Upload 6 files

plotcom/README.md

@@ -0,0 +1,36 @@
+# Plot Completion Dataset
+
+### Data Example
+
+```
+{
+	"story": 
+		"在一个金碧辉煌的天国王宫里住着两位公主，她们都非常的美丽善良。<MASK>大王就跟国王说，你要是不把你的女儿交出来我就杀了你，正当国王在想办法的时候，一个白马王子出现了，他直接把山寨的大王杀了，跟国王说，我要娶两位公主，国王也同意让女儿嫁了白马王子。从此以后，两位天国的公主与白马王子过上了幸福的日子。", 
+	"plot": 
+		"有一次，山寨的大王要来天国王宫里提亲，可是国王并不同意把自己的女儿嫁给山寨的大王做山寨夫人。"
+}
+```
+
+- "story" (`str`)：input story，`<MASK>` means the position of the removed sentence.
+- "plot" (`str`)：the removed sentence.
+
+  
+
+### Evaluation
+
+The prediction result should have the same format with `test.jsonl`
+
+```shell
+python eval.py prediction_file test.jsonl
+```
+
+
+
+We use bleu and distinct as the evaluation metrics. The output of the script `eval.py` is a dictionary as follows:
+
+```python
+{'bleu-1': '_', 'bleu-2': '_', 'bleu-3': '_', 'bleu-4': '_', 'distinct-1': '_', 'distinct-2': '_',  'distinct-3': '_', 'distinct-4': '_'}
+```
+
+- Dependencies: jieba=0.42.1, nltk=3.6.2, numpy=1.20.3
+

plotcom/eval.py

@@ -0,0 +1,91 @@
+import json
+import argparse
+import sys
+import numpy as np
+import jieba
+import nltk
+from nltk.translate.bleu_score import sentence_bleu, SmoothingFunction
+from nltk import ngrams
+
+def bleu(data):
+    """
+    compute rouge score
+    Args:
+        data (list of dict including reference and candidate):
+    Returns:
+            res (dict of list of scores): rouge score
+    """
+
+    res = {}
+    for i in range(1, 5):
+        res["sentence-bleu-%d"%i] = []
+        res["corpus-bleu-%d"%i] = nltk.translate.bleu_score.corpus_bleu([[d["reference"].strip().split()] for d in data], [d["candidate"].strip().split() for d in data], weights=tuple([1./i for j in range(i)]))
+    for tmp_data in data:
+        origin_candidate = tmp_data['candidate']
+        origin_reference = tmp_data['reference']
+        assert isinstance(origin_candidate, str)
+        if not isinstance(origin_reference, list):
+            origin_reference = [origin_reference]
+
+        for i in range(1, 5):
+            res["sentence-bleu-%d"%i].append(sentence_bleu(references=[r.strip().split() for r in origin_reference], hypothesis=origin_candidate.strip().split(), weights=tuple([1./i for j in range(i)]))) 
+
+    for key in res:
+        if "sentence" in key:
+            res[key] = np.mean(res[key])
+        
+    return res
+
+
+
+def distinct(eval_data):
+    result = {}
+    for i in range(1, 5):
+        all_ngram, all_ngram_num = {}, 0.
+        for k, tmp_data in enumerate(eval_data):
+            ngs = ["_".join(c) for c in ngrams(tmp_data["candidate"].strip().split(), i)]
+            all_ngram_num += len(ngs)
+            for s in ngs:
+                if s in all_ngram:
+                    all_ngram[s] += 1
+                else:
+                    all_ngram[s] = 1
+        result["distinct-%d"%i] = len(all_ngram) / float(all_ngram_num)
+    return result
+
+
+
+def load_file(filename):
+    data = []
+    with open(filename, "r") as f:
+        for line in f.readlines():
+            data.append(json.loads(line))
+        f.close()
+    return data
+
+def proline(line):
+    return " ".join([w for w in jieba.cut("".join(line.strip().split()))])
+
+
+def compute(golden_file, pred_file, return_dict=True):
+    golden_data = load_file(golden_file)
+    pred_data = load_file(pred_file)
+
+    if len(golden_data) != len(pred_data):
+        raise RuntimeError("Wrong Predictions")
+
+    eval_data = [{"reference": proline(g["plot"]), "candidate": proline(p["plot"])} for g, p in zip(golden_data, pred_data)]
+    res = bleu(eval_data)
+    res.update(distinct(eval_data))
+    for key in res:
+        res[key] = "_"
+    return res
+
+def main():
+    argv = sys.argv
+    print("预测结果：{}, 测试集: {}".format(argv[1], argv[2]))
+    print(compute(argv[2], argv[1]))
+
+
+if __name__ == '__main__':
+    main()