Papers › Iterative Hierarchical Attention for Answering Complex Questions over Long Documents

Iterative Hierarchical Attention for Answering Complex Questions over Long Documents

1 Jun 2021arXiv:2106.00200archive 2025-07-28

Haitian Sun, William W. Cohen, Ruslan Salakhutdinov

We propose a new model, DocHopper, that iteratively attends to different parts of long, hierarchically structured documents to answer complex questions. Similar to multi-hop question-answering (QA) systems, at each step, DocHopper uses a query q to attend to information from a document, combines this ``retrieved'' information with q to produce the next query. However, in contrast to most previous multi-hop QA systems, DocHopper is able to ``retrieve'' either short passages or long sections of the document, thus emulating a multi-step process of ``navigating'' through a long document to answer a question. To enable this novel behavior, DocHopper does not combine document information with q by concatenating text to the text of q, but by combining a compact neural representation of q with a compact neural representation of a hierarchical part of the document, which can potentially be quite large. We experiment with DocHopper on four different QA tasks that require reading long and complex documents to answer multi-hop questions, and show that DocHopper achieves state-of-the-art results on three of the datasets. Additionally, DocHopper is efficient at inference time, being 3--10 times faster than the baselines.

PaperPDFConference PDF

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

No code repository is listed for this paper in the archive or in Syntology's graph.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Multi-hop Question AnsweringQuestion AnsweringSentence Embedding

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Question Answering ConditionalQA DocHopper Conditional (answers) 42.0 / 46.4 #2 of 3 Archive leaderboard report
Question Answering ConditionalQA DocHopper Conditional (w/ conditions) 3.1 / 3.8 #2 of 3 Archive leaderboard report
Question Answering ConditionalQA DocHopper Overall (answers) 40.6 / 45.2 #2 of 3 Archive leaderboard report
Question Answering ConditionalQA DocHopper Overall (w/ conditions) 31.9 / 36.0 #2 of 3 Archive leaderboard report
Question Answering HybridQA DocHopper ANS-EM 46.3 #3 of 4 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

AdamAttentionAttention DropoutBARTBERTBPEDense ConnectionsDropoutLayer NormalizationLinear LayerLinear Warmup With Linear DecayMulti-Head AttentionRAGResidual ConnectionSoftmaxWeight DecayWordPiece

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections