private void Populate()
{
tsvs.Clear();
string rootpath = System.IO.Path.GetDirectoryName(Application.dataPath);
findDir = System.IO.Path.Combine(AppSetting.localDataPath, "setting"); //捕鱼
var allFiles = Directory.GetFiles(findDir, "*.*", SearchOption.AllDirectories);
foreach (var file in allFiles)
{
TSVFile tsv = new TSVFile(file);
tsv.hash = GetHash(tsv.fileName); //捕鱼
Debug.LogError(tsv.fileName + " " + file);
Debug.LogError(tsv.hash);
tsvs.Add(tsv);
}
}
static void Main(string[] args)
{
string inputExtraction = args[0];
string outputExtraction = args[1];
string word2vecBinaryModel = args[2];
Dictionary <String, float[]> model = LoadWord2vecModel(word2vecBinaryModel);
using (StreamReader sr = new StreamReader(TSVFile.OpenInputTSVStream(inputExtraction)))
using (StreamWriter sw = new StreamWriter(TSVFile.OpenOutputTSVStream(outputExtraction, false)))
{
double[] feature_weight = new double[300] {
0.2538461826219447, -0.0686701626700071, -0.09006037319865194, 0.010241813154946948, 0.16611995837541962, -0.3200690030466925, -0.05746840591717508, -0.1927495240028911, -0.22660920941591495, -0.10381030156269831, -0.14736031889966342, 0.11536268609662927, 0.12119522536615442, -0.039806620784286324, -0.0025101378952937688, -0.2153229411692349, -0.06793695499553809, -0.19982183784820795, 0.1785592770014214, -0.03728167079366318, -0.04866069425856831, -0.05480345738563925, 0.03806922005174796, -0.18659513727092658, 0.035261132976043434, 0.09538052854804377, 0.06910223780419475, -0.026332687914170854, 0.0009809842277506747, 0.1347575137793603, -0.0681431589269305, -0.03505422921691469, 0.04904192967538813, 0.1599604306864603, 0.05009272261731429, -0.1258245585154701, -0.12267849745633382, -0.06445273569908316, 0.04455768851052863, 0.1456388688807849, -0.057093726230681846, -0.12936979341629265, -0.13638922771686002, -0.08939319319834929, -0.011476289596134033, 0.15948959216655081, 0.18045344609112682, 0.1715445691915844, 0.11689097401556907, 0.00357715055288158, 0.1033659184460083, -0.13710154342921194, -0.13340939222299508, -0.13417341504660002, 0.02397604688687353, 0.08238485079744025, 0.05072189657209351, 0.24972006746824357, -0.04953889910780581, 0.1987699083332565, -0.06863651671206697, 0.0708081998661408, -0.01772202917173831, -0.17708075170242213, 0.1092504029870327, -0.1615226924410854, -0.002824149896208, -0.15871479301424538, 0.09551884065988868, 0.22499299478770737, 0.042810565305358976, -0.12060475691977857, -0.1649414248444354, 0.0950155892901811, -0.007700405282429591, -0.14405241847707634, 0.09016637986593633, 0.0, 0.03517828854534869, 0.08261696667192023, -0.012270322184217588, -0.14966906300820781, 0.049029023162959526, -0.146348688381927, -0.10657611733470904, -0.10655303765811588, 0.20842227542584651, 0.21350690259445912, -0.19801641517204105, -0.20697618723363664, -0.08597897404990515, 0.05420691069740911, -0.11987485077792477, -0.19602894653596667, -0.30559499902745363, 0.04073064278074961, 0.2212026667741698, -0.05832500695868755, -0.10394692395991723, 0.3695558830689171, -0.005793209310414317, 0.01903310756683418, 0.11051303508366477, -0.10246715824550284, -0.03186798882681539, 0.2014930754053219, -0.10764951320539465, 0.11619974102593565, -0.06571329645418186, 0.05213321854838495, -0.1451942065295862, -0.0031139050846557895, -0.021807048007396152, 0.06812344103068556, -0.31649584251145874, 0.20166552377872646, 0.2216094587519973, -0.13333208957167494, -0.10996714815341177, -0.1571511611578555, 0.011146248503686357, -0.21279734217335172, -0.3086844176648976, 