我正在建立一个项目，该项目使用LSTM从输入的医学报告中检测表型文本（甚至RNN也可以使用）。我通常使用一些标记为数据集的文本。只是对于如何使用HPO数据集来训练我的模型没有正确的方向。 HPO数据集链接（https://raw.githubusercontent.com/obophenotype/human-phenotype-ontology/master/hp.obo）。 数据样本：
HP:0000002 abnormality of body height
HP:0000003 multicystic renal dysplasia
HP:0000003 multicystic kidney dysplasia
HP:0000003 multicystic dysplastic kidney
HP:0000003 multicystic kidneys
HP:0000008 abnormality of female internal genitalia
HP:0000009 functional abnormality of the bladder
HP:0000009 poor bladder function
HP:0000010 urinary tract infections recurrent
HP:0000010 recurrent utis
HP:0000010 recurrent urinary tract infections
HP:0000010 frequent urinary tract infections
我已经对推文进行了情感分析，并且已经有了开源的标签数据集。下面的代码是LPO在HPO上的非常糟糕的实现。
text_words = df1['Text']
max_words = 20000
tokenizer = Tokenizer(num_words=max_words, oov_token='UNK')
tokenizer.fit_on_texts(list(text_words))
tokenizer.word_index = {e:i for e,i in tokenizer.word_index.items() if i <= max_words} # <= because tokenizer is 1 indexed
tokenizer.word_index[tokenizer.oov_token] = max_words + 1
maxlen = 100
X = pad_sequences(tokenized_train, maxlen = maxlen)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2)
model = Sequential()
model.add(Embedding(max_words, output_dim=128, mask_zero=True))
model.add(LSTM(60))
model.add(Dropout(0.1))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
output = model.fit(X_train, y_train, batch_size=128, epochs=5, validation_split=0.2)
我得到的准确性很低。 期望的解决方案是知道如何使用LSTM或RNN处理此类数据集。