$field[4]; }
    1170. 

  1170.     }
    1171. 

  1171.   }
    1172. 

  1172. 

  1173.   return $coverage;
    1174. 

  1174. }
    1175. 

  1175. 

  1176. // annotate an inpute sentence
    1177. 

  1177. function input_annotation($sentence,$input,$segmentation,$filter) {
    1178. 

  1178.   global $biconcor;
    1179. 

  1179.   list($words,$coverage_vector) = split("\t",$input);
    1180. 

  1180. 

  1181.   # get information from line in input annotation file
    1182. 

  1182.   $coverage = array();
    1183. 

  1183.   foreach (split(" ",$coverage_vector) as $item) {
    1184. 

  1184.     if (preg_match("/[\-:]/",$item)) {
    1185. 

  1185.       list($from,$to,$corpus_count,$ttable_count,$ttable_entropy) = preg_split("/[\-:]/",$item);
    1186. 

  1186.       $coverage[$from][$to]["corpus_count"] = $corpus_count;
    1187. 

  1187.       $coverage[$from][$to]["ttable_count"] = $ttable_count;
    1188. 

  1188.       $coverage[$from][$to]["ttable_entropy"] = $ttable_entropy;
    1189. 

  1189.     }
    1190. 

  1190.   }
    1191. 

  1191.   $word = split(" ",$words);
    1192. 

  1192. 

  1193.   # compute the display level for each input phrase
    1194. 

  1194.   for($j=0;$j<count($word);$j++) {
    1195. 

  1195.     $box[] = array();
    1196. 

  1196.     $separable[] = 1;
    1197. 

  1197.   }
    1198. 

  1198.   $max_level = 0;
    1199. 

  1199.   for($length=1;$length<=7;$length++) {
    1200. 

  1200.     for($from=0;$from<count($word)-($length-1);$from++) {
    1201. 

  1201.       $to = $from + ($length-1);
    1202. 

  1202.       if (array_key_exists($from,$coverage) &&
    1203. 

  1203.           array_key_exists($to,$coverage[$from]) &&
    1204. 

  1204.           array_key_exists("corpus_count",$coverage[$from][$to])) {
    1205. 

  1205.         $level=0;
    1206. 

  1206. 	$available = 0;
    1207. 

  1207. 	while(!$available) {
    1208. 

  1208. 	  $available = 1;
    1209. 

  1209. 	  $level++;
    1210. 

  1210. 	  for($j=$from;$j<=$to;$j++) {
    1211. 

  1211.             if (array_key_exists($level,$box) &&
    1212. 

  1212.                 array_key_exists($j,$box[$level])) {
    1213. 

  1213. 	      $available = 0;
    1214. 

  1214. 	    }
    1215. 

  1215.           }
    1216. 

  1216.         }
    1217. 

  1217. 	for($j=$from;$j<=$to;$j++) {
    1218. 

  1218. 	  $box[$level][$j] = $to;
    1219. 

  1219. 	}
    1220. 

  1220. 	$max_level = max($max_level,$level);
    1221. 

  1221. 	for($j=$from+1;$j<=$to;$j++) {
    1222. 

  1222. 	  $separable[$j] = 0;
    1223. 

  1223. 	}
    1224. 

  1224.       }
    1225. 

  1225.     }
    1226. 

  1226.   }
    1227. 

  1227.   $separable[count($word)] = 1;
    1228. 

  1228. 

  1229.   # display input phrases
    1230. 

  1230.   $sep_start = 0;
    1231. 

  1231.   for($sep_end=1;$sep_end<=count($word);$sep_end++) {
    1232. 

  1232.     if ($separable[$sep_end] == 1) {
    1233. 

  1233.       # one table for each separable block
    1234. 

  1234.       print "<table cellpadding=1 cellspacing=0 border=0 style=\"display: inline;\">";
    1235. 

  1235.       for($level=$max_level;$level>=1;$level--) {
    1236. 

  1236.         # rows for phrase display
    1237. 

  1237. 	print "<tr style=\"height:5px;\">";
    1238. 

  1238. 	for($from=$sep_start;$from<$sep_end;$from++) {
    1239. 

  1239. 	  if (array_key_exists($from,$box[$level])) {
    1240. 

  1240. 	    $to = $box[$level][$from];
    1241. 

  1241.             $size = $to - $from + 1;
    1242. 

  1242. 	    if ($size == 1) {
    1243. 

  1243. 	      print "<td><div style=\"height:0px; opacity:0; position:relative; z-index:-9;\">".$word[$from];
    1244. 

  1244.             }
    1245. 

  1245. 	    else {
    1246. 

