Word Disambiguation Research Articles

Calculation of Sentence Semantic Similarity Based on Syntactic Structure

Combined with the problem of single direction of the solution of the existing sentence similarity algorithms, an algorithm for sentence semantic similarity based on syntactic structure was proposed. Firstly, analyze the sentence constituent, then through analysis convert sentence similarity into words similarity on the basis of syntactic structure, then convert words similarity into concept similarity through words disambiguation, and, finally, realize the semantic similarity comparison. It also gives the comparison rules in more detail for the modifier words in the sentence which also have certain contributions to the sentence. Under the same test condition, the experiments show that the proposed algorithm is more intuitive understanding of people and has higher accuracy.

Semi-Supervised Word Sense Disambiguation via Context Weighting

Word sense disambiguation as a central research topic in natural language processing can promote the development of many applications such as information retrieval, speech synthesis, machine translation, summarization and question answering. Previous approaches can be grouped into three categories: supervised, unsupervised and knowledge-based. The accuracy of supervised methods is the highest, but they suffer from knowledge acquisition bottleneck. Unsupervised method can avoid knowledge acquisition bottleneck, but its effect is not satisfactory. With the built-up of large-scale knowledge, knowledge-based approach has attracted more and more attention. This paper introduces a new context weighting method, and based on which proposes a novel semi-supervised approach for word sense disambiguation. The significant contribution of our method is that thesaurus and machine learning techniques are integrated in word sense disambiguation. Compared with the state of the art on the test data of the English all words disambiguation task in Sensaval-3, our method yields obvious improvements over existing methods in nouns, adjectives and verbs disambiguation.

Parallel corpora make sense

We present a multilingual approach to Word Sense Disambiguation (WSD), which automatically assigns the contextually appropriate sense to a given word. Instead of using a predefined monolingual sense-inventory, we use a language-independent framework by deriving the senses of a given word from word alignments on a multilingual parallel corpus, which we made available for corpus linguistics research. We built five WSD systems with English as the input language and translations in five supported languages (viz. French, Dutch, Italian, Spanish and German) as senses. The systems incorporate both binary translation features and local context features. The experimental results are very competitive, which confirms our initial hypothesis that each language contributes to the disambiguation of polysemous words. Because our system extracts all information from the parallel corpus, it offers a flexible language-independent approach, which implicitly deals with the sense distinctness issue and allows us to bypass the knowledge acquisition bottleneck for WSD.

Context Effects in the Processing of Phonolexical Ambiguity in L2

In order to comprehend speech, listeners have to combine low‐level phonetic information about the incoming auditory signal with higher‐order contextual information to make a lexical selection. This requires stable phonological categories and unambiguous representations of words in the mental lexicon. Unlike native speakers, second language (L2) speakers, who perceive nonnative sounds through the prism of their first language (L1), operate with fuzzy phonological categories, which lead to phonologically ambiguous lexical representations (e.g., the words rock and lock can be confused if phonological representations for /r/ and /l/ are not sufficiently robust). The present study uses the AX discrimination task to establish the degree of sensitivity of L2 listeners to the Russian hard/soft phonological contrast. The same phonological contrasts are then used in the stimuli for the second task—listening comprehension task with word identification—to mark semantic, syntactic, and morphological distinctions in words. The goal of the study is to examine the contributions and relative efficiency of different contextual constraints (semantic, syntactic, and morphological) to the resolution of phonolexical ambiguity in L2 auditory sentence processing. The results suggest that when L2 phonological contrasts present a discriminability problem and create phonolexical ambiguity, L2 listeners rely on morphological constraints for disambiguation of word forms and syntactic constraints for disambiguation of words belonging to different parts of speech to a greater extent than on semantic constraints for disambiguation of nouns in the same form.

Comparison of Association Ratio in English and Polish Languages

This paper compares the results of using association ratio for a positional language, English, and for an inflexional language, Polish. The main goal is proposing a way for automatic or semi-automatic creation of a context database that would serve as a training or reference data for word disambiguation algorithms. Association ratio proves to be useful for this purpose. The tests show that the performance of the association ratio does not depend on a specific language.

