Review-Based Recommender Systems (RBRS) have attracted increasing research interest due to their ability to alleviate well-known cold-start problems. RBRS utilizes reviews to construct the user and items representations. However, in this paper, we argue that such a reliance on reviews may instead expose systems to the risk of being shilled. To explore this possibility, in this paper, we propose the first generation-based model for shilling attacks against RBRSs. Specifically, we learn a fake review generator through reinforcement learning, which maliciously promotes items by forcing prediction shifts after adding generated reviews to the system. By introducing the auxiliary rewards to increase text fluency and diversity with the aid of pre-trained language models and aspect predictors, the generated reviews can be effective for shilling with high fidelity. Experimental results demonstrate that the proposed framework can successfully attack three different kinds of RBRSs on the Amazon corpus with three domains and Yelp corpus. Furthermore, human studies also show that the generated reviews are fluent and informative. Finally, equipped with Attack Review Generators (ARGs), RBRSs with adversarial training are much more robust to malicious reviews.

Abstract While sentence simplification is an active research topic in NLP, its adjacent tasks of sentence complexification and same-level paraphrasing are not. To train models on all three tasks, we present two new unsupervised datasets. We compare these datasets, one labeled by a weak classifier and the other by a rule-based approach, with a single supervised dataset. Using these three datasets for training, we perform extensive experiments on both multitasking and prompting strategies. Compared to other systems trained on unsupervised parallel data, models trained on our weak classifier labeled dataset achieve state-of-the-art performance on the ASSET simplification benchmark. Our models also outperform previous work on sentence-level targeting. Finally, we establish how a handful of Large Language Models perform on these tasks under a zero-shot setting.

This paper describes the process of extracting and filtering Wikipedia revision history as a resource for grammatical error correction (GEC). Edits in Wikipedia revision history vary widely, including grammatical error corrections, information supplements, format amendments, and even vandalism. To extract only GEC-related revisions, we use an automated error annotation toolkit, ERRANT1, and extend it to process large data in parallel efficiently. With error-type analysis, we can then identify GEC-related edits and omit other unrelated edits (i.e., only the correction parts are reserved). The resulting corpus is - to our knowledge - the largest publicly available corpus of parallel possibly erroneous and correct sentences with error type labels.

In this paper, we present a observation that a word with some special sense should appear in specific category of articles, and this appearance will “take” the word to its Wiki sense [1] by Wikipedia link graph. This property can help to solve word sense disambiguation problem using Wikipedia.

This paper describes WriteAhead2, an interactive writing environment that provides lexical and grammatical suggestions for second language learners, and helps them write fluently and avoid common writing errors. The method involves learning phrase templates from dictionary examples, and extracting grammar patterns with example phrases from an academic corpus. At run-time, as the user types word after word, the actions trigger a list after list of suggestions. Each successive list contains grammar patterns and examples, most relevant to the half-baked sentence. WriteAhead2 facilitates steady, timely, and spot-on interactions between learner writers and relevant information for effective assisted writing. Preliminary experiments show that WriteAhead2 has the potential to induce better writing and improve writing skills.

The syntactic ambiguity of a transitive verb (Vt) followed by a noun (N) has long been a problem in Chinese parsing. In this paper, we propose a classifier to resolve the ambiguity of Vt-N structures. The design of the classifier is based on three important guidelines, namely, adopting linguistically motivated features, using all available resources, and easy in-tegration into a parsing model. The lin-guistically motivated features include semantic relations, context, and morpho-logical structures; and the available re-sources are treebank, thesaurus, affix da-tabase, and large corpora. We also pro-pose two learning approaches that resolve the problem of data sparseness by auto-parsing and extracting relative knowledge from large-scale unlabeled data. Our experiment results show that the Vt-N classifier outperforms the cur-rent PCFG parser. Furthermore, it can be easily and effectively integrated into the PCFG parser and general statistical pars-ing models. Evaluation of the learning approaches indicates that world knowledge facilitates Vt-N disambigua-tion through data selection and error cor-rection. 1

