Retrieval and novelty detection at the sentence level

An information-pattern-based approach to novelty detection

The detection of new information in a document stream is an important component of many potential applications. In this paper, a new novelty detection approach based on the identification of sentence level patterns is proposed. Given a user's information need, some patterns in sentences such as combinations of query words, named entities and phrases, may contain more important and relevant information than single words. Therefore, the proposed novelty detection approach focuses on the identification of previously unseen query-related patterns in sentences. Specifically, a query is preprocessed and represented with patterns that include both query words and required answer types. These patterns are used to retrieve sentences, which are then determined to be novel if it is likely that a new answer is present. An analysis of patterns in sentences was performed with data from the TREC 2002 novelty track and experiments on novelty detection were carried out on data from the TREC 2003 and 2004 novelty tracks. The experimental results show that the proposed pattern-based approach significantly outperforms all three baselines in terms of precision at top ranks.

L1 norm based KPCA for novelty detection

Factual inconsistencies existed in the output of abstractive summarization models with original documents are frequently presented. Fact consistency assessment requires the reasoning capability to find subtle clues to identify whether a model-generated summary is consistent with the original document. This paper proposes a fine-grained two-stage Fact Consistency assessment framework for Summarization models (SumFC). Given a document and a summary sentence, in the first stage, SumFC selects the top-K most relevant sentences with the summary sentence from the document. In the second stage, the model performs fine-grained consistency reasoning at the sentence level, and then aggregates all sentences' consistency scores to obtain the final assessment result. We get the training data pairs by data synthesis and adopt contrastive loss of data pairs to help the model identify subtle cues. Experiment results show that SumFC has made a significant improvement over the previous state-of-the-art methods. Our experiments also indicate that SumFC distinguishes detailed differences better.

Most current image captioning systems focus on describing general image content, and lack background knowledge to deeply understand the image, such as exact named entities or concrete events. In this work, we focus on the entity-aware news image captioning task which aims to generate informative captions by leveraging the associated news articles to provide background knowledge about the target image. However, due to the length of news articles, previous works only employ news articles at the coarse article or sentence level, which are not fine-grained enough to refine relevant events and choose named entities accurately. To overcome these limitations, we propose an Information Concentrated Entity-aware news image CAPtioning (ICECAP) model, which progressively concentrates on relevant textual information within the corresponding news article from the sentence level to the word level. Our model first creates coarse concentration on relevant sentences using a cross-modality retrieval model and then generates captions by further concentrating on relevant words within the sentences. Extensive experiments on both BreakingNews and GoodNews datasets demonstrate the effectiveness of our proposed method, which outperforms other state-of-the-arts. The code of ICECAP is publicly available at https://github.com/HAWLYQ/ICECAP.

Novelty detection is concerned with identifying abnormal system behaviours and abrupt changes from one regime to another. This paper proposes an on-line (causal) novelty detection method capable of detecting both outliers and regime change points in sequential time-series data. Our approach is based on a Kalman filter in order to model time-series data and extreme value theory is used to compute a novelty measure in a principled manner. The proposed approach is shown to be effective via experiments on several real-world data sets.

Documents discussing public affairs, common themes, interesting products, and so on, are reported and distributed on the Web. Positive and negative opinions embedded in documents are useful references and feedbacks for governments to improve their services, for companies to market their products, and for customers to purchase their objects. Web opinion mining aims to extract, summarize, and track various aspects of subjective information on the Web. Mining subjective information enables traditional information retrieval (IR) systems to retrieve more data from human viewpoints and provide information with finer granularity. Opinion extraction identifies opinion holders, extracts the relevant opinion sentences, and decides their polarities. Opinion summarization recognizes the major events embedded in documents and summarizes the supportive and the nonsupportive evidence. Opinion tracking captures subjective information from various genres and monitors the developments of opinions from spatial and temporal dimensions. To demonstrate and evaluate the proposed opinion mining algorithms, news and bloggers' articles are adopted. Documents in the evaluation corpora are tagged in different granularities from words, sentences to documents. In the experiments, positive and negative sentiment words and their weights are mined on the basis of Chinese word structures. The f‐measure is 73.18% and 63.75% for verbs and nouns, respectively. Utilizing the sentiment words mined together with topical words, we achieve f‐measure 62.16% at the sentence level and 74.37% at the document level.

Sentence level novelty detection aims at spotting sentences with novel information from an ordered sentence list. In the task, sentences appearing later in the list with no new meanings are eliminated. For the task of novelty detection, the contributions of this paper are three-fold. First, conceptually, this paper reveals the computational nature of the task currently overlooked by the Novelty community--Novelty as a combination of partial overlap (PO) and complete overlap (CO) relations between sentences. We define partial overlap between two sentences as a sharing of common facts, while complete overlap is when one sentence covers all of the meanings of the other sentence. Second, technically, a novel approach, the selected pool method is provided which follows naturally from the PO-CO computational structure. We provide formal error analysis for selected pool and methods based on this PO-CO framework. We address the question how accurate must the PO judgments be to outperform the baseline pool method. Third, experimentally, results were presented for all the three novelty datasets currently available. Results show that the selected pool is significantly better or no worse than the current methods, an indication that the term overlap criterion for the PO judgments could be adequately accurate.

