In this study, we introduce and evaluate a novel extractive text summarization methodology, “SummarEyes,” based on the visual interaction of the user with the text, using eye-tracking data, as opposed to the traditional approaches based on analysis of textual content only. We conducted a large-scale user study aiming to collect eye-tracking data while reading the text to be summarized. We utilized various user’s implicit attention metrics to generate novel eye-tracking-based text summarization models and compared them both to eye-tracking models typically using only a single feature of the gaze duration and to traditional, as well as state-of-the-art summarization methods, based solely on textual features. The models’ quality was evaluated in terms of ROUGE scores using intrinsic evaluation on the datasets we had generated, relating gaze behavior to personalized and DUC gold-standard summaries. The experimental results showed that “SummarEyes” significantly outperformed the other summarizers in predicting both the user’s personalized summarization and the generic gold standard summaries. With the increasing availability of eye-tracking technology, this research can lead to a new generation of effective user-centric text summarization tools.

Yonatan Bitton, Gabriel Stanovsky, Roy Schwartz, Michael Elhadad. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021.

We present a semantic role labeling resource for Hebrew built semi-automatically through annotation projection from English. This corpus is derived from the multilingual OpenSubtitles dataset and includes short informal sentences, for which reliable linguistic annotations have been computed. We provide a fully annotated version of the data including morphological analysis, dependency syntax and semantic role labeling in both FrameNet and ProbBank styles. Sentences are aligned between English and Hebrew, both sides include full annotations and the explicit mapping from the English arguments to the Hebrew ones. We train a neural SRL model on this Hebrew resource exploiting the pre-trained multilingual BERT transformer model, and provide the first available baseline model for Hebrew SRL as a reference point. The code we provide is generic and can be adapted to other languages to bootstrap SRL resources.

OBJECTIVE: In Hebrew online health communities, participants commonly write medical terms that appear as transliterated forms of a source term in English. Such transliterations introduce high variability in text and challenge text-analytics methods. To reduce their variability, medical terms must be normalized, such as linking them to Unified Medical Language System (UMLS) concepts. We present a method to identify both transliterated and translated Hebrew medical terms and link them with UMLS entities. MATERIALS AND METHODS: We investigate the effect of linking terms in Camoni, a popular Israeli online health community in Hebrew. Our method, MDTEL (Medical Deep Transliteration Entity Linking), includes (1) an attention-based recurrent neural network encoder-decoder to transliterate words and mapping UMLS from English to Hebrew, (2) an unsupervised method for creating a transliteration dataset in any language without manually labeled data, and (3) an efficient way to identify and link medical entities in the Hebrew corpus to UMLS concepts, by producing a high-recall list of candidate medical terms in the corpus, and then filtering the candidates to relevant medical terms. RESULTS: We carry out experiments on 3 disease-specific communities: diabetes, multiple sclerosis, and depression. MDTEL tagging and normalizing on Camoni posts achieved 99% accuracy, 92% recall, and 87% precision. When tagging and normalizing terms in queries from the Camoni search logs, UMLS-normalized queries improved search results in 46% of the cases. CONCLUSIONS: Cross-lingual UMLS entity linking from Hebrew is possible and improves search performance across communities. Annotated datasets, annotation guidelines, and code are made available online (https://github.com/yonatanbitton/mdtel).

Query-Focused Summarization (QFS) summarizes a document cluster in response to a specific input query. QFS algorithms must combine query relevance assessment, central content identification, and redundancy avoidance. Frustratingly, state of the art algorithms designed for QFS do not significantly improve upon generic summarization methods, which ignore query relevance, when evaluated on traditional QFS datasets. We hypothesize this lack of success stems from the nature of the dataset. We define a task-based method to quantify topic concentration in datasets, i.e., the ratio of sentences within the dataset that are relevant to the query, and observe that the DUC 2005, 2006 and 2007 datasets suffer from very high topic concentration. We introduce TD-QFS, a new QFS dataset with controlled levels of topic concentration. We compare competitive baseline algorithms on TD-QFS and report strong improvement in ROUGE performance for algorithms that properly model query relevance as opposed to generic summarizers. We further present three new and simple QFS algorithms, RelSum, ThresholdSum, and TFIDF-KLSum that outperform state of the art QFS algorithms on the TD-QFS dataset by a large margin.

