Chatty Goose

Multi-stage Conversational Passage Retrieval: An Approach to Fusing Term Importance Estimation and Neural Query Rewriting

Installation

1. Make sure Java 11+ and Python 3.7+ are installed

2. Install the chatty-goose PyPI module

pip install chatty-goose

3. If you are using T5 or BERT, make sure to install PyTorch 1.4.0 - 1.7.1 using your specific platform instructions. Note that PyTorch 1.8 is currently incompatible due to the transformers version we currently use. Also make sure to install the corresponding torchtext version.

4. Download the English model for spaCy

python -m spacy download en_core_web_sm

Quickstart Guide

The following example shows how to initialize a searcher and build a ConversationalQueryRewriter agent from scratch using HQE and T5 as first-stage retrievers, and a BERT reranker. To see a working example agent, see chatty_goose/agents/chat.py.

First, load a searcher

from pyserini.search import SimpleSearcher

# Option 1: load a prebuilt index
searcher = SimpleSearcher.from_prebuilt_index("INDEX_NAME_HERE")
# Option 2: load a local Lucene index
searcher = SimpleSearcher("PATH_TO_INDEX")

searcher.set_bm25(0.82, 0.68)

Next, initialize one or more first-stage CQR retrievers

from chatty_goose.cqr import Hqe, Ntr
from chatty_goose.settings import HqeSettings, NtrSettings

hqe = Hqe(searcher, HqeSettings())
ntr = Ntr(NtrSettings())

Load a reranker

from chatty_goose.util import build_bert_reranker

reranker = build_bert_reranker()

Create a new RetrievalPipeline

from chatty_goose.pipeline import RetrievalPipeline

rp = RetrievalPipeline(searcher, [hqe, ntr], searcher_num_hits=50, reranker=reranker)

And we're done! Simply call rp.retrieve(query) to retrieve passages, or call rp.reset_history() to reset the conversational history of the retrievers.

Running Experiments

1. Clone the repo and all submodules (git submodule update --init --recursive)

2. Clone and build Anserini for evaluation tools

3. Install dependencies

pip install -r requirements.txt

4. Follow the instructions under docs/cqr_experiments.md to run experiments using HQE, T5, or fusion.

Example Agent

To run an interactive conversational search agent with ParlAI, simply run chat.py. By default, we use the CAsT 2019 pre-built Pyserini index, but it is possible to specify other indexes using the --from_prebuilt flag. See the file for other possible arguments:

python -m chatty_goose.agents.chat

Alternatively, run the agent using ParlAI's command line interface:

python -m parlai interactive --model chatty_goose.agents.chat:ChattyGooseAgent

We also provide instructions to deploy the agent to Facebook Messenger using ParlAI under examples/messenger.