Citation Context

... ∑n i=1(T (qi) − yi)2. So, GBRT depends on three parameters: the learning rate α, the depth of the tree, d, and the number of iterations, k. We set α = 0.02 and k = 1000, and, following Hastie et al. =-=[13]-=-, we set d = 4. Finally, Mohan et al. [29] show that, since RF is resistant to overfitting and also often outperforms GBRT, the RF predictions can be used as starting point for GBRT. By doing so, GBRT...

Citation Context

...the relative number of concepts in which q occurs, which is defined as IDF (q) = log (|C|/df (q)), where |C| indicates the total number of concepts and df (q) the number of concepts in which q occurs =-=[21]-=-. The subscript f denotes the field of the Wikipedia articles used, see above. WIG(q) indicates the weighted information gain [39], which can be considered a predictor of the retrieval performance of ...

Citation Context

...g this setting on the linking effectiveness. In recent years, gradient boosted regression trees (GBRTs) have been established as the de facto state-of-the-art learning paradigm for web search ranking =-=[7, 12, 29]-=-. It is a point-wise learning to rank algorithm that predicts the relevance score of a result to a query by minimizing a loss function (e.g., the squared loss) using stochastic gradient descent. It is...

Citation Context

...2012 ACM 978-1-4503-0747-5/12/02 ...$10.00. valuable sources for many kinds of analyses, including online reputation management, news and trend detection, and targeted marketing and customer services =-=[4, 18, 32, 35]-=-. Searching and mining microblog streams offers interesting technical challenges, because of the sheer volume of the data, its dynamic nature, the creative language usage, and the length of individual...

Citation Context

... (4) As to the concept fields we use either the full Wikipedia article, its title, or its incoming anchor texts. To combine the rankings produced by each constituent n-gram of a tweet, we use combMNZ =-=[33]-=-. CombMNZ is a result list merging method and a variant of CombSUM—which sums a document’s scores from all lists where it was retrieved. CombMNZ multiplies the CombSUM score by the number of lists tha...

Citation Context

...icles. With over 3.5 million articles, Wikipedia has become a rich source of knowledge and a common target for linking; automatic linking approaches using Wikipedia have met with considerable success =-=[14, 25, 27, 28]-=-. Most, if not all, of the linking methods assume that the input text is relatively clean and grammatically correct and that it provides sufficient context for the purposes of identifying concepts. Mi...

Citation Context

...guage usage, and the length of individual posts [17, 22]. In many microblog search scenarios the goal is to find out what people are saying about concepts such as products, brands, persons, et cetera =-=[31]-=-. Here, it is important to be able to accurately retrieve tweets that are on topic, including all possible naming and other lexical variants. So, it is common to manually construct lengthy keyword que...

Citation Context

...icles. With over 3.5 million articles, Wikipedia has become a rich source of knowledge and a common target for linking; automatic linking approaches using Wikipedia have met with considerable success =-=[14, 25, 27, 28]-=-. Most, if not all, of the linking methods assume that the input text is relatively clean and grammatically correct and that it provides sufficient context for the purposes of identifying concepts. Mi...

Citation Context

...n seen as a way of providing semantics to digital items. The idea has been used for different media types (such as text [27, 28] and multimedia [34]) and for different text genres (such as news pages =-=[9]-=-, queries [23], archives [6], and radiology reports [14]). A simple and frequently taken approach for linking text to concepts is to perform lexical matching between (parts of the text) and the concep...

Citation Context

...ing amount of attention in recent years. Starting from the domain of named entity recognition (NER), current approaches establish links not just to entity types, but to the actual entities themselves =-=[15, 20, 30]-=-. Instead of merely identifying types, we also aim to disambiguate the found concepts and link them to Wikipedia articles. With over 3.5 million articles, Wikipedia has become a rich source of knowled...

Citation Context

... [23], archives [6], and radiology reports [14]). A simple and frequently taken approach for linking text to concepts is to perform lexical matching between (parts of the text) and the concept titles =-=[10, 26]-=-, an approach related to keyword-based interfaces to databases [38]. However, merely matching an input text with concept titles suffers from many drawbacks, including ambiguity (where different concep...

Citation Context

...2012 ACM 978-1-4503-0747-5/12/02 ...$10.00. valuable sources for many kinds of analyses, including online reputation management, news and trend detection, and targeted marketing and customer services =-=[4, 18, 32, 35]-=-. Searching and mining microblog streams offers interesting technical challenges, because of the sheer volume of the data, its dynamic nature, the creative language usage, and the length of individual...

