From bbb33206eedac63078cb1355c12e1e96c3138343 Mon Sep 17 00:00:00 2001 From: brightmart Date: Thu, 20 Dec 2018 23:06:31 +0800 Subject: [PATCH 01/13] comment out test() under p1_HierarchicalAttention_model_transformer.py comment out test() under p1_HierarchicalAttention_model_transformer.py --- .../p1_HierarchicalAttention_model_transformer.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/a05_HierarchicalAttentionNetwork/p1_HierarchicalAttention_model_transformer.py b/a05_HierarchicalAttentionNetwork/p1_HierarchicalAttention_model_transformer.py index 71dd2b5..718d415 100644 --- a/a05_HierarchicalAttentionNetwork/p1_HierarchicalAttention_model_transformer.py +++ b/a05_HierarchicalAttentionNetwork/p1_HierarchicalAttention_model_transformer.py @@ -472,4 +472,4 @@ def get_input_y(i,input_x,batch_size): input_y = np.array(input_y) return input_y -test() \ No newline at end of file +#test() From b74e0ef82b1375707dc15729b882f342fd90efee Mon Sep 17 00:00:00 2001 From: brightmart Date: Sat, 9 Feb 2019 02:55:13 +0800 Subject: [PATCH 02/13] Update README.md --- README.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/README.md b/README.md index 012dee0..ec6873b 100644 --- a/README.md +++ b/README.md @@ -5,6 +5,8 @@ the purpose of this repository is to explore text classification methods in NLP #### Update: try bert model for multi-label classification, please check session #Models Detail, 3) Bert. + +https://github.com/brightmart/nlp_chinese_corpus Google's BERT achieved new state of art result on more than 10 tasks in NLP using pre-train in language model then From 05218348abc25dbbf974b733cdc5658feb5a9e6b Mon Sep 17 00:00:00 2001 From: brightmart Date: Sat, 9 Feb 2019 02:56:33 +0800 Subject: [PATCH 03/13] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index ec6873b..a6a3513 100644 --- a/README.md +++ b/README.md @@ -6,7 +6,7 @@ the purpose of this repository is to explore text classification methods in NLP try bert model for multi-label classification, please check session #Models Detail, 3) Bert. -https://github.com/brightmart/nlp_chinese_corpus +project with large amount of chinese corpus for nlp is available now!' Google's BERT achieved new state of art result on more than 10 tasks in NLP using pre-train in language model then From 21d772a69798405f314e254acf454d19ad50c5b0 Mon Sep 17 00:00:00 2001 From: brightmart Date: Sat, 9 Feb 2019 02:58:16 +0800 Subject: [PATCH 04/13] Update README.md --- README.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index a6a3513..9479405 100644 --- a/README.md +++ b/README.md @@ -1,12 +1,12 @@ Text Classification ------------------------------------------------------------------------- -the purpose of this repository is to explore text classification methods in NLP with deep learning. +The purpose of this repository is to explore text classification methods in NLP with deep learning. #### Update: -try bert model for multi-label classification, please check session #Models Detail, 3) Bert. +Try bert model for multi-label classification, please check session #Models Detail, 3) Bert. -project with large amount of chinese corpus for nlp is available now!' +Large Amount of Chinese Corpus for NLP available!' Google's BERT achieved new state of art result on more than 10 tasks in NLP using pre-train in language model then From 0134a06606667d087fa9927a3ad5958c77c0b7f5 Mon Sep 17 00:00:00 2001 From: brightmart Date: Sat, 9 Feb 2019 02:58:42 +0800 Subject: [PATCH 05/13] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 9479405..c8f0a9a 100644 --- a/README.md +++ b/README.md @@ -6,7 +6,7 @@ The purpose of this repository is to explore text classification methods in NLP Try bert model for multi-label classification, please check session #Models Detail, 3) Bert. -Large Amount of Chinese Corpus for NLP available!' +Large Amount of Chinese Corpus for NLP Available! Google's BERT achieved new state of art result on more than 10 tasks in NLP using pre-train in language model then From a6715eaf03a2c32ee28dbf5227427809070bf455 Mon Sep 17 00:00:00 2001 From: gunjianpan Date: Fri, 12 Apr 2019 01:40:24 +0800 Subject: [PATCH 06/13] =?UTF-8?q?