// generate a result set from two sets of datapoints of which the first set contains all
// datapoints with other datapoints in the buffer zone and the of which the second set
// contains all datapoints without other datapoints in the buffer zone
dataset generateSet(dataset& withNearbyDataset, dataset& standaloneDataset) {
random_device rd;
mt19937 rng(rd());
dataset remainingDataset(withNearbyDataset.begin(), withNearbyDataset.end());
dataset resultSet(standaloneDataset.begin(), standaloneDataset.end());
while (remainingDataset.size() != 0) {
// create iterator
dataset::iterator it = remainingDataset.begin();
// generate random index
uniform_int_distribution<int> uni(0, (int)remainingDataset.size());
int r = uni(rng);
// pick random datapoint by advancing the iterator to the random position
advance(it, r % remainingDataset.size());
// add picked datapoint to result list
resultSet.insert(*it);
// remove all datapoints within buffer zone if still in remaining dataset
for (dataset::iterator j = it->buffer.begin(); j != it->buffer.end(); ++j) {
dataset::iterator tmp = remainingDataset.find(*j);
if (tmp != remainingDataset.end()) {
remainingDataset.erase(tmp);
}
}
// remove picked datapoint from remaining list
remainingDataset.erase(remainingDataset.find(*it));
}
return resultSet;
}
F foreach_well(const dataset& data, F fn, std::string id_field)
{
const auto& id = data.at(id_field);
std::size_t begin_rec = 0, end_rec = 0;
for (std::size_t i = 0; i < id.size(); ++i) {
if (id[i] != id[begin_rec] || i == id.size() - 1) {
if (i == id.size() - 1)
end_rec = i;
dataset well;
std::for_each(data.begin(), data.end(),
[&](const std::pair<std::string,
std::vector<std::string>>& column)
{
well[column.first] = std::vector<std::string>(
column.second.data() + begin_rec,
column.second.data() + end_rec + 1);
}
);
fn(well);
begin_rec = i;
}
end_rec = i;
}
return fn;
}
double chunk_evaluate_ioe1(dataset & data, chunkset & chunks) {
vector<int> human_chk_count;
vector<int> model_chk_count;
vector<int> match_chk_count;
int i;
int num_chunks = chunks.size();
for (i = 0; i < num_chunks; i++) {
human_chk_count.push_back(0);
model_chk_count.push_back(0);
match_chk_count.push_back(0);
}
dataset::iterator datait;
for (datait = data.begin(); datait != data.end(); datait++) {
for (i = 0; i < num_chunks; i++) {
human_chk_count[i] += count_chunks_ioe1(1, *datait, chunks[i][0], chunks[i][1]);
model_chk_count[i] += count_chunks_ioe1(2, *datait, chunks[i][0], chunks[i][1]);
match_chk_count[i] += count_matching_chunks_ioe1(*datait, chunks[i][0], chunks[i][1]);
}
}
printf("\tChunk-based performance evaluation:\n\n");
printf("\t\tChunk\tManual\tModel\tMatch\tPre.(%)\tRec.(%)\tF1-Measure(%)\n");
printf("\t\t-----\t------\t-----\t-----\t-------\t-------\t-------------\n");
int count = 0;
double pre = 0.0, rec = 0.0, f1 = 0.0;
double total1_pre = 0.0, total1_rec = 0.0, total1_f1 = 0.0;
double total2_pre = 0.0, total2_rec = 0.0, total2_f1 = 0.0;
int total_human = 0, total_model = 0, total_match = 0;
for (i = 0; i < num_chunks; i++) {
if (model_chk_count[i] > 0) {
pre = (double)match_chk_count[i] / model_chk_count[i];
total_model += model_chk_count[i];
total1_pre += pre;
} else {
pre = 0.0;
}
if (human_chk_count[i] > 0) {
rec = (double)match_chk_count[i] / human_chk_count[i];
total_human += human_chk_count[i];
total1_rec += rec;
count++;
} else {
rec = 0.0;
}
total_match += match_chk_count[i];
if (pre + rec > 0) {
f1 = (double) 2 * pre * rec / (pre + rec);
} else {
f1 = 0.0;
}
printf("\t\t%s\t%d\t%d\t%d\t%6.2f\t%6.2f\t%6.2f\n",
chunks[i][2].c_str(), human_chk_count[i], model_chk_count[i],
match_chk_count[i], pre * 100, rec * 100, f1 * 100);
}
printf("\t\t-----\t------\t-----\t-----\t-------\t-------\t-------------\n");
if (count > 0) {
total1_pre /= count;
total1_rec /= count;
if (total1_pre + total1_rec > 0) {
total1_f1 = 2 * total1_pre * total1_rec / (total1_pre + total1_rec);
}
