Home page

Question_answer_pair (QAP) LF

Question_answer_pair (QAP) is a relation that is used to link an answer to the question it is an answer to. This comes up a lot in the STAC corpus, as many offers are in the form of questions.

LL (linguistic source -> linguistic target)

Rule(s):

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
def LF_QAP_LL_case1(row):
    l=0
    if row.source_dialogue_act == 'Offer' \
    and row.target_dialogue_act in ['Refusal', 'Other', 'Counteroffer', 'Accept']:
        l=1
    return l

def LF_QAP_LL_case2(row):
    l=0
    if row.source_dialogue_act == 'Counteroffer' \
    and row.target_dialogue_act in ['Offer', 'Counteroffer', 'Accept', 'Refusal']:
        l=1
    return l

def LF_QAP_LL_case3(row):
    l=0  
    if row.source_surface_act == 'Question' \
    and row.target_surface_act in ['Assertion', 'Accept', 'Request']:
        l=1
    return l

Application of rules to candidates:

 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
#1 -- get dialogues
finals = []
dialogues = cands.dialogue_num.drop_duplicates()

for d in tqdm(dialogues):
    
    #keep track of dus involved in a seq as source or target -- on du cannot be source or target for > 1 rel
    memo_target = [] 
    open_questions = defaultdict(list)
 
    # 2 -- get segment list
    seg_list = get_seg_list(cands[cands.dialogue_num == d])

    # 3 -- create seg pairs list
    seg_pairs = []
    for i, s in enumerate(seg_list):
        for n in [j for j in reversed(range((i+1)-20, i+1)) if j>=0]:
            try:
                seg_pairs.append((seg_list[n], seg_list[i+1]))   
            except IndexError:
                pass
            
    # 4 -- for each pair, pull row and append rules

    for s in seg_pairs:
        row = cands[(cands.source_id == s[0]) & (cands.target_id == s[1])]
        if row.empty:
            continue
        else:
            r_index = row.index[0]
            row = row.iloc[0]
            
        #since we know no Result relations are backwards, eliminate immediately   
        backwards = cands[(cands.source_id == s[1]) & (cands.target_id == s[0])]
        if backwards.shape[0] > 0:
            b_index = backwards.index[0]
            finals.append((b_index, -1))
            
        
        #we only look at ling --> ling cases
        
        if row.source_type != 'Segment' or row.target_type != 'Segment':
            
            link = 0  
            
        else: 
        
            if LF_QAP_LL_case1(row) or LF_QAP_LL_case2(row) or LF_QAP_LL_case3(row):
                if row.source_emitter == row.target_emitter \
                or row.source_emitter != row.target_addressee \
                or row.target_emitter in open_questions[row.source_id] \
                or row.target_id in memo_target:
                    link = 0
                else:
                    link = 1
                    open_questions[row.source_id].append(row.target_emitter)
                    memo_target.append(row.target_id)
                    
            else:
                link = 0
    
    
        finals.append((r_index, link))

NLNL (non-linguistic source -> non-linguistic target)

Rule(s):

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
def LF_QAP_NL_NL_case1(row):
    #target = X ended their turn to most recent 'X' eeu
    l=0
    if madeanoffertotrade(row.source_text) and traded(row.target_text):
        if row.source_text.split(' ')[0] == row.target_text.split(' ')[0]:
                l=1
    else:
            l=0
    return l

def LF_QAP_NL_NL_case2(row):
    #target = X ended their turn to most recent 'X' eeu
    l=0
    if madeanoffertotrade(row.source_text) and XrejectedTradeOffer(row.target_text):
        if row.source_text.split(' ')[0] == row.target_addressee:
                l=1
    else:
            l=0
    return l

Application of rules to candidates:

 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
#1 -- get dialogues
finals = []
dialogues = cands.dialogue_num.drop_duplicates()
#dialogues = [98]

for d in tqdm(dialogues):
    
    #keep track of dus involved in a seq as source or target -- on du cannot be source or target for > 1 seq rel
    memo_source = []
    memo_target = []
 
    # 2 -- get segment list
    seg_list = get_seg_list(cands[cands.dialogue_num == d])

    # 3 -- create seg pairs list
    seg_pairs = []
    for i, s in enumerate(seg_list):
        for n in [j for j in reversed(range((i+1)-20, i+1)) if j>=0]:
            try:
                seg_pairs.append((seg_list[n], seg_list[i+1]))   
            except IndexError:
                pass
            
    # 4 -- for each pair, pull row and append rules

    for s in seg_pairs:
        row = cands[(cands.source_id == s[0]) & (cands.target_id == s[1])]
        if row.empty:
            continue
        else:
            r_index = row.index[0]
            row = row.iloc[0]
        
        # make sure that if any backwards links exist they are also 0
            
        backwards = cands[(cands.source_id == s[1]) & (cands.target_id == s[0])]
        if backwards.shape[0] > 0:
            b_index = backwards.index[0]
            finals.append((b_index, -1))
            
        
        #we only look at ling --> ling cases
        
        if row.source_type != 'NonplayerSegment' or row.target_type != 'NonplayerSegment':
            
            link = 0  
            
        else: 
            
            if LF_QAP_NL_NL_case1(row) or LF_QAP_NL_NL_case2(row):
                if row.target_id in memo_target or row.source_id in memo_source:
                    link = 0
                else:
                    link = 1
                    memo_target.append(row.target_id)
                    memo_source.append(row.source_id) 
            else:
                link = 0
               
        finals.append((r_index, link))