Natural Language Processing based Soft Computing Techniques

paper presents the implementation of soft computing (SC) techniques in the field of natural language processing. An attempt is made to design and implement an automatic tagger that extract a free text and then tag it. The part of speech taggers (POS) is the process of categorization words based on their meaning, functions and types (noun, verb, adjective, etc). Two stages tagging system based MPL, FRNN and SVM are implemented and designed. The system helps to classify words and assign the correct POS for each of them. The taggers are tested using two different languages (Arabic and Hindi). The Word disambiguation issue has been solved successfully for Arabic text. Experience has shown that the proposed taggers achieved a great accuracy (99%).

Lexicon structure and the disambiguation of novel words: Evidence from bilingual infants

In ambiguous word learning situations, infants can use systematic strategies to determine the referent of a novel word. One such heuristic is disambiguation. By age 16–18months, monolinguals infer that a novel noun refers to a novel object rather than a familiar one (Halberda, 2003), while at the same age bilinguals and trilinguals do not reliably show disambiguation (Byers-Heinlein & Werker, 2009; Houston-Price, Caloghiris, & Raviglione, 2010). It has been hypothesized that these results reflect a unique aspect of the bilingual lexicon: bilinguals often know many translation equivalents, cross-language synonyms such as English dog and Mandarin gǒu. We studied the role of vocabulary knowledge in the development of disambiguation by relating 17–18month-old English–Chinese bilingual infants’ performance on a disambiguation task to the percentage of translation equivalents in their comprehension vocabularies. Those bilingual infants who understood translation equivalents for more than half the words in their vocabularies did not show disambiguation, while infants who knew a smaller proportion of translation equivalents showed disambiguation just as same-aged monolinguals do. These results demonstrate that the structure of the developing lexicon plays a key role in infants’ use of disambiguation.

Unsupervised approximate-semantic vocabulary learning for human action and video classification

The paper presents a novel unsupervised contextual spectral (CSE) framework for human action and video classification. Similar to textual words, the visual word (a mid-level semantic) representation of an image or video contains a combination of synonymous words which give rise to the ambiguity of the representation. To narrow the semantic gap between visual words (mid-level semantic representation) and high-level semantics, we propose a high level representation called approximate-semantic descriptor. The experimental results show that the proposed approach for visual words disambiguation could improve the subsequent classification performance. In the paper, the approximate-semantic descriptor learning is formulated as a spectral clustering problem, such that semantically associated visual words are placed closely in low-dimensional semantic space and then clustered into one approximate-semantic descriptor. Specifically, the high level representation of human action videos is learnt by capturing the inter-video context of mid-level semantics via a non-parametric correlation measure. Experiments on four standard datasets demonstrate that our approach can achieve significantly improved results with respect to the state of the art, particularly for unconstrained environments.

Fast mapping, slow learning: Disambiguation of novel word–object mappings in relation to vocabulary learning at 18, 24, and 30 months

When hearing a novel name, children tend to select a novel object rather than a familiar one, a bias known as disambiguation. Using online processing measures with 18-, 24-, and 30-month-olds, we investigate how the development of this bias relates to word learning. Children’s proportion of looking time to a novel object after hearing a novel name related to their success in retention of the novel word, and also to their vocabulary size. However, skill in disambiguation and retention of novel words developed gradually: 18-month-olds did not show a reliable preference for the novel object after labeling; 24-month-olds reliably looked at a novel object on Disambiguation trials but showed no evidence of retention; and 30-month-olds succeeded on Disambiguation trials and showed only fragile evidence of retention. We conclude that the ability to find the referent of a novel word in ambiguous contexts is a skill that improves from 18 to 30months of age. Word learning is characterized as an incremental process that is related to – but not dependent on – the emergence of disambiguation biases.

Why Clowns Taste Funny: The Relationship between Humor and Semantic Ambiguity

What makes us laugh? One crucial component of many jokes is the disambiguation of words with multiple meanings. In this functional MRI study of normal participants, the neural mechanisms that underlie our experience of getting a joke that depends on the resolution of semantically ambiguous words were explored. Jokes that contained ambiguous words were compared with sentences that contained ambiguous words but were not funny, as well as to matched verbal jokes that did not depend on semantic ambiguity. The results confirm that both the left inferior temporal gyrus and left inferior frontal gyrus are involved in processing the semantic aspects of language comprehension, while a more widespread network that includes both of these regions and the temporoparietal junction bilaterally is involved in processing humorous verbal jokes when compared with matched nonhumorous material. In addition, hearing jokes was associated with increased activity in a network of subcortical regions, including the amygdala, the ventral striatum, and the midbrain, that have been implicated in experiencing positive reward. Moreover, activity in these regions correlated with the subjective ratings of funniness of the presented material. These results allow a more precise account of how the neural and cognitive processes that are involved in ambiguity resolution contribute to the appreciation of jokes that depend on semantic ambiguity.

