QRec-NLI: Interactive Data Analysis with Next-step Natural Language Query Recommendation

Video demo | Code | Authors: Xingbo Wang, Furui Cheng, Yong Wang, Ke Xu, Jiang Long, Hong Lu, Huamin Qu

QRec-NLI is a natural language interface (NLI) with a step-wise query recommendation module to guide visual data analysis, where data is stored across multiple tables in SQL databases. Please checkout our paper for more details.

User workflow: (1) A user first makes a series of natural language data queries by selecting from system recommendations step by step. The results of data queries are presented in visualizations. (2) After several steps of data exploration, the user chooses some desired insights (shown in visualizations) from previous queries, and organizes them into a dashboard by direct manipulation.

The system adopts a log-based approach to generate semantically-relevant and context-aware query suggestions that adapt to a user's interested domains based on all queries made by current user and reference queries in query databases.

How to run the system

System interface of QRec-NLI.

The environment setups include frontend (vue@2.6.11, d3v5), and backend (python 3.7 or above).

1. Install python packages (suggest using conda for package management):

   pip install -r requirements.txt
   

2. Download data & model (we use SmBop as text2sql model and Spider as reference query databases)

   python download_model_data.py
   

3. set up frontend

   cd frontend
   npm install
   npm run serve
   

4. set up backend

   cd backend
   python run-data-backend.py
   

How to cite

If this paper and tool helps your research projects, please considering citing our paper:

@article{qrecNLI2022,
  title={Interactive Data Analysis with Next-step Natural Language Query Recommendation},
  author={Wang, Xingbo and Cheng, Furui and Wang, Yong and Xu, Ke and Long, Jiang and Lu, Hong and Qu, Huamin},
  journal={arXiv preprint arXiv:2201.04868},
  year={2022}
}