Update summarization is a form of multidocument summarization where a document set must be summarized in the context of other documents assumed to be known.Efficient update summarization must focus on identifying new information and avoiding repetition of known information.In Query-focused summarization, the task is to produce a summary as an answer to a given query.We introduce a new task, Query-Chain Summarization, which combines aspects of the two previous tasks: starting from a given document set, increasingly specific queries are considered, and a new summary is produced at each step.This process models exploratory search: a user explores a new topic by submitting a sequence of queries, inspecting a summary of the result set and phrasing a new query at each step.We present a novel dataset comprising 22 querychains sessions of length up to 3 with 3 matching human summaries each in the consumerhealth domain.Our analysis demonstrates that summaries produced in the context of such exploratory process are different from informative summaries.We present an algorithm for Query-Chain Summarization based on a new LDA topic model variant.Evaluation indicates the algorithm improves on strong baselines.

BACKGROUND: The increasing availability of Electronic Health Record (EHR) data and specifically free-text patient notes presents opportunities for phenotype extraction. Text-mining methods in particular can help disease modeling by mapping named-entities mentions to terminologies and clustering semantically related terms. EHR corpora, however, exhibit specific statistical and linguistic characteristics when compared with corpora in the biomedical literature domain. We focus on copy-and-paste redundancy: clinicians typically copy and paste information from previous notes when documenting a current patient encounter. Thus, within a longitudinal patient record, one expects to observe heavy redundancy. In this paper, we ask three research questions: (i) How can redundancy be quantified in large-scale text corpora? (ii) Conventional wisdom is that larger corpora yield better results in text mining. But how does the observed EHR redundancy affect text mining? Does such redundancy introduce a bias that distorts learned models? Or does the redundancy introduce benefits by highlighting stable and important subsets of the corpus? (iii) How can one mitigate the impact of redundancy on text mining? RESULTS: We analyze a large-scale EHR corpus and quantify redundancy both in terms of word and semantic concept repetition. We observe redundancy levels of about 30% and non-standard distribution of both words and concepts. We measure the impact of redundancy on two standard text-mining applications: collocation identification and topic modeling. We compare the results of these methods on synthetic data with controlled levels of redundancy and observe significant performance variation. Finally, we compare two mitigation strategies to avoid redundancy-induced bias: (i) a baseline strategy, keeping only the last note for each patient in the corpus; (ii) removing redundant notes with an efficient fingerprinting-based algorithm. (a)For text mining, preprocessing the EHR corpus with fingerprinting yields significantly better results. CONCLUSIONS: Before applying text-mining techniques, one must pay careful attention to the structure of the analyzed corpora. While the importance of data cleaning has been known for low-level text characteristics (e.g., encoding and spelling), high-level and difficult-to-quantify corpus characteristics, such as naturally occurring redundancy, can also hurt text mining. Fingerprinting enables text-mining techniques to leverage available data in the EHR corpus, while avoiding the bias introduced by redundancy.

We present a constituency parsing system for Modern Hebrew. The system is based on the PCFG-LA parsing method of Petrov et al. 2006 , which is extended in various ways in order to accommodate the specificities of Hebrew as a morphologically rich language with a small treebank. We show that parsing performance can be enhanced by utilizing a language resource external to the treebank, specifically, a lexicon-based morphological analyzer. We present a computational model of interfacing the external lexicon and a treebank-based parser, also in the common case where the lexicon and the treebank follow different annotation schemes. We show that Hebrew word-segmentation and constituency-parsing can be performed jointly using CKY lattice parsing. Performing the tasks jointly is effective, and substantially outperforms a pipeline-based model. We suggest modeling grammatical agreement in a constituency-based parser as a filter mechanism that is orthogonal to the grammar, and present a concrete implementation of the method. Although the constituency parser does not make many agreement mistakes to begin with, the filter mechanism is effective in fixing the agreement mistakes that the parser does make. These contributions extend outside of the scope of Hebrew processing, and are of general applicability to the NLP community. Hebrew is a specific case of a morphologically rich language, and ideas presented in this work are useful also for processing other languages, including English. The lattice-based parsing methodology is useful in any case where the input is uncertain. Extending the lexical coverage of a treebank-derived parser using an external lexicon is relevant for any language with a small treebank.