=F0=9F=99=8A=20fix=20macro=20f1?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- a02_TextCNN/p7_TextCNN_train.py | 259 ++++++++------------------------ 1 file changed, 60 insertions(+), 199 deletions(-) diff --git a/a02_TextCNN/p7_TextCNN_train.py b/a02_TextCNN/p7_TextCNN_train.py index 0f50670..ff901c4 100644 --- a/a02_TextCNN/p7_TextCNN_train.py +++ b/a02_TextCNN/p7_TextCNN_train.py @@ -15,6 +15,7 @@ import h5py import os import random +from numba import jit #configuration FLAGS=tf.app.flags.FLAGS @@ -54,8 +55,7 @@ def main(_): #print some message for debug purpose print("trainX[0:10]:", trainX[0:10]) print("trainY[0]:", trainY[0:10]) - train_y_short = get_target_label_short(trainY[0]) - print("train_y_short:", train_y_short) + print("train_y_short:", trainY[0]) #2.create session. config=tf.ConfigProto() @@ -128,203 +128,64 @@ def main(_): # 在验证集上做验证，报告损失、精确度 -def do_eval(sess,textCNN,evalX,evalY,num_classes): - evalX=evalX[0:3000] - evalY=evalY[0:3000] - number_examples=len(evalX) - eval_loss,eval_counter,eval_f1_score,eval_p,eval_r=0.0,0,0.0,0.0,0.0 - batch_size=1 - label_dict_confuse_matrix=init_label_dict(num_classes) - for start,end in zip(range(0,number_examples,batch_size),range(batch_size,number_examples,batch_size)): - feed_dict = {textCNN.input_x: evalX[start:end], textCNN.input_y_multilabel:evalY[start:end],textCNN.dropout_keep_prob: 1.0, +def do_eval(sess, textCNN, evalX, evalY, num_classes): + evalX = evalX[0:3000] + evalY = evalY[0:3000] + number_examples = len(evalX) + eval_loss, eval_counter, eval_f1_score, eval_p, eval_r = 0.0, 0, 0.0, 0.0, 0.0 + batch_size = FLAGS.batch_size + predict = [] + + for start, end in zip(range(0, number_examples, batch_size), range(batch_size, number_examples, batch_size)): + ''' evaluation in one batch ''' + feed_dict = {textCNN.input_x: evalX[start:end], textCNN.input_y_multilabel: evalY[start:end], textCNN.dropout_keep_prob: 1.0, textCNN.is_training_flag: False} - curr_eval_loss, logits= sess.run([textCNN.loss_val,textCNN.logits],feed_dict)#curr_eval_acc--->textCNN.accuracy - predict_y = get_label_using_logits(logits[0]) - target_y= get_target_label_short(evalY[start:end][0]) - #f1_score,p,r=compute_f1_score(list(label_list_top5), evalY[start:end][0]) - label_dict_confuse_matrix=compute_confuse_matrix(target_y, predict_y, label_dict_confuse_matrix) - eval_loss,eval_counter=eval_loss+curr_eval_loss,eval_counter+1 - - f1_micro,f1_macro=compute_micro_macro(label_dict_confuse_matrix) #label_dict_accusation is a dict, key is: accusation,value is: (TP,FP,FN). where TP is number of True Positive - f1_score=(f1_micro+f1_macro)/2.0 - return eval_loss/float(eval_counter),f1_score,f1_micro,f1_macro - -####################################### -def compute_f1_score(predict_y,eval_y): - """ - compoute f1_score. - :param logits: [batch_size,label_size] - :param evalY: [batch_size,label_size] - :return: - """ - f1_score=0.0 - p_5=0.0 - r_5=0.0 - return f1_score,p_5,r_5 - -def compute_f1_score_removed(label_list_top5,eval_y): - """ - compoute f1_score. - :param logits: [batch_size,label_size] - :param evalY: [batch_size,label_size] - :return: - """ - num_correct_label=0 - eval_y_short=get_target_label_short(eval_y) - for label_predict in label_list_top5: - if label_predict in eval_y_short: - num_correct_label=num_correct_label+1 - #P@5=Precision@5 - num_labels_predicted=len(label_list_top5) - all_real_labels=len(eval_y_short) - p_5=num_correct_label/num_labels_predicted - #R@5=Recall@5 - r_5=num_correct_label/all_real_labels - f1_score=2.0*p_5*r_5/(p_5+r_5+0.000001) - return f1_score,p_5,r_5 - -random_number=1000 -def compute_confuse_matrix(target_y,predict_y,label_dict,name='default'): - """ - compute true postive(TP), false postive(FP), false negative(FN) given target lable and predict label - :param target_y: - :param predict_y: - :param label_dict {label:(TP,FP,FN)} - :return: macro_f1(a scalar),micro_f1(a scalar) - """ - #1.get target label and predict label - if random.choice([x for x in range(random_number)]) ==1: - print(name+".target_y:",target_y,";predict_y:",predict_y) #debug purpose - - #2.count number of TP,FP,FN for each class - y_labels_unique=[] - y_labels_unique.extend(target_y) - y_labels_unique.extend(predict_y) - y_labels_unique=list(set(y_labels_unique)) - for i,label in enumerate(y_labels_unique): #e.g. label=2 - TP, FP, FN = label_dict[label] - if label in predict_y and label in target_y:#predict=1,truth=1 (TP) - TP=TP+1 - elif label in predict_y and label not in target_y:#predict=1,truth=0(FP) - FP=FP+1 - elif label not in predict_y and label in target_y:#predict=0,truth=1(FN) - FN=FN+1 - label_dict[label] = (TP, FP, FN) - return label_dict - -def compute_micro_macro(label_dict): - """ - compute f1 of micro and macro - :param label_dict: - :return: f1_micro,f1_macro: scalar, scalar - """ - f1_micro = compute_f1_micro_use_TFFPFN(label_dict) - f1_macro= compute_f1_macro_use_TFFPFN(label_dict) - return f1_micro,f1_macro - -def compute_TF_FP_FN_micro(label_dict): - """ - compute micro FP,FP,FN - :param label_dict_accusation: a dict. {label:(TP, FP, FN)} - :return:TP_micro,FP_micro,FN_micro - """ - TP_micro,FP_micro,FN_micro=0.0,0.0,0.0 - for label,tuplee in label_dict.items(): - TP,FP,FN=tuplee - TP_micro=TP_micro+TP - FP_micro=FP_micro+FP - FN_micro=FN_micro+FN - return TP_micro,FP_micro,FN_micro -def compute_f1_micro_use_TFFPFN(label_dict): - """ - compute f1_micro - :param label_dict: {label:(TP,FP,FN)} - :return: f1_micro: a scalar - """ - TF_micro_accusation, FP_micro_accusation, FN_micro_accusation =compute_TF_FP_FN_micro(label_dict) - f1_micro_accusation = compute_f1(TF_micro_accusation, FP_micro_accusation, FN_micro_accusation,'micro') - return f1_micro_accusation - -def compute_f1_macro_use_TFFPFN(label_dict): - """ - compute f1_macro - :param label_dict: {label:(TP,FP,FN)} - :return: f1_macro - """ - f1_dict= {} - num_classes=len(label_dict) - for label, tuplee in label_dict.items(): - TP,FP,FN=tuplee - f1_score_onelabel=compute_f1(TP,FP,FN,'macro') - f1_dict[label]=f1_score_onelabel - f1_score_sum=0.0 - for label,f1_score in f1_dict.items(): - f1_score_sum=f1_score_sum+f1_score - f1_score=f1_score_sum/float(num_classes) - return f1_score - -small_value=0.00001 -def compute_f1(TP,FP,FN,compute_type): - """ - compute f1 - :param TP_micro: number.e.g. 200 - :param FP_micro: number.e.g. 200 - :param FN_micro: number.e.g. 200 - :return: f1_score: a scalar - """ - precison=TP/(TP+FP+small_value) - recall=TP/(TP+FN+small_value) - f1_score=(2*precison*recall)/(precison+recall+small_value) - - if random.choice([x for x in range(500)]) == 1:print(compute_type,"precison:",str(precison),";recall:",str(recall),";f1_score:",f1_score) - - return f1_score -def init_label_dict(num_classes): - """ - init label dict. this dict will be used to save TP,FP,FN - :param num_classes: - :return: label_dict: a dict. {label_index:(0,0,0)} - """ - label_dict={} - for i in range(num_classes): - label_dict[i]=(0,0,0) - return label_dict - -def get_target_label_short(eval_y): - eval_y_short=[] #will be like:[22,642,1391] - for index,label in enumerate(eval_y): - if label>0: - eval_y_short.append(index) - return eval_y_short - -#get top5 predicted labels -def get_label_using_logits(logits,top_number=5): - # index_list=np.argsort(logits)[-top_number:] - #vindex_list=index_list[::-1] - y_predict_labels = [i for i in range(len(logits)) if logits[i] >= 0.50] # TODO 0.5PW e.g.