We introduce a method for learning to predict text and grammatical construction in a computer-assisted translation and writing framework. In our approach, predictions are offered on the fly to help the user make appropriate lexical and grammar choices during the translation of a source text, thus improving translation quality and productivity. The method involves automatically generating general-to-specific word usage summaries (i.e., writing suggestion module), and automatically learning high-confidence word- or phrase-level translation equivalents (i.e., translation suggestion module). At runtime, the source text and its translation prefix entered by the user are broken down into n-grams to generate grammar and translation predictions, which are further combined and ranked via translation and language models. These ranked prediction candidates are iteratively and interactively displayed to the user in a pop-up menu as translation or writing hints. We present a prototype writing assistant, TransAhead , that applies the method to a human-computer collaborative environment. Automatic and human evaluations show that novice translators or language learners substantially benefit from our system in terms of translation performance (i.e., translation accuracy and productivity) and language learning (i.e., collocation usage and grammar). In general, our methodology of inline grammar and text predictions or suggestions has great potential in the field of computer-assisted translation, writing, or even language learning.

We introduce a method for learning to translate out-of-vocabulary (OOV) words. The method focuses on combining sublexical/constituent translations of an OOV to generate its translation candidates. In our approach, wildcard searches are formulated based on our OOV analysis, aimed at maximizing the probability of retrieving OOVs’ sublexical translations from existing resources of Machine Translation (MT) systems. At run-time, translation candidates of the unknown words are generated from their suitable sublexical translations and ranked based on monolingual and bilingual information. We have incorporated the OOV model into a state-of-the-art machine translation system and experimental results show that our model indeed helps to ease the impact of OOVs on translation quality, especially for sentences containing more OOVs (significant improvement).

This paper introduces a method for the semi-automatic generation of grammar test items by applying Natural Language Processing (NLP) techniques. Based on manually-designed patterns, sentences gathered from the Web are transformed into tests on grammaticality. The method involves representing test writing knowledge as test patterns, acquiring authentic sentences on the Web, and applying generation strategies to transform sentences into items. At runtime, sentences are converted into two types of TOEFL-style question: multiple-choice and error detection. We also describe a prototype system FAST (Free Assessment of Structural Tests). Evaluation on a set of generated questions indicates that the proposed method performs satisfactory quality. Our methodology provides a promising approach and offers significant potential for computer assisted language learning and assessment.

This paper introduces a method for computational analysis of move structures in abstracts of research articles. In our approach, sentences in a given abstract are analyzed and labeled with a specific move in light of various rhetorical functions. The method involves automatically gathering a large number of abstracts from the Web and building a language model of abstract moves. We also present a prototype concordancer, CARE, which exploits the move-tagged abstracts for digital learning. This system provides a promising approach to Web-based computer-assisted academic writing.

Named entity (NE) extraction is one of the fundamental tasks in natural language processing (NLP). Although many studies have focused on identifying NEs within monolingual documents, aligning NEs in bilingual documents has not been investigated extensively due to the complexity of the task. In this article we introduce a new approach to aligning bilingual NEs in parallel corpora by incorporating statistical models with multiple knowledge sources. In our approach, we model the process of translating an English NE phrase into a Chinese equivalent using lexical translation/transliteration probabilities for word translation and alignment probabilities for word reordering. The method involves automatically learning phrase alignment and acquiring word translations from a bilingual phrase dictionary and parallel corpora, and automatically discovering transliteration transformations from a training set of name-transliteration pairs. The method also involves language-specific knowledge functions, including handling abbreviations, recognizing Chinese personal names, and expanding acronyms. At runtime, the proposed models are applied to each source NE in a pair of bilingual sentences to generate and evaluate the target NE candidates; the source and target NEs are then aligned based on the computed probabilities. Experimental results demonstrate that the proposed approach, which integrates statistical models with extra knowledge sources, is highly feasible and offers significant improvement in performance compared to our previous work, as well as the traditional approach of IBM Model 4.

This paper presents a new approach to bitext correspondence problem (BCP) of noisy bilingual corpora based on image processing (IP) techniques. By using one of several ways of estimating the lexical translation probability (LTP) between pairs of source and target words, we can turn a bitext into a discrete gray-level image. We contend that the BCP, when seen in the light, bears a striking resemblance to the line detection problem in IP. Therefore, BCPs, including sentence and word alignment, can benefit from a wealth of effective, well established IP techniques, including convolution-based filters, texture analysis and Hough transform. This paper describes a new program, PlotAlign that produces a word-level bitext map for noisy or non-literal bitext, based on these techniques.