Extending the language modeling framework for sentence retrieval to include local context

Abstract

Employing effective methods of sentence retrieval is essential for many tasks in Information Retrieval, such as summarization, novelty detection and question answering. The best performing sentence retrieval techniques attempt to perform matching directly between the sentences and the query. However, in this paper, we posit that the local context of a sentence can provide crucial additional evidence to further improve sentence retrieval. Using a Language Modeling Framework, we propose a novel reformulation of the sentence retrieval problem that extends previous approaches so that the local context is seamlessly incorporated within the retrieval models. In a series of comprehensive experiments, we show that localized smoothing and the prior importance of a sentence can improve retrieval effectiveness. The proposed models significantly and substantially outperform the state of the art and other competitive sentence retrieval baselines on recall-oriented measures, while remaining competitive on precision-oriented measures. This research demonstrates that local context plays an important role in estimating the relevance of a sentence, and that existing sentence retrieval language models can be extended to utilize this evidence effectively.

A recursive TF-ISF Based Sentence Retrieval Method with Local Context

“…Tasks like novelty detection, question answering, summarization, opinion mining and information provenance make use of sentence retrieval [1], [2]. Sentence retrieval methods are usually simple adaptations of document retrieval methods where sentences are treated as documents [3]- [5].…”

“…Sentence retrieval methods are usually simple adaptations of document retrieval methods where sentences are treated as documents [3]- [5]. The state of the art and most successful models for sentence retrieval are the vector space model [1], [6], [7] and the language model approach with local context [1], [2]. The vector space model for sentence retrieval uses the term frequency -inverse sentence frequency method (TF-ISF) which is analogous to the traditional term frequency -inverse document frequency method (TF-IDF) used for document retrieval.…”

“…When it comes to the language model approach for sentence retrieval at first it was implemented as a trivial adaptation from document retrieval to sentence retrieval, with sentences treated as documents [7]. In the recent works language model approach was successfully modified to take into account the context of sentence, like the document which contains it or some number of previous or next sentences [1], [2], [8]. The TF-ISF method was also modified to take into account context, but unsuccessfully without statistically significant improvements [1].…”

“…In the recent works language model approach was successfully modified to take into account the context of sentence, like the document which contains it or some number of previous or next sentences [1], [2], [8]. The TF-ISF method was also modified to take into account context, but unsuccessfully without statistically significant improvements [1]. Fernandez sees the reason for using context of sentences in the vocabulary mismatch problem rooted in the sparsity of sentences [1].…”

“…The TF-ISF method was also modified to take into account context, but unsuccessfully without statistically significant improvements [1]. Fernandez sees the reason for using context of sentences in the vocabulary mismatch problem rooted in the sparsity of sentences [1]. He also thinks that the context of sentences contains a content that is not encoded in the sentence, but that is related to the sentence and it is important not just because of vocabulary mismatch.…”