Spaces:
Runtime error
Runtime error
mertbozkurt
commited on
Commit
•
2e7d21f
1
Parent(s):
3858ca8
add
Browse files- chatbot.py +169 -0
- chatbot_streamlit.py +28 -0
- intents.json +43 -0
- model/checkpoint +2 -0
- model/model.tflearn.data-00000-of-00001 +0 -0
- model/model.tflearn.index +0 -0
- model/model.tflearn.meta +0 -0
- model/training_data +0 -0
- nltk.txt +1 -0
- requirements.txt +6 -0
chatbot.py
ADDED
@@ -0,0 +1,169 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import nltk
|
2 |
+
import numpy as np
|
3 |
+
import tflearn
|
4 |
+
import tensorflow as tf
|
5 |
+
import random
|
6 |
+
import json
|
7 |
+
import nltk
|
8 |
+
from nltk.stem.lancaster import LancasterStemmer
|
9 |
+
nltk.download('punkt')
|
10 |
+
stemmer = LancasterStemmer()
|
11 |
+
|
12 |
+
# import our chat-bot intents file
|
13 |
+
with open('intents.json') as json_data:
|
14 |
+
intents = json.load(json_data)
|
15 |
+
|
16 |
+
bot_name = 'Kevin'
|
17 |
+
words = []
|
18 |
+
classes = []
|
19 |
+
documents = []
|
20 |
+
ignore_words = ['?']
|
21 |
+
|
22 |
+
# loop through each sentence in our intents patterns
|
23 |
+
for intent in intents['intents']:
|
24 |
+
for pattern in intent['patterns']:
|
25 |
+
# tokenize each word in the sentence
|
26 |
+
w = nltk.word_tokenize(pattern)
|
27 |
+
# add to our words list
|
28 |
+
words.extend(w)
|
29 |
+
# add to documents in our corpus
|
30 |
+
documents.append((w, intent['tag']))
|
31 |
+
# add to our classes list
|
32 |
+
if intent['tag'] not in classes:
|
33 |
+
classes.append(intent['tag'])
|
34 |
+
|
35 |
+
# stem and lower each word and remove duplicates
|
36 |
+
words = [stemmer.stem(w.lower()) for w in words if w not in ignore_words]
|
37 |
+
words = sorted(list(set(words)))
|
38 |
+
|
39 |
+
# remove duplicates
|
40 |
+
classes = sorted(list(set(classes)))
|
41 |
+
|
42 |
+
print (len(documents), "documents")
|
43 |
+
print (len(classes), "classes", classes)
|
44 |
+
print (len(words), "unique stemmed words", words)
|
45 |
+
# create our training data
|
46 |
+
training = []
|
47 |
+
output = []
|
48 |
+
# create an empty array for our output
|
49 |
+
output_empty = [0] * len(classes)
|
50 |
+
|
51 |
+
# training set, bag of words for each sentence
|
52 |
+
for doc in documents:
|
53 |
+
# initialize our bag of words
|
54 |
+
bag = []
|
55 |
+
# list of tokenized words for the pattern
|
56 |
+
pattern_words = doc[0]
|
57 |
+
# stem each word
|
58 |
+
pattern_words = [stemmer.stem(word.lower()) for word in pattern_words]
|
59 |
+
# create our bag of words array
|
60 |
+
for w in words:
|
61 |
+
bag.append(1) if w in pattern_words else bag.append(0)
|
62 |
+
|
63 |
+
# output is a '0' for each tag and '1' for current tag
|
64 |
+
output_row = list(output_empty)
|
65 |
+
output_row[classes.index(doc[1])] = 1
|
66 |
+
|
67 |
+
training.append([bag, output_row])
|
68 |
+
|
69 |
+
# shuffle our features and turn into np.array
|
70 |
+
random.shuffle(training)
|
71 |
+
training = np.array(training)
|
72 |
+
|
73 |
+
# create train and test lists
|
74 |
+
train_x = list(training[:,0])
|
75 |
+
train_y = list(training[:,1])
|
76 |
+
|
77 |
+
# Build neural network
|
78 |
+
net = tflearn.input_data(shape=[None, len(train_x[0])])
|
79 |
+
net = tflearn.fully_connected(net, 8)
|
80 |
+
net = tflearn.fully_connected(net, 8)
|
81 |
+
net = tflearn.fully_connected(net, len(train_y[0]), activation='softmax')
|
82 |
+
net = tflearn.regression(net)
|
83 |
+
|
84 |
+
# Define model and setup tensorboard
|
85 |
+
model = tflearn.DNN(net, tensorboard_dir='tflearn_logs')
|
86 |
+
|
87 |
+
#if u need u can fit the model
|
88 |
+
