Subjective Induction QA

Citation Author(s):
Yufeng
Zhang
Submitted by:
Yufeng Zhang
Last updated:
Sun, 12/08/2024 - 09:03
DOI:
10.21227/ht8z-n304
Data Format:
.pkl
Research Article Link:
Answering Subjective Induction Questions on Products by Summarizing Multi-sources Multi-viewpoints Knowledge
License:
Creative Commons Attribution
7 Views
Categories:
Artificial Intelligence
Keywords:
question and answer; subjective qa
0
0 ratings - Please login to submit your rating.
ACCESS DATASET CITE SHARE/EMBED

Abstract 

We proposes a new task in the field of Answering Subjective Induction Question on Products (SUBJPQA). The answer to this kind of question is non-unique, but can be interpreted from many perspectives. For example, the answer to ‘whether the phone is heavy’ has a variety of different viewpoints. A satisfied answer should be able to summarize these subjective opinions from multiple sources and provide objective knowledge, such as the weight of a phone. That is quite different from the traditional QA task, in which the answer to a factoid question is unique and can be found from a single data source. To address this new task, we propose a three-steps method. We first retrieve all answer-related clues from multiple knowledge sources on facts and opinions. The implicit commonsense facts are also collected to supplement the necessary but missing contexts. We then capture their relevance with the questions by interactive attention. Next, we design a reinforcement-based summarizer to aggregate all these knowledgeable clues. Based on a templatecontrolled decoder, we can output a comprehensive and multiperspective answer. Due to the lack of a relevant evaluated benchmark set for the new task, we construct a large-scale dataset, named SupQA, consisting of 48,352 samples across 15 product domains. Evaluation results show the effectiveness of our approach.

Instructions: 

We utilized a typical SupQA dataset in the field of S{\scriptsize{UBJ}}PQA which contains 48,352 samples across 15 product domains. Each sample includes a subjective question, product descriptions, attributes, multiple user reviews with diversified sentiments, and a multi-perspective answer summarizing objective facts and subjective reviews. Each sample in the dataset contains the following fields:

  • category: The general product category.

  • ASIN: The unique Amazon product identifier.

  • question: The question asked about the product.

  • q_sum: A short summary of the question.

  • multi_answers: Multiple answers to the question provided by different users in community question answering.

  • bm25_score_answers: The BM25 relevance score of each answer.

  • polarity_answers: The sentiment polarity (positive/ negative/ neutral) of each answer.

  • objective_answer: The gold objective answer to the question.

  • subjective_answer: The gold subjective answer to the question.

  • use_template: A template for generating the subjective answer summary.

  • product_info: Additional structured info about the product including title, description, feature bullets, attributes, etc.

  • product_reviews: Review texts for the product.

  • question_length: Number of characters in the question.

  • objective_answer_length: Length of characters in the objective answer.

  • subjective_answer_length: Length of characters in the subjective answer.

  • review_length: Length of characters in each review.

  • product_reviews_scores: Score/ rating for each review (bm25).

  • sentiment_scores: Sentiment score for each review (1-5), in which 1-2 represent the negative, 3 means the netrual and 4-5 represent the positive.

  • knowledge: Relevant commonsense knowledge for each question from LLMs (GPT-3.5 turbo).

The key field information contained in each example in the dataset is described in detail above.

Dataset Files

LOGIN TO ACCESS DATASET FILES

Documentation

AttachmentSize
File README.md2.36 KB

 

QUESTIONS?