printf("\t\tAvg1.\t\t\t\t%6.2f\t%6.2f\t%6.2f\n",
total1_pre * 100, total1_rec * 100, total1_f1 * 100);
}
if (total_model > 0) {
total2_pre = (double)total_match / total_model;
}
if (total_human > 0) {
total2_rec = (double)total_match / total_human;
}
if (total2_pre + total2_rec > 0) {
total2_f1 = 2 * total2_rec * total2_pre / (total2_rec + total2_pre);
}
printf("\t\tAvg2.\t%d\t%d\t%d\t%6.2f\t%6.2f\t%6.2f\n\n",
total_human, total_model, total_match,
total2_pre * 100, total2_rec * 100, total2_f1 * 100);
return total2_f1 * 100;
}
// call this to compute precision, recall, and F1-measure
void evaluate(dataset & data, string & chunktype, labelset & labels, chunkset & chunks) {
map<string, int> lbstr2int;
map<int, string> lbint2str;
vector<int> human_lb_count, model_lb_count, human_model_lb_count;
int i;
int num_labels = labels.size();
for (i = 0; i < num_labels; i++) {
lbstr2int.insert(pair<string, int>(labels[i], i));
lbint2str.insert(pair<int, string>(i, labels[i]));
human_lb_count.push_back(0);
model_lb_count.push_back(0);
human_model_lb_count.push_back(0);
}
// start to count
dataset::iterator datait;
sequence::iterator seqit;
for (datait = data.begin(); datait != data.end(); datait++) {
for (seqit = datait->begin(); seqit != datait->end(); seqit++) {
int label = str_2_int(lbstr2int, (*seqit)[seqit->size() - 2]);
int model_label = str_2_int(lbstr2int, (*seqit)[seqit->size() - 1]);
if (label >= 0 && label < num_labels) {
human_lb_count[label]++;
}
if (model_label >= 0 && model_label < num_labels) {
model_lb_count[model_label]++;
}
if (label == model_label && label >= 0 && label < num_labels) {
human_model_lb_count[label]++;
}
}
}
// print out
printf("\tLabel-based performance evaluation:\n\n");
printf("\t\tLabel\tManual\tModel\tMatch\tPre.(%)\tRec.(%)\tF1-Measure(%)\n");
printf("\t\t-----\t------\t-----\t-----\t-------\t-------\t-------------\n");
int count = 0;
double precision = 0.0, recall = 0.0, f1, total1_pre = 0.0,
total1_rec = 0.0, total1_f1 = 0.0, total2_pre = 0.0,
total2_rec = 0.0, total2_f1 = 0.0;
int total_human = 0, total_model = 0, total_match = 0;
for (i = 0; i < num_labels; i++) {
if (model_lb_count[i] > 0) {
precision = (double)human_model_lb_count[i] / model_lb_count[i];
total_model += model_lb_count[i];
total1_pre += precision;
} else {
precision = 0.0;
}
if (human_lb_count[i] > 0) {
recall = (double)human_model_lb_count[i] / human_lb_count[i];
total_human += human_lb_count[i];
total1_rec += recall;
count++;
} else {
recall = 0.0;
}
total_match += human_model_lb_count[i];
if (recall + precision > 0) {
f1 = (double) 2 * precision * recall / (precision + recall);
} else {
f1 = 0;
}
char buff[50];
sprintf(buff, "%d", i);
string strlabel = int_2_str(lbint2str, i);
if (strlabel != "") {
sprintf(buff, "%s", strlabel.c_str());
}
printf("\t\t%s\t%d\t%d\t%d\t%6.2f\t%6.2f\t%6.2f\n",
buff, human_lb_count[i], model_lb_count[i], human_model_lb_count[i],
precision * 100, recall * 100, f1 * 100);
}
total1_pre /= count;
total1_rec /= count;
total1_f1 = 2 * total1_pre * total1_rec / (total1_pre + total1_rec);
// print the average performance
total2_pre = (double)total_match / total_model;
total2_rec = (double)total_match / total_human;
total2_f1 = 2 * total2_pre * total2_rec / (total2_pre + total2_rec);
printf("\t\t-----\t------\t-----\t-----\t-------\t-------\t-------------\n");
printf("\t\tAvg1.\t\t\t\t%6.2f\t%6.2f\t%6.2f\n", total1_pre * 100, total1_rec * 100,
total1_f1 * 100);
printf("\t\tAvg2.\t%d\t%d\t%d\t%6.2f\t%6.2f\t%6.2f\n\n", total_human, total_model, total_match,
total2_pre * 100, total2_rec * 100, total2_f1 * 100);
if (chunks.size() <= 0) {
return;
}
// chunk based evaluation
if (chunktype == "IOB1") {
chunk_evaluate_iob1(data, chunks);
}
if (chunktype == "IOB2") {
chunk_evaluate_iob2(data, chunks);
}
if (chunktype == "IOE1") {
chunk_evaluate_ioe1(data, chunks);
}
if (chunktype == "IOE2") {
chunk_evaluate_ioe2(data, chunks);
}
}