Bug fixes

retrieval/models/QueryLikelihood.py

@@ -19,7 +19,10 @@ def _score_passage(self, pid: int, query_words: list[str]) -> float:
         probabilities = []
         for word in query_words:
             probabilities.append(self._probability(pid, word))
-        return math.log(np.prod(probabilities))
+        try:
+            return math.log(np.prod(probabilities))
+        except ValueError:
+            return 0.0
 
     def _probability(self, pid: int, word: str) -> float:
         tf = self._index[word].get_posting(pid).freq if word in self._collection[pid] else 0

retrieval/models/VectorSpace.py

@@ -29,6 +29,8 @@ def _similarity(self, pid: int, query_words: list[str]) -> float:
         counter = Counter(query_words)
         query_length = sum(counter.values())
         for word in query_words:
+            if word not in self._index:
+                continue
             tfidf = tf_idf(counter[word], query_length, self._index[word].doc_freq,
                            self._collection_length)
             if word in vocab:

start.py

@@ -30,8 +30,10 @@ def main():
     data = ''
     for qid, passages in results.items():
         for rank, (pid, score) in enumerate(passages.items()):
-            data += f"{qid}\t{'A1'}\t{pid}\t{rank}\t{format(score, '.2f')}\t{model}{smoothing}\n"
-    write_txt(f'results/{model}.txt', data)
+            if rank >= 100:
+                break
+            data += f"{qid}\t{'A1'}\t{pid}\t{rank+1}\t{format(score, '.2f')}\t{model}{smoothing}\n"
+    write_txt(f'results/{model}{smoothing}.txt', data)
 
 
 if __name__ == '__main__':