Search

Choice overload describes a situation in which a person has difficulty in making decisions due to too many options. We examine choice overload when displaying related-article recommendations in digital libraries, and examine the effectiveness of recommendation algorithms in this domain. We first analyzed existing digital libraries, and found that only 30% of digital libraries show related-article recommendations to their users. Of these libraries, the majority (74%) displays 3–5 related articles; 28% of them display 6–10 related articles; and no digital library displayed more than ten related-article recommendations. We then conducted our experimental evaluation through GESIS' digital library Sowiport, with recommendations delivered by recommendations-as-a-service provider Mr. DLib. We use four metrics to analyze 41.3 million delivered recommendations: click-through rate (CTR), percentage of clicked recommendation sets (clicked set rate, CSR), average clicks per clicked recommendation set (ACCS), and time to first click (TTFC), which is the time between delivery of a set of recommendations to the first click. These metrics help us to analyze choice overload and can yield evidence for finding the ideal number of recommendations to display. We found that with increasing recommendation set size, i.e., the numbers of displayed recommendations, CTR decreases from 0.41% for one recommendation to 0.09% for 15 recommendations. Most recommendation sets only receive one click. ACCS increases with set size, but increases more slowly for six recommendations and more. When displaying 15 recommendations, the average clicks per set is at a maximum (1.15). Similarly, TTFC increases with larger recommendation set size, but increases more slowly for sets of more than five recommendations. While CTR and CSR do not indicate choice overload, ACCS and TTFC point toward 5–6 recommendations as being optimal for Sowiport. Content-based filtering yields the highest CTR with 0.118%, while stereotype recommendations yield the highest ACCS (1.28). Stereotype recommendations also yield the highest TTFC. This means that users take more time before clicking stereotype recommendations when compared to recommendations based on other algorithms. [ABSTRACT FROM AUTHOR]

Scientific documents generally contain multiple mathematical expressions in them. Detecting inline mathematical expressions are one of the most important and challenging tasks in scientific text mining. Recent works that detect inline mathematical expressions in scientific documents have looked at the problem from an image processing perspective. There is little work that has targeted the problem from NLP perspective. Towards this, we define a few features and applied Conditional Random Fields (CRF) to detect inline mathematical expressions in scientific documents. Apart from this feature based approach, we also propose a hybrid algorithm that combines Bidirectional Long Short Term Memory networks (Bi‐LSTM) and feature‐based approach for this task. Experimental results suggest that this proposed hybrid method outperforms several baselines in the literature and also individual methods in the hybrid approach. [ABSTRACT FROM AUTHOR]

• Unsupervised word sense disambiguation technique for diverse NLP tasks. • Word embeddings combining sense disambiguation, improving its vector representation. • Synset embedding models can help to leverage the disambiguation of polysemous words. • Multi-sense embedding models provide better representation than single-sense ones. • Recurrent synset-embedding models can improve the quality of word representations. Natural Language Understanding has seen an increasing number of publications in the last few years, especially after robust word embeddings models became prominent, when they proved themselves able to capture and represent semantic relationships from massive amounts of data. Nevertheless, traditional models often fall short in intrinsic issues of linguistics, such as polysemy and homonymy. Any expert system that makes use of natural language in its core, can be affected by a weak semantic representation of text, resulting in inaccurate outcomes based on poor decisions. To mitigate such issues, we propose a novel approach called Most Suitable Sense Annotation (MSSA) , that disambiguates and annotates each word by its specific sense, considering the semantic effects of its context. Our approach brings three main contributions to the semantic representation scenario: (i) an unsupervised technique that disambiguates and annotates words by their senses, (ii) a multi-sense embeddings model that can be extended to any traditional word embeddings algorithm, and (iii) a recurrent methodology that allows our models to be re-used and their representations refined. We test our approach on six different benchmarks for the word similarity task, showing that our approach can produce state-of-the-art results and outperforms several more complex state-of-the-art systems. [ABSTRACT FROM AUTHOR]

Current mathematical search systems allow math expressions within a document to be queried using math expressions and keywords. To accept such queries, math search systems must index both math expressions and textual information in documents. Each indexed math expression is usually associated with all the words in its surrounding context within a given window size. However, we found that this context is often ineffective for explaining math expressions in scientific papers. The meaning of an expression is usually defined in the early part of a document, and the meaning of each symbol contained in the expression can be useful for explaining the entire expression. This explanation may not be captured within the context of a math expression, unless we set the context to have a very wide window size. However, widening the window size also increases the proportion of words that are unrelated to the expression. This paper proposes the use of dependency relationships between math expressions to enrich the textual information of each expression. We examine the influence of this enrichment in a math search system. The experimental results show that significantly better precision can be obtained using the enriched textual information rather than the math expressions' own textual information. This indicates that the enrichment of textual information for each math expression using dependency relationships enhances the math search system. [ABSTRACT FROM AUTHOR]

Formulaic expressions, such as ' in this paper we propose ', are helpful for authors of scholarly papers because they convey communicative functions; in the above, it is ' showing the aim of this paper '. Thus, resources of formulaic expressions, such as a dictionary, that could be looked up easily would be useful. However, forms of formulaic expressions can often vary to a great extent. For example, ' in this paper we propose ', ' in this study we propose ' and ' in this paper we propose a new method to ' are all regarded as formulaic expressions. Such a diversity of spans and forms causes problems in both extraction and evaluation of formulaic expressions. In this paper, we propose a new approach that is robust to variation of spans and forms of formulaic expressions. Our approach regards a sentence as consisting of a formulaic part and non-formulaic part. Then, instead of trying to extract formulaic expressions from a whole corpus, by extracting them from each sentence, different forms can be dealt with at once. Based on this formulation, to avoid the diversity problem, we propose evaluating extraction methods by how much they convey specific communicative functions rather than by comparing extracted expressions to an existing lexicon. We also propose a new extraction method that utilises named entities and dependency structures to remove the non-formulaic part from a sentence. Experimental results show that the proposed extraction method achieved the best performance compared to other existing methods. • Formulaic expression extraction using named entities and dependency structure. • Sentence retrieval task as an extrinsic evaluation for formulaic expressions. • Changing weights on word embeddings captures characteristics of extraction methods. [ABSTRACT FROM AUTHOR]

Abstract Sentence compression aims to shorten a sentence into a compression while remaining grammatical and preserving the underlying meaning of the original sentence. Previous works have recognized that linguistic features such as parts-of-speech tags and dependency labels are helpful to compression generation. In this work, we introduce a gating mechanism and propose a gated neural network that selectively exploits linguistic knowledge for deletion-based sentence compression. An extensive experiment was conducted on four downstream datasets, showing that the proposed gated neural network method leads to better compression upon both automatic metrics and human evaluation, compared to previous competitive compression methods. We also observed that the generated compression by the proposed gated neural network share more grammatical relations in common with the ground-truth compression than the baseline method, indicating that important grammatical relations, such as subject or object of a sentence, are more likely to be kept in the compression by the proposed method. Furthermore, visualization analysis is conducted to explore the selective use of linguistic features, suggesting that the gate mechanism could condition the predicted compression on different linguistic features. [ABSTRACT FROM AUTHOR]