  1246. 		$color = coverage_color($coverage[$from][$to]);
    1247. 

  1247. 		$phrase = "";
    1248. 

  1248. 		$highlightwords = "";
    1249. 

  1249.                 $lowlightwords = "";
    1250. 

  1250. 		for($j=$from;$j<=$to;$j++) {
    1251. 

  1251. 		  if ($j>$from) { $phrase .= " "; }
    1252. 

  1252. 		  $phrase .= $word[$j];
    1253. 

  1253.                   $highlightwords .= " document.getElementById('inputword-$i-$j').style.backgroundColor='#ffff80';";
    1254. 

  1254.                   $lowlightwords .= " document.getElementById('inputword-$i-$j').style.backgroundColor='".coverage_color($coverage[$j][$j])."';";
    1255. 

  1255. 		}
    1256. 

  1256. 	        print "<td colspan=$size><div style=\"background-color: $color; height:3px;\" onmouseover=\"show_word_info($sentence,".$coverage[$from][$to]["corpus_count"].",".$coverage[$from][$to]["ttable_count"].",".$coverage[$from][$to]["ttable_entropy"]."); this.style.backgroundColor='#ffff80';$highlightwords\" onmouseout=\"hide_word_info($sentence); this.style.backgroundColor='$color';$lowlightwords;\"".($biconcor?" onclick=\"show_biconcor($sentence,'".base64_encode($phrase)."');\"":"").">";
    1257. 

  1257.             }
    1258. 

  1258.             print "</div></td>";
    1259. 

  1259. 	    $from += $size-1;
    1260. 

  1260. 	  }
    1261. 

  1261. 	  else {
    1262. 

  1262. 	    print "<td><div style=\"height:".($from==$to ? 0 : 3)."px;\"></div></td>";
    1263. 

  1263. 	  }
    1264. 

  1264. 	}
    1265. 

  1265. 	print "</tr>\n";
    1266. 

  1266.       }
    1267. 

  1267.       # display input words
    1268. 

  1268.       print "<tr><td colspan=".($sep_end-$sep_start)."><div style=\"position:relative; z-index:1;\">";
    1269. 

  1269.       for($j=$sep_start;$j<$sep_end;$j++) {
    1270. 

  1270.         if ($segmentation && array_key_exists($j,$segmentation["input_start"])) {
    1271. 

  1271.           $id = $segmentation["input_start"][$j];  
    1272. 

  1272.           print "<span id=\"input-$sentence-$id\" style=\"border-color:#000000; border-style:solid; border-width:1px;\" onmouseover=\"highlight_phrase($sentence,$id);\" onmouseout=\"lowlight_phrase($sentence,$id);\">";
    1273. 

  1273.         }
    1274. 

  1274.         if (array_key_exists($j,$coverage)) {
    1275. 

  1275.           $color = coverage_color($coverage[$j][$j]);
    1276. 

  1276.           $cc = $coverage[$j][$j]["corpus_count"];
    1277. 

  1277.           $tc = $coverage[$j][$j]["ttable_count"];
    1278. 

  1278.           $te = $coverage[$j][$j]["ttable_entropy"];
    1279. 

  1279.         }
    1280. 

  1280.         else { # unknown words
    1281. 

  1281. 	  $color = '#ffffff';
    1282. 

  1282.           $cc = 0; $tc = 0; $te = 0;
    1283. 

  1283.         }
    1284. 

  1284.         print "<span id=\"inputword-$sentence-$j\" style=\"background-color: $color;\" onmouseover=\"show_word_info($sentence,$cc,$tc,$te); this.style.backgroundColor='#ffff80';\" onmouseout=\"hide_word_info($sentence);  this.style.backgroundColor='$color';\"".($biconcor?" onclick=\"show_biconcor($sentence,'".base64_encode($word[$j])."');\"":"").">";
    1285. 

  1285. 	if ($word[$j] == $filter) {
    1286. 

  1286. 	  print "<b><font color=#ff0000>".$word[$j]."</font></b>";
    1287. 

  1287. 	}
    1288. 

  1288. 	else {
    1289. 

  1289. 	  print $word[$j];
    1290. 

  1290. 	}
    1291. 

  1291. 	print "</span>";
    1292. 

  1292.         if ($segmentation && array_key_exists($j,$segmentation["input_end"])) {
    1293. 

  1293.           print "</span>";
    1294. 

  1294.         }
    1295. 

  1295.         print " ";
    1296. 

  1296.       }
    1297. 

  1297.       print "</div></td></tr>\n";
    1298. 

  1298.       print "</table>\n";
    1299. 

  1299.       $sep_start = $sep_end;
    1300. 

  1300.     }
    1301. 

