Question Answering on SQuAD

نویسنده

  • Chenjie Yang
چکیده

In this project, we exploit several deep learning architectures in Question Answering field, based on the newly released Stanford Question Answering dataset (SQuAD)[7]. We introduce a multi-stage process that encodes context paragraphs at different levels of granularity, uses co-attention mechanism to fuse representations of questions and context paragraphs, and finally decodes the co-attention vectors to get the answers. Our best model gets 62.23% F1 score and 48.72% EM score on the test set.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Rolling Deep with the SQuAD: Question Answering

Stanford Question Answering Dataset (SQuAD) is a ”reading comprehension dataset”. It is composed of passages of text, a question about that text, and an answer directly from the text. It allows groups to test their question-answering systems against other submissions as well as a Human Performance benchmark of 82.3 F1, 91.221 EM scores. Our task is to build our own question-answering system usi...

متن کامل

SQuAD Question Answering Dataset: CS224N Assn 4

We solve the contextual question answering problem, which is an essential part in many automated question-answering datasets. Recently the SQuAD dataset [1] was uploaded and there were several deep learning approaches proposed to solve this. We implement a modified version of one of them, the Dynamic Coattention model as well as simple baseline.

متن کامل

SQuAD Question Answering Problem : A match-lstm implementation

As a first experience in neural network modeling, and tensor flow usage, we try to implement a match-LSTM attention model, which is known for achieving interesting results on the SQuaD dataset leader board.

متن کامل

Co-Dependent Attention on SQuAD

In the realm of natural language processing, machine comprehension of textual documents is an incredibly important problem that presents various challenges and difficulties. A benchmark dataset for question answering named SQuAD is comprised of around a hundred thousand question-answer pairs, along with context paragraphs for each. The answer to each question is a span within the context, and i...

متن کامل

Analysis of Wikipedia-based Corpora for Question Answering

This paper gives comprehensive analyses of corpora based on Wikipedia for several tasks in question answering. Four recent corpora are collected, WIKIQA, SELQA, SQUAD, and INFOBOXQA, and first analyzed intrinsically by contextual similarities, question types, and answer categories. These corpora are then analyzed extrinsically by three question answering tasks, answer retrieval, selection, and ...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2017