Citation Context

...h as those from Wikipedia entries and news articles. For comparative purposes, we include their method as one of the baselines and show that it does not perform well on tweets. Ferragina and Scaiella =-=[11]-=- propose an approach similar to [28], but with an explicit focus on short texts. They incorporate a voting scheme as well as pruning of n-grams unrelated to the input text. Since this method is geared...

Citation Context

...2.1 Linking Text Links to a knowledge structure are often seen as a way of providing semantics to digital items. The idea has been used for different media types (such as text [27, 28] and multimedia =-=[34]-=-) and for different text genres (such as news pages [9], queries [23], archives [6], and radiology reports [14]). A simple and frequently taken approach for linking text to concepts is to perform lexi...

Citation Context

...ing amount of attention in recent years. Starting from the domain of named entity recognition (NER), current approaches establish links not just to entity types, but to the actual entities themselves =-=[15, 20, 30]-=-. Instead of merely identifying types, we also aim to disambiguate the found concepts and link them to Wikipedia articles. With over 3.5 million articles, Wikipedia has become a rich source of knowled...

Citation Context

...umber of concepts and df (q) the number of concepts in which q occurs [21]. The subscript f denotes the field of the Wikipedia articles used, see above. WIG(q) indicates the weighted information gain =-=[39]-=-, which can be considered a predictor of the retrieval performance of a query. It uses the set of all candidate concepts retrieved for this n-gram, Cq , and determines the relative probability of q oc...

Citation Context

... occurrence of q in separate concept fields, the position of the first occurrence of the n-gram, the distance between the first and last occurrence, and various IR-based measures [21]. Of these, RIDF =-=[8]-=- is the difference between expected and observed IDF for a concept, which is defined as RIDF (c, q) = log ( |C| df (q) ) + log ( 1− exp (−n(q, C) |C| )) . We also consider whether the title of the Wik...

Citation Context

...ion we lift in our work, enabling us to add semantics to tweets without hashtags. Previous work has shown that hashtag usage is quite low and differs a lot per country and language [36]. Huang et al. =-=[16]-=- analyze the semantics of hashtags in more detail and reveal that hashtagging in Twitter is more commonly used to join public discussions than to organize content for future retrieval. In order to ver...

Citation Context

...hing and mining microblog streams offers interesting technical challenges, because of the sheer volume of the data, its dynamic nature, the creative language usage, and the length of individual posts =-=[17, 22]-=-. In many microblog search scenarios the goal is to find out what people are saying about concepts such as products, brands, persons, et cetera [31]. Here, it is important to be able to accurately ret...

Citation Context

... meaning” to text in general [24] or to text contained in tweets. Liu et al. [20] focus on NER on tweets and use a semi-supervised learning framework to identify four types of entities. Benson et al. =-=[3]-=- try to match tweets to “records.” These records are, for example, artist-venue pairs and can be obtained from sources like music guides. They train a model that extracts artists and venues from tweet...

Citation Context

...g this setting on the linking effectiveness. In recent years, gradient boosted regression trees (GBRTs) have been established as the de facto state-of-the-art learning paradigm for web search ranking =-=[7, 12, 29]-=-. It is a point-wise learning to rank algorithm that predicts the relevance score of a result to a query by minimizing a loss function (e.g., the squared loss) using stochastic gradient descent. It is...

Citation Context

...g this setting on the linking effectiveness. In recent years, gradient boosted regression trees (GBRTs) have been established as the de facto state-of-the-art learning paradigm for web search ranking =-=[7, 12, 29]-=-. It is a point-wise learning to rank algorithm that predicts the relevance score of a result to a query by minimizing a loss function (e.g., the squared loss) using stochastic gradient descent. It is...

Citation Context

...xchanges. As Twitter has grown, novel language use and standards such as mentions (to reference another user), hashtags (to refer to a topic), and retweets (similar to an e-mail forward) have emerged =-=[37]-=-. Various authors have attempted to “give meaning” to text in general [24] or to text contained in tweets. Liu et al. [20] focus on NER on tweets and use a semi-supervised learning framework to identi...

Citation Context

...ly taken approach for linking text to concepts is to perform lexical matching between (parts of the text) and the concept titles [10, 26], an approach related to keyword-based interfaces to databases =-=[38]-=-. However, merely matching an input text with concept titles suffers from many drawbacks, including ambiguity (where different concepts with the same label can be confused) and a possible lack of spec...