BACKGROUND: The OMIM database is a tool used daily by geneticists. Syndrome pages include a Clinical Synopsis section containing a list of known phenotypes comprising a clinical syndrome. The phenotypes are in free text and different phrases are often used to describe the same phenotype, the differences originating in spelling variations or typing errors, varying sentence structures and terminological variants.These variations hinder searching for syndromes or using the large amount of phenotypic information for research purposes. In addition, negation forms also create false positives when searching the textual description of phenotypes and induce noise in text mining applications. DESCRIPTION: Our method allows efficient and complete search of OMIM phenotypes as well as improved data-mining of the OMIM phenome. Applying natural language processing, each phrase is tagged with additional semantic information using UMLS and MESH. Using a grammar based method, annotated phrases are clustered into groups denoting similar phenotypes. These groups of synonymous expressions enable precise search, as query terms can be matched with the many variations that appear in OMIM, while avoiding over-matching expressions that include the query term in a negative context. On the basis of these clusters, we computed pair-wise similarity among syndromes in OMIM. Using this new similarity measure, we identified 79,770 new connections between syndromes, an average of 16 new connections per syndrome. Our project is Web-based and available at http://fohs.bgu.ac.il/s2g/csiomim CONCLUSIONS: The resulting enhanced search functionality provides clinicians with an efficient tool for diagnosis. This search application is also used for finding similar syndromes for the candidate gene prioritization tool S2G.The enhanced OMIM database we produced can be further used for bioinformatics purposes such as linking phenotypes and genes based on syndrome similarities and the known genes in Morbidmap.

This book comes with "batteries included" (a reference to the phrase often used to explain the popularity of the Python programming language).It is the companion book to an impressive open-source software library called the Natural Language Toolkit (NLTK), written in Python.NLTK combines language processing tools (tokenizers, stemmers, taggers, syntactic parsers, semantic analyzers) and standard data sets (corpora and tools to access the corpora in an efficient and uniform manner).Although the book builds on the NLTK library, it covers only a relatively small part of what can be done with it.The combination of the book with NLTK, a growing system of carefully designed, maintained, and documented code libraries, is an extraordinary resource that will dramatically influence the way computational linguistics is taught.The book attempts to cater to a large audience: It is a textbook on computational linguistics for science and engineering students; it also serves as practical documentation for the NLTK library, and it finally attempts to provide an introduction to programming and algorithm design for humanities students.I have used the book and its earlier on-line versions to teach advanced undergraduate and graduate students in computer science in the past eight years.The book adopts the following approach: r It is first a practical approach to computational linguistics.It provides readers with practical skills to solve concrete tasks related to language.r It is a hands-on programming text: The ultimate goal of the book is to empower students to write programs that manipulate textual data and perform empirical experiments on large corpora.Importantly, NLTK includes a large set of corpora-this is one of the most useful and game-changing contributions of the toolkit.r It is principled: It exposes the theoretical underpinnings-both computational and linguistic-of the algorithms and techniques that are introduced.r It attempts to strike a pragmatic balance between theory and applications.The goal is to introduce "just enough theory" to fit in a single semester course for advanced undergraduates, while still leaving room for practical programming and experimentation.r It aims to make working with language pleasurable.

Morphological disambiguation is the process of assigning one set of morphological features to each individual word in a text. When the word is ambiguous (there are several possible analyses for the word), a disambiguation procedure based on the word context must be applied. This paper deals with morphological disambiguation of the Hebrew language, which combines morphemes into a word in both agglutinative and fusional ways. We present an un-supervised stochastic model - the only resource we use is a morphological analyzer-which deals with the data sparseness problem caused by the affixational morphology of the Hebrew language.We present a text encoding method for languages with affixational morphology in which the knowledge of word formation rules (which are quite restricted in Hebrew) helps in the disambiguation. We adapt HMM algorithms for learning and searching this text representation, in such a way that segmentation and tagging can be learned in parallel in one step. Results on a large scale evaluation indicate that this learning improves disambiguation for complex tag sets. Our method is applicable to other languages with affix morphology.