[2,12,13,10] - if len(y_predict_labels) < 1: y_predict_labels = [np.argmax(logits)] - - return y_predict_labels - -#统计预测的准确率 -def calculate_accuracy(labels_predicted, labels,eval_counter): - label_nozero=[] - #print("labels:",labels) - labels=list(labels) - for index,label in enumerate(labels): - if label>0: - label_nozero.append(index) - if eval_counter<2: - print("labels_predicted:",labels_predicted," ;labels_nozero:",label_nozero) - count = 0 - label_dict = {x: x for x in label_nozero} - for label_predict in labels_predicted: - flag = label_dict.get(label_predict, None) - if flag is not None: - count = count + 1 - return count / len(labels) - -################################################## + current_eval_loss, logits = sess.run( + [textCNN.loss_val, textCNN.logits], feed_dict) + predict += logits[0] + eval_loss += current_eval_loss + eval_counter += 1 + + if not FLAGS.multi_label_flag: + predict = [int(ii > 0.5) for ii in predict] + _, _, f1_macro, f1_micro, _ = fastF1(predict, evalY) + f1_score = (f1_micro+f1_macro)/2.0 + return eval_loss/float(eval_counter), f1_score, f1_micro, f1_macro + +@jit +def fastF1(result, predict): + ''' f1 score ''' + true_total, r_total, p_total, p, r = 0, 0, 0, 0, 0 + total_list = [] + for trueValue in range(6): + trueNum, recallNum, precisionNum = 0, 0, 0 + for index, values in enumerate(result): + if values == trueValue: + recallNum += 1 + if values == predict[index]: + trueNum += 1 + if predict[index] == trueValue: + precisionNum += 1 + R = trueNum / recallNum if recallNum else 0 + P = trueNum / precisionNum if precisionNum else 0 + true_total += trueNum + r_total += recallNum + p_total += precisionNum + p += P + r += R + f1 = (2 * P * R) / (P + R) if (P + R) else 0 + print(id2rela[trueValue], P, R, f1) + total_list.append([P, R, f1]) + p /= 6 + r /= 6 + micro_r = true_total / r_total + micro_p = true_total / p_total + macro_f1 = (2 * p * r) / (p + r) if (p + r) else 0 + micro_f1 = (2 * micro_p * micro_r) / (micro_p + + micro_r) if (micro_p + micro_r) else 0 + print('P: {:.2f}%, R: {:.2f}%, Micro_f1: {:.2f}%, Macro_f1: {:.2f}%'.format( + p*100, r*100, micro_f1 * 100, macro_f1*100)) + return p, r, macro_f1, micro_f1, total_lists def assign_pretrained_word_embedding(sess,vocabulary_index2word,vocab_size,textCNN,word2vec_model_path): import word2vec # we put import here so that many people who do not use word2vec do not need to install this package. you can move import to the beginning of this file. @@ -390,4 +251,4 @@ def load_data(cache_file_h5py,cache_file_pickle): print("INFO. cache file load successful...") return word2index, label2index,train_X,train_Y,vaild_X,valid_Y,test_X,test_Y if __name__ == "__main__": - tf.app.run() \ No newline at end of file + tf.app.run() From 03219171ef2085cefd7750eea34b0550d479f079 Mon Sep 17 00:00:00 2001 From: Jason Cooke Date: Fri, 9 Aug 2019 09:52:29 +1200 Subject: [PATCH 07/13] docs: fix typos --- README.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index c8f0a9a..c29bac9 100644 --- a/README.md +++ b/README.md @@ -537,7 +537,7 @@ III.Attention Mechanism: 1) transfer encoder input list and hidden state of decoder -2) calculate similiarity of hidden state with each encoder input, to get possibility distribution for each encoder input. +2) calculate similarity of hidden state with each encoder input, to get possibility distribution for each encoder input. 3) weighted sum of encoder input based on possibility distribution. @@ -617,7 +617,7 @@ Model Structure: 2) Dynamic memory: -a. compute gate by using 'similiarity' of keys,values with input of story. +a. compute gate by using 'similarity' of keys,values with input of story. b. get candidate hidden state by transform each key,value and input. @@ -709,7 +709,7 @@ During the process of doing large scale of multi-label classification, serveral It depend the task you are doing. From the task we conducted here, we believe that ensemble models based on models trained from multiple features including word, character for title and description can help to reach very high accuarcy; However, in some cases,as just alphaGo Zero demonstrated, algorithm is more important then data or computational power, in fact alphaGo Zero did not use any humam data. 2) Is there a ceiling for any specific model or algorithm? -The answer is yes. lots of different models were used here, we found many models have similiar