# model.fit(train_x, train_y, n_epoch=1000, batch_size=8, show_metric=True)
|
89 |
+
|
90 |
+
|
91 |
+
# restore all of our data structures
|
92 |
+
import pickle
|
93 |
+
data = pickle.load( open( "model/training_data", "rb" ) )
|
94 |
+
words = data['words']
|
95 |
+
classes = data['classes']
|
96 |
+
train_x = data['train_x']
|
97 |
+
train_y = data['train_y']
|
98 |
+
|
99 |
+
|
100 |
+
|
101 |
+
#we have saved model on local
|
102 |
+
# load our saved model
|
103 |
+
model.load('model/model.tflearn')
|
104 |
+
def clean_up_sentence(sentence):
|
105 |
+
# tokenize the pattern
|
106 |
+
sentence_words = nltk.word_tokenize(sentence)
|
107 |
+
# stem each word
|
108 |
+
sentence_words = [stemmer.stem(word.lower()) for word in sentence_words]
|
109 |
+
return sentence_words
|
110 |
+
|
111 |
+
# return bag of words array: 0 or 1 for each word in the bag that exists in the sentence
|
112 |
+
def bow(sentence, words, show_details=False):
|
113 |
+
# tokenize the pattern
|
114 |
+
sentence_words = clean_up_sentence(sentence)
|
115 |
+
# bag of words
|
116 |
+
bag = [0]*len(words)
|
117 |
+
for s in sentence_words:
|
118 |
+
for i,w in enumerate(words):
|
119 |
+
if w == s:
|
120 |
+
bag[i] = 1
|
121 |
+
if show_details:
|
122 |
+
print ("found in bag: %s" % w)
|
123 |
+
|
124 |
+
return(np.array(bag))
|
125 |
+
|
126 |
+
# create a data structure to hold user context
|
127 |
+
context = {}
|
128 |
+
|
129 |
+
ERROR_THRESHOLD = 0.25
|
130 |
+
def classify(sentence):
|
131 |
+
# generate probabilities from the model
|
132 |
+
results = model.predict([bow(sentence, words)])[0]
|
133 |
+
# filter out predictions below a threshold
|
134 |
+
results = [[i,r] for i,r in enumerate(results) if r>ERROR_THRESHOLD]
|
135 |
+
# sort by strength of probability
|
136 |
+
results.sort(key=lambda x: x[1], reverse=True)
|
137 |
+
return_list = []
|
138 |
+
for r in results:
|
139 |
+
return_list.append((classes[r[0]], r[1]))
|
140 |
+
# return tuple of intent and probability
|
141 |
+
return return_list
|
142 |
+
|
143 |
+
|
144 |
+
|
145 |
+
def response(sentence, userID='123', show_details=False):
|
146 |
+
results = classify(sentence)
|
147 |
+
# if we have a classification then find the matching intent tag
|
148 |
+
if results:
|
149 |
+
# loop as long as there are matches to process
|
150 |
+
while results:
|
151 |
+
for i in intents['intents']:
|
152 |
+
# find a tag matching the first result
|
153 |
+
if i['tag'] == results[0][0]:
|
154 |
+
# set context for this intent if necessary
|
155 |
+
if 'context_set' in i:
|
156 |
+
if show_details: print ('context:', i['context_set'])
|
157 |
+
context[userID] = i['context_set']
|
158 |
+
|
159 |
+
# check if this intent is contextual and applies to this user's conversation
|
160 |
+
if not 'context_filter' in i or \
|
161 |
+
(userID in context and 'context_filter' in i and i['context_filter'] == context[userID]):
|
162 |
+
if show_details: print ('tag:', i['tag'])
|
163 |
+
# a random response from the intent
|
164 |
+
return random.choice(i['responses'])
|
165 |
+
#print(random.choice(i['responses']))
|
166 |
+
|
167 |
+
results.pop(0)
|
168 |
+
|
169 |
+
|
chatbot_streamlit.py
ADDED
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
from streamlit_chat import message as st_message
|
3 |
+
from chatbot import response, bot_name
|
4 |
+
|
5 |
+
|
6 |
+
st.set_page_config(
|
7 |
+
page_title="AI- Kevin",
|
8 |
+
page_icon=":robot:"
|
9 |
+
)
|
10 |
+
|
11 |
+
if "history" not in st.session_state:
|
12 |
+
st.session_state.history = []
|
13 |
+
|
14 |
+
st.title(bot_name)
|
15 |
+
|
16 |
+
def ol():
|
17 |
+
user_message = st.session_state.input_text
|
18 |