Word sense disambiguation for event trigger word detection in biomedicine

This paper describes a method for detecting event trigger words in biomedical text based on a word sense disambiguation (WSD) approach. We first investigate the applicability of existing WSD techniques to trigger word disambiguation in the BioNLP 2009 shared task data, and find that we are able to outperform a traditional CRF-based approach for certain word types. On the basis of this finding, we combine the WSD approach with the CRF, and obtain significant improvements over the standalone CRF, gaining particularly in recall.

A New Intelligent Topic Extraction Model on Web

We tackle the problem of topic extraction on Web. In this paper, we propose an approach to implementing ontology-based data access in WordNet with the distinguishing feature of optimizing density-based clustering OPTICS algorithm (DBCO) to extract topics. Our solution has the following two desirable properties: i) it uses WordNet for word sense disambiguation of words in the learning resources documents and ii) it mapping the data space of the original method to a vector space of sentence, improving the original OPTICS algorithm. We outline the interface between our scheme and the current data Web, and show that, in contrast to the existing approaches, no exponential blowup is produced by the DBCO. Based on the experiments with a number of real-world data sets of 310 users in three study sites, we demonstrate that topic extraction in the proposed approach is efficient, especially for large-scale web learning resources. According to the user ratings data of four learning sites in the 150 days, the average rate of increase of user rating after the system is used reaches 25.18%.

Automatic web pages hierarchical classification using dynamic domain ontologies

The use of ontologies for knowledge representation has had a fast increase in the last years and they are used in several application context. One of these challenging applications is the web. Managing large amount of information on internet needs more efficient and effective methods and techniques for mining and representing information. In this article, we present a methodology for automatic topic annotation of web pages. We describe an algorithm for words disambiguation using an apposite metric for measuring the semantic relatedness and we show a technique which allows to detect the topic of the analysed document using ontologies extracted from a knowledge base. The strategy is implemented in a system where these information are used to build a topic hierarchy automatically created and not a priori defined for classifying web pages. Experimental results are presented and discussed in order to measure the effectiveness of our approach.

Cyber Swarm optimization for general keyboard arrangement problem

Traditional QWERTY keyboard has been known to be an ill design due to its low typing speed, fatigue-inducement, no reference to learnability, to name a few. Many researchers have put their effort on optimizing the character arrangement of the keyboard by taking into account the co-occurrence frequency of characters in words, typing ergonomics, and word disambiguation effectiveness. However, most of the existing works addressed the problem for a single objective instead of multiple ones, and the design of customized keyboards for motor-impaired users is often overlooked. In this paper, we propose a general mathematical model that integrates varying keyboard arrangement context. A Cyber Swarm method considering multiple objectives is proposed and it can create additional benefits that cannot be obtained by traditional methods. Our Cyber Swarm method accommodates ergonomic criteria and disambiguation/prediction effectiveness simultaneously. Experimental results manifest that the Cyber Swarm keyboard outperforms several benchmark keyboards and other competing algorithms. We also show an illustrative example for preliminary keyboard shape design which could be very useful in customized keyboard production for motor-impaired users whose physical capacity has been evaluated a priori. Finally, an empirical experiment involving human subjects is performed in order to analyze the feedbacks from humans’ experiences in using the new keyboard layouts. Relevance to industry Using the proposed general framework for keyboard arrangement, various scenarios (for example, single-character or multi-character keys, single-finger or multi-finger typing) raised in industrial production can be all taken into account without the hassle to change solution methods for different settings. Hence, the time and costs in industrial production for customized keyboard is significantly reduced.

Word sense disambiguation based on improved vector space model

To increase the word disambiguation accuracy,a word disambiguation solution based on improved Vector Space Model (VSM) was presented.Since the algorithm takes account of grammar,morphology and semantic and calculates the context similarity requiring the character vector abstraction,the algorithm is able to achieve better results by using collocation constraint.The open test precision can reach 80%.The result shows that the method can fully describe the features of context,and is beneficial to further semantic parsing.