Cross-lingual natural language processing relies on translation, either by humans or machines, at different levels, from translating training data to translating test sets. However, compared to original texts in the same language, translations possess distinct qualities referred to as translationese. Previous research has shown that these translation artifacts influence the performance of a variety of cross-lingual tasks. In this work, we propose a novel approach to reducing translationese by extending an established bias-removal technique. We use the Iterative Null-space Projection (INLP) algorithm, and show by measuring classification accuracy before and after debiasing, that translationese is reduced at both sentence and word level. We evaluate the utility of debiasing translationese on a natural language inference (NLI) task, and show that by reducing this bias, NLI accuracy improves. To the best of our knowledge, this is the first study to debias translationese as represented in latent embedding space.

Self-regulation and early writing: A longitudinal examination from preschool through first grade

The present paper aims to investigate how Egyptian bilingual social media users resort to code-mixing in their daily use of social media. The aim is to uncover the reasons behind code- mixing practices among this class of educated Egyptians in a monolingual country. In doing so, the paper attempts to demonstrate the nature of their linguistic endeavors and the role of code- mixing in identifying social media texts. Data for the present paper was collected from a sample of social media texts after permission to participants’ accounts was requested. The study followed a qualitative research approach where the data was analyzed to identify the common factors of code- mixing, and find out the reasons why some bilinguals engage in the phenomenon and practice it in their social media texts. The theoretical framework within which this study is conducted corresponds to Muysken (2000); his classification of code-mixing phenomena at the sentence level (i.e. insertion, alternation, and congruent lexicalization) is followed here. The results revealed that most of the bilingual educated Egyptians who resort to code mixing view it as a facilitating and a time-saving strategy especially when expressing complex feelings, emotions and opinions. The paper claims, contrary to previous research, that the occurrence of a word or a structure from one language into the context of another language has to be referred to as code-mixing rather than code switching. I claim here that code switching is one of the processes of code mixing which is similar to alternation and it has to be embedded under the general frame of code mixing. The study proposes two types of code mixed texts that are referred to as Multiple Code-Mixed Texts (MCMTs) and Transliterated Code-Mixed Texts (TCMTs). The results also indicated that a high percentage of the participants, though extensively involved in code-mixing, view it as a negative influence on their learning of English as a second language.

BOOK NOTICES 883 tic) approach to meaning, 'which leaves the bulk of "explanation" to the pragmatics' (93). Ideally, the basis for pragmatic approaches to meaning should be the utterance within its situational context. This, however, is at odds with the fact that most analyses are confined to the sentence level—a problem the author himself realizes. A related problem is that W commonly invokes 'default', 'core' or (proto)- 'typical' meanings of his examples. As with many semantic analyses, it is often hard to see how W arrives at such default interpretations since he largely dismisses the 'chimera of intuition' (219) as the basis for his investigation. The potential of analyzing frequency patterns to that end, a method which would have conformed more closely to W's theoretical preliminaries , has apparently been overlooked. In the empirical chapters, many individual proposals and example sentences are taken up that are already discussed in the three introductory chapters. This is perhaps the greatest weakness of the study. It fails to provide exhaustive discussions of previous proposals in the relevant place. The book has its merits, though. Given the enormous literature and the terminological profusion in the field, a consistent presentation ofthe semantics of related modals and penphrastics alongside a critical review of the relevant literature is an achievement in itself. Second and more rewarding, unlike many previous researchers, W does notjump to grand conclusions . Instead he disproves, or at least questions, some strong claims that have so far gone unchallenged . W has done a lot of the tedious work of testing previous hypotheses against empirical data. This is also one reason why the book's main results echo scholars from the 1960s and 1970s: Modals are unmarked , less specific semantically, and generally more subjective relative to their corresponding periphrastics , which in their turn fill the syntactic and semantic gaps left by the modals' defective paradigms . Unfortunately, there is another reason why W is right in being so cautious: the scarcity of new, conclusive data offered in this study. It is not always satisfying to read that matters are complex. More hints at alternative explanations would have been welcome. Nor are the results produced convincing throughout. 'Confirmation' of rather uncontroversial issues such as the relative formality of the will future when compared to the be going to future is gathered from two text samples that are by no means balanced or even representative (19Of.). Finally, it seems rather premature to rule out differences between American and British English in the relevant area when this has not been investigated. In sum, this book provides a very readable account of the state of the art, but it leaves open many questions for further research into English modals and periphrastics. [Manfred Krug, University of Freiburg .] The Nilo-Saharan languages: A comparative essay. By Lionel M. Bender. (Lincom handbooks in linguistics 6.) Munich & Newcastle: Lincom Europa, 1996. Pp. 253. After the preface and acknowledgments, m which the National Science Foundation is recognized as his primary source ofsupport, Bender surveys the historical development of African linguistics with special emphasis on the Nilo-Saharan language grouping, in which earlier and different theories of relationships are exhaustively discussed. In Section 2.2.2, (Proto-segments and modern reflexes , 66-7), B offers a conservative reconstruction of Proto-Nilo-Saharan phonemes and then carefully supports his reconstruction with sections devoted to labials, alveolare or dentals, palatals, velars, and vowels, which first list the phoneme then conveniently key its occurrence to the roots discussed in the following pages, e.g. *g-: #35, 36, 121, 124 (all families) . . . *-g: #111. An appreciated hallmark of B's general approach is facilitating investigation of any of his assumptions through copious cross-referencing. His approach is consistently conservative and, as a result, very convincing . He presents several lists of isoglosses, graded for reliability according to their attestation m two major groupings: Outliers (Songai, Sanaran, Kuhak) and Satellite Core (all others). He characterizes as 'excellent' those isoglosses which 'are found in all the Outliers (A, B, K) and in Satellite-Core.' In this category, 16 items are presented. A typical entry (77) is: '3. SS40: work = cultivate, builds make = mold = create2, change3, hoe4, dig5(8)', which he then details...

Selective attention (SA) is an important strategy for language comprehension both in L1 and L2. Much of the existing research on L2 listening and reading has identified SA as a distinctive ability of effective and skilled L2 listeners and readers. However, extensive and intensive research on this specific strategy is relatively sparse. Therefore, this article provides an in-depth review of previous research involving the SA mechanism in both L1 and L2 language comprehension. As a result, this article identifies that there are two-levels of SA existing in previous academic research: the word level and sentence level SA. They may represent different cognitive processes as well as different SA strategies in real practice. However, this kind of classification on SA has not been fully recognized and specifically pinpointed by previous literature, since previous studies focused on only one type of SA while ignoring the other. Therefore, this article proposes to classify these two kinds of SA and then reviews previous literature according to this novel classification. In addition, this article summarizes the most frequently investigated modulators of SA as well as commonly used research methods for SA-related research. Some possible research gaps are pinpointed accordingly. Finally, this article illustrates several concepts and models that could explain the SA mechanism from psychological and linguistic perspectives, which could serve as a theoretical framework in future study. This article may offer some inspirations for future academic research in this field and listening or reading practices in the real world.

Implicit in the design of many intelligibility and perception studies is the assumption that noise masks all sentence stimuli equivalently, at least when they are controlled for length and structure. For example, controlled and normed sentences, such as the IEEE Harvard set, are typically thought to be equally affected by masker-noise. However, previous research has established that spoken stimuli of different structures, lengths, or complexities, are differentially masked (e.g. nonsense versus real words, words with different usage frequencies), therefore it is possible that masking affects even controlled sentences differentially . Using the UAW speech intelligibility dataset we analyzed Levenshtein Distance values based on transcriptions from over 900 native listeners of English to over 604 sentences from the UWNU IEEE sentence corpus. Each sentence was presented in noise at three SNRs ( + 2, 0, −2 dB). We find that the sentences are not equally intelligible. Moreover, there is an interaction with SNR level and sentence. In other words, the intelligibility of the sentences is impacted differentially by masker-noise. The results of this study suggest that, as researchers using speech stimuli, we should recognize that there are many sentence level factors that may introduce variance or otherwise affect the outcomes of our studies.

We propose a group recommender system considering the recommendation quantity and repeat purchasing by using the existing collaborative filtering algorithm in order to optimize the offline physical store inventories. This research is the first of its kind to consider recommendation quantity and repetitive recommendations when creating group recommender systems. In offline stores, physical limitations result in the ability to display only a limited number of items. Quantity and selection of the item is an important decision for offline stores. In this paper, we suggest applying the user-based recommender system, which is capable of determining the best suited recommendation items for each store. This model is evaluated by the MAE, precision, recall, and F1 measures, and shows higher performance than the baseline model. A new performance evaluation measure is also suggested in this research. New quantity precision, quantity recall, and quantity F1 measures consider a penalty for a shortage or excess of the recommendation quantity. Novelty is defined as the proportion of items that the consumer may not have experienced in the recommendation list. Through the use of this novelty measure, we assess the new profit creation effect of the suggested model. Finally, previous research focused on recommendations for online customers, however, we expanded the recommender system to incorporate offline stores. This research is not only an academic contribution to the marketing field, but also practical contribution to offline stores through the usability of a developed offline shopping algorithm.