随着游戏市场的崛起，游戏相关从业人员急需了解玩家对游戏的实际体验，以便有针对性地指导游戏运营和开发。因此可基于深度学习模型对玩家评论进行情感分析，做网络游戏舆情态势分析项目。

主流互联网游戏评论情感态势分析项目

应用场景：随着游戏市场的崛起，游戏相关从业人员急需了解玩家对游戏的实际体验，以便有针对性地指导游戏运营和开发。与此同时，在舆情信息监测的实际业务中，也存在着信息处理效率过低以及分析结果过于主观等问题。因此，为了解决游戏评论体量大、更新快、含义不清的问题，可基于深度学习模型对玩家评论进行情感分析，做网络游戏舆情态势分析项目。

项目简介：本次项目利用情感分析预训练模型SKEP完成模型训练与预测，利用爬虫程序爬取3个主流平台（TapTap、bilibili、豆瓣）的三个主流互联网游戏（《原神》、《王者荣耀》、《和平精英》）的用户评论进行数据分析，最后利用pyqt5的qtdeisgner进行ui设计，完成数据可视化。

#引入需要的库import paddlenlp as ppnlpimport paddlefrom paddlenlp.datasets import load_datasetfrom paddlenlp.datasets import MapDatasetfrom paddlenlp.data import Stack, Pad, Tupleimport paddle.nn.functional as Fimport numpy as npimport pandas as pdfrom functools import partialimport os登录后复制 ? ?

采用数据集为aistdio上“TapTap游戏评论”数据集，链接：https://aistudio.baidu.com/aistudio/datasetdetail/101119

包含TapTap上约300款游戏的标签评论，可用于情感分析的应用，共4888个数据示例

采用该数据集优点：涉及游戏种类较多，训练数据样本较多

缺点：缺乏目前主流互联网游戏相关评论，网络热词较少覆盖，无阴阳怪气评论

#读取数据集并划分训练集与验证集df=pd.read_csv('data/data109273/taptap_review_ready.csv')#读取数据集的评论内容review = df.review.tolist()#读取数据集的情感指数sentiment = df.sentiment.tolist()full_data=[]for i in range(len(review)): #将每条评论内容与情感指数形成字典并添加到full_data列表 dic = {'text':review[i],'label':sentiment[i]} full_data.append(dic)print(len(full_data))#数据集切分train_ds = MapDataset(full_data[:4400])dev_ds = MapDataset(full_data[4400:4880])label_list = ['0', '1']print(len(train_ds))print(train_ds[3])登录后复制 ? ? ? ?

48884400{'text': '对我来说，这种游戏让我感觉很好。刚开始是因为同学推荐我玩，我玩了一下，哇，真的很好，然后，最近没怎么登录游戏，怎么关服了？这就很难受。希望反斗联盟快速上架，我想玩。等待这种事。对我来说度日如年。我就是觉得那个空投好像有点问题。每次那个空投差不多都在对面基地附近。对面英雄又有点难对付。抢又抢不着。这就特别难受了。', 'label': 1}登录后复制 ? ? ? ?In [5]

from paddlenlp.transformers import SkepForSequenceClassification, SkepTokenizer# 加载skep模型model = SkepForSequenceClassification.from_pretrained(pretrained_model_name_or_path=”skep_ernie_1.0_large_ch“, num_classes=len(label_list))# 加载模型对应的Tokenizer，用于数据预处理tokenizer = SkepTokenizer.from_pretrained(pretrained_model_name_or_path=”skep_ernie_1.0_large_ch“)登录后复制 ? ? ? ?

[2022-07-16 15:24:57,173] [ INFO] - Downloading https://paddlenlp.bj.bcebos.com/models/transformers/skep/skep_ernie_1.0_large_ch.pdparams and saved to /home/aistudio/.paddlenlp/models/skep_ernie_1.0_large_ch[2022-07-16 15:24:57,176] [ INFO] - Downloading skep_ernie_1.0_large_ch.pdparams from https://paddlenlp.bj.bcebos.com/models/transformers/skep/skep_ernie_1.0_large_ch.pdparams100%|██████████| 1238309/1238309 [00:17<00:00, 71093.85it/s]W0716 15:25:14.762745 151 device_context.cc:404] Please NOTE: device: 0, GPU Compute Capability: 7.0, Driver API Version: 11.2, Runtime API Version: 10.1W0716 15:25:14.766620 151 device_context.cc:422] device: 0, cuDNN Version: 7.6.[2022-07-16 15:25:22,647] [ INFO] - Downloading skep_ernie_1.0_large_ch.vocab.txt from https://paddlenlp.bj.bcebos.com/models/transformers/skep/skep_ernie_1.0_large_ch.vocab.txt100%|██████████| 55/55 [00:00<00:00, 10636.12it/s]登录后复制 ? ? ? ?In [6]

