RelationalIntentionGraph/answer_process.py at main · HKUST-KnowComp/RelationalIntentionGraph · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
import pandas as pd
import os
import random
import time
import json
import pickle
import numpy as np
# from sentence_transformers import SentenceTransformer, util

##### for generate triple from answer######
def get_triple(model_name):


    triple_output_file = model_name + "_triple.txt"
    input_file = model_name + "_answer.txt"


    current_i = 0
    intentions = []
    needs = []
    final_dict = {}
    flag = False
    flag_not_error = True


    with open(input_file,'r') as f:
        for lines in f.readlines():
            if lines == '\n':continue
            elif lines[:7].lower() == 'session':
                current_session = lines.strip().split(':')[1].strip()
            elif lines[:5].lower() == 'error': flag_not_error=False
            elif lines[:5].lower()=='needs':
                flag = True
            elif lines[:9].lower()=='intention':
                print("intention: ", lines)
                try:
                    answer = lines.strip().split(':')[1].strip()
                except:continue
                if flag:
                    needs.append(answer)
                else:
                    intentions.append(answer)
            elif lines[:5]=='-----':
                if flag_not_error:
                    final_dict[current_i]=[current_session,intentions,needs]
                current_i+=1
                intentions = []
                needs = []
                flag_not_error = True
                flag = False

    with open(triple_output_file,'w') as f:
        for i in final_dict:
            sessions,intentions,needs = final_dict[i]
            print(sessions,intentions,needs)
            if len(needs)<len(intentions): needs.extend("null" for _ in range(len(intentions)-len(needs)))
            for idx,intention in enumerate(intentions):
                f.write(str(sessions)+'\t'+str(intention)+'\t'+str(needs[idx])+'\n')


##### for session idx #######
def get_session_idx_for_transformer():
    result_file = 'session.pkl'
    session_all =[]
    s = []
    last_session = 'start'
    curr_id = 0
    with open('result_triple.txt','r') as f:
        for i,lines in enumerate(f.readlines()):
            lists = lines.strip().split('\t')
            session = lists[0]
            if last_session == 'start':
                last_session = session
                s.append(i)
            elif last_session ==session:
                s.append(i)
            else:
                session_all.append(s)
                s = [i]
                last_session = session
        session_all.append(s)

    with open(result_file,'wb') as f:
        pickle.dump(session_all,f)

#### for all intentions ####
def get_intention_sentence_for_transformer():
    result_file = 'sentences.pkl'
    session_all =[]

    with open('result_triple.txt','r') as f:
        for lines in f.readlines():
            lists = lines.strip().split('\t')
            intentions = lists[1]
            session_all.append(intentions)

    with open(result_file,'wb') as f:
        pickle.dump(session_all,f)

### get graph ####
def cal_intention_sim():
    with open('sentences.pkl','rb') as f:
        sentences = pickle.load(f)

    model = SentenceTransformer('all-mpnet-base-v2')
    ## try other training dataset

    #Encode all sentences
    embeddings = model.encode(sentences)

    #Compute cosine similarity between all pairs
    cos_sim = util.cos_sim(embeddings, embeddings)
    mat = cos_sim.numpy()

    with open('session.pkl','rb') as f:
        sen_idx = pickle.load(f)

    all_row = []
    all_col = []
    for idxs in sen_idx:
        for i in range(len(idxs)-1):
            for j in range(i+1,len(idxs)):
                all_row.append(idxs[i])
                all_col.append(idxs[j])

    mat[all_row,all_col] = 0
    mat[all_col,all_row]=0
    final_mat = mat - np.eye(len(mat))
    with open('intention_score_mpnet.pkl','wb') as f:
        pickle.dump(final_mat,f)


if __name__ == '__main__':
    for model_name in ['gpt3.5', 'gpt4']:
        get_triple(model_name)

        print('triple done' + model_name)