Citation Context

...ay of providing semantics to digital items. The idea has been used for different media types (such as text [27, 28] and multimedia [34]) and for different text genres (such as news pages [9], queries =-=[23]-=-, archives [6], and radiology reports [14]). A simple and frequently taken approach for linking text to concepts is to perform lexical matching between (parts of the text) and the concept titles [10, ...

Citation Context

...icles. With over 3.5 million articles, Wikipedia has become a rich source of knowledge and a common target for linking; automatic linking approaches using Wikipedia have met with considerable success =-=[14, 25, 27, 28]-=-. Most, if not all, of the linking methods assume that the input text is relatively clean and grammatically correct and that it provides sufficient context for the purposes of identifying concepts. Mi...

Citation Context

...clude the former as a feature in our framework and evaluate the latter as a baseline. Kwak et al. [18] show that hashtags are good indicators to detect events and trending topics and Laniado and Mika =-=[19]-=- explore the use of hashtags in Twitter and the relation to (Freebase) concepts. Using manual annotations, they find that about half of the hashtags can be mapped to Freebase concepts, most of them be...

Citation Context

... [23], archives [6], and radiology reports [14]). A simple and frequently taken approach for linking text to concepts is to perform lexical matching between (parts of the text) and the concept titles =-=[10, 26]-=-, an approach related to keyword-based interfaces to databases [38]. However, merely matching an input text with concept titles suffers from many drawbacks, including ambiguity (where different concep...

Citation Context

...2012 ACM 978-1-4503-0747-5/12/02 ...$10.00. valuable sources for many kinds of analyses, including online reputation management, news and trend detection, and targeted marketing and customer services =-=[4, 18, 32, 35]-=-. Searching and mining microblog streams offers interesting technical challenges, because of the sheer volume of the data, its dynamic nature, the creative language usage, and the length of individual...

Citation Context

...icles. With over 3.5 million articles, Wikipedia has become a rich source of knowledge and a common target for linking; automatic linking approaches using Wikipedia have met with considerable success =-=[14, 25, 27, 28]-=-. Most, if not all, of the linking methods assume that the input text is relatively clean and grammatically correct and that it provides sufficient context for the purposes of identifying concepts. Mi...

Citation Context

...rieve tweets that are on topic, including all possible naming and other lexical variants. So, it is common to manually construct lengthy keyword queries that (hopefully) capture all possible variants =-=[2]-=-. We propose an alternative approach, namely to determine what a microblog post is about by automatically identifying concepts in them. We take a concept to be any item that has a unique and unambiguo...

Citation Context

... semantics to digital items. The idea has been used for different media types (such as text [27, 28] and multimedia [34]) and for different text genres (such as news pages [9], queries [23], archives =-=[6]-=-, and radiology reports [14]). A simple and frequently taken approach for linking text to concepts is to perform lexical matching between (parts of the text) and the concept titles [10, 26], an approa...

Citation Context

...entions (to reference another user), hashtags (to refer to a topic), and retweets (similar to an e-mail forward) have emerged [37]. Various authors have attempted to “give meaning” to text in general =-=[24]-=- or to text contained in tweets. Liu et al. [20] focus on NER on tweets and use a semi-supervised learning framework to identify four types of entities. Benson et al. [3] try to match tweets to “recor...

Citation Context

...2012 ACM 978-1-4503-0747-5/12/02 ...$10.00. valuable sources for many kinds of analyses, including online reputation management, news and trend detection, and targeted marketing and customer services =-=[4, 18, 32, 35]-=-. Searching and mining microblog streams offers interesting technical challenges, because of the sheer volume of the data, its dynamic nature, the creative language usage, and the length of individual...

Citation Context

...hing and mining microblog streams offers interesting technical challenges, because of the sheer volume of the data, its dynamic nature, the creative language usage, and the length of individual posts =-=[17, 22]-=-. In many microblog search scenarios the goal is to find out what people are saying about concepts such as products, brands, persons, et cetera [31]. Here, it is important to be able to accurately ret...

Citation Context

... tweets, an assumption we lift in our work, enabling us to add semantics to tweets without hashtags. Previous work has shown that hashtag usage is quite low and differs a lot per country and language =-=[36]-=-. Huang et al. [16] analyze the semantics of hashtags in more detail and reveal that hashtagging in Twitter is more commonly used to join public discussions than to organize content for future retriev...