Revert "Adds doc 2 speech node to keyword search and plays audio in UI"

This reverts commit 1d466d71500aa4ad0916150db61be4b47f859cc9.

.gitignore

@@ -128,6 +128,4 @@ dmypy.json
 # Pyre type checker
 .pyre/
 
-.vscode/
-
-data/audio/
+.vscode/

core/pipelines.py

@@ -2,19 +2,14 @@
 Haystack Pipelines
 """
 
-from pathlib import Path
 from haystack import Pipeline
 from haystack.document_stores import InMemoryDocumentStore
 from haystack.nodes.retriever import DensePassageRetriever, TfidfRetriever
 from haystack.nodes.preprocessor import PreProcessor
 from haystack.nodes.ranker import SentenceTransformersRanker
-from haystack.nodes.audio.document_to_speech import DocumentToSpeech
-import os
 
-data_path = 'data/'
-os.makedirs(data_path, exist_ok=True)
 
-def keyword_search(index="documents", split_word_length=100, audio_output=False):
+def keyword_search(index="documents", split_word_length=100):
     """
     **Keyword Search Pipeline**
 
@@ -49,13 +44,6 @@ def keyword_search(index="documents", split_word_length=100, audio_output=False)
     index_pipeline.add_node(
         document_store, name="DocumentStore", inputs=["TfidfRetriever"]
     )
-    
-    if audio_output:
-        doc2speech = DocumentToSpeech(
-            model_name_or_path="espnet/kan-bayashi_ljspeech_vits",
-            generated_audio_dir=Path(data_path + 'audio'),
-        )
-        search_pipeline.add_node(doc2speech, name='DocumentToSpeech', inputs=['TfidfRetriever'])
 
     return search_pipeline, index_pipeline
 

core/search_index.py

@@ -32,17 +32,13 @@ def search(queries, pipeline):
         for res in matches:
             if not score_is_empty:
                 score_is_empty = True if res.score is None else False
-            match = {
-                "text": res.content,
-                "id": res.meta["id"],
-                "fragment_id": res.id,
-            }
-            if not score_is_empty:
-                match.update({'score': res.score})
-            if hasattr(res, 'content_audio'):
-                match.update({'content_audio': res.content_audio})
             query_results.append(
-                match
+                {
+                    "text": res.content,
+                    "score": res.score,
+                    "id": res.meta["id"],
+                    "fragment_id": res.id,
+                }
             )
         if not score_is_empty:
             query_results = sorted(

interface/components.py

@@ -59,10 +59,8 @@ def component_show_search_result(container, results):
             st.markdown(f"### Match {idx+1}")
             st.markdown(f"**Text**: {document['text']}")
             st.markdown(f"**Document**: {document['id']}")
-            if 'score' in document:
+            if document["score"] is not None:
                 st.markdown(f"**Score**: {document['score']:.3f}")
-            if 'content_audio' in document:
-                st.audio(str(document['content_audio']))
             st.markdown("---")
 
 

requirements.txt

@@ -5,8 +5,4 @@ black==22.8.0
 plotly==5.10.0
 newspaper3k==0.2.8
 PyPDF2==2.10.7
-pytesseract==0.3.10
-soundfile==0.10.3.post1
-espnet
-pydub==0.25.1
-espnet_model_zoo==0.1.7
+pytesseract==0.3.10