We present a method for Noun Phrase chunking in Hebrew. We show that the traditional definition of base-NPs as non-recursive noun phrases does not apply in Hebrew, and propose an alternative definition of Simple NPs. We review syntactic properties of Hebrew related to noun phrases, which indicate that the task of Hebrew SimpleNP chunking is harder than base-NP chunking in English. As a confirmation, we apply methods known to work well for English to Hebrew data. These methods give low results (F from 76 to 86) in Hebrew. We then discuss our method, which applies SVM induction over lexical and morphological features. Morphological features improve the average precision by ~0.5%, recall by ~1%, and F-measure by ~0.75, resulting in a system with average performance of 93% precision, 93.4% recall and 93.2 F-measure.

End-user computing is needed in creative artistic applications or integrated editing environments, where the activity cannot be planned in advance. Following the paper by Orlarey et al., concrete abstractions (abstractions from examples) are suggested as a new mode for function definition, appropriate for end-user editor programmability. For certain applications, the direct, associative, not planned in advance character of concrete abstraction plays a qualitative role in the mere ability to specify abstractions. In this paper, we propose to use concrete abstraction as a general tool for end-user programmability in editors. We distinguish two kinds of abstractions: value abstraction and structure abstraction, and explain how they can be combined. We describe a framework of historical editing that is based on a double view, in which the two abstraction kinds are combined. Finally, BOOMS, an implemented prototype for such an editing framework, is described. BOOMS is a domain-independent toolkit, with three sample instantiations. We believe that the proposed framework captures the conceptualization operation that characterizes creative, associative work types and addresses the needs for end-user computing in integrated environments.

This paper presents the integration of a large-scale, reusable lexicon for generation with the FUF/SURGE unification-based syntactic realizer. The lexicon was combined from multiple existing resources in a semi-automatic process. The integration is a multi-step unification process. This integration allows the reuse of lexical, syntactic, and semantic knowledge encoded in the lexicon in the development of lexical chooser module in a generation system. The lexicon also brings other benefits to a generation system: for example, the ability to generate many lexical and syntactic paraphrases and the ability to avoid non-grammatical output.

We present a method to automatically generate a concise summary by identifying and synthesizing similar elements across related text from a set of multiple documents. Our approach is unique in its usage of language generation to reformulate the wording of the summary.

Syntactic realization grammars have traditionally attempted to accept inputs with the highest possible level of abstraction, in order to facilitate the work of the components (sentence planner) preparing the input.Recently, the search for higher abstraction has been, however, challenged (E1hadad and Robin, 1996)(Lavoie and Rambow, 1997)(Busemann andHoracek, 1998).In this paper, we contribute to the issue of selecting the "ideal" abstraction level in the input to syntactic realization grammar by considering the case of partitives and possessives in a bilingual Hebrew-English generation grammar.In the case of bilingual generation, the ultimate goal is to provide a single input structure, where only the openclass lexical entries are specific to the language.In that case, the minimal abstraction required must cover the different syntactic constraints of the two languages.We present a contrastive analysis of the syntactic realizations of possessives and partitives in Hebrew and English and conclude by presenting an input specification for complex NPs which is slightly more abstract than the one used in SURGE.We define two main features -possessor and rejLset, and • discuss how the grammar handles complex syntactic co-occurrence phenomena based on this input.We conclude by evaluating how the resulting input specification language is appropriate for both languages.

Functional Unification Grammars (FUGs) are popular for natural language applications because the formalism uses very few primitives and is uniform and expressive. In our work on text generation, we have found that it also has annoying limitations: it is not suited for the expression of simple, yet very common, taxonomic relations and it does not allow the specification of completeness conditions. We have implemented an extension of traditional functional unification. This extension addresses these limitations while preserving the desirable properties of FUGs. It is based on the notions of typed features and typed constituents. We show the advantages of this extension in the context of a grammar used for text generation.