0.13960434294636065, -0.17544726875147848, -0.26752957670639677, -0.18539788931921058, 0.12403277495848146, -0.13146737000714312, 0.045057415657836365, 0.09549312508753022, 0.1282164084042249, -0.1852291525033577, -0.12363997752903899, 0.025274005243707477, -0.1432111161966799, 0.028109335662460138, -0.19241225973649792, 0.25063543765733615, -0.024721741876937833, 0.06764354492418954, -0.029344067911690565, -0.03026715835880135, -0.2206575941082282, -0.21532023620068988, 0.031279984478324654, 0.21756237100709266, 0.09470035331599212, -0.07611208412946659, 0.09019794038772162, -0.1083477012384884, -0.17691437850337238, -0.1879493155919005, -0.006056503489002652, -0.1404576364445292, 0.11202378571327978, 0.04439694489311643, -0.16093301008062177, 0.03988707430368873, -0.0015753594664960938, -0.04629162013281132, 0.06962310554652107, 0.0533513404393514, 0.15868486162453477, -0.02931484289440335, 0.009453368678482111, -0.038394404037557554, 0.061849554088233484, 0.0, -0.006642123233885528, 0.008869541483276837, -0.2053017895318143, -0.20969205915119835, -0.010971307237947456, -0.09785662435795048, -0.23031486653075323, 0.05387902285400744, -0.03590919473861954, 0.20686936388824181, -0.11193556823296877, 0.023196225103504914, 0.14765682560593255, -0.2612241571497319, 0.13992584571874067, -0.00890802236698661, -0.10109580710689492, 0.1433220484486877, -0.2785484734344504, 0.23486021428980686, -0.05438133348268732, -0.08227834239204994, 0.262134353420296, 0.039216623584098435, 0.29641728240268195, 0.18816167568582823, -0.009999733229932221, -0.029847182257469417, 0.034313840286451604, -0.16388570990350332, -0.035028095836567304, 0.003648822232978713, 0.055995982015074844, -0.015422474389861099, 0.2317731051717546, 0.0042360297926754605, 0.018260764542252533, 0.06469089965819264, 0.18536546202138415, 0.06514611562185264, -0.00239938170349781, 0.03314845714693998, -0.05145930970774032, -0.01087456297094482, -0.03217405832674326, -0.018547721106957266, 0.13324243317501508, -0.2305852610248915, -0.1545052789766213, 0.08686151415013053, -0.020826343844124167, 0.06525417114290195, -0.08861328325945618, 0.06215810645144515, -0.17174651019914483, -0.0715576951991129, 0.014493799522224881, 0.011214157751636698, 0.12174100394428665, 0.049637394007593545, 0.18049744595730924, 0.005001785036087733, -0.09329506938347155, 0.18559212312749554, -0.15239136908391027, -0.004325657683385898, 0.012133537097089604, 0.036038759959274545, 0.10471249983988289, -0.04725311662558393, 0.20631377417775082, -0.07397148558402832, -0.01047857915026429, 0.0645852939696661, -0.033278644546580284, 0.04357401529335759, -0.31183745887279035, 0.17123628011259803, 0.1818856907218545, -0.09675749934956455, 0.188981562180895, -0.08189464986182685, -0.03936709371880006, 0.14385154415969278, 0.0711579375736589, -0.06888724935373179, 0.04168909711619999, -0.01258340882825076, 0.05300427771251361, 0.11186829284302975, 0.11895481404113507, 0.03375407989670895, 0.19704550656217293, 0.07172560346412016, 0.03141101182352664, 0.08349284322149193, -0.14155144458842578, 0.07030418504419357, 0.07908101773813546, 0.2723930142314312, 0.13541112124844049, 0.07688919954539378, -0.008367473340056695, -0.035704947828022575, 0.29221989404594206, 0.037628554097414224, -0.0981772269673222, 0.2817900398981206, -0.05345188651367586, -0.013043812239508066, -0.2334262446413108, 0.150450791729621, -0.08971941701284769, -0.13387407121713765, 0.04619728510762145, -0.11815924390613752, 0.15364599731826173, 0.048670112770281666, -0.12594078551832524, -0.020983153233194673, -0.034821815325370174, -0.1715868593419294, -0.0792120422152978, 0.16840214358196495, 0.11721076757088779, 0.06484404805630939, 0.05906618529852858, -0.3429377835341666, -0.09679992436156105, 0.2426323808019555, -0.02137471106044561
