INSPIRED: A Transparent Dialogue Dataset for Interactive Semantic Parsing

Existing studies on semantic parsing focus primarily on mapping a natural-language utterance to a corresponding logical form in one turn. However, because natural language can contain a great deal of ambiguity and variability, this is a difficult challenge. In this work, we investigate an interactive semantic parsing framework that explains the predicted logical form step by step in natural language and enables the user to make corrections through natural-language feedback for individual steps. We focus on question answering over knowledge bases (KBQA) as an instantiation of our framework, aiming to increase the transparency of the parsing process and help the user appropriately trust the final answer. To do so, we construct INSPIRED, a crowdsourced dialogue dataset derived from the ComplexWebQuestions dataset.

This repository will contain the dataset and code for our paper Towards Transparent Interactive Semantic Parsing via Step-by-Step Correction.

Data

Dataset Download

The dataset can be downloaded under this path: ./data/dataset.jsonl

Data Structure

In the dataset file, each line is a dictionary with several keys:

{
    "id": "ID number",
    "cwq_question": "Original complex question in CWQ dataset",
    "rephrased_question": "Rephrased complex question by workers",
    "rephrased_question_label": " 'Replacement' or 'Alternative' ",
    "question": "If rephrased_question_label is marked as 'Replacement', set the value the same as rephrased_question; Otherwise, set it the same as cwq_question",
    "final_answer": "Final answer for the complex question",
    "gold_parse": "Gold sparql query for complex question",
    "preprocessed_gold_parse": "Preprocessed gold parse with entities and prefix replaced",
    "predicted_parse": "Predicted sparql query by initial semantic parser",
    "gold_sub_lfs": "A list of gold sub-logical forms after decomposition",
    "pred_sub_lfs": "A list of predicted sub-logical forms after decomposition",
    "gold_sub_qs": [
        {
          "sub_id": "ID of sub questions",
          "sub_question": "Rephrased sub question",
          "temp_sub_question": "Templated sub question for gold sub-logical form",
          "answer": "Answer for each sub question",
        }, "..."], 
    "pred_sub_qs": [
        {
          "sub_id": "ID of sub questions",
          "sub_question": "Rephrased sub question",
          "temp_sub_question": "Templated sub question for predicted sub-logical form",
          "answer": "Answer for each sub question",
        }, "..."], 
    "feedback": "A list of human feedback"
    
}

INSPIRED: A Transparent Dialogue Dataset for Interactive Semantic Parsing

Related tags

Overview

INSPIRED: A Transparent Dialogue Dataset for Interactive Semantic Parsing

Data

Dataset Download

Data Structure

Owner

Exploring Image Deblurring via Blur Kernel Space (CVPR'21)

This repository accompanies our paper “Do Prompt-Based Models Really Understand the Meaning of Their Prompts?”

Code for the ICML 2021 paper: "ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision"

Fast Scattering Transform with CuPy/PyTorch

Source code of "Hold me tight! Influence of discriminative features on deep network boundaries"

Open-AI's DALL-E for large scale training in mesh-tensorflow.

Demo code for ICCV 2021 paper "Sensor-Guided Optical Flow"

Extreme Lightwegith Portrait Segmentation

Reinforcement Learning Theory Book (rus)

Code for the paper "Implicit Representations of Meaning in Neural Language Models"

A unet implementation for Image semantic segmentation

Evaluating different engineering tricks that make RL work

Python periodic table module

"MST++: Multi-stage Spectral-wise Transformer for Efficient Spectral Reconstruction" (CVPRW 2022) & (Winner of NTIRE 2022 Challenge on Spectral Reconstruction from RGB)

Direct Multi-view Multi-person 3D Human Pose Estimation

The official implementation for "FQ-ViT: Fully Quantized Vision Transformer without Retraining".

Official Implementation (PyTorch) of "Point Cloud Augmentation with Weighted Local Transformations", ICCV 2021

Federated Learning - Including common test models for federated learning, like CNN, Resnet18 and lstm, controlled by different parser

This tool uses Deep Learning to help you draw and write with your hand and webcam.

Heat transfer problemas solved using python