Utilize existing FAQs for QA about haystack

[tholor]

Merged with #44

Basic example:

document_store = ElasticsearchDocumentStore(host="localhost", username="", password="",
                                            index="document", text_field="answer",
                                            embedding_field="question_emb",
                                            embedding_dim=768,
                                            excluded_meta_data=["question_emb"])

retriever = ElasticsearchRetriever(document_store=document_store, embedding_model="bert-base-cased", gpu=False)

# Get dataframe with columns "question", "answer" and some custom metadata
df = pd.read_csv("data/covid/faq_covidbert.csv")
df.fillna(value="", inplace=True)


# prepare docs for indexing (list of dicts)
docs_to_index = []
doc_id = 1
for idx, row in df.iterrows():
    d = row.to_dict()
    d = {k:v.strip() for k, v in d.items()}
    d["document_id"] = idx
    # add embedding
    question_embedding = retriever.create_embedding(row["question"])
    d["question_emb"] = question_embedding
    docs_to_index.append(d)
 
# Index to ES
document_store.write_documents(docs_to_index)


# Use Finder to get answer
finder = Finder(reader=None, retriever=retriever)
prediction = finder.get_answers_via_similar_questions(question="How is the virus spreading?", top_k_retriever=10)
print_answers(prediction, details="all")

Will add a tutorial soon ....