  1301.   }
    1302. 

  1302.   print "<br>";
    1303. 

  1303. }
    1304. 

  1304. 

  1305. // color-coded coverage stats (corpus count, ttable count, entropy)
    1306. 

  1306. function coverage_color($phrase) {
    1307. 

  1307.   $corpus_count = 255 - 10 * log(1 + $phrase["corpus_count"]);
    1308. 

  1308.   if ($corpus_count < 128) { $corpus_count = 128; }
    1309. 

  1309.   $cc_color = dechex($corpus_count / 16) . dechex($corpus_count % 16);
    1310. 

  1310. 

  1311.   $ttable_count = 255 - 20 * log(1 + $phrase["ttable_count"]);
    1312. 

  1312.   if ($ttable_count < 128) { $ttable_count = 128; }
    1313. 

  1313.   $tc_color = dechex($ttable_count / 16) . dechex($ttable_count % 16);
    1314. 

  1314. 

  1315.   $ttable_entropy = 255 - 32 * $phrase["ttable_entropy"];
    1316. 

  1316.   if ($ttable_entropy < 128) { $ttable_entropy = 128; }
    1317. 

  1317.   $te_color = dechex($ttable_entropy / 16) . dechex($ttable_entropy % 16);
    1318. 

  1318. 

  1319. //  $color = "#". $cc_color . $te_color .  $tc_color; # reddish browns with some green
    1320. 

  1320. //  $color = "#". $cc_color . $tc_color .  $te_color; # reddish brown with some blueish purple
    1321. 

  1321.   $color = "#". $te_color . $cc_color .  $tc_color; # pale green towards red
    1322. 

  1322. //  $color = "#". $te_color . $tc_color .  $cc_color; # pale purple towards red
    1323. 

  1323. //  $color = "#". $tc_color . $te_color .  $cc_color; // # blue-grey towards green
    1324. 

  1324. //  $color = "#". $tc_color . $cc_color .  $te_color; // # green-grey towards blue
    1325. 

  1325. 

  1326.   return $color;
    1327. 

  1327. }
    1328. 

  1328. 

  1329. // annotate an output sentence
    1330. 

  1330. function output_annotation($sentence,$system,$segmentation) {
    1331. 

  1331.   #$color = array("#FFC0C0","#FFC0FF","#C0C0FF","#C0FFFF","#C0FFC0");
    1332. 

  1332.   $color = array("#c0c0c0","#e0e0ff","#b0b0ff","#8080ff","#4040ff");
    1333. 

  1333.   $word = split(" ",$system);
    1334. 

  1334. 

  1335.   for($j=0;$j<count($word);$j++) {
    1336. 

  1336.     list($surface,$correct) = split("\|", $word[$j]);
    1337. 

  1337.     if ($segmentation && array_key_exists($j,$segmentation["output_start"])) {
    1338. 

  1338.       $id = $segmentation["output_start"][$j];
    1339. 

  1339.       print "<span id=\"output-$sentence-$id\" style=\"border-color:#000000; border-style:solid; border-width:1px;\" onmouseover=\"highlight_phrase($sentence,$id);\" onmouseout=\"lowlight_phrase($sentence,$id);\">";
    1340. 

  1340.     }
    1341. 

  1341.     print "<span style=\"background-color: $color[$correct]\">$surface</span>";
    1342. 

  1342.     if ($segmentation && array_key_exists($j,$segmentation["output_end"])) {
    1343. 

  1343.       print "</span>";
    1344. 

  1344.     }
    1345. 

  1345.     print " ";
    1346. 

  1346.   }
    1347. 

  1347. }
    1348. 

  1348. 

  1349. function annotation_hierarchical($sentence,$segmentation,$segmentation_out,$node) {
    1350. 

  1350.     print "<script language=\"javascript\">\n";
    1351. 

  1351.     print "max_depth[$sentence] = ".strlen($segmentation[0]["brackets"]).";\n";
    1352. 

  1352.     print "span_count_out[$sentence] = ".count($segmentation_out).";\n";
    1353. 

  1353.     print "span_count_in[$sentence] = ".count($segmentation).";\n";
    1354. 

  1354.     print "nodeIn[$sentence] = [];\n";
    1355. 

  1355.     print "nodeOut[$sentence] = [];\n";
    1356. 

  1356.     print "nodeChildren[$sentence] = [];\n";
    1357. 

  1357.     for($n=0;$n<count($node);$n++) {
    1358. 

  1358. 	print "nodeIn[$sentence].push({ start: ".$node[$n]['start_div_in'].", end: ".$node[$n]['…
    1359.

Large files files are truncated, but you can click here to view the full file