1 |
#!/usr/bin/perl -w |
2 |
|
3 |
|
4 |
###################################### |
5 |
# |
6 |
# A simple search engine program |
7 |
# |
8 |
###################################### |
9 |
|
10 |
|
11 |
use strict; |
12 |
use CGI::Carp qw(fatalsToBrowser); |
13 |
use Semantic::API; |
14 |
use CGI; |
15 |
use Data::Dump qw/dump/; |
16 |
use EPrints; |
17 |
use Cwd qw/abs_path/; |
18 |
use KinoSearch::Simple; |
19 |
use lib '/home/dpavlin/stem-hr/'; |
20 |
use StemHR; |
21 |
|
22 |
my $debug = 1; |
23 |
|
24 |
my $abs_path; |
25 |
|
26 |
BEGIN { |
27 |
use CGI::Carp qw(carpout); |
28 |
|
29 |
$abs_path = abs_path( $0 ); |
30 |
$abs_path =~ s!/[^/]*$!/!; #!fix-vim |
31 |
|
32 |
open(my $log_fh, '>>', "$abs_path/log") or |
33 |
die("Unable to open $abs_path/log: $!\n"); |
34 |
carpout($log_fh); |
35 |
} |
36 |
|
37 |
|
38 |
############################################################# |
39 |
my $COLLECTION = 'EPrints'; |
40 |
my ( @TERMS, @RESULTS ); |
41 |
my ( $RESULTS_TO_DISPLAY, $TERMS_TO_DISPLAY ) = ( 20, 20 ); |
42 |
############################################################# |
43 |
|
44 |
|
45 |
############################### |
46 |
# CGI Variables |
47 |
############################### |
48 |
my $cgi = new CGI; |
49 |
my $start = $cgi->param( 'start' ) || 0; |
50 |
my $query = $cgi->param( 'query' ) || ''; |
51 |
my $similar = $cgi->param( 'similar' ) || ''; |
52 |
my $slogovi = $cgi->param( 'slogovi' ) || ''; |
53 |
my $stem = $cgi->param( 'stem' ) || ''; |
54 |
my $similar_to_kino = $cgi->param( 'similar_to_kino' ); |
55 |
|
56 |
my $charset='iso-8859-2'; |
57 |
|
58 |
my $full_query = $query; |
59 |
$full_query .= " " . EPrints->slogovi( $query ) if ($slogovi); |
60 |
$full_query .= " " . StemHR->stem( $query ) if ($stem); |
61 |
|
62 |
############################## |
63 |
# Start the HTML output |
64 |
############################## |
65 |
print "Content-type: text/html; charset=$charset\n\n"; |
66 |
print qq|<?xml version="1.0" encoding="$charset"?> |
67 |
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN" |
68 |
"http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd"> |
69 |
<html xmlns="http://www.w3.org/1999/xhtml"> |
70 |
<head> |
71 |
<meta http-equiv="content-type" content="text/html; charset=$charset" /> |
72 |
<title>Semantic Search Engine</title> |
73 |
</head> |
74 |
<body> |
75 |
<form method="get" action=""> |
76 |
<p> |
77 |
Enter bunch of related terms to documents you are trying to find: |
78 |
<br/><input type="text" name="query" value="$query" size="80"> |
79 |
<br/><input type="submit" /> |
80 |
|, |
81 |
$cgi->checkbox( -name => 'stem' ), |
82 |
$cgi->checkbox( -name => 'slogovi' ), |
83 |
$cgi->checkbox( -name => 'similar_to_kino', -label=>'similar to fulltext' ), |
84 |
qq| |
85 |
</p>\n|; |
86 |
|
87 |
|
88 |
|
89 |
########################## |
90 |
# Do the actual search |
91 |
########################## |
92 |
if( $query || $similar ) { |
93 |
|
94 |
# Create collection-based objects |
95 |
my $semantic = Semantic::API::Search->new( storage => 'sqlite', |
96 |
database => "$abs_path/eprints.db", |
97 |
collection => $COLLECTION ); |
98 |
|
99 |
my $score_ponder; |
100 |
|
101 |
my ($results, $terms); |
102 |
if ( $query ) { |
103 |
my $kino = KinoSearch::Simple->new( |
104 |
path => 'kinoindex/', |
105 |
language => 'ru', |
106 |
); |
107 |
|
108 |
my $kino_query = join(" AND ", split(/\s+/, $query ) ); |
109 |
|
110 |
my $total_hits = $kino->search( |
111 |
query => $kino_query, |
112 |
offset => 0, |
113 |
num_wanted => 10, |
114 |
); |
115 |
|
116 |
my @similar_ids; |
117 |
my ( $max, $min ) = ( 0,0 ); |
118 |
|
119 |
print qq|<div style="width:20%; float:right;">KinoSearch hits $total_hits for $kino_query\n<ol>|; |
120 |
while ( my $hit = $kino->fetch_hit_hashref ) { |
121 |
print qq|<li><a href="#id_$hit->{id}">$hit->{title}</a> $hit->{score}</li>\n|; |
122 |
if ( $similar_to_kino ) { |
123 |
push @similar_ids, $hit->{id}; |
124 |
} else { |
125 |
$score_ponder->{ $hit->{id} } = $hit->{score}; |
126 |
$min = $hit->{score} if ( $hit->{score} < $min ); |
127 |
$max = $hit->{score} if ( $hit->{score} > $max ); |
128 |
} |
129 |
} |
130 |
|
131 |
my $d = $max - $min; |
132 |
|
133 |
map { |
134 |
$score_ponder->{ $_ } -= $min; |
135 |
$score_ponder->{ $_ } /= $d; |
136 |
warn "score_ponder $_ = ", $score_ponder->{$_} if $debug; |
137 |
} keys %$score_ponder; |
138 |
|
139 |
if ( $similar_to_kino ) { |
140 |
print qq|</ol></div>|; |
141 |
($results, $terms) = $semantic->find_similar( @similar_ids ); |
142 |
} else { |
143 |
print qq|</ol>score range: $min - $max</div>|; |
144 |
($results, $terms) = $semantic->semantic_search( $full_query ); |
145 |
} |
146 |
|
147 |
} else { |
148 |
($results, $terms) = $semantic->find_similar( $similar ); |
149 |
} |
150 |
|
151 |
warn "results = ",dump( $results ) if $debug; |
152 |
warn "terms = ",dump( $terms ) if $debug; |
153 |
|
154 |
################################## |
155 |
# TERM BASED CALCULATIONS |
156 |
################################## |
157 |
my @sorted_terms = sort { $terms->{$b} <=> $terms->{$a} } keys %$terms; |
158 |
my @top_terms = splice( @sorted_terms, $start, $TERMS_TO_DISPLAY ); |
159 |
|
160 |
warn "top_terms = ", dump( @top_terms ) if $debug; |
161 |
|
162 |
print "<p>Full query: $full_query</p>\n"; |
163 |
print "<p>Related Terms: ". ( join ", ", @top_terms ) ."</p>\n"; |
164 |
print "<hr />\n"; |
165 |
|
166 |
|
167 |
################################## |
168 |
# DOCUMENT BASED CALCULATIONS |
169 |
################################## |
170 |
|
171 |
print "<p>Result Count: ".(scalar keys %$results)."</p>\n"; |
172 |
|
173 |
my @sorted_results = sort { |
174 |
$results->{$b} <=> $results->{$a} |
175 |
} map { |
176 |
$results->{$_} *= ( $score_ponder->{$_} || 0.1 ); $_; |
177 |
} keys %$results; |
178 |
my @display_results = splice( @sorted_results, $start, $RESULTS_TO_DISPLAY ); |
179 |
|
180 |
warn "display results = ", dump( @display_results ) if $debug; |
181 |
|
182 |
################################## |
183 |
# Access the storage engine to |
184 |
# retrieve the title and text |
185 |
################################## |
186 |
my $i = 1 + $start; |
187 |
print $semantic->paginate( "?query=$query;similar=$similar;stem=$stem;slogovi=$slogovi", $start, scalar keys %$results, $RESULTS_TO_DISPLAY); |
188 |
foreach my $id ( @display_results ){ |
189 |
EPrints->id( $id ); |
190 |
print "<p><a name=\"id_$id\"/>$i. <b>", EPrints->lookup( 'title' ), "</b>"; |
191 |
print "| score: <em>", sprintf("%.2f",$results->{$id}), "</em> | id: $id | <a href=\"?similar=$id\">similar</a> | "; |
192 |
my ($type,$uri) = EPrints->fulltext; |
193 |
print qq|<a href="$uri">$type</a>|; |
194 |
print "</p>\n"; |
195 |
print "<p>"; |
196 |
# print $semantic->summarize($id); |
197 |
print "</p><p>Keywords: ", EPrints->lookup('keywords'), "</p><p>"; |
198 |
print "<small>", EPrints->lookup('abstract'), "</small>"; |
199 |
print "</p>\n"; |
200 |
$i++; |
201 |
} |
202 |
|
203 |
} |
204 |
|
205 |
|
206 |
print "</body>\n</html>\n"; |