fix text

app.py

@@ -150,9 +150,10 @@ def predict_zh(text: str) -> List:
 with gr.Blocks() as demo:
     gr.Markdown(
         """
-        ## ChatGPT Detector 🔬 (Linguistic version)
+        ## ChatGPT Detector 🔬 (Linguistic version / 语言学版)
+
         Visit our project on Github: [chatgpt-comparison-detection project](https://github.com/Hello-SimpleAI/chatgpt-comparison-detection)<br>
-        欢迎在 Github 上关注我们的 [ChatGPT 对比与检测项目](https://github.com/Hello-SimpleAI/chatgpt-comparison-detection)
+        欢迎在 Github 上关注我们的 [ChatGPT 对比与检测项目](https://github.com/Hello-SimpleAI/chatgpt-comparison-detection)<br>
         We provide three kinds of detectors, all in Bilingual / 我们提供了三个版本的检测器，且都支持中英文:
         - [QA version / 问答版](https://huggingface.co/spaces/Hello-SimpleAI/chatgpt-detector-qa)<br>
             detect whether an **answer** is generated by ChatGPT for certain **question**, using PLM-based classifiers / 判断某个**问题的回答**是否由ChatGPT生成，使用基于PTM的分类器来开发;
@@ -160,20 +161,6 @@ with gr.Blocks() as demo:
             detect whether a piece of text is ChatGPT generated, using PLM-based classifiers / 判断**单条文本**是否由ChatGPT生成，使用基于PTM的分类器来开发;
         - [**Linguistic version / 语言学版** (👈 Current / 当前使用)](https://huggingface.co/spaces/Hello-SimpleAI/chatgpt-detector-ling)<br>
             detect whether a piece of text is ChatGPT generated, using linguistic features / 判断**单条文本**是否由ChatGPT生成，使用基于语言学特征的模型来开发;
-        
-        ## Introduction:
-        Two Logistic regression models trained with two kinds of features:
-        1. [GLTR](https://aclanthology.org/P19-3019) Test-2, Language model predict token rank top-k buckets, top 10, 10-100, 100-1000, 1000+.
-        2. PPL-based, text ppl, `avg` & `max` & `std` of sentence ppls, `avg` & `max` &`std` of timestep ppls.
-
-        English LM is [GPT2-small](https://huggingface.co/gpt2).
-
-        ## 介绍:
-        两个逻辑回归模型, 分别使用以下两种特征:
-        1. [GLTR](https://aclanthology.org/P19-3019) Test-2, 每个词的语言模型预测排名分桶, top 10, 10-100, 100-1000, 1000+.
-        2. 基于语言模型困惑度 (PPL), text ppl, `avg` & `max` & `std` of sentence ppls, `avg` & `max` &`std` of timestep ppls.
-
-        中文语言模型使用 闻仲 [Wenzhong-GPT2-110M](https://huggingface.co/IDEA-CCNL/Wenzhong-GPT2-110M).
 
         """
     )
@@ -181,6 +168,13 @@ with gr.Blocks() as demo:
     with gr.Tab("English"):
         gr.Markdown(
             """
+            ## Introduction:
+            Two Logistic regression models trained with two kinds of features:
+            1. [GLTR](https://aclanthology.org/P19-3019) Test-2, Language model predict token rank top-k buckets, top 10, 10-100, 100-1000, 1000+.
+            2. PPL-based, text ppl, sentence ppl, etc.
+
+            English LM is [GPT2-small](https://huggingface.co/gpt2).
+
             Note: Providing more text to the `Text` box can make the prediction more accurate!
             """
         )
@@ -199,6 +193,13 @@ with gr.Blocks() as demo:
     with gr.Tab("中文版"):
         gr.Markdown(
             """
+            ## 介绍:
+            两个逻辑回归模型, 分别使用以下两种特征:
+            1. [GLTR](https://aclanthology.org/P19-3019) Test-2, 每个词的语言模型预测排名分桶, top 10, 10-100, 100-1000, 1000+.
+            2. 基于语言模型困惑度 (PPL), 整个文本的PPL、单个句子的PPL等特征.
+
+            中文语言模型使用 闻仲 [Wenzhong-GPT2-110M](https://huggingface.co/IDEA-CCNL/Wenzhong-GPT2-110M).
+
             注意: 在`文本`栏中输入更多的文本，可以让预测更准确哦！
             """
         )