Author name cluster

Grace Hui Yang

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

3 papers

1 author row

AAAI Conference 2020 Conference Paper

Corpus-Level End-to-End Exploration for Interactive Systems

Zhiwen Tang
Grace Hui Yang

A core interest in building Artiﬁcial Intelligence (AI) agents is to let them interact with and assist humans. One example is Dynamic Search (DS), which models the process that a human works with a search engine agent to accomplish a complex and goal-oriented task. Early DS agents using Reinforcement Learning (RL) have only achieved limited success for (1) their lack of direct control over which documents to return and (2) the difﬁculty to recover from wrong search trajectories. In this paper, we present a novel corpus-level endto-end exploration (CE3) method to address these issues. In our method, an entire text corpus is compressed into a global low-dimensional representation, which enables the agent to gain access to the full state and action spaces, including the under-explored areas. We also propose a new form of retrieval function, whose linear approximation allows end-to-end manipulation of documents. Experiments on the Text REtrieval Conference (TREC) Dynamic Domain (DD) Track show that CE3 outperforms the state-of-the-art DS systems.

PDF Details

AAAI Conference 2019 Conference Paper

DeepTileBars: Visualizing Term Distribution for Neural Information Retrieval

Zhiwen Tang
Grace Hui Yang

Most neural Information Retrieval (Neu-IR) models derive query-to-document ranking scores based on term-level matching. Inspired by TileBars, a classical term distribution visualization method, in this paper, we propose a novel Neu- IR model that handles query-to-document matching at the subtopic and higher levels. Our system first splits the documents into topical segments, “visualizes” the matchings between the query and the segments, and then feeds an interaction matrix into a Neu-IR model, DeepTileBars, to obtain the final ranking scores. DeepTileBars models the relevance signals occurring at different granularities in a document’s topic hierarchy. It better captures the discourse structure of a document and thus the matching patterns. Although its design and implementation are light-weight, DeepTileBars outperforms other state-of-the-art Neu-IR models on benchmark datasets including the Text REtrieval Conference (TREC) 2010-2012 Web Tracks and LETOR 4. 0.

PDF Details

RLDM Conference 2017 Conference Abstract

Modeling Exploration of Intrinsically Diverse Search Tasks as Markov Decision Processes

Razieh Rahimi
Grace Hui Yang

Online learning of ranking models based on user interactions has attracted considerable attention in recent years. However, these models are mainly designed to optimize search results for one-time queries, while intrinsically diverse search tasks, where the goal is to explore all aspects of an information need, generally require search page navigation. The goal in such search scenarios is to utilize user interactions with results in order to optimize the whole search session. To provide the user with the next page of search results, the search system, on the one hand, should exploit all feedback information about the query, obtained from the user. On the other hand, to fully satisfy the user’s information need, the search system needs to cover all aspects of the information need in the search results, while the available feedback may be on some, but not all, aspects. The search system thus needs to explore and provide a diversified search results. Therefore, in each interaction, the search system has to resolve the exploration-exploitation tradeoff. We hypothesize that this tradeoff depends on the state of the search session, defined by feedback information obtained from user interactions. We thus propose a new approach that casts the defined search task as a Markov decision process (MDP) whose parameters are initially unknown to the search engine, and should be learned through user interactions. Keywords: Multi-page exploratory search, Markov decision process Acknowledgements This research is supported by DARPA grant FA8750-14-2-0226 and NSF CAREER AWARD IIS-1453721.

PDF Details