};
TSVReader tsvReader = new TSVReader(sr, true, false);
// write new header
String header = tsvReader.GetHeaderLine();
Boolean has_title_snippet = false;
if (header.Contains("\tTitle\t"))
{
has_title_snippet = true;
}
header = header.Replace("Title\tSnippet\t", "");
header += "\ttermweight_wordsfound";
sw.WriteLine(header);
TSVLine line;
int count1 = 0;
while ((line = tsvReader.ReadLine()) != null)
{
count1++;
Dictionary <String, Double> term_weight_map = new Dictionary <String, Double>();
String query = line.GetFeatureValueString("Query");
String passage = line.GetFeatureValueString("Passage");
String[] terms = query.Split(' ');
// calculate wordvec of every term
Dictionary <String, float[]> term_word2vec = new Dictionary <String, float[]>();
for (int i = 0; i < terms.Length; i++)
{
String term = terms[i];
if (model.ContainsKey(term))
{
float[] vec = model[term];
if (!term_word2vec.ContainsKey(term))
{
term_word2vec.Add(term, vec);
}
}
}
// calculate average word2vec
float[] average = new float[300];
int count = 0;
foreach (String temp_term in term_word2vec.Keys)
{
float[] vec = term_word2vec[temp_term];
for (int i = 0; i < 300; i++)
{
average[i] += vec[i];
}
count++;
}
if (count > 0)
{
for (int i = 0; i < 300; i++)
{
average[i] /= count;
}
}
// calculate relative term word2vec to average
Dictionary <String, float[]> term_feature_vector = new Dictionary <String, float[]>();
foreach (String temp_term in term_word2vec.Keys)
{
float[] vec = term_word2vec[temp_term];
for (int i = 0; i < 300; i++)
{
vec[i] -= average[i];
}
term_feature_vector.Add(temp_term, vec);
}
// get term weight
foreach (String temp_term in term_feature_vector.Keys)
{
float[] vec = term_feature_vector[temp_term];
double termweight = 0;
for (int i = 0; i < 300; i++)
{
termweight += vec[i] * feature_weight[i];
}
termweight = Math.Exp(termweight) / (1 + Math.Exp(termweight));
term_weight_map.Add(temp_term, termweight);
}
// normalize the term weight : sum is 1
double sum = 0;
int all_count = 0;
foreach (String temp_term in term_weight_map.Keys)
{
sum += term_weight_map[temp_term];
}
if (sum == 0)
{
sum = 1;
}
Dictionary <String, double> normlized_term_weight_map = new Dictionary <String, double>();
if (sum != 0)
{
foreach (String temp_term in term_weight_map.Keys)
{
double normalize_termweight = term_weight_map[temp_term];
normalize_termweight /= sum;
normlized_term_weight_map.Add(temp_term, normalize_termweight);
}
}
String term_weight_str = "{";
foreach (String temp_term in term_weight_map.Keys)
{
term_weight_str += temp_term;
term_weight_str += ":";
term_weight_str += term_weight_map[temp_term];
term_weight_str += ",";
}
if (term_weight_str.EndsWith(","))
{
term_weight_str = term_weight_str.Remove(term_weight_str.Length - 1, 1);
}
term_weight_str += "}";
String[] lineitems = line.GetWholeLineString().Split('\t');
String out_line = "";
for (int j = 0; j < lineitems.Length; j++)
{
if (has_title_snippet)
{
if (j == 2 || j == 3)
{
continue;
}
}
out_line += lineitems[j];
if (j < (lineitems.Length - 1))
{
out_line += "\t";
}
}
//out_line += "\t"+ term_weight_str;
int wordsfound = GetTermWeightWordsfound(query, term_weight_map, passage);
out_line += "\t" + wordsfound;
sw.WriteLine(out_line);
}
}
}
