Microblog Retrieval Using Ensemble of Feature Sets through Supervised Feature Selection

Abstract

Abu Nowshed CHY †a) , Md Zia ULLAH †b) , Nonmembers, and Masaki AONO †c) , Member SUMMARYMicroblog, especially twitter, has become an integral part of our daily life for searching latest news and events information. Due to the short length characteristics of tweets and frequent use of unconventional abbreviations, content-relevance based search cannot satisfy user's information need. Recent research has shown that considering temporal and contextual aspects in this regard has improved the retrieval performance significantly. In this paper, we focus on microblog retrieval, emphasizing the alleviation of the vocabulary mismatch, and the leverage of the temporal (e.g., recency and burst nature) and contextual characteristics of tweets. To address the temporal and contextual aspect of tweets, we propose new features based on query-tweet time, word embedding, and query-tweet sentiment correlation. We also introduce some popularity features to estimate the importance of a tweet. A three-stage query expansion technique is applied to improve the relevancy of tweets. Moreover, to determine the temporal and sentiment sensitivity of a query, we introduce query type determination techniques. After supervised feature selection, we apply random forest as a feature ranking method to estimate the importance of selected features. Then, we make use of ensemble of learning to rank (L2R) framework to estimate the relevance of query-tweet pair. We conducted experiments on TREC Microblog 2011 and 2012 test collections over the TREC Tweets2011 corpus. Experimental results demonstrate the effectiveness of our method over the baseline and known related works in terms of precision at 30 (P@30), mean average precision (MAP), normalized discounted cumulative gain at 30 (NDCG@30), and R-precision (R-Prec) metrics. key words: microblog search, temporal information retrieval, query expansion, feature selection, learning to rank, time-aware ranking IntroductionNowadays, microblog web sites are not only the places in maintaining the social relationships, but also act as a valuable information source. Everyday lots of users turn into microblog sites for sharing their views, opinions, experiences, important news, and also want to get some information what is happening around the world. Among several microblog sites, Twitter * is now the most popular, where lots of users post tweets whenever a notable event occurs. That is why; information retrieval in twitter has made a hit with a lot of complaisance. By searching tweets, users find temporally relevant information, such as breaking news and real-time events [1]. That means, freshness (i.e. recency) of the tweet with respect to query time is an important factor of rele- vance. Another important characteristic of twitter is that people tends to post about a topic within a specific period of time (i.e. bursty nature). For example, when the breakup news of famous band "White Stripes" published on 2nd Feb, 2011, many people post tweets about this topic on that day. That is why; posts that are generate...

Query Expansion for Microblog Retrieval Focusing on an Ensemble of Features

“…Among several query expansion methods, pseudo-relevance feedback (PRF) based approaches were widely studied in microblog retrieval [4], [5], [6], [11]. The PRF approach assumes that top-ranked documents of the original query based on initial retrieval results are relevant and contain terms related to the query intent to augment the query representation.…”

“…Zingla et al [6] proposed a technique that extracted semantically related expansion terms from Wikipedia, DBpedia, and unstructured texts. Chy et al [11] used a PRF based simple query expansion by leveraging external resources and focused mainly on re-ranking the initial retrieved tweets based on several features by estimating the relevance of query-tweet pair. In contrast, we focus on exploiting several term relevance features to select the candidate expansion terms for query expansion and use a simple retrieval model.…”

“…In Web search, learning-to-rank (L2R) methods had been used by several researchers [9], [10], [41], [42] to rank the candidate expansion terms. In addition, supervised feature selection approaches based on learning-to-rank algorithm [10], [43] and elastic-net regularization method [11] were employed by researchers to select the effective set of features.…”

“…Moreover, supervised feature selection and a state-of-the-art machine learning technique is also applied to learn the feature weight in a learning-to-rank (L2R) framework. Experimental results with the TREC microblog collections show that our method improves the retrieval performance over the baseline and some other competitive query expansion methods [4], [7], [8], [11], [12], [13], [14].…”

Corpus Expansion for Neural CWS on Microblog-Oriented Data with <i>λ</i>-Active Learning Approach

“…Recently, microblog-oriented NLP tasks have grabbed considerable attention of researchers, such as microblogoriented sentiment analysis [11], named entity recognition from microblogs [12]- [14], microblog retrieval [15]. CWS is the prerequisite step of microblog-oriented NLP tasks.…”