#数据预处理def convert_example(example,tokenizer,label_list,max_seq_length=128,is_test=False): if is_test: text = example['text'] else: text = example['text'] label = example['label'] #tokenizer.encode方法实现切分token，映射token ID以及拼接特殊token encoded_inputs = tokenizer.encode(text=text, max_seq_len=max_seq_length) input_ids = encoded_inputs[”input_ids“] token_type_ids = encoded_inputs[”token_type_ids“] if not is_test: label_map = {} for (i, l) in enumerate(label_list): label_map[l] = i # label：情感极性类别 label = np.array([label], dtype=”int64“) return input_ids, token_type_ids, label else: return input_ids, token_type_ids#数据迭代器def create_dataloader(dataset, trans_fn=None, mode='train', batch_size=1, use_gpu=False, pad_token_id=0, batchify_fn=None): if trans_fn: dataset = dataset.map(trans_fn, lazy=True) if mode == 'train' and use_gpu: sampler = paddle.io.DistributedBatchSampler(dataset=dataset, batch_size=batch_size, shuffle=True) else: shuffle = True if mode == 'train' else False sampler = paddle.io.BatchSampler(dataset=dataset, batch_size=batch_size, shuffle=shuffle) #生成一个取样器 dataloader = paddle.io.DataLoader(dataset, batch_sampler=sampler, return_list=True, collate_fn=batchify_fn) return dataloader#将数据处理成模型可读入的数据格式trans_fn = partial(convert_example, tokenizer=tokenizer, label_list=label_list, max_seq_length=128, is_test=False)# 将数据组成批量式数据batchify_fn = lambda samples, fn=Tuple(Pad(axis=0,pad_val=tokenizer.pad_token_id), Pad(axis=0, pad_val=tokenizer.pad_token_id), Stack(dtype=”int64“)):[data for data in fn(samples)]#batch_size批量数据大小batch_size=64#训练集迭代器train_loader = create_dataloader(train_ds, mode='train', batch_size=batch_size, batchify_fn=batchify_fn, trans_fn=trans_fn)#验证集迭代器dev_loader = create_dataloader(dev_ds, mode='dev', batch_size=batch_size, batchify_fn=batchify_fn, trans_fn=trans_fn)登录后复制 ? ?In [7]

#设置训练参数import timefrom utils import evaluate# 训练轮次epochs = 3# 训练过程中保存模型参数的文件夹ckpt_dir = ”skep_ckpt“# len(train_loader)一轮训练所需要的step数num_training_steps = len(train_loader) * epochs# Adam优化器optimizer = paddle.optimizer.AdamW( learning_rate=2e-5, parameters=model.parameters())# 交叉熵损失函数criterion = paddle.nn.loss.CrossEntropyLoss()# accuracy评价指标metric = paddle.metric.Accuracy()登录后复制 ? ?In [12]

#模型训练global_step = 0tic_train = time.time()for epoch in range(1, epochs + 1): for step, batch in enumerate(train_loader, start=1): input_ids, token_type_ids, labels = batch # 喂数据给model logits = model(input_ids, token_type_ids) # 计算损失函数值 loss = criterion(logits, labels) # 预测分类概率值 probs = F.softmax(logits, axis=1) # 计算acc correct = metric.compute(probs, labels) metric.update(correct) acc = metric.accumulate() global_step += 1 if global_step % 10 == 0: print( ”global step %d, epoch: %d, batch: %d, loss: %.5f, accu: %.5f, speed: %.2f step/s“ % (global_step, epoch, step, loss, acc, 10 / (time.time() - tic_train))) tic_train = time.time() # 反向梯度回传，更新参数 loss.backward() optimizer.step() optimizer.clear_grad() if global_step % 100 == 0: save_dir = os.path.join(ckpt_dir, ”model_%d“ % global_step) if not os.path.exists(save_dir): os.makedirs(save_dir) # 评估当前训练的模型 evaluate(model, criterion, metric, dev_loader) # 保存当前模型参数等 model.save_pretrained(save_dir) # 保存tokenizer的词表等 tokenizer.save_pretrained(save_dir)登录后复制 ? ? ? ?

