Add a document_name field in answers

haystack/database/elasticsearch.py

@@ -120,5 +120,11 @@ def query(self, query, top_k=10, candidate_doc_ids=None):
         meta_data = []
         for hit in result:
             paragraphs.append(hit["_source"][self.text_field])
-            meta_data.append({"paragraph_id": hit["_id"], "document_id": hit["_source"][self.doc_id_field]})
+            meta_data.append(
+                {
+                    "paragraph_id": hit["_id"],
+                    "document_id": hit["_source"][self.doc_id_field],
+                    "document_name": hit["_source"][self.name_field],
+                }
+            )
         return paragraphs, meta_data

haystack/finder.py

@@ -36,8 +36,15 @@ def get_answers(self, question, top_k_reader=1, top_k_retriever=10, filters=None
         # 3) Apply reader to get granular answer(s)
         logger.info(f"Applying the reader now to look for the answer in detail ...")
         results = self.reader.predict(question=question,
-                                      paragrahps=paragraphs,
+                                      paragraphs=paragraphs,
                                       meta_data_paragraphs=meta_data,
                                       top_k=top_k_reader)
 
+        # Add corresponding document_name if an answer contains the document_id (only supported in FARMReader)
+        for ans in results["answers"]:
+            document_name = next(
+                (meta["document_name"] for meta in meta_data if meta["document_id"] == ans["document_id"]), None
+            )
+            ans["document_name"] = document_name
+
         return results

haystack/reader/farm.py

@@ -190,7 +190,8 @@ def predict(self, question, paragraphs, meta_data_paragraphs=None, top_k=None, m
         for paragraph, meta_data in zip(paragraphs, meta_data_paragraphs):
             cur = {"text": paragraph,
                    "questions": [question],
-                   "document_id": meta_data["document_id"]
+                   "document_id": meta_data["document_id"],
+                   "document_name": meta_data["document_name"],
             }
             input_dicts.append(cur)
 

haystack/reader/transformers.py

@@ -43,7 +43,7 @@ def __init__(
         #TODO param to modify bias for no_answer
 
 
-    def predict(self, question, paragrahps, meta_data_paragraphs=None, top_k=None):
+    def predict(self, question, paragraphs, meta_data_paragraphs=None, top_k=None):
         """
         Use loaded QA model to find answers for a question in the supplied paragraphs.
 
@@ -76,7 +76,7 @@ def predict(self, question, paragrahps, meta_data_paragraphs=None, top_k=None):
 
         # get top-answers for each candidate passage
         answers = []
-        for p in paragrahps:
+        for p in paragraphs:
             query = {"context": p, "question": question}
             predictions = self.model(query, topk=self.n_best_per_passage)
             # assemble and format all answers