+
res= response(user_message)
|
19 |
+
st.session_state.history.append({"message": user_message, "is_user": True})
|
20 |
+
st.session_state.history.append({"message": res, "is_user": False})
|
21 |
+
|
22 |
+
#user_message = st.session_state.input_text
|
23 |
+
#result = model.generate(**inputs)
|
24 |
+
|
25 |
+
st.text_input("Ask me about AI", key="input_text", on_change=ol)
|
26 |
+
|
27 |
+
for chat1 in st.session_state.history:
|
28 |
+
st_message(**chat1) # unpacking
|
intents.json
ADDED
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{"intents": [
|
2 |
+
{"tag": "greeting",
|
3 |
+
"patterns": ["Hi", "How are you", "Is anyone there?", "Hello", "Good day"],
|
4 |
+
"responses": ["Hello, thanks for visiting", "Good to see you again", "Hi there, how can I help?"],
|
5 |
+
"context_set": ""
|
6 |
+
},
|
7 |
+
{"tag": "goodbye",
|
8 |
+
"patterns": ["Bye", "See you later", "Goodbye"],
|
9 |
+
"responses": ["See you later, thanks for visiting", "Have a nice day", "Bye! Come back again soon."]
|
10 |
+
},
|
11 |
+
{"tag": "thanks",
|
12 |
+
"patterns": ["Thanks", "Thank you", "That's helpful"],
|
13 |
+
"responses": ["Happy to help!", "Any time!", "My pleasure"]
|
14 |
+
},
|
15 |
+
{"tag": "ai",
|
16 |
+
"patterns": ["What is ai?", "Do you know ai?", "Can you explain ai?" ],
|
17 |
+
"responses": ["Artificial intelligence (AI) refers to the simulation of human intelligence", " AI is machines that are programmed to think like humans and mimic their actions."]
|
18 |
+
},
|
19 |
+
{"tag": "overfitting",
|
20 |
+
"patterns": ["What is the overfitting?", "Do you know overfitting?" ],
|
21 |
+
"responses": ["Overfitting is occurs when a statistical model fits exactly against its training data.", " Overfitting occurs when the model has a high variance"]
|
22 |
+
},
|
23 |
+
{"tag": "machine",
|
24 |
+
"patterns": ["Can machines think?" ],
|
25 |
+
"responses": ["Yes i think so " ]
|
26 |
+
},
|
27 |
+
{"tag": "underfitting",
|
28 |
+
"patterns": ["What is the underfitting?", "Do you know underfitting?"],
|
29 |
+
"responses": ["Underfitting is data model is unable to capture the relationship between the input and output variables accurately", "When the model performs poorly on the training data you have underfitting "]
|
30 |
+
},
|
31 |
+
{"tag": "nlp",
|
32 |
+
"patterns": ["What is nlp ?", "Can you explain nlp ?", "Do you know nlp?" ],
|
33 |
+
"responses": ["Do you mean Natural language processing "],
|
34 |
+
"context_set": "nlp"
|
35 |
+
},
|
36 |
+
{"tag": "yes",
|
37 |
+
"patterns": ["yes", "sure", "absolutely"],
|
38 |
+
"responses": ["NLP is the ability of a computer program to understand human language as it is spoken and written", "NLP is a collective term referring to automatic computational processing of human languages"],
|
39 |
+
"context_filter": "nlp"
|
40 |
+
}
|
41 |
+
]
|
42 |
+
}
|
43 |
+
|
model/checkpoint
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
model_checkpoint_path: "/content/model.tflearn"
|
2 |
+
all_model_checkpoint_paths: "/content/model.tflearn"
|
model/model.tflearn.data-00000-of-00001
ADDED
Binary file (5.13 kB). View file
|
|
model/model.tflearn.index
ADDED
Binary file (887 Bytes). View file
|
|
model/model.tflearn.meta
ADDED
Binary file (102 kB). View file
|
|
model/training_data
ADDED
Binary file (2.91 kB). View file
|
|
nltk.txt
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
punkt
|
requirements.txt
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
streamlit == 1.11.0
|
2 |
+
streamlit_chat == 0.0.2.1
|
3 |
+
nltk == 3.5
|
4 |
+
numpy == 1.23.1
|
5 |
+
tflearn == 0.5.0
|
6 |
+
tensorflow-cpu==2.9.1
|