WARNING:root:DataLoader reader thread raised an exception.Exception in thread Thread-5:Traceback (most recent call last): File ”/opt/conda/envs/python35-paddle120-env/lib/python3.7/threading.py“, line 926, in _bootstrap_inner self.run() File ”/opt/conda/envs/python35-paddle120-env/lib/python3.7/threading.py“, line 870, in run self._target(*self._args, **self._kwargs) File ”/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dataloader/dataloader_iter.py“, line 192, in _thread_loop six.reraise(*sys.exc_info()) File ”/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/six.py“, line 719, in reraise raise value File ”/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dataloader/dataloader_iter.py“, line 160, in _thread_loop batch = self._dataset_fetcher.fetch(indices) File ”/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dataloader/fetcher.py“, line 106, in fetch data = [self.dataset[idx] for idx in batch_indices] File ”/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dataloader/fetcher.py“, line 106, in <listcomp> data = [self.dataset[idx] for idx in batch_indices] File ”/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddlenlp/datasets/dataset.py“, line 181, in __getitem__ idx]) if self._transform_pipline else self.new_data[idx] File ”/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddlenlp/datasets/dataset.py“, line 172, in _transform data = fn(data) File ”/tmp/ipykernel_3624/1567443302.py“, line 6, in convert_example text = example['text']TypeError: tuple indices must be integers or slices, not str登录后复制 ? ? ? ?---------------------------------------------------------------------------SystemError?Traceback (most recent call last)/tmp/ipykernel_3624/864604143.py?in??2?tic_train?=?time.time()?3?for?epoch?in?range(1,?epochs?+?1):?----> 4for?step,?batch?in?enumerate(train_loader,?start=1):?5?input_ids,?token_type_ids,?labels?=?batch?6??# 喂数据给model?/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dataloader/dataloader_iter.py?in?__next__(self)?195??try:?196??if?in_dygraph_mode():?--> 197?data?=?self._reader.read_next_var_list()?198?data?=?_restore_batch(data,?self._structure_infos.pop(0))?199?else:?SystemError: (Fatal) Blocking queue is killed because the data reader raises an exception. [Hint: Expected killed_ != true, but received killed_:1 == true:1.] (at /paddle/paddle/fluid/operators/reader/blocking_queue.h:166)In [8]

# 加载训练好的模型参数params_path = save_dir+'/model_state.pdparams'if params_path and os.path.isfile(params_path): # 加载模型参数 state_dict = paddle.load(params_path) model.set_dict(state_dict) print(”Loaded parameters from %s“ % params_path)登录后复制 ? ? ? ?

Loaded parameters from skep_ckpt/model_200/model_state.pdparams登录后复制 ? ? ? ?In [9]

#定义测试集数据的处理函数def convert_example(example,tokenizer,label_list,max_seq_length=512,is_test=False): encoded_inputs = tokenizer(text=example, max_seq_len=max_seq_length) input_ids = np.array(encoded_inputs[”input_ids“], dtype=”int64“) token_type_ids = np.array(encoded_inputs[”token_type_ids“], dtype=”int64“) return input_ids, token_type_ids登录后复制 ? ?In [10]

#定义预测函数def predict(model, data, tokenizer, label_map, batch_size=1): examples = [] for text in data: input_ids, token_type_ids = convert_example(text,tokenizer,label_list=label_map.values(),max_seq_length=512,is_test=True) examples.append((input_ids, token_type_ids)) #划分数据 batches = [ examples[idx:idx + batch_size] for idx in range(0, len(examples), batch_size) ] batchify_fn = lambda samples, fn=Tuple( Pad(axis=0, pad_val=tokenizer.pad_token_id), # input ids Pad(axis=0, pad_val=tokenizer.pad_token_type_id), # token type ids ): [data for data in fn(samples)] #返回结果集 results = [] model.eval() for batch in batches: input_ids, token_type_ids = batchify_fn(batch) input_ids = paddle.to_tensor(input_ids) token_type_ids = paddle.to_tensor(token_type_ids) logits = model(input_ids, token_type_ids) probs = F.softmax(logits, axis=1) idx = paddle.argmax(probs, axis=1).numpy() idx = idx.tolist() labels = [label_map[i] for i in idx] results.extend(labels) return results登录后复制 ? ?

测试集数据：

爬虫程序获取游戏评论，也是本次项目分析数据的来源

爬虫程序放在spider文件夹下：py文件为爬虫源代码，csv文件为平台内游戏评论数据

爬取数据

#加载测试集import csvwith open('spider/data/bilibili_yuanshen.csv','r',encoding='gbk',errors='ignore')as f: cs =list(csv.reader(f))#定义存储评论列表comments=[]#csv文件含有表头，故i初始化为1#评论文本在csv文件第三列，故选择cs[i][2]for i in range(1,len(cs)): comments.append(cs[i][2])print(len(comments))print(comments[1])登录后复制 ? ? ? ?

997还一起冒险呢？你有本事放??进去啊？不公测又不给内测资格，??梦中去冒险啊？登录后复制 ? ? ? ?In [12]

#对测试集数据情感分析label_map = {0: '0', 1: '1'}results = predict(model,comments,tokenizer,label_map,batch_size=1)#统计积极评论数量count = 0for result in results: if result == '1': count =count + 1print(count)#计算积极评论比例positive_ratio = count / len(results)print(positive_ratio)登录后复制 ? ? ? ?

5020.5035105315947843登录后复制 ? ? ? ?

情感分析结果：

封装为exe文件：

使用pytrhon的pyinstaller直接生成exe文件，需要的相关文件

效果展示：