Lexical and Affective Prosody in Children With High-Functioning Autism

To investigate the perception and production of lexical stress and processing of affective prosody in adolescents with high-functioning autism (HFA). We hypothesized preserved processing of lexical and affective prosody but atypical lexical prosody production. Sixteen children with HFA and 15 typically developing (TD) peers participated in 3 experiments that examined the following: (a) perception of affective prosody (Experiment 1), (b) lexical stress perception (Experiment 2), and (c) lexical stress production (Experiment 3). In Experiment 1, participants labeled sad, happy, and neutral spoken sentences that were low-pass filtered, to eliminate verbal content. In Experiment 2, participants disambiguated word meanings based on lexical stress (HOTdog vs. hot DOG). In Experiment 3, participants produced these words in a sentence completion task. Productions were analyzed with acoustic measures. Accuracy levels showed no group differences. Participants with HFA could determine affect from filtered sentences and disambiguate words on the basis of lexical stress. They produced appropriately differentiated lexical stress patterns but demonstrated atypically long productions, indicating reduced ability in natural prosody production. Children with HFA were as capable as their TD peers in receptive tasks of lexical stress and affective prosody. Prosody productions were atypically long, despite accurate differentiation of lexical stress patterns. Future research should use larger samples and spontaneous versus elicited productions.

Using Sense Clustering for the Disambiguation of Words (pp. 23-28)

Using Sense Clustering for the Disambiguation of Words (pp. 23-28)

Utilizing Clues in Syntactic Relationship for Automatic Target Word Sense Disambiguation

Multiple translations to the target language are due to several meanings of source words and various target word equivalents, depending on the context of the source word. Thus, an automated approach is presented for resolving target-word selection, based on word-to-sense and sense-to-word relationship between source words and their translations, using syntactic relationships (subject-verb, verb-object, adjective-noun). Translation selection proceeds from sense disambiguation of source words based on knowledge from a bilingual dictionary and word similarity measures from WordNet, and selection of a target word using statistics from a target language corpus. Test results using English to Tagalog translations show an overall 64% accuracy for selecting word translation with a standardized precision of at least 80% for generating expected translations using 200 sentences with ambiguous words (an average of 4 senses) in three categories: nouns, verbs, and adjectives. This system is tested on 145,746 word pairs in syntactic relationships that are extracted from target corpora (with 317,113 words). Sense profile, with 2681 entries for source words, is built from an existing bilingual dictionary that includes clues for disambiguation and target translations. The results show an improvement on the performance of the method with the utilization of syntactic information in resolving target-word ambiguity. Further improvements include the integration of other content words and syntactic categories, the addition of reliable clues for sense disambiguation, and the integration of smoothing techniques.

A WSD Model for Corpus Construction

A common difficulty in the annotation of large corpora is the disambiguation of the meanings of polysemous words, a task which currently depends on unassisted manual checking by human experts. In this paper, we propose RFR-SUM, an approach to the automatic disambiguation of polysemous Chinese words that calculates and then sums the collocational strengths of words in the local context of a given word. The approach is efficient in that it learns from a small annotated corpus. It can also be used to support manual checking by automatically identifying sentences where usages are borderline and require a human decision. Testing has shown that RFR-SUM outperforms two other commonly used WSD approaches.

Using Mutual Information Criterion to Design an Efficient Phoneme Set for Chinese Speech Recognition

Chinese is a representative tonal language, and it has been an attractive topic of how to process tone information in the state-of-the-art large vocabulary speech recognition system. This paper presents a novel way to derive an efficient phoneme set of tone-dependent units to build a recognition system, by iteratively merging a pair of tone-dependent units according to the principle of minimal loss of the Mutual Information (MI). The mutual information is measured between the word tokens and their phoneme transcriptions in a training text corpus, based on the system lexical and language model. The approach has a capability to keep discriminative tonal (and phoneme) contrasts that are most helpful for disambiguating homophone words due to lack of tones, and merge those tonal (and phoneme) contrasts that are not important for word disambiguation for the recognition task. This enables a flexible selection of phoneme set according to a balance between the MI information amount and the number of phonemes. We applied the method to traditional phoneme set of Initial/Finals, and derived several phoneme sets with different number of units. Speech recognition experiments using the derived sets showed its effectiveness.