def get_golden_ranking(question):
ideal_answer = unicode(question.ideal_answer[0]).encode("ascii", 'ignore')
sentences = question.sentences
ranked_sentences = RM.get_ranked_sentences(question_text=ideal_answer,
sentences=sentences,
retrieval_algo='BM25')
return ranked_sentences
def get_golden_ranking(question):
ideal_answer = question['ideal_answer']
sentences = RM.get_sentences(question['snippets'])
sentences = RM.preprocess_sentences(sentences)
ranked_sentences = RM.get_ranked_sentences(question_text=ideal_answer,
sentences=sentences,
retrieval_algo='BM25')
return ranked_sentences
def ranked_sentences(self):
scores = {}
for sentence, score in retrieval_model.get_ranked_sentences(
self.question, self.snippet_sentences, BM25):
scores[sentence] = [score]
for sentence, score in retrieval_model.get_ranked_sentences(
self.question, self.snippet_sentences, INDRI):
scores[sentence].append(score)
all_sentences = [(bm25_score, indri_score, s)
for s, [bm25_score, indri_score] in scores.items()]
all_sentences = sorted(all_sentences, reverse=True)
sentences = [{
TEXT: s,
BM25: bm25_score,
INDRI: indri_score
} for (bm25_score, indri_score, s) in all_sentences]
return sentences
def create_feature_vectors(question):
sentences = set(question.sentences)
feature_vectors = []
# ranked sentences also gives a score which can be used as feature
ranked_sentences_bm25 = RM.get_ranked_sentences(question_text=question.question_text, sentences=sentences,
retrieval_algo='BM25')
# ranked sentences also gives a score which can be used as feature
ranked_sentences_Indri = RM.get_ranked_sentences(question_text=question.question_text, sentences=sentences,
retrieval_algo='Indri')
# TO DO @Gabe, above results can be used for LeToR
return feature_vectors