For a general user, easy access to vast amounts of online information available on past events has made retrospection much harder. We propose a problem of automatic event digest generation to aid effective and efficient retrospection. For this, in addition to text, a digest should maximize the reportage of time, geolocations, and entities to present a holistic view on the past event of interest. We propose a novel divergence-based framework that selects excerpts from an initial set of pseudo-relevant documents, such that the overall relevance is maximized, while avoiding redundancy in text, time, geolocations, and named entities, by treating them as independent dimensions of an event. Our method formulates the problem as an Integer Linear Program (ILP) for global inference to diversify across the event dimensions. Relevance and redundancy measures are defined based on JS-divergence between independent query and excerpt models estimated for each event dimension. Elaborate experiments on three real-world datasets are conducted to compare our methods against the state-of-the-art from the literature. Using Wikipedia articles as gold standard summaries in our evaluation, we find that the most holistic digest of an event is generated with our method that integrates all event dimensions. We compare all methods using standard Rouge-1, -2, and -SU4 along with Rouge-NP, and a novel weighted variant of Rouge.

Event Digest: A Holistic View on Past Events

Organizing the structure of fixed-length text summaries for events is important for their coherence and readability. However, typical measures used for evaluation in text summarization tasks often ignore the structure. In this paper, we conduct an empirical study on a crowdsourcing platform to get insights into regularities that make a text summary coherent and readable. For this, we generate four variants of human-written text summaries with 10 sentences for 100 seminal events, and conduct three experiments. Experiment 1 and 2 focus on analyzing the impact of sentence ordering and proximity between originally occurring adjacent sentences, respectively. Experiment 3 analyzes the feasibility of conducting such a study on a crowdsourcing platform. We release our data to facilitate future work like designing dedicated measures to evaluate summary structures.

How Do Order and Proximity Impact the Readability of Event Summaries

Today in this Big Data era, overwhelming amounts of textual information across different sources with a high degree of redundancy has made it hard for a consumer to retrospect on past events. A plausible solution is to link semantically similar information contained across the different sources to enforce a structure thereby providing multiple access paths to relevant information. Keeping this larger goal in view, this work uses Wikipedia and online news articles as two prominent yet disparate information sources to address the following three problems: • We address a linking problem to connect Wikipedia excerpts to news articles by casting it into an IR task. Our novel approach integrates time, geolocations, and entities with text to identify relevant documents that can be linked to a given excerpt. • We address an unsupervised extractive multi-document summarization task to generate a fixed-length event digest that facilitates efficient consumption of information contained within a large set of documents. Our novel approach proposes an ILP for global inference across text, time, geolocations, and entities associated with the event. • To estimate temporal focus of short event descriptions, we present a semi-supervised approach that leverages redundancy within a longitudinal news collection to estimate accurate probabilistic time models. Extensive experimental evaluations demonstrate the effectiveness and viability of our proposed approaches towards achieving the larger goal.

/pdf/leveraging-semantic-annotations-for-event-focused-search-3fczbduv7p.pdf

Leveraging Semantic Annotations for Event-focused Search & Summarization

This paper presents a novel approach to query-focused multi-document summarization. As a good biased summary is expected to keep a balance among query relevance, content salience and information diversity, the approach first makes use of both the content feature and the relationship feature to select a number of sentences via the co- training based semi-supervised learning, which can identify the query relevant sentences beyond a single point of view. Then the ranking algorithm based on Markov chain random walks is employed on the relevant sentences by encouraging content salience and information diversity in a unified framework. The final summary focusing on the integration of relevance, salience and diversity is created after several sentences with the highest overall ranking scores are extracted. We performed experiments on DUC2007 dataset and the evaluation results show that the proposed approach can achieve significant improvement over standard baseline approaches and gain comparable performance to the state-of-the-art systems.

Hai Wang

Papers

Query-Focused Multi-Document Summarization Using Co-Training Based Semi-Supervised Learning