1from tensorflow.keras.models import load_model
2from tensorflow.keras.preprocessing.sequence import pad_sequences
3
4model = load_model("best_model.h5")
5
6# You'll need the word list and ID mappings.
7# See the GitHub repo for full chat script.Performance & Quirks
8The model generates grammatically correct English sentences but has no semantic understanding. It tends to:
9
10Apologize excessively ("I am sorry.")
11
12Obsess over the word "seat"
13
14Insert "?" randomly
15
16Get stuck in repetitive loops
17
18Example Conversation
19text
20User: How are you?
21Bot: I am a student in the mood for a while .
22User: What are you studying?
23Bot: I am applying for a vacation .
24User: Did you get the vacation?
25Bot: Yes , I am going to get a lot of the ? .
26User: Where will you go?
27Bot: I don't know what's a ? .
28User: Seat?
29Bot: Yes , I am going to get a lot of the ? .
30User: ?
31Bot: I don't know what's a seat .
32Why Does It Behave Like This?
33Small training: Only 10 epochs on CPU.
34
35Limited context: 50-token window.
36
37No pretrained embeddings: Learned everything from scratch.
38
39Data bias: The word "seat" appears frequently in DailyDialog.
40
41Full Source Code
42All preprocessing, training, and chat scripts are available on GitHub:
43
44👉 zyxdtt/Fun-Project/tree/main/RNN_AI-bot
45
46There you'll find:
47
48preprocessing.cpp – C++ data pipeline
49
50RNN_train.py – Model training code
51
52chat_test.py – Interactive chat script
53
54Chat screenshots showing its hilarious responses
55
56Fun Fact
57This is the first AI project completed by a university freshman. It may be dumb, but it's 100% my own work — from data processing to deployment.
58
59This bot proves that perfect grammar doesn't equal intelligence. It's a lesson in overfitting, data bias, and the importance of semantics. 😅