performances, even though there are quite different in structure. In some extent, the difference of performance is not so big. +The answer is yes. lots of different models were used here, we found many models have similar performances, even though there are quite different in structure. In some extent, the difference of performance is not so big. 3) Is case study of error useful? I think it is quite useful especially when you have done many different things, but reached a limit. For example, by doing case study, you can find labels that models can make correct prediction, and where they make mistakes. And to imporove performance by increasing weights of these wrong predicted labels or finding potential errors from data. From a86149e3892b493ebe41820653ffb054cd26ee20 Mon Sep 17 00:00:00 2001 From: gunjianpan Date: Fri, 16 Aug 2019 00:47:15 +0800 Subject: [PATCH 08/13] =?UTF-8?q?=F0=9F=A4=90=20fix=20textCNN=20eval=20pro?= =?UTF-8?q?cess?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- a02_TextCNN/p7_TextCNN_train.py | 38 ++++++++++++++++----------------- 1 file changed, 19 insertions(+), 19 deletions(-) diff --git a/a02_TextCNN/p7_TextCNN_train.py b/a02_TextCNN/p7_TextCNN_train.py index ff901c4..bd2dc1c 100644 --- a/a02_TextCNN/p7_TextCNN_train.py +++ b/a02_TextCNN/p7_TextCNN_train.py @@ -133,31 +133,34 @@ def do_eval(sess, textCNN, evalX, evalY, num_classes): evalY = evalY[0:3000] number_examples = len(evalX) eval_loss, eval_counter, eval_f1_score, eval_p, eval_r = 0.0, 0, 0.0, 0.0, 0.0 - batch_size = FLAGS.batch_size + batch_size = 1 predict = [] - for start, end in zip(range(0, number_examples, batch_size), range(batch_size, number_examples, batch_size)): + for start, end in zip(range(0, number_examples, batch_size), range(batch_size, number_examples + batch_size, batch_size)): ''' evaluation in one batch ''' - feed_dict = {textCNN.input_x: evalX[start:end], textCNN.input_y_multilabel: evalY[start:end], textCNN.dropout_keep_prob: 1.0, + feed_dict = {textCNN.input_x: evalX[start:end], + textCNN.input_y_multilabel: evalY[start:end], + textCNN.dropout_keep_prob: 1.0, textCNN.is_training_flag: False} current_eval_loss, logits = sess.run( [textCNN.loss_val, textCNN.logits], feed_dict) - predict += logits[0] + predict = [*predict, np.argmax(np.array(logits[0]))] eval_loss += current_eval_loss eval_counter += 1 + evalY = [np.argmax(ii) for ii in evalY] if not FLAGS.multi_label_flag: predict = [int(ii > 0.5) for ii in predict] - _, _, f1_macro, f1_micro, _ = fastF1(predict, evalY) - f1_score = (f1_micro+f1_macro)/2.0 - return eval_loss/float(eval_counter), f1_score, f1_micro, f1_macro + _, _, f1_macro, f1_micro, _ = fastF1(predict, evalY, num_classes) + f1_score = (f1_micro + f1_macro) / 2.0 + return eval_loss / float(eval_counter), f1_score, f1_micro, f1_macro @jit -def fastF1(result, predict): +def fastF1(result: list, predict: list, num_classes: int): ''' f1 score ''' true_total, r_total, p_total, p, r = 0, 0, 0, 0, 0 total_list = [] - for trueValue in range(6): + for trueValue in range(num_classes): trueNum, recallNum, precisionNum = 0, 0, 0 for index, values in enumerate(result): if values == trueValue: @@ -174,18 +177,15 @@ def fastF1(result, predict): p += P r += R f1 = (2 * P * R) / (P + R) if (P + R) else 0 - print(id2rela[trueValue], P, R, f1) total_list.append([P, R, f1]) - p /= 6 - r /= 6 - micro_r = true_total / r_total - micro_p = true_total / p_total + p, r = np.array([p, r]) / num_classes + micro_r, micro_p = true_total / np.array([r_total, p_total]) macro_f1 = (2 * p * r) / (p + r) if (p + r) else 0 - micro_f1 = (2 * micro_p * micro_r) / (micro_p + - micro_r) if (micro_p + micro_r) else 0 - print('P: {:.2f}%, R: {:.2f}%, Micro_f1: {:.2f}%, Macro_f1: {:.2f}%'.format( - p*100, r*100, micro_f1 * 100, macro_f1*100)) - return p, r, macro_f1, micro_f1, total_lists + micro_f1 = (2 * micro_p * micro_r) / (micro_p + micro_r) if (micro_p + micro_r) else 0 + accuracy = true_total / len(result) + print('P: {:.2f}%, R: {:.2f}%, Micro_f1: {:.2f}%, Macro_f1: {:.2f}%, Accuracy: {:.2f}'.format( + p * 100, r * 100, micro_f1 * 100, macro_f1 * 100, accuracy * 100)) + return p, r, macro_f1, micro_f1, total_list def assign_pretrained_word_embedding(sess,vocabulary_index2word,vocab_size,textCNN,word2vec_model_path): import word2vec # we put import here so that many people who do not use word2vec do not need to install this package. you can move import to the beginning of this file. From 8b17e3c4b6d689e48660bc999e2e910166ec4714 Mon Sep 17 00:00:00 2001 From: brightmart Date: Tue, 3 Sep 2019 23:46:36 +0800 Subject: [PATCH 09/13] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index c8f0a9a..d3d8cde 100644 --- a/README.md +++ b/README.md @@ -4,7 +4,7 @@ The purpose of this repository is to explore text classification methods in NLP #### Update: -Try bert model for multi-label classification, please check session #Models Detail, 3) Bert. +Releasing Pre-trained Model of RoBERTa_Chinese Trained with 30G Raw Chinese Corpus, 24 layers and 12 layers, 2019-Sep-8, This Weekend! Large Amount of Chinese Corpus for NLP Available! From cac14a0a522dff2ce95eef9818aa990ef53df571 Mon Sep 17 00:00:00 2001 From: brightmart Date: Sun, 8 Sep 2019 00:57:46 +0800 Subject: [PATCH 10/13] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index d3d8cde..4e767b9 100644 --- a/README.md +++ b/README.md @@ -4,7 +4,7 @@ The purpose of this repository is to explore text classification methods in NLP #### Update: -Releasing Pre-trained Model of RoBERTa_Chinese Trained with 30G Raw Chinese Corpus, 24 layers and 12 layers, 2019-Sep-8, This Weekend! +Releasing Pre-trained Model of RoBERTa_Chinese Trained with 30G Raw Chinese Corpus, 24 layers and 12 layers, Target to match State of the Art result in Chinese, 2019-Sep-8, This Weekend! Large Amount of Chinese Corpus for NLP Available! From 8bfdbab24a165a9c968b1c838ebdb322fadac319 Mon Sep 17 00:00:00 2001 From: b Date: Wed, 11 Sep 2019 23:17:48 +0800 Subject: [PATCH 11/13] update --- README.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/README.md b/README.md index ea07aeb..9fcacfe 100644 --- a/README.md +++ b/README.md @@ -54,6 +54,8 @@ question(as query), then ask the model to predict an answer; if you feed story s classification task. +To discuss ML/DL/NLP problems and get tech support from each other, you can join QQ group: 836811304 + Models: ------------------------------------------------------------------------- From b4101fe8d1f64f8675e610d5237180f7940cd61e Mon Sep 17 00:00:00 2001 From: b Date: Wed, 18 Sep 2019 23:59:41 +0800 Subject: [PATCH 12/13] update --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 9fcacfe..3f4bcff 100644 --- a/README.md +++ b/README.md @@ -139,7 +139,7 @@ if you use python3, it will be fine as long as you change print/try catch functi TextCNN model is already transfomed to python 3.6 -Sample data: cached file +Sample data: cached file of baidu or Google Drive:send me an email ------------------------------------------------------------------------------------------------------- to help you run this repository, currently we re-generate training/validation/test data and vocabulary/labels, and saved From e09b47e21de7cac85a01904926bf6af73782f9c1 Mon Sep 17 00:00:00 2001 From: brightmart Date: Sat, 28 Sep 2019 15:48:42 +0800 Subject: [PATCH 13/13] Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 3f4bcff..e438c99 100644 --- a/README.md +++ b/README.md @@ -4,7 +4,7 @@ The purpose of this repository is to explore text classification methods in NLP #### Update: -Releasing Pre-trained Model of RoBERTa_Chinese Trained with 30G Raw Chinese Corpus, 24 layers and 12 layers, Target to match State of the Art result in Chinese, 2019-Sep-8, This Weekend! +Releasing Pre-trained Model of ALBERT_Chinese Training with 30G+ Raw Chinese Corpus, xxlarge, xlarge and more, Target to match State of the Art performance in Chinese, 2019-Oct-7, During the National Day of China! Large Amount